Full text of Economic Policy Review (Federal Reserve Bank of New York) : Volume 11, Number 2 : Urban Dynamics in New York City

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Erica L. Groshen and Giorgio Topa

Conference Overview
and Summary of Papers
1. Introduction

2. Spatial Dynamics and Growth

O

ver the past two decades, researchers and practitioners
alike have increasingly focused their attention on
cities. This attention arises for a variety of reasons. Urban
agglomerations can be seen as laboratories for studying the
mechanisms of sustained economic growth, the dynamics of
economic activities, and the trajectories of immigration flows.
By the same token, cities are also viewed as volatile and fragile
organisms that can rise and decline dramatically over a short
time span. New York City, in particular, has weathered longrun adverse trends as well as sudden unanticipated shocks.
To promote the discussion of these important processes,
in April 2005 the Federal Reserve Bank of New York organized
a conference on “Urban Dynamics in New York City.” The goal
of the conference was threefold: to examine the historical
transformations of the engine-of-growth industries in New
York and distill the main determinants of the city’s historical
dominance as well as the challenges to its continued success;
to study the nature and evolution of immigration flows into
New York; and to analyze recent trends in a range of socioeconomic outcomes, both for the general population and
recent immigrants more specifically.

New York City has demonstrated remarkable growth over
the past four centuries. Edward L. Glaeser offers an in-depth
historical account of the major contributors to the city’s
economic dominance over such a long period. The first of the
three central themes identified by Glaeser is the importance of
geography in determining New York’s early success. The city
enjoyed a natural advantage provided by its port and by its
proximity to the Hudson River and a water-borne connection
to the Great Lakes. The second theme is the value of simple
transportation cost and scale economies. The rise of
manufacturing in the city, observes Glaeser, hinged on New
York’s place at the center of a large transport hub and the
benefits afforded by that prime location. Lastly, the author
describes the city’s clear advantage in facilitating information
flows and face-to-face interactions. The fast and convenient
dissemination of knowledge, for example, has been essential to
the success of information-intensive industries such as
finance—the undisputed engine of growth in New York’s more
recent history.
The discussion by J. Vernon Henderson complements
Glaeser by emphasizing two other themes that have been
instrumental in the city’s success. One is the role played by
New York’s vibrant ethnic neighborhoods in providing

Erica L. Groshen is an assistant vice president and Giorgio Topa
a senior economist at the Federal Reserve Bank of New York.
<erica.groshen@ny.frb.org>
<giorgio.topa@ny.frb.org>

The views summarized are those of the presenters and do not necessarily
reflect the position of the Federal Reserve Bank of New York or the
Federal Reserve System.

FRBNY Economic Policy Review / December 2005

1

networks of contacts to new immigrants. These networks have
made it easier and more desirable for immigrants to stay in
New York. The other is the importance of the knowledge
spillovers that arise from the city’s dense centers of commercial
activities. Knowledge spillovers are vital not only to the health
of finance, notes Henderson, but also to the health of other
innovative New York City industries, such as fashion,
advertising, and the arts.
Another key aspect of New York’s dynamism is the city’s
entrepreneurs. Stuart S. Rosenthal and William C. Strange
analyze the geography of entrepreneurship in the New York
metro area to uncover its determinants. They find that births
of new establishments and the number of jobs in new
establishments—their measures of entrepreneurial activity—
are positively affected by the density of local employment and
even more so by the amount of local employment in the
entrepreneur’s own industry. Interestingly, the most powerful
effects are for the smallest distances—within a city block or so.
The results obtained by Rosenthal and Strange provide fresh
evidence on the importance of very local agglomeration
economies to sustained growth.
Robert Inman’s commentary argues that the very local
nature of the agglomeration economies identified by the
Rosenthal-Strange analysis suggests that economic
development policies can be locally designed and, more
significantly, locally funded. Countrywide or statewide
policies, according to Inman, should then be limited to projects
that have clear effects on multiple communities.
To advance the understanding of the dynamics of city
growth, Andrew F. Haughwout and Bess Rabin examine the
response of New York City’s economy to an exogenous,
unanticipated, and large—yet localized—shock. Specifically,
they study the response in terms of the spatial distribution of
activities following the September 11 terrorist attacks. The
authors identify several patterns: long-run demand for city
locations relative to locations elsewhere in the country was
hardly affected; after a temporary weakening, long-run
demand for residential space in Lower Manhattan
strengthened; and both short- and long-run demand for office
space weakened in Lower Manhattan while it strengthened in
Midtown. Haughwout and Rabin conclude that the city’s
economy was remarkably resilient to the shock, and that the
shock itself only accelerated a preexisting trend that was
making Lower Manhattan a mixed-use community as offices
gravitated toward Midtown, to be replaced by residences
and shops. They also suggest that government activities and
announcements can serve as valuable coordination tools in
the presence of agglomeration economies.

2

Conference Overview and Summary of Papers

An alternative and complementary explanation for the
attacks’ relatively minor impact on the city economy is put
forth in the remarks by Stephen L. Ross. The shock was small
compared with the total stock of commercial real estate in the
New York metro area, Ross argues. Furthermore, the relatively
high mobility of workers and firms throughout the area
enabled the shock to be absorbed fairly quickly.

3. The Making of a World Metropolis
In the sessions’ keynote address, Kenneth T. Jackson offers his
insight into the characteristics that continue to make New York
a unique and vibrant city. He observes that New York is very
different from other American cities in the sense that wealth is
concentrated in its center, Manhattan, rather than in its
suburbs; its population density is several times that of most
U.S. cities; and the density is increasing rather than declining
over time. Another unique characteristic of New York is its
openness to newcomers, whether they take the form of new
ideas, new communities, or new religious groups. The constant
inflow of innovations embodied by newcomers, explains
Jackson, has enabled the city to reinvent itself amid such
economic challenges as the decline of its port and of
manufacturing in general. Jackson adds that a long history
of diversity has made New York a haven for dissent and
tolerance—a characteristic that he views as one of the city’s
fundamental strengths.

4. Immigration
The nature and evolution of immigration flows into the
New York metro area offer myriad avenues of research.
George J. Borjas focuses on immigration trends from 1970 to
2000, characterizing the skill levels and earnings of immigrant
workers in the New York area relative to those of immigrants
who settle elsewhere in the United States and to those of native
New Yorkers. He finds that in terms of educational attainment
over the thirty-year period, skill levels increased more for
native- and foreign-born workers in the New York metro area
than for their counterparts elsewhere in the country. Over the
same period, though, the skill gap between New York native
and immigrant workers has widened. Wages reflect the same
pattern: immigrant wages have risen in New York relative
to other areas of the country, but they have fallen relative to

those of New York natives. Borjas’ results also reveal that
immigrants in New York are substantially more skilled than
immigrants in Los Angeles or Miami.
The immigrant population in New York is remarkably
diverse relative to other immigrant populations in the United
States. Stephen J. Trejo, in his commentary, suggests that a
large share of the skill differential between immigrants in New
York and those elsewhere can be explained by differences in
national origins. He places Borjas’ findings in the larger context
of optimal immigration policy, touching upon questions of the
optimal skill mix of immigrants to the United States as well as
the spatial distribution of immigrants within the country.
Focusing on the socioeconomic achievements of secondgeneration immigrants, John Mollenkopf sheds light on the
intergenerational trajectories of immigrant groups, linking the
experiences of U.S.-born children of immigrants to those of
their parents. He paints a varied picture. Children of South
American, Dominican, and West Indian immigrant families
fare slightly better on a range of outcome measures than do
children growing up in very similar native Puerto Rican or
African American families. Moreover, second-generation
Chinese and Russians have made extraordinary educational
progress vis-à-vis their parental backgrounds. These two
groups in fact have outdistanced the native white children who
grew up and stayed in New York, even after the author controls
for parental background. Mollenkopf’s findings suggest that
intergenerational transmission strategies interact with
perceptions about race and neighborhood conditions in
complex ways when determining second-generation
immigrant trajectories.
A reductive view of segmented immigrant assimilation
revolving only around race and ethnicity warrants caution,
observes Douglas S. Massey. His comments on Mollenkopf
identify a variety of factors that can also play important roles in
shaping intergenerational trajectories. Massey points to the
original motivation for migration, the immigrant’s legal status,
and the characteristics of the residential location in which the
immigrant family resides as the most notable factors.

5. Socioeconomic Outcomes
The relationship between immigration and health outcomes
motivates the work by Guillermina Jasso, Douglas S. Massey,
Mark R. Rosenzweig, and James P. Smith. The authors employ

a novel data set on new legal immigrants to the United States to
study health trajectory from the beginning of the immigration
process and continuing after arrival in the United States. This
approach enables the authors to identify three distinct sources
of health change: visa stress, migration stress, and U.S.
exposure. Jasso et al. find that the combined effects on health
outcomes of visa stress and migration stress are negative, while
the pure effect of U.S. exposure is positive, especially for men.
Weight measures are found to increase with time in the
country, suggesting a role for environmental and dietary
influences. In addition, the study finds that immigrants in New
York tend to be healthier on arrival relative to immigrants who
settle elsewhere and that their subsequent trajectories do not
differ significantly from those of other immigrants.
Adriana Lleras-Muney discusses biases that could affect
the Jasso et al. analysis, including cultural differences across
countries of origin and recollection bias. Should one, she asks,
provide special health services to particular immigrant groups
during the immigration process? Can one disentangle the
impact of changes in job and earnings upon arrival from that
of environmental conditions? As these questions suggest,
Lleras-Muney argues that the authors’ findings must be viewed
in the broader context of immigration and health policy.
Pursuing a different line of inquiry, Amy Ellen Schwartz and
Leanna Stiefel offer a rich portrait of changing educational
outcomes and public education in New York City. One of their
most striking results is that children of immigrants tend to
perform better than native children on several standardized
tests, despite their less favorable initial background. Moreover,
this “immigrant advantage” tends to increase in higher grades.
Their finding that immigrant students of Russian or Chinese
descent perform especially well is consistent with Mollenkopf’s
results. Furthermore, Schwartz and Stiefel conclude that
several recent reforms to the New York City public school
system—aimed at, among other things, improving resource
allocation and opening new and smaller schools—have had
slightly positive effects on the test scores of immigrant and
native children alike.
Dalton Conley adds a few cautionary notes to the SchwartzStiefel paper. A study of the peer effects of immigrants on
native-born students, he contends, would be useful for gaining
a better understanding of the overall impact of immigrant
students on the New York City public school system. Attrition
out of the system could bias the “immigrant advantage” results.
With respect to the effects of school reform, Conley observes
that such reforms could be endogenous to school quality.

FRBNY Economic Policy Review / December 2005

3

6. Conclusion
How does a large urban agglomeration such as New York City
survive, even thrive, in an ever-changing environment? How
does this dynamic affect a city’s population and institutions?
The papers and discussions from this conference consider these
two fundamental questions from a variety of perspectives.
A central theme that emerges is the importance of “openness,”
both to new ideas and to newcomers. A degree of openness

and the cross-fertilization it allows seem essential to ensuring
a city’s ability to reinvent itself in the face of adverse circumstances. With this openness, however, come challenges,
including the need for institutions to coordinate individual
actions and integrate newcomers in a productive way.
Challenges like this and the ways in which cities meet them
will no doubt command the attention of future researchers
on urban dynamics.

The views summarized are those of the presenters and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
4

Conference Overview and Summary of Papers

Edward L. Glaeser

Urban Colossus: Why Is
New York America’s
Largest City?
1. Introduction

F

or 200 years, New York City has been the largest city in the
nation, and it continues to outperform most cities that
were once its competitors. In the 1990s, the city’s population
grew by 9 percent and finally passed the eight-million mark.
New York is the only one of the sixteen largest cities in the
northeastern or midwestern United States with a larger
population today than it had fifty years ago. Its economy
remains robust. Payroll per employee is more than $80,000 per
year in Manhattan’s largest industry and almost $200,000 per
year in its second-largest industry.
All cities, even New York, go through periods of crisis and
seeming rebirth, and New York certainly went through a real
crisis in the 1970s. However, while the dark periods for Boston,
Chicago, or Washington, D.C., lasted for thirty or fifty years,
New York’s worst period lasted for less than a decade. While
Boston’s history is one of ongoing crises and reinvention
(Glaeser 2005), New York’s is one of almost unbroken
triumph. The remarkable thing about New York is its ability to
thrive despite the massive technological changes that
challenged every other dense city built around public
transportation.
What explains New York’s ongoing ability to dominate
America’s urban landscape? In this paper, we explore the
economic history of the city and argue that three themes
emerge. First, New York’s emergence as the nation’s premier

Edward L. Glaeser is a professor of economics at Harvard University and
director of Harvard’s A. Alfred Taubman Center for State and Local
Government.
<eglaeser@harvard.edu>

port was not the result of happenstance followed by lemminglike agglomeration. While there are limits to geographic
determinism, the clear superiority of New York’s port in terms
of its initial depth, the Hudson River and its location, and the
other advantages provided by the water-borne connection to
the Great Lakes ensured that this port would be America’s port.
In this case, geography really was destiny, and the significance
of trade and immigration to the early republic ensured that
New York would dominate.
The second theme to emerge from New York’s history is the
importance of simple transportation cost and scale economies.
The rise of the city’s three great manufacturing industries in the
nineteenth century—sugar refining, publishing, and the
garment trade—depended on New York’s place at the center of
a transport hub. In all three industries, manufacturing
transformed products from outside the United States into
finished goods to be sold within the country. Because New
York was a hub and products were dispersed throughout the
country and the world after entry into that hub, it made perfect
sense to perform the manufacturing in the city.
The tendency of people to attract more people is the central
idea of urban economics, and nowhere is that idea more
obvious than in America’s largest city. New York’s initial
advantage as a port then attracted manufacturing and services
to cater to the mercantile firms and to take advantage of their
low shipping costs. The traditional model of this phenomenon
(Krugman 1991) emphasizes that scale matters because it

The author thanks the Taubman Center for State and Local Government.
Joshua Samuelson provided excellent research assistance. Stanley Engerman
provided guidance on sugar. The views expressed are those of the author and
do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System.

FRBNY Economic Policy Review / December 2005

7

allows manufacturers to save on the costs of supplying goods to
residents of the city. But the history of New York suggests that
this phenomenon was less important than the advantage of
producing in a central location for export elsewhere.
Obviously, scale economies were also important; otherwise,
there would be no incentive to centralize manufacturing.
New York’s growth in the early nineteenth century was
driven by the rise of manufacturing in the city, which itself
depended on New York’s primacy as a port. New York’s growth
in the late nineteenth century owed at least as much to its role
as the entryway for immigrants into the United States. Indeed,
the basic industrial structure of New York remained
remarkably consistent between 1860 and 1910 while the scale
increased enormously. Immigrants stayed in New York in port
for “consumption” reasons. Ethnic neighborhoods made the
transition to the New World easier, and New York as a city
acquired over time a remarkable capacity to cater to immigrant
needs. However, immigrants also stayed because the traditional
New York industries, especially the garment trade, were able to
increase in scale to accommodate extra labor without a huge
drop in wages.
In the mid-twentieth century, a large number of
technological changes challenged cities throughout the United
States. Declining transport costs reduced the advantages of
access to waterways. The air conditioner helped move citizens
west and south. The automobile and the truck enabled the
population to disperse from city centers to outlying areas.
Almost all of America’s biggest cities declined—sometimes
precipitously—over the past fifty years in response to the

shock. Eight of the ten largest U.S. cities in 1930 have a smaller
population today than they did then (Table 1). New York and
Los Angeles are the exceptions.
New York’s remarkable survival is a result of its dominance
in the fields of finance, business services, and corporate
management. Forty years ago, Chinitz (1961) described New
York as a model of diversity in comparison with industrial
Pittsburgh. New York in 2005 does not look nearly as diverse.
Today, 28 percent of Manhattan’s payroll goes to workers in a
single three-digit industry; 56 percent goes to workers in four
three-digit industries. New York’s twentieth-century success
primarily reflects an ability to attract and retain a single
industry, and the city’s future appears to be linked to a
continuing ability to hold that industry.
The attraction of finance and business services to New York
reflects the city’s advantages in facilitating face-to-face contact
and the spread of information. Transportation costs for goods
have declined by 95 percent over the twentieth century (Glaeser
and Kohlhase 2004), but there has been no comparable
reduction in the cost of moving people. After all, the primary
cost involved in the movement of people is the opportunity
cost of time, which rises with wages. For this reason, cities,
which represent the elimination of physical distance between
people, still excel in delivering services. In addition, as the
demand for timely information rises, the proximity that
facilitates the flow of that information continues to be critical.
The success of finance and business services on the island of
Manhattan hinges critically on the advantage that the island has
in bringing people together and speeding the flow of knowledge.

Table 1

Growth in Top Ten U.S. Cities by 1930 Population
Percentage Growth in Population
Population in
1930

1950-60

1960-70

1970-80

1980-90

1990-2000

Population in
2000

New York
Chicago
Philadelphia
Detroit
Los Angeles
Cleveland
St. Louis
Baltimore
Boston
Pittsburgh

6,930,446
3,376,438
1,950,961
1,568,662
1,238,048
900,429
821,960
804,874
781,188
669,817

-0.01
-0.02
-0.03
-0.10
0.26
-0.04
-0.12
-0.01
-0.13
-0.11

0.01
-0.05
-0.03
-0.09
0.14
-0.14
-0.17
-0.04
-0.08
-0.14

-0.10
-0.11
-0.13
-0.20
0.05
-0.24
-0.27
-0.13
-0.12
-0.17

0.04
-0.07
-0.06
-0.15
0.17
-0.12
-0.12
-0.06
0.02
-0.13

0.09
0.04
-0.04
-0.07
0.06
-0.05
-0.12
-0.12
0.03
-0.10

8,008,278
2,896,016
1,517,550
951,270
3,694,820
478,403
348,189
651,154
589,141
334,563

United States

151,325,798

0.19

0.13

0.11

0.09

0.13

281,421,906

City

Source: U.S. Census Bureau, U.S. Census of Population.

8

Urban Colossus

These advantages are the result of scale and density, which
themselves result from New York’s unique history. The vast
number of people crammed together on a narrow island is
what makes Manhattan an information hub. The flow of ideas
has been exacerbated by the tendency of highly skilled people
and industries to locate in the city, which is natural, given that
density and idea flows appear to complement one another. The
most visible result of New York’s strength as a conduit for
information is its penchant for information-intensive
industries, such as finance or publishing, to locate in the city.
While New York’s ability to weather past challenges has
been remarkable, we cannot be certain that its future success
is assured. New York’s importance as a port is long past. The
declining transport costs of moving goods indicate that the
scale advantages remain important only in services. Even in this
area, technological changes may reduce New York’s
transportation cost advantages. In the long run, New York
City’s success depends on its advantage in transmitting
knowledge quickly. This advantage may also be eroded by
changes in information technology; however, in the short run,
information technology may increase the value of face-to-face
interaction and make New York stronger, not weaker (Gaspar
and Glaeser 1998).

2. The Early City: 1624-1790
The traditional story of New York’s origin is that in 1626, the
island of Manhattan was bought by Peter Minuit from the
Lenapes for “sixty guilders worth of trade goods” (Burrows and
Wallace 1999, p. 23). New Amsterdam was founded by the
Dutch West India Company as a trading post oriented toward
the lucrative fur trade. As Burrows and Wallace (p. 23) explain,
the fur trade involved two exchanges: “In the first, European
traders and coastal Algonkians exchanged manufactured goods
for wampum; in the second, European traders used wampum
(and manufactured goods) to obtain first at Fort Orange
[Albany].” Manhattan’s location—a deep-water port at the
heart of the Hudson—made it an ideal center for commerce,
connecting Europeans, coastal native Americans who dealt in
wampum, and upriver native Americans who had access to
furs.
Manufacturing had a place in New York from its inception.
An essential part of trade with the natives was the production
of manufactured goods, and these were cheaper to make in
New Amsterdam than to import from the Netherlands.
Agglomeration in a city was natural because of the gains from
centralized commerce and because there was substantial risk
from ongoing battles with natives. A significant advantage of

Lower Manhattan was that it was easier to defend because it
was surrounded on three sides by water.
The Dutch colonies of New Netherlands were not solely furtrading outposts. Land was abundant, and a steady stream of
settlers acquired land (sometimes vast tracts of it such as
Rensselaerswyck) and began making basic agricultural
products like bread, corn, and meat. The density of settlers was
much lower than it was in Massachusetts, but gradually the
New Amsterdam area also developed an agricultural hinterland
that could both feed the traders and seamen in the city and
begin to export basic foodstuffs to more colonies that exported
cash crops.
In 1664, the town was conquered by the English and
renamed New York. The city was conquered, but the English
were able to keep the city only by giving the Dutch West India
Company the more lucrative colony of Surinam. The
integration of New York with the English colonies increased
the potential for trading opportunities, and the population of
the city surged to approximately 3,000 in 1680 (Burrows and
Wallace 1999) and 5,000 in 1698 (Kantrowitz 1995). While
many Dutch merchants continued to trade with the
Netherlands and the Dutch colonies, a growing group of
English merchants and laborers came to the city as well.
During this period, New York’s trade became primarily
oriented toward the West Indies. The primary exports of the
port were bread and flour, made from wheat grown in the
farms of New York, Connecticut, and New Jersey. This model
of selling foodstuffs to the colonies, which had cash crops that
could be sold back in Europe, had been pioneered by
Bostonians in the late 1630s, but New Yorkers (and
Philadelphians) had several significant advantages over the
Boston merchants. The land in New York and Pennsylvania
was better than the land in Massachusetts. The Hudson and
Delaware rivers were longer, bigger rivers than the Charles.
Indeed, the one long river in New England, the Connecticut,
suffered from heavy silt that formed a sandbar near its mouth.
New York’s Dutch heritage gave it an advantage over
Philadelphia in dealing with the Dutch colonies in the
Caribbean.
New York also offered one more striking advantage over
Boston: its ethnic heterogeneity and religious tolerance.
Boston’s Puritan heritage carried both advantages and
disadvantages. The strong religious community invested in
education and generally proved able to organize the city and
provide basic public goods. Quaker Philadelphia may have
been more tolerant than Puritan Boston, but it was still
fundamentally a faith-based colony. In contrast, New York was
irreligious from the start, and there were fewer barriers against
Jewish or Catholic immigrants. Commercial interests ensured
that New York City was unusually tolerant relative to other

FRBNY Economic Policy Review / December 2005

9

colonies and relative to England itself. New York’s place as a
haven for America’s ethnically heterogeneous immigrants
made the city a magnet for immigrants from its earliest years.
Despite these advantages, the growth of New York during its
first 130 years was relatively modest. Generally, New York was
America’s third or fourth busiest port. In tonnage, it lagged
behind Boston and Charleston in the early eighteenth century
and behind Boston and Philadelphia in the late colonial period.
Boston had a stronger maritime tradition; Philadelphia had a
more developed hinterland. As of 1753, Manhattan had 13,000
inhabitants, making it one of the colonies’ bigger cities, but
hardly a dominant metropolis.
The French and Indian War ended the French presence in
Canada and increased the relative value of New York’s access
through the Hudson to the north. The Revolutionary War had
an even more remarkable effect on New York City. The port
was the only large city that remained in British hands
throughout the war. While combat was certainly disruptive, the
port’s activity also expanded as it provided entry and exit for
military men and material. Perhaps just as important, Boston
and Philadelphia’s long-term reputations as centers of
revolution meant that New York would end up being the
preferred delivery point for British goods coming into the new
republic.
As of 1786, Manhattan had 23,614 residents. In the first
American census, the City of New York had 33,131 residents.
Over the entire 1698-1786 period, the population of
Manhattan had grown by 1.8 percent annually. This increase is
impressive, but ultimately it is far less impressive than the
growth of Philadelphia over the same period. Even though
New York was larger than Philadelphia in 1790, Philadelphia
was a newer city and it had been bigger than New York for
many years during the eighteenth century. When the U.S.
Constitution was signed in 1789, New York was an important
port, but its rise to dominance was still ahead.

war-torn period between 1810 and 1820, New York grew by
more than 50 percent per decade. Except for the period when
New York’s population soared because of the incorporation of
Brooklyn, the city would never grow by comparable rates again.
By 1860, New York was far and away the biggest and most
important city in the United States, with almost 250,000 more
residents than Philadelphia. Over the 140 years since then,
New York’s preeminence among American cities has never
been challenged. In a sense, the key to understanding New
York’s tremendous success lies in understanding the 1790-1860
period.
There are two distinct but closely related growth processes
that occurred over this period. First, the port of New York came
to dominate American shipping and immigration completely.
Second, New York exploded as a manufacturing town, as
industries such as sugar, publishing, and most importantly the
garment trade clustered around the port. The growth of New
York City’s port seems like an almost inevitable result of New
York’s clear geographic advantages (especially when nature was
helped along by the Erie Canal). The growth of manufacturing
in the city informs us about the nature of agglomeration
economies and transportation costs.
Albion (1970) describes the increased use of New York City
as a dumping ground for European goods. The Napoleonic
Wars (and the War of 1812) had severely curtailed trade
between the United States and the United Kingdom. As soon as
peace was declared, British merchantmen with millions of
dollars of goods hastened to America to finally sell these wares.
The merchantmen packed large ships and came to New York to

Chart 1

Growth of New York City and Manhattan
Populations
Population in millions
8

3. The Rise to Dominance: 1790-1860

6

If the growth of New York City prior to 1790 was impressive,
the expansion over the next seventy years was nothing short of
spectacular. Chart 1 depicts the growth of New York City’s
population since 1790 and the growth of Manhattan’s
population since 1900. Chart 2 shows the growth of New York
City and Manhattan as a share of the U.S. population. Between
1790 and 1860, New York City’s population rose from 33,131
to 813,669. The annual rate of increase rose from 1.8 percent to
4.7 percent. Chart 3 presents the time path of the decadal
growth rates of New York City. During every decade, except the

4

10

Urban Colossus

New York City

Manhattan
2

0
1800

1850

1900

1950

Source: U.S. Census Bureau (for city population, 1790-1990: <http://
www.census.gov/population/www/documentation/twps0027.html>;
for borough population, 1900-90: <http://www.census.gov/
population/cencounts/ny190090.txt>).

2000

Chart 2

Growth of New York City and Manhattan
Populations as a Share of U.S. Population
Annual rate of increase (percent)
6
New York City/
United States
4

2
Manhattan
relative population
0
1900

1850

1800

1950

2000

Source: U.S. Census Bureau, U.S. Census of Population.

drop their wares, which were then shipped throughout the
republic. This basic pattern became the model for trade with
Europe over the nineteenth and early twentieth centuries.
At the end of the colonial period, Boston, not New York, was
America’s premier port. Between 1790 and 1820, New York
came to supersede Boston and ultimately attracted a large
number of Boston merchants and sailors into its harbor. From
1820 to 1860, New York completely surpassed its northern

competition in terms of trade. Chart 4 shows the time path of
annual imports, measured in dollars, between 1821 and 1860.
At the start of the period, New York’s exports were $13 million
and Boston’s were $12 million. By the end of the period, New
York’s exports were $145 million and Boston’s were $17 million.
As the chart shows, New Orleans, not Boston or Philadelphia,
rivaled New York City by the mid-nineteenth century.
What changed? Why had the harbors of Boston and
Philadelphia been good enough to be the leading ports of the
colonial era, but not good enough to maintain their strength
over the nineteenth century? There are actually two different
sets of answers to this question. First, there are the technical
factors that make New York a somewhat superior port. Second,
there are the economic factors that translated this modest
geographic superiority into complete mercantile dominance.
We start with New York’s geographic advantages.
One advantage was New York’s central location. While
Boston is at the northern edge of the United States, New York
is in the center. For ships from England and elsewhere trying to
make a single delivery to the colonies, New York offered a
better location because it would be cheaper to ship goods from
there to the southern colonies or Philadelphia than from
Boston. One of the great advantages of the Constitution over
the Articles of Confederation is that the Constitution
significantly reduced the barriers to interstate trade. As these
barriers fell, the possibility for interstate trade rose and the
advantage of a location near the center of the colonies
increased.

Chart 3

Population Growth Rates of New York City
by Decade

Chart 4

Exports from Principal Ports, 1821-60

Percentage growth by decade
1.5

Millions of dollars
150

1900

New York City
1.0
100

1800
1810

1830 1850
1840 1860

0.5
1910
1820

18801890
1870

50
2000
19401950
1990
19601970
1980

0
1800

1850

1900

New Orleans

1930
1920

1950

Boston
0
Philadelphia

2000
1820

Source: U.S. Census Bureau, U.S. Census of Population (<http://www
.census.gov/population/www/documentation/twps0027.html>).

1830

1840

1850

1860

Source: Historical Statistics of the United States.

FRBNY Economic Policy Review / December 2005

11

A second advantage was New York’s large river, which
facilitated shipping deep into the American continent. The
Charles quickly becomes narrow and shallow and is less than
100 miles long. The Hudson is longer than 300 miles and is
extremely navigable. The Erie Canal connects the Hudson to
the Great Lakes system, which enables goods to travel from the
American heartland to Europe completely by water. In an age
when water-borne transport was far cheaper than land
transport, New York’s access to canals, lakes, and rivers gave it
a significant edge over most competitors.
Philadelphia shared some of New York’s advantages of
centrality and water access to the interior. Of course,
Philadelphia’s connection with Pittsburgh and the west used
both rail and water, and as such was decidedly more difficult to
travel than New York’s pure water connection. Moreover, New
York enjoys a third advantage over Philadelphia: direct access
to the ocean. The port of Philadelphia is more than 100 miles
from the Atlantic, whereas the port of New York is less than 20
miles from the ocean. As such, a European ship looking to save
time and money would naturally be attracted to New York. The
ports along the Chesapeake Bay, such as Baltimore, also
suffered from a greater distance to the ocean.
Finally, New York’s port is also superb in terms of its
combination of depth, shelter, and freedom from ice. New
York harbor is protected from the ocean by Staten Island and
the Brooklyn peninsula. It is much deeper than the harbors of
Boston or Philadelphia—a factor that became increasingly
important as ship tonnage increased starting in the 1790s.
Finally, New York harbor is less prone to ice than either Boston
or Philadelphia. The advantage over Philadelphia occurs
because despite Philadelphia’s more southern locale, its
location on a river makes its water freeze faster.
These advantages were significant, but they implied only
that New York would be the first among equals. The city’s
remarkable dominance over America’s exports requires more
explanation. Why did New York end up having five or six times
the exports of Boston and twenty-five times the exports of
Philadelphia in 1860? This question lies at the essence of the
agglomeration economies behind cities.
The rise of New York City as the dominant port can be seen
as an early example of a hub-and-spoke transportation
network. In the earliest period of colonial history, the
dominant form of transportation between the New and Old
Worlds consisted of point-to-point transport, where bales of
tobacco were picked up in Virginia and transported to
England. But point-to-point transport was plagued by a
problem: the exporting areas did not import nearly enough
goods from England to fill the ships on their voyage to the
Americas. First, the southern plantation owners generally
maintained a large current account surplus that was offset

12

Urban Colossus

either by capital accumulation or by paying debts on the
purchase of land and slaves. Second, the manufactured goods
that were imported from the Old World used much less space
than the tobacco or cotton that was exported. Third, the
southern plantation owners found it increasingly efficient to
buy from New World producers of manufactured goods or
food and avoid the lengthy Atlantic trip.
The lack of southern imports can be seen from Chart 5,
which shows the imports and exports of New York and
New Orleans. Throughout the 1821-60 period, the New York
harbor imported more than it exported. This pattern reflected
the general tendency of America to run a current account
deficit that was offset by shipments of bullion back to the Old
World. Throughout the same period, New Orleans maintained
a staggering current account surplus. By 1860, New Orleans
exported $107 million of goods and imported $22 million of
goods. In a sense, this imbalance made it somewhat amazing
that New Orleans’ port could thrive as an export market,
despite the enormous advantage of being at the mouth of the
Mississippi.
This lack of coincidence of wants was solved in the
eighteenth century by the early “triangle” trade, in which
manufactured goods in England were brought to Africa and
traded for slaves, which were in turn brought to the Caribbean
and the South. The ships reloaded with plantation produce that
was then sent to England. But this triangle could hardly survive
the elimination of the slave trade in 1808. Moreover, the
elimination of the slave trade coincided with an enormous
increase in the production of cotton following Eli Whitney’s
invention of the cotton gin in 1794. At the same time as the

Chart 5

Exports and Imports of New York and New Orleans
Millions of dollars
300
New York City imports

200
New York City exports

100

0

New Orleans imports

1820

1830

New Orleans exports

1840

Source: Historical Statistics of the United States.

1850

1860

South had more and more to export, the importation of slaves
became illegal.
The “cotton triangle” in New York City solved this problem.
Cotton was shipped to New York and was transferred from
coastal ships to trans-Atlantic lines. Manufactured goods, often
made in the city, went south. Ships coming to New York were
filled with imported goods from the Old World. Ships leaving
the city were filled with cotton and other basic commodities
being shipped east. While the New York port of the eighteenth
century had focused on shipping flour grown in the vicinity of
the harbor, the port of the nineteenth century became a
conduit through which a large amount of the colonies’ trade
would pass.
The cotton triangle is just one example of New York
becoming a hub connecting two spokes. Obviously, New York
also connected the river, lake, and canal traffic from the west
with the trans-Atlantic ships to the New World. Tobacco
products from the South came to New York from Baltimore
and other, more southern, ports. More surprisingly, New York
also served as a hub for goods from Philadelphia and even
Boston. For example, Boston textile producers would often
ship their wares to New York to be sold in that large entrepot
to buyers from across the country. Similarly, Philadelphia
shipped coal from the Pennsylvania anthracite mines up to
Manhattan.
The increasing attractiveness of hub-and-spoke shipping
owed much to changes in shipping technology. Two large
changes occurred, which added advantages to having a focal
port. First, trans-Atlantic ships became increasingly large over
the early nineteenth century. For example, Albion (1970,
p. 398) reports that in 1834, 1,950 vessels entered New York
harbor carrying 465,000 tons of cargo. In 1860, 3,982 vessels
entered the harbor carrying 1,983,000 tons of cargo. The
average tonnage per ship entering the harbor increased from
238 to 498 tons of cargo over that twenty-six-year period. The
rise in ship size is particularly clear when considering the
packet lines that provided regular service from New York to
Liverpool. In the early 1820s, these ships typically carried
between 300 and 400 tons. By 1838, 1,000 tons became normal
and the Amazon carried 1,771 tons in 1854 (Albion 1970).
These large ships provided great scale economies in the
sense that they required smaller crews per ton. Furthermore,
they were generally safer and faster than their smaller
predecessors. However, large ships created an indivisibility that
makes the gains from a centralized port obvious. While small
ships could readily go point-to-point, dropping their small
cargoes at disparate locations, large ships needed a market that
could accept their bigger cargoes. This created a centralizing
tendency, just as scale economies and indivisibilities do in
standard models of economic geography (Krugman 1991).

This effect is exactly parallel to the tendency to use the largest
planes only for travel between the largest airports. These bigger
ships also increased the advantage inherent in New York’s
deeper harbor. Although Philadelphia could readily compete in
handling the shallow draft ships of the eighteenth century, the
Delaware was simply not deep enough to handle regular
commerce with the largest ships of the nineteenth century.
The second significant change of the nineteenth century was
an increase in specialized shipping, which was itself a byproduct of the increased use of large ships for trans-Atlantic
crossings. In a small-ship world, the ships that plied the coastal
trade and the ships that crossed the ocean were not all that
different. However, the rise of big ships meant that it became
efficient to use different ships to carry goods up and down the
American coast and to carry goods across the Atlantic. Small
ships are far more appropriate for picking up smaller cargoes
and carrying them on shallower waters. Big ships had more of
a risk of running aground, and could not be used to pick up the
smaller cargoes being shipped to and from the disparate
settlements of the young republic. Instead, it increasingly made
sense to use smaller ships, such as schooners, to ply the coastal
trade. These ships would then bring their cargoes to New York
and be consolidated into larger cargoes carried in big ships for
the trans-Atlantic crossing.
These technological advantages were further abetted by
learning-by-doing, specialized investment in port-related
infrastructure, and the agglomeration of manufacturing
(described in the next section). There is little doubt that New
York gradually acquired an unequal set of skills and institutions
that supported large-scale trade. Its auction houses and
insurance system became the largest in the Americas. New York
invested in its wharves, further enhancing its port. Indeed, the
Erie Canal should also be seen as a form of port-related
investment that further exacerbated its initial advantages. As
trade became more intricate and as financial transactions
became larger, gains to specialization increased. As such, the
initial advantage that New York had because of its deep harbor
and central location ultimately translated into massive
dominance as a port.
The rise of the New York port does not illustrate a random
accident leading to geographic concentration. New York was
the best port in the United States and it should have been the
largest. However, its rise does show the conditions under which
an initial advantage, which might have been slight, translates
into vast scale. Probably the most important reason for
centralization was the mismatch between supply and demand,
especially in the southern colonies. This mismatch in New
York’s case, as in most cases, led to the advantages of a large
market that eliminated the need for bilateral commodity
transactions. A secondary factor was the changes in technology

FRBNY Economic Policy Review / December 2005

13

that create larger boats and benefits from specialization. These
changes also created scale economies in the port. Finally, these
advantages were further advanced by trade-specific
infrastructure and trade-specific human capital, which became
increasingly important in the more complicated world of the
nineteenth century.

3.1 The Rise of the Manufacturing City
Although the rise of New York City as a port is a striking
example of agglomeration economies at work, the majority of
New York’s burgeoning population was not involved either
directly in commerce or in the maritime trades. While Boston
specialized in seafaring men, New York’s population
increasingly engaged in manufacturing. As early as 1820,
New York had 9,523 workers in manufacturing and 3,142 in
commerce. By 1850, there were 43,340 people in manufacturing and 11,360 in commerce. New York’s port may have
been the catalyst for the city’s rise, but New Yorkers were far
more likely to be involved in producing manufactured goods
than in working on the ships themselves.
Drennan and Matson (1995) include data from the census
of manufacturers in various decades. The dominant industries,
measured by value, are generally sugar refining, printing and
publishing, and the garment industry. In the 1810 economic
census, sugar refining was the largest industry, and it was
responsible for more than one-third of the value of total
manufactured products in the city. In 1870, sugar would be the
second-largest industry, by value, in New York City and the
largest industry in Kings County (Brooklyn). Even in 1900,
sugar was the second-largest industry in the city. Needless to
say, sugar’s dominance did not continue into the twentieth
century.
The sugar industry began in New York in the eighteenth
century, when Nicholas Bayard opened the first sugar refinery
in the city in 1730. Several other refineries followed and in the
nineteenth century, the Havemeyers began refining in
Brooklyn. Sugar refining, certainly relative to the garment
industry, was highly capital intensive for its day. The refineries
were large industrial undertakings that produced vast returns
for early industrialists.
New York’s dominant role in the sugar industry resulted
from its trade with the West Indies, which increasingly
specialized in sugar production in the 1750s and 1760s. During
this period, New York flour was shipped to the Caribbean and
raw sugar was one of the commodities that returned in the
holds of the ships. The raw sugar would be refined in New York
and consumed in the city, or shipped elsewhere. This pattern

14

Urban Colossus

would continue after the Revolutionary War, when New York’s
central role as the hub of a trading network meant that sugar
passed through the city on its way both to Europe and to
markets within the United States.
But why was New York the natural place to refine sugar? In
principle, sugar could have been refined in the West Indies at
the final point of consumption. In the case of some
commodities, processing removes so much weight that it is
generally efficient to engage in processing at source. Indeed,
even in the case of sugar, it would have been madness to ship
untouched sugar cane up to New York for processing without
first turning the sugar cane into raw sugar. The excess weight
would have badly compromised profits, and even more
important, unprocessed sugar cane rots quickly.
While initial processing must be done soon after the cane is
cut to avoid rot and close to the sugar plantation to avoid the
carrying of excess weight, sugar refining occurs “close to where
the sugar is to be consumed” (Galloway 1989, p. 17). Galloway
writes, “the fundamental reason for the separation of the final
stage in the manufacture of sugar—refining—from the cane
fields, a separation that in the western world dates back several
hundred years, lies in the fact that crystals of sugar coalesce
during the human conditions of a long sea voyage, and so any
imported refined sugar would have had to have been reworked
if customers were to have received the top quality.” Galloway
also emphasizes the lack of cheap fuel for refining in the
tropics, and he might have also stressed the high cost of labor
in the tropics that was skilled enough to run refineries.
Sugar refining occurred in North America rather than in the
Caribbean because of high transport costs, but sugar refining
occurred in New York rather than in small towns throughout
the country because of scale economies. By the standards of
early-nineteenth-century industry, sugar refining involved
large infrastructure investment and significant fixed costs.
Sugar refineries were among the largest factories of this early
period. These scale economies meant that it was impractical to
spread sugar refineries throughout the colonies in every town
or village. The technology of sugar production almost dictated
that sugar refining occur in a central location close to most
centers of consumption, and New York City was an ideal
central location.
The strength of the sugar industry in New York therefore
owes everything to the city’s role as a shipping hub connecting
Caribbean ports both with the American hinterland and with
European final consumers. The scale economies in sugar
refining are strong enough that it makes sense to centralize, and
centralized production is most efficient if it occurs in the port
through which the sugar is passing anyway. The growth of
sugar manufacturing shows a basic pattern for the growth of

New York as a manufacturing center. Trade brought raw
commodities through the city. In cases where manufacturing in
the initial agricultural area was inefficient, but where it made
sense to manufacture in a single place, this gateway city was the
natural site to create finished products.
While the sugar refining industry produced a great deal of
value, it generally only included a modest number of New
Yorkers. For example, in 1860, the economic census of
manufacturers reported 1,494 employees in sugar refining in
New York City making more than $19,000,000 of products. By
contrast, the garment industry employed 26,857 workers in
that same year and produced $22,320,769 of goods. From the
mid-nineteenth century through 1970, the garment trade
remained New York City’s dominant manufacturing industry,
at least in terms of total employment. In 1860, almost
30 percent of New York City manufacturing employment was
in the garment industry. In 1900, 19 percent of New York’s
manufacturing employment was in that sector. In 1940 and
1967, 27 percent of manufacturing employment was in
garments.
New York was generally a diversified economy, but to the
extent that one industry dominated the city for a century, it was
the garment trade. The basic economics of the nineteenthcentury New York garment industry are not so different than
the economics of the sugar refining industry. The essence of
this industry is turning cloth into clothing. Cloth was generally
produced in textile mills, either in England or later in the textile
mills of New England. As was the case with sugar, cloth and silk
came through Manhattan. Similarly, there was a strong
economic rationale to have manufacturing centered at the port
of entry.
The starting point for the textile trade was England’s
commercial dominance as an exporter of wool and cotton
cloth. This dominance was historical, but at the end of the
eighteenth century, early industrialization gave English
manufacturers a huge advantage in the production of textiles.
This advantage, and the general importance of clothing in
budgets, meant that in the first half of the nineteenth century,
“textiles amounted to nearly 60 percent of England’s domestic
exports and about one-third of the imports of the United
States” (Albion 1970, p. 58). This trade increasingly came
through New York with the city’s dominance of trans-Atlantic
shipping. In 1860, more than 80 percent of the nation’s textiles
entered through New York. In the same year, wool, cotton, and
silk goods accounted for 37 percent of the imports coming into
the harbor.
England was the only producer sending textiles into
America through New York harbor. The city was also the
entryway for silks from France and even China. As New
England mills began production and competed with English

producers, even they found themselves shipping cloth to
Manhattan to take advantage of this central market. The vast
flow of cloth into Manhattan was the natural result of New
York’s dominance as a port and textile’s dominance as an item
of trade.
In the early part of the nineteenth century, this trade did not
create a garment industry. In the 1810 economic census, New
York City had significant tanneries and hatteries, but not a
significant garment trade. Fifty years later, the garment
industry had become the city’s largest industry. The big change
occurred because of the rise of the ready-to-wear industry. In
1810, cloth was turned into clothing by tailors, seamstresses,
and by the end users themselves. There were no factories for the
production of clothes. When clothes were made-to-measure,
there was no place for centralized production of garments. At
the start of the nineteenth century, therefore, New York’s
garment industry consisted mainly of tailors catering to the
local population.
Over the nineteenth century, there were changes both in
demand and production technology that turned New York into
a center of ready-to-wear clothes. On the demand side, the
rising slave population of the South had a demand for
extremely cheap, ready-to-wear clothing. George Opdyke
began the manufacture of ready-to-wear clothing in New York
in 1831, catering to the market in New Orleans. The changes in
production technology included the development of the
factory system, and even more important, Elias Howe’s
invention of the sewing machine in 1846. Mechanization
greatly decreased the costs of mass production relative to
custom tailoring and furthered the rise of the ready-to-wear
garment industry.
Once such an industry existed, and given that there were
substantial scale economies in the production of clothes due to
machinery and specialized human capital, it is hardly
surprising that this industry centered in New York City. Given
that the cloth came into that city, there was no reason to wait
until the cloth reached its final destination before transforming
it into shirts and pants. There would be few advantages to
making ready-to-make clothes in disparate locations rather
than in one centralized locale.
As with sugar, we must ask why manufacturing did not
occur in the place where the raw material was first produced,
which in this case was England. First, while England had a long
history of cloth production, it had no history of producing
ready-to-make clothes. No place did in 1830. As a result,
England had no natural advantage in this form of
manufacturing. New York manufacturers had the advantage of
better knowledge of local demand, and could therefore cater to
local tastes. They had access to relatively inexpensive labor
from the increasing immigrant populations. In short, there

FRBNY Economic Policy Review / December 2005

15

were probably only mild advantages to centralizing ready-tomake clothing in New York rather than in London, but these
small advantages were enough for this industry to be located on
the American side of the Atlantic.
Another important point about the garment trade, which
helps explain its 100-year dominance in New York, is that
among manufacturing industries, its need for physical space
and power was quite mild. Textile mills themselves were more
efficient on a grand scale, and in the first part of the nineteenth
century, the mills needed water power. As a result, they were
generally located away from urban areas along the banks of
rivers like the Merrimack. By contrast, the garment trade
involved human beings and relatively small sewing machines.
In many cases, working women could contract work to be done
in their own apartments. This was the ideal industry for a city
where land was expensive.
Over the decades, New York developed an increasing
human and physical infrastructure that supported the
continuing presence of the garment trade even after the port’s
primacy had passed. Factories were built to cater to this trade.
Singer came to New York to popularize his adaptation of the
Howe sewing machine. An entire section of the city (the
Garment District) became oriented toward clothing
production, and a network of spatially proximate suppliers
catered to this industry. Perhaps even more important, the
city’s industry attracted skilled workers who created a powerful
agglomerating force that trained new workers and attracted
entrepreneurs. There was an initial comparative advantage in
manufacturing garments that came from New York’s port, but
this advantage produced an agglomeration that kept the
industry in the city.
The third-largest manufacturing industry in the city in 1860
was printing and publishing. As late as the 1960s, publishing
would be a distant second to garment manufacturing in its
share of New York employment. Only in the past thirty years
has publishing passed garment manufacturing to become New
York’s largest manufacturing industry. Still, value added per
worker was generally much higher in this industry than in the
garment trade. Moreover, the rise of New York publishing
suggests the increasing role of New York as a city centered
around the transfer of ideas.
Somewhat surprisingly, the early development of New
York’s publishing trade was also linked to the city’s role as a
port connecting America with the Old World. In the early
nineteenth century “the big money, however, came from
pirated copies of English authors (who didn’t yet have to be
paid royalties because the United States government refused to
as yet to recognize foreign copyrights)” (Burrows and Wallace

16

Urban Colossus

1999, p. 441). As such, there was a huge advantage in this
industry to being the first printer with a copy of the latest
London sensation and “printers and book dealers in New York
and Philadelphia competed furiously to bring out the first
American editions of new English novels” (Burrows and
Wallace, p. 441).
In this competitive atmosphere, being at the center of the
trans-Atlantic trade offered a crucial advantage. New York
printers would have been capable of receiving new novels
from England more quickly and regularly than their
Philadelphia competitors because of the more frequent
sea traffic between New York and Liverpool. The closer
connections between New York and England also ensured
a steadier infusion of information about the latest books.
New York’s production advantages were complemented by
the advantages in distributing to western consumers via the
Erie Canal.
As in the case of the garment trade, this initial advantage
stuck because of specialized human capital and the advantages
that came from local agglomeration economies. New York
attracted networks of suppliers and tradesmen who catered to
the book producers. Book sellers from around the country
would come to New York for book fairs to get access to the
latest novels. Eventually, the combination of high costs of land
and low transport costs would push the printing presses
themselves off of Manhattan, but to this day, there is a strong
community of publishing houses in Manhattan connecting
with authors and potential customers.
While publishing English novels was one part of the early
success of Manhattan publishing, news was the other
cornerstone of this industry. Information was extremely
valuable to the growing mercantile economy, and most of the
early papers focused on providing this information. Scale
economies in this industry also meant that New York had a
disproportionate number of newspapers. As the news became
entertainment, and even entertainment for the masses, scale
economies and New York’s large population ensured that the
city would remain a center for newspaper production.
The central lesson of the rise of New York in the early
nineteenth century is that manufacturing congregated around
a port. Changes in transportation technologies turned New
York into the preeminent port of the United States. This meant
that raw inputs, including sugar, cloth, and even English
novels, came into the city first. The first manufacturing
industries were based on these raw inputs. As scale economies
rose with industrialization, production was increasingly
centralized in the one place that welcomed the nation’s imports
of these inputs.

4. The Immigrant City: 1860-1920
While New York City was the largest city in the country in
1860, it would continue to grow significantly over the next
ninety years. Over this period, the population of the city
increased from 813,000 to 7.9 million. Much of this increase
reflected the incorporation of the outer boroughs into New
York City, but even Manhattan’s population continued to grow
until 1920. As shown in Chart 2, New York reached its peak
relative to the U.S. population as a whole in 1940, when
5.6 percent of the U.S. population lived in the city. Manhattan
was at its largest relative to the nation in 1910, when almost
3 percent of the U.S. population lived on the island.
During this amazing period, the basic structure of the New
York economy was remarkably static. The city remained
primarily manufacturing-oriented. In 1910, there were 873,497
employees in manufacturing, 40 percent of New York’s total.
Trade and transportation had slightly more than 500,000
employees and domestic service included more than 330,000
workers. The primary export industries were manufactured
goods and the transportation sector. New York’s port remained
the biggest in the nation during this era.
Even more remarkable, the composition of manufacturing
employment remained constant across industries. The
garment trade declined somewhat as a share of overall
employment, but it remained New York’s dominant industry.
Sugar refining, printing, tobacco, and bread all remained big
products. In the first half of the nineteenth century, New York’s
population explosion was connected with a radical
restructuring of the city economy and the rise of
manufacturing. In the second half of the nineteenth century,
New York’s population increases continued despite the fact
that the basic structure of production remained remarkably
constant.
Still, there were trends that supported the growth of New
York’s industries, particularly the garment trade, during this
period. Demand for finished clothing increased steadily as
populations and incomes rose in the country as a whole. Input
prices dropped significantly over the 1870-90 period. For
example, the Warren and Pearson index of the wholesale cost
of textiles shows a 20 percent decline relative to the Bureau of
Labor Statistics’ consumer price index during these years. As
the South recovered from the Civil War, cotton in particular
became less expensive: the cost per pound of raw cotton fell
from 29 cents in 1869 to 11 cents in 1890. Wool dropped from
90 cents per pound in 1870 to less than 40 cents in the mid1890s.
Despite the continuing strengths of New York City’s
industries, it would be a mistake to ignore the explosion of

immigration to America from Europe. Chart 6 shows the levels
of immigration into the United States by decadal frequencies
between 1820 and 1970. Prior to 1841, annual immigration had
always been below 90,000. Except for the five years between
1849 and 1854, immigration never passed 250,000 per year
until 1865.
After the Civil War, as the chart shows, immigration began
to soar. There were almost 400,000 immigrants in 1870. There
were 450,000 immigrants in 1880, 1890, and 1900; between
1903 and 1914, there were almost 12 million immigrants. The
overwhelming share of these immigrants entered the United
States through the port of New York City. Again, New York’s
dominance as a port meant that it was the center for the import
of America’s most significant economic input: its labor force.
The rise in immigration is probably best seen as the result of
declining transportation costs in trans-Atlantic passenger
travel. Just as improvements in shipping ensured that New
York captured a larger share of the goods shipped into the
United States in the early nineteenth century, continuing
improvements in sea travel meant that New York was able to
retain an increasingly large group of immigrants. These
reductions in travel costs were accompanied by political
problems in European countries like Russia that terrorized
their Jewish citizens with pogroms and by a continuing gap
between high American wages and worse economic prospects
in the poorer European countries. Accompanying these factors
was the phenomenon of chain migration, in which an initial
group of immigrants made it more socially comfortable for
later immigrants to follow.

Chart 6

Immigration to the United States
Number of immigrants
1,000,000

500,000

0
1800

1850

1900

1950

2000

Source: Historical Statistics of the United States.

FRBNY Economic Policy Review / December 2005

17

The reason for the vast number of immigrants who stayed in
New York, and who continued to settle (at least temporarily) in
the city, can be understood as the result of four factors. First,
transportation costs for internal transport within the United
States were still high enough to make it cheaper to just stay in
New York. This factor would have been particularly important
for immigrants from poorer countries such as Italy, AustriaHungary, and Russia, who were frequently stretched to their
financial limits by the trans-Atlantic journey itself. After
making the long and costly trip across the ocean, many
immigrants simply did not want to spend the time and money
to travel further.
Second, New York’s economy may have kept its basic
structure over this period, but it still showed a remarkable
ability to increase its scale with the influx of new labor. The
rising American population meant that demand for garments
continued to rise, and there was nothing intrinsic to the
production process that limited even more production within
the city. The garment industry was also special in the sense that
it relied on skills that were more prevalent among immigrants
than the skills required in more advanced industries.
Third, improvements in transportation technologies for
within-city transport increasingly made development out of
the boroughs feasible. New York began its omnibus routes in
the 1820s. Streetcars and the subway line soon followed. The
introduction of the automobile was soon accompanied by that
of the bus. Public transportation made it possible for new
immigrants to occupy the outlying boroughs and commute
into the city.
Fourth, and perhaps most significantly, the city itself
acquired considerable immigrant-specific social and political
infrastructure that made, and continues to make, New York a
magnet for immigration. The most important form of this
infrastructure may be large communities of immigrants from
specific countries. These communities allowed new
immigrants to come to New York while continuing to speak
their own language. In these areas, suppliers provided
commodities that were closer to those that the immigrants had
consumed in their home countries. It was certainly easier for a
Jewish Orthodox immigrant to keep kosher in the Lower East
Side of Manhattan than in rural Minnesota.
Immigrants provided the voting base for Tammany Hall
during this period, and city services as a result were oriented
toward immigrant needs. This meant that judges were quick to
approve naturalization and that the city machine stood ready
to provide patronage and emergency supplies to new arrivals.
Churches and synagogues were built to cater to the growing
immigrant population. Indeed, New York had been an
immigrant town well before the Civil War, so there was a long

18

Urban Colossus

tradition of providing economic services and employment to
new arrivals.
Did the flow of immigrants in the late nineteenth century
mean that New York City’s labor supply was outstripping labor
demand? Long time series on wages for the city are not
available, but we can show the time path of average wages (in
2005 dollars) for production workers in manufacturing for
New York State and the nation as a whole (Chart 7). If New
York’s growth primarily reflects labor supply, we would expect
wages in the city to fall relative to wages in the nation as a
whole. If New York’s growth reflects labor demand, we would
expect wages in the city to increase.
Chart 7 shows that from 1870 to 1890, manufacturing wages
were rising in the United States as a whole, and the New York
State wage premium increased from 7 percent to 13 percent.
Labor supply may have been increasing during this period, but
labor demand in both New York and the nation was increasing
even faster. From 1890 to 1914, real manufacturing wages in
New York State declined and the New York State wage
premium fell back to only 3 percent. This period of declining
real wages in the state corresponds with the period when
immigration truly exploded. These figures suggest that during
the twenty-five years after the Civil War, labor demand
increases outpaced labor supply, especially in New York,
perhaps as a result of declining costs of inputs and rising
demand in the country as a whole. Changes in transportation
technology made it increasingly possible for manufacturers to
locate in the city and sell their wares throughout the world.
New industrial technologies and products also strengthened
the local economy. New York remained innovative, and this

Chart 7

Average Manufacturing Wages in New York State
and the United States
2005 dollars
20,000

New York State
15,000
United States
10,000

0
1850

1900

Source: U.S. Census Bureau, U.S. Census of Population.

1950

characteristic helped to ensure that rising population levels did
not push wages down precipitously.
However, between 1890 and 1914, the growth of the city had
more to do with the immigrant shock to labor supply than with
increases in labor demand. Nonetheless, the driving force
behind the rise of New York City’s population and the
continuing growth of the city’s economy was the steady influx
of immigrants between 1890 and 1920. The immigrants came
to America because of higher wages, better safety, and cheaper
ocean travel. They stayed in New York for the same reasons that
cotton and sugar were processed in the city: because of lower
transportation costs and because New York specialized in
imports.

5. The Rise of the Information City:
1920-2000
New York’s immigrant boom ended with the national
restriction on immigration in 1921. The quota law drove
immigration down significantly and ended the prewar
explosion of immigration to the island of Manhattan. For the
first time in decades, the foreign born would represent a
declining share of New York’s population.
This negative shock was accompanied by a pair of
technological shocks that would hurt almost all of America’s
larger cities. First, the rise of the automobile made cities such as
New York, which had been built around older transportation

technologies, somewhat obsolete. Automobiles, at least in lowdensity, car-oriented areas, are much faster means of travel
than public transportation. The average commute by car in the
United States is twenty-three minutes, compared with fortyseven minutes for public transportation. New York and other
cities are built at higher densities to take advantage of public
transportation and to allow travelers to walk from public
transport stops to their final destination. Car-based
communities are built at much lower densities to allow
automobiles to drive without congestion and to allow the
consumption of more land.
Second, the rise of the truck led to a spectacular decline in
transportation costs and a decrease in the need for high-density
work environments. Glaeser and Kohlhase (2004) estimate that
the real cost of transportation declined by 95 percent over the
twentieth century. As such, cities like New York that were built
to take advantage of transportation technologies lost this
comparative advantage. Moreover, the truck does not require
the same centralized infrastructure as the older form of
shipping technology does. This meant that manufacturing no
longer needed to cluster around a port or a train station. Over
the twentieth century, manufacturing left large cities and is
now generally located in medium-density countries (Glaeser
2005). Chart 8 presents a long time series of the share of
national manufacturing employment that was located in New
York State; Chart 9 shows the decline in manufacturing both in
New York City as a whole and in Manhattan after 1949.
These shocks impacted New York City just as they did all of
America’s major cities. Table 1 shows the time path of
population levels (after 1950) for the ten largest cities in the

Chart 8

Manufacturing Employment in New York State
Relative to the United States

Chart 9

Manufacturing Employment over Time
in New York City and Manhattan

Share of manufacturing (in New York)
0.20

1850

1880

Number employed

1890

1,000,000
1860

1870

New York City

1949

1900
1914
1904
1909
1919

0.15

800,000

1929 1939

1.10

600,000

1958

Manhattan

1967

400,000

1977
1987

0.05

1997

1850

1900

1950

Source: U.S. Census Bureau, U.S. Census of Population and
U.S. Manufacturing Census.

2000

200,000
1940

1960

1980

2000

Source: U.S. Census Bureau, Statistical Abstract of the United States
(1946, 1956, 1967, 1977, 1983, 1994, 2000).

FRBNY Economic Policy Review / December 2005

19

United States in 1930. Every city but Los Angeles lost
population in the 1950s and the 1970s. Every city but New York
and Los Angeles lost population in the 1960s. Every city but
New York, Boston, and Los Angeles lost population in the
1980s. In the 1990s, New York, Chicago, Boston, and Los
Angeles all managed to lose population. The figures in the table
show the generally declining period experienced after World
War II by all major cities as transportation technologies made
high-density living in traditional manufacturing towns
relatively much less attractive.
Table 1 makes it clear that the remarkable thing about New
York City is not its postwar decline, but rather its success
relative to other older cities. Only in the 1970s did New York
lose more than 1 percent of its population. Even in that decade,
it lost the least amount of population of any of these cities
(again, except for Los Angeles). New York–oriented writers
often emphasize the city’s big problems during the 1970s, but
such a focus ignores the fact that almost every other traditional
city fared far worse during this period. The era of Lindsay and
Beame may have had its problems, but New York was in much
better shape than either Detroit or Philadelphia during the
same period.
After World War II, New York had many of the same
problems that plagued other large cities. Crime skyrocketed
between 1960 and 1975, and the increase in crime made wider
social problems more visible. Bad urban governance, which in
most cases had been going on for decades, became more
obvious during a period of urban decline when steadily
increasing tax receipts could not hide waste and

mismanagement. Furthermore, decaying infrastructure made
the city seem grungy.
However, New York survived these problems better than its
peers did mainly because its economy remained more robust.
While the economies of Philadelphia, Detroit, and Pittsburgh
never truly survived the collapse of local manufacturing, New
York (like Boston) has reinvented itself over the past eighty
years as a service city increasingly oriented around finance and
corporate management. New York continues to boom to this
day primarily because of finance and business services.
Table 2 shows the 2002 distribution of employment in
Manhattan. Twenty-eight percent of the city’s payroll is in a
single three-digit industry: security, commodity contracts, and
like activity. This level of concentration is even higher than the
commitment of the city to the garment trade during the height
of that industry. Another 28.5 percent of total payroll is in three
other industries: business, scientific, and services (mostly
lawyers and accountants); credit intermediation; and company
management. Together, the four industries account for
56.6 percent of total payroll in Manhattan. When Chinitz
(1961) compared agglomeration in New York and Pittsburgh,
he emphasized the remarkably diverse nature of the New York
economy. This is no longer the case. Manhattan employment is
remarkably dependent on finance, business management, and
business services.
This is not true in the city’s outlying boroughs, which
employ primarily in nontraded service sectors. Tables 3 and 4
show the importance of health care, for example, in the
economies of Brooklyn and Queens. Both boroughs also have

Table 2

Employment in Manhattan, 2002

Three-Digit Industry Name
Professional, scientific, and technical services (541)
Security, commodity contracts,
and like activity (523)
Administrative and support services (561)
Food services and drinking places (722)
Educational services (611)
Credit intermediation and related activities (522)
Management of companies and enterprises (551)
Hospitals (622)
Religious, grantmaking, civil, professional,
and like activity (813)
Ambulatory health care services (621)

Employment

Share of Total
(1.99 Million)

Payroll
(Thousands
of Dollars)

Share of Total
(150 Billion)

Payroll/Worker

261,157

0.131

21,389,318

0.143

81,902

210,960
142,796
107,778
94,945
90,105
84,821
73,230

0.106
0.072
0.054
0.048
0.045
0.043
0.037

42,107,893
5,521,745
2,208,254
3,764,351
11,191,706
10,059,521
4,320,883

0.281
0.037
0.015
0.025
0.075
0.067
0.029

199,601
38,669
20,489
39,648
124,207
118,597
59,004

67,823
67,399

0.034
0.034

2,955,000
2,660,933

0.020
0.018

43,569
39,480

Source: U.S. Census Bureau, 2002 County Business Patterns for New York, New York (<http://www.census.gov/epcd/cbp/map/02data/36/061.txt>).

20

Urban Colossus

Table 3

Employment in Brooklyn, 2002

Three-Digit Industry Name
Ambulatory and health care services (621)
Hospitals (622)
Social assistance (624)
Educational services (611)
Food services and drinking places (722)
Administrative and support services (561)
Nursing and residential care facilities (623)
Special trade contractors (235)
Wholesale trade, nondurable goods (422)
Professional, scientific, and technical services (541)

Employment

Share of Total
(435,948)

54,537
45,098
21,891
21,145
18,395
17,997
16,849
14,976
14,852
14,474

0.125
0.103
0.050
0.049
0.042
0.041
0.038
0.034
0.034
0.033

Payroll
(Thousands
of Dollars)
1,682,173
2,315,354
498,796
500,278
261,438
434,805
542,854
613,787
492,365
497,593

Share of Total
(13.9 Billion)

Payroll/Worker

0.121
0.166
0.036
0.036
0.019
0.031
0.039
0.044
0.035
0.036

30,845
51,341
22,785
23,659
14,212
24,160
32,219
40,985
33,151
34,378

Source: U.S. Census Bureau, 2002 County Business Patterns for Kings, New York (<http://www.census.gov/epcd/cbp/map/02data/36/047.txt>).

Table 4

Employment in Queens, 2002
Three-Digit Industry Name
Ambulatory and health care services (621)
Special trade contractors (235)
Air transportation (481)
Food services and drinking places (722)
Hospitals (622)
Administrative and support services (561)
Nursing and residential care facilities (623)
Professional, scientific, and technical services (541)
Wholesale trade, durable goods (421)
Educational services (611)

Employment

Share of Total
(468,585)

Payroll
(Thousands
of Dollars)

Share of Total
(16.8 Billion)

Payroll/Worker

37,272
29,330
27,502
26,680
24,729
21,818
16,215
14,329
13,661
13,513

0.080
0.063
0.059
0.057
0.053
0.047
0.035
0.031
0.029
0.029

1,146,772
1,541,310
1,448,255
401,915
1,288,459
506,225
537,169
477,570
601,030
389,995

0.068332
0.091841
0.086296
0.023949
0.076774
0.030164
0.032008
0.028457
0.035813
0.023238

30,768
52,551
52,660
15,064
52,103
23,202
33,128
33,329
43,996
28,861

Source: U.S. Census Bureau, 2002 County Business Patterns for Queens, New York (<http://www.census.gov/epcd/cbp/map/02data/36/081.txt>).

export sectors, such as Queens’ airport industry, but these are
much smaller economic areas and are much more oriented
toward providing services to the residents of the greater New
York area.
New York’s move into finance and management is not really
paralleled by any of the other older cities. Perhaps the closest
parallel to New York is Chicago, which, during the past decade,
has somewhat remade itself around business services. Boston’s
post-1980 renaissance is completely different and should be
seen as the result of small-scale entrepreneurship in a number
of disparate, high-human-capital sectors. The other large cities
are still in decline and cannot be said to have found any

meaningful replacement for the manufacturing firms that once
employed thousands of their citizens.
The success of New York as a financial city suggests three
questions. How did New York become the financial capital of
the world? Why has New York’s dominance managed to
expand in the modern era? Will New York manage to continue
to survive on the basis of its financial industries?
Unsurprisingly, the origins of New York’s financial
community lie in its role as a port. The financial sector on Wall
Street has its origins as an organization designed around
sharing risk on sea voyages. This financial community
branched into government securities in the 1790s. In the early

FRBNY Economic Policy Review / December 2005

21

nineteenth century, New York was a close rival to Philadelphia
as a center for trading stocks and bonds.
Eventually, New York replaced Philadelphia for at least
three reasons. New York’s greater connection to England
became increasingly important in the late nineteenth century
as English capital financed American development. New York’s
greater size meant that there were more companies in New
York, which had a direct, local market for financing. Finally,
the great incentive to agglomerate in finance comes from the
desire for the latest information. In no other industry are the
returns to knowing the latest fact greater; this meant that once
New York had a slight edge, the edge turned into a complete
preponderance as the financial community came to the city to
obtain access to the latest information.
The rise to world dominance by New York’s financial
community was a twentieth-century phenomenon that
followed the decline of New York as a port. Instead, there are
two major agglomeration economies at work. The first is the
role of the dense city as a center for idea flows. The high value
of knowledge meant that being in the city was particularly
valuable. New York’s high density levels, which ended up being
unattractive for most manufacturing firms, may have even
helped New York finance continue to thrive because those high
density levels are particularly conducive to chance meetings,
regular exchanges of new ideas, and the general flow of
information.
Chart 10 depicts the rising share of U.S. and New York City
employment in finance, insurance, and real estate. The
concentration of New York City in this sector is much lower
than the concentration of Manhattan in this sector, and the
concentration of employment is much lower than the
concentration of payroll. Nonetheless, the city has much more
of its employment in this area than does the United States as a
whole. Furthermore, both city and national data show that this
sector is increasing employment. Somewhat surprisingly, the
decade in which the share of New York City employment in
this sector increased the most was the 1970s. In 1970, 7.4
percent of the city’s employment was in this sector; by 1980,
12 percent was in the sector. This change reflected both the
increase of finance and the decline of other industries, such as
manufacturing. As such, it may make sense to date New York’s
dependence on this sector to 1980.
New York’s high density levels and massive scale drove its
success as a center of business services. The cost of delivering
manufactured goods depends only on transportation
technology, but the cost of delivering services depends both on
technology and on the value of the time involved by the
participants in the transaction. Because services are by
definition face-to-face, during an era of rising wages there is an
increased incentive to agglomerate these activities. This simple

22

Urban Colossus

Chart 10

Share of Finance, Insurance, and Real Estate (FIRE)
Employment in New York City and the United States
Percent
15

FIRE as share of
New York City
employment

10

FIRE as share of
U.S. employment

5

0
1940

1960

1980

2000

Source: U.S. Census Bureau, U.S. Census of Population.

argument can explain why New York was able to thrive at the
same time that its manufacturing base was fleeing. Services
replaced manufacturing because of the transportation cost
advantages of locating in a large, dense city.
The flow of information and the ability to buy and sell
business services are the reasons why Manhattan has survived
as the center of world finance. But if finance had remained at
its 1940 level, it would have had no effect on the long-run
fortunes of New York. The city’s great fortune was that at the
same time that it was suffering from an exodus of the garment
trade, the international financial sector boomed. Individuals
saved and invested more. Improvements in communication
technology and changes in regulation made it increasingly
attractive for people to become involved in New York’s formal
economic markets. Firms had an ongoing demand for
financing. The industry soared and New York was its center.
However, it is less obvious that this trend will continue. New
York City is still the epicenter for the transmission of new ideas
in finance, but the past fifteen years have seen a remarkable
growth of cutting-edge financial institutions in the caroriented edge cities surrounding the metropolis. Some of the
more famous and infamous financial market participants have
been located far from Manhattan (Warren Buffett in Omaha,
Peter Lynch in Boston, Michael Milken in Los Angeles). As
important as face-to-face contact appears to be, information
technologies have made major inroads, and the continuing
economic vitality of New York City is less obvious than it was
fifteen years ago.
A final point on the future of New York worth emphasizing
is that the city recently has made remarkable progress in
changing itself from a relatively unattractive to a relatively

attractive place to live. In 1970, real wages in New York were
quite high, which was necessary to compensate workers for
crime and other problems associated with the city. In 2000, real
wages were much lower. Nominal wages have risen, reflecting
in part the continuing vitality of the financial sector, but prices
have risen even more. This rise in real wages relates to the
increasing demand for New York as a consumer city. If the city
is able to continue to attract financial professionals who want
the excitement of New York, then it can thrive from the labor
supply just as it did during the period of immigration of the late
nineteenth century.

6. Conclusion
In Glaeser (2005), we argue that the long-term success of
Boston reflects a process of ongoing reinvention, whereby
smart entrepreneurs react to a continuing set of crises by
discovering new ways to turn a profit and still live in that city.
New York’s history is far more continuous, more stable, and
more triumphant. The city’s rise to dominance occurs during
the early nineteenth century and is driven primarily by New
York’s advantages as a port. Manufacturing, immigration, and
even finance followed from this maritime supremacy. The
ultimate success of New York comes from its role as the center
of the global trading network.
There are several lessons for urban and regional economics
from the economic history of New York City. First, there is
something to be said for geographic determinism. New York
City should have had the biggest harbor and it did. However,
we cannot appreciate the full extent of the city’s dominance

without understanding that agglomeration economies and
New York’s rise to dominance as a port are associated with the
increasing scale of ships and the benefits of specialization.
A second lesson from New York is that transportation costs
really matter. The city’s port status obviously came about in
large part because of these advantages, but its role as a center
for immigration and as a sugar refinery also came about largely
because of cost savings that resulted from reduced
transportation costs. This point may be less relevant today in
the manufacturing sector, but the ongoing importance of
transportation costs in business services helps explain New
York’s continuing strength in that area.
A third lesson is the obvious importance of what Henderson
(1977) calls localization economies. Generally speaking, every
industry has some form of very specific industry-related needs
that were met by agglomeration in New York. Indeed, even the
concentration of immigrants tends to suggest a benefit from
very particular groups of immigrants locating near one
another. These agglomeration economies helped ensure that
initial transportation-cost-based agglomerations did not
disappear as transportation costs fell.
A fourth and final lesson is that New York’s success for
centuries has been connected to its edge as an idea city.
Publishing centered in New York because people there could
read the latest books from England more quickly. Sugar
refining and the garment trade were located in New York, as
opposed to places that made primary products, in part because
of the information gains offered by the city. Finally, and most
spectacularly, for almost 200 years, the success of New York’s
financial sector owes a great deal to the city’s role as a place
where the latest news can be picked up quickly.

FRBNY Economic Policy Review / December 2005

23

References

Albion, R. 1970. The Rise of New York Port [1815-1860].
New York: Scribners.

Glaeser, E. L. 2005. “Reinventing Boston: 1620-2003.” Journal of
Economic Geography 5, no. 2 (April): 119-53.

Burrows, E., and M. Wallace. 1999. Gotham: A History of
New York City to 1898. New York: Oxford University Press.

Glaeser, E. L., and J. Kohlhase. 2004. “Cities, Regions, and the Decline
of Transport Costs.” Papers in Regional Science 83, no. 1
(January): 197-228.

Chinitz, B. J. 1961. “Contrasts in Agglomeration: New York and
Pittsburgh.” American Economic Review 51, no. 2 (May):
279-89.

Henderson, J. V. 1977. Economic Theory and the Cities.
New York: Academic Press.

Drennan, M., and C. Matson. 1995. “Economy.” In K. T. Jackson, ed.,
The Encyclopedia of New York City, 358-62. New Haven:
Yale University Press.

Kantrowitz, N. 1995. “Population.” In K. T. Jackson, ed.,
The Encyclopedia of New York City, 920-4.
New Haven: Yale University Press.

Galloway, J. H. 1989. The Sugar Cane Industry: An Historical
Geography from Its Origins to 1914. Cambridge: Cambridge
University Press.

Krugman, P. 1991. “Increasing Returns and Economic Geography.”
Journal of Political Economy 99, no. 3 (June): 483-99.

Gaspar, J., and E. L. Glaeser. 1998. “Information Technology and the
Future of Cities.” Journal of Urban Economics 43, no. 1
(January): 136-56.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
24

Urban Colossus

J. Vernon Henderson

Commentary

he paper by Edward L. Glaeser offers an insightful and
entertaining overview of almost four centuries of the
economic growth of New York City. First, I will address some
of the themes I took away from the history sections of the
paper. Then I will turn to the modern era and comment on a
basic point I think the paper misses in the description of both
the historical record and the modern era: the role of New
York’s vibrant neighborhoods.
The author’s first theme is New York City’s four-centuries-plus
record of sustained economic and population growth. During this
period, New York has outperformed Boston (and Philadelphia) to
become the nation’s leading city and metropolitan area.
Geography has played a key role. New York has a great natural
harbor connected to a long navigable river, the Hudson,
something that Boston lacks. In addition, New York’s central
location on the east coast offered advantages over Boston’s
periphery location to the north. At the dawn of major
industrialization, New York was the hub of what emerged as a
hub-and-spoke transport system stretching up and down the coast
and inland. Moreover, given New York’s initial transport cost
advantages at the time and its slightly larger population, the city
benefited from a noticeable “home-market effect,” as described in
the recent economic-geography literature. For industries
exhibiting scale economies, a larger home market becomes a
source of local demand that helps escalate local production scale.
Glaeser describes how these advantages helped New York
become America’s center for manufacturing in the sugar,

T

garment, and publishing industries well into the twentieth
century. Even today, New York’s presence in two of these
industries continues, evidenced by a high concentration of
firms engaged in the haute-couture fashion and upscale
magazine publishing industries. For publishing, New York’s
initial advantage was its high-volume port connections with
England, allowing the city to receive most first copies of new
books published there. New York publishers could then pirate
versions of these books for sale in the city’s local market and the
rest of the country. As for pirating, it had none of those Puritan
scruples slowing down commerce.
The author also describes the waves of immigration from
Europe to New York, which swelled the city’s population and
helped meet the demand for workers in the factories and
centers of commerce. The reason why so many immigrants
chose to stay in New York, however, is not fully explained;
Glaeser contends that they “did not want to spend the time and
money to travel further.” While it is plausible to believe that an
immigrant who had traveled for months would be tempted to
stop where the boat dropped him off, one could argue that he
would only stay for an extended time if New York’s advantages
made doing so attractive. Besides a strong labor market
demand, New York City offered immigrants ethnic
neighborhoods. Each neighborhood had a network of contacts
to aid in finding housing and jobs. These rich and colorful
neighborhoods and networks are well described in historical
accounts as well as in the literature set in New York. The

J. Vernon Henderson is the Eastman Professor of Political Economy
at Brown University.
<j_henderson@brown.edu>

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

25

richness of the city’s dense neighborhoods is a theme that
continues into the modern era.
Glaeser’s discussion of the modern era focuses on New York
City as a financial capital where face-to-face contact and
immediacy of information are critical. The point he misses is
that this immediacy of information and ability to make face-toface contact exist in other industries as well, especially those
where the creative side is critical. The table shows some of the
major industries in New York. For each industry, the location
quotient is given—the ratio of New York’s (Manhattan
County’s) share of national employment to its share of total
national private employment. For example, for
headquarters, the location quotient is 1.6, that is, New
York’s 3 percent share of headquarters employment divided
by its 1.83 percent share of all national employment.
The table reveals that with the exception of FIRE (finance,
insurance, and real estate), New York is not really a headquarters
capital—its share of national headquarters employment (and
its share of headquarters establishments) is only modestly
above its share of general employment. The literature asserting
that the city’s economic base is driven by headquarters is
misguided. What New York does have is finance, as emphasized by
Glaeser, and services, such as advertising. New York is by far the
nation’s leading advertising agency city, with an even greater
concentration of sales than employment. New York City is also a
center for the arts. These activities have two key features: first, they
are New York’s leading exports; second, the creative activities such

Location Quotients of Selected Major Industries
in New York
Industry
Headquarters
FIRE headquarters
Financial services
Securities brokers
Business services
Advertising
Arts

1997 Quotient
1.6
5.5
6.4
13.4
4.1
8.0
3.8+

Source: Aarland et al. (2005).
Notes: The location quotient is defined as New York’s (Manhattan
County’s) share of national employment in industry x divided by its
share of total national employment. The numerator of the arts quotient
excludes arts employment in the twelfth and fifteenth congressional
districts. FIRE is finance, insurance, and real estate. The arts category
includes performing arts, publishing, museums, and broadcasting.

26

Commentary

as advertising, theater, and fashion are located in dense
neighborhoods, where people in these businesses interact at the
neighborhood level. New York’s success today is based in large
part on its dense commercial neighborhoods, where face-to-face
meetings and the exchange of information are essential.
Consider advertising. New York has more than 1,000
advertising agencies. These agencies are clustered throughout
southern Manhattan, although some clusters remain on or
near Madison Avenue. For these agencies, networking with
others in the creative design of ad campaigns is critical.
Networking can be formal, such as asking another agency to
contribute work for a campaign, or informal, such as
exchanging ideas over coffee or lunch.
The key questions are why advertisers are so concentrated in
New York and what role New York’s neighborhoods play. We do
not know all the answers, but several things are apparent. One
attraction of New York is its “buzz”—a great labor market of
young, creative, and ambitious people. For advertisers, the people
they sell to—broadcasters—are there as well. But what New York
also offers is an array of dense advertising agency neighborhoods
from which to choose. Arzaghi and Henderson (2005) uncover
two fascinating aspects of these neighborhoods. The implied
benefits of having more neighbors nearby dissipate over space
incredibly quickly. The authors find strong positive effects of
having more neighbors within 500 meters, some between 500 and
750 meters, but none beyond 750 meters. Clusters of advertising
agencies on one side of Manhattan do not network with agencies
on the other side. What are firms willing to pay to be at the center
of the action of a large cluster of agencies rather than isolated on
the fringes? In 1992, the average monthly rent for Class A office
space in southern Manhattan was $28 per square foot. The typical
advertising agency was willing to pay $10 more per square foot per
month for an increase of up to 50 neighbors nearby, which is close
to the maximum number of neighbors in any one census tract. But
then who pays the higher rent and locates in the dense clusters, and
who operates more in isolation?
Arzaghi and Henderson find spatial separation in the local
market. The highest-quality firms are the ones willing to pay
the most to be at the center of big clusters, while lower quality
firms operate on the fringes. Agencies in New York move
within the city, with new firms spinning off from old ones, on
an ongoing basis. For example, a new agency can set up on the
fringes of Manhattan, develop its talent and potential, and
move to the center of a large cluster where it pays higher rent.
Some employees will then spin off their own firm and move to
another cluster, and so on.

Part of the lifeblood of New York City is its vibrant
neighborhoods and dense centers of activity. A century or two
ago, part of this vitality was manifested by waves of immigrants
who clustered in the networks of their own ethnic neighborhoods. Today, some of the city’s vitality is manifested by the

different clusters of advertising agencies, fashion designers, and
artists scattered throughout New York. As long as those
individuals engaged in certain creative commercial activities
require face-to-face networking, New York will offer the dense,
vibrant neighborhoods that can help them to succeed.

FRBNY Economic Policy Review / December 2005

27

References

Aarland, K., J. Davis, J. V. Henderson, and Y. Ono. 2005. “Spatial
Organization of Firms.” Federal Reserve Bank of Chicago
Working Paper no. 2003-30. Revised May 2005.

Arzaghi, M., and J. V. Henderson. 2005. “Networking Off Madison
Avenue.” Brown University working paper, March. Available at
<http://www.econ.brown.edu/faculty/henderson/madison.pdf>.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
28

Commentary

Stuart S. Rosenthal and William C. Strange

The Geography of
Entrepreneurship in the
New York Metropolitan Area
1. Introduction
New York will be a great place when they finish it.
- Popular saying

N

ew York City is often used as a paradigm for all that is
urban. For instance, the analysis of New York in Jacobs
(1969) is explicitly presented as bearing on fundamental
aspects of urbanization in general, not just on New York. This
approach is easy to understand. Cities are defined by their scale
and density, and among the cities in the United States, New
York has the most: the most employment, the most population,
the most density. Almost any urban phenomenon that one
might want to study is present in New York, and New York’s
size means that the phenomenon in question is magnified and
thus easier to understand. This magnification makes the study
of New York an essential part of the study of cities in general,
and it is why the particular discussions of New York in Hoover
and Vernon (1959), Vernon (1960), and Chinitz (1961) have
had such long-lasting general impact on urban economics.
This paper also looks at New York as an urban paradigm.
Our focus is on New York’s constant change, as captured in the
famous unattributed quote above. The central aspect of New
York’s dynamism that we consider is entrepreneurship.
Specifically, we focus on the geography of entrepreneurship,
examining how the levels and character of nearby economic

Stuart S. Rosenthal is a professor of economics at Syracuse University;
William C. Strange is the RioCan Real Estate Investment Trust Professor
of Real Estate and Urban Economics at the University of Toronto.
<ssrosent@maxwell.syr.edu>
<wstrange@rotman.utoronto.ca>

activity influence the births of new establishments and the scale
at which they operate.
This paper builds primarily on research on agglomeration
economies. Much of the empirical work on agglomeration has
sought to estimate the effect on productivity of an
establishment’s local environment. The estimation has
sometimes involved direct estimates of productivity
(Henderson 2003) and has sometimes involved estimating
correlates of productivity, including wages (Glaeser and Mare
2001) and growth (Henderson, Kuncoro, and Turner 1995).1
Our paper is concerned with two productivity correlates:
establishment births and new-establishment employment.
Prior work on agglomeration and births has established the
importance of the metropolitan environment (Carlton 1983).
Rosenthal and Strange (2003) show that agglomeration effects
attenuate geographically for six standard industrial
classification (SIC) industries—software (SIC 7371-73, 75),
food products (SIC 20), apparel (SIC 23), printing and
publishing (SIC 27), fabricated metal (SIC 34), and machinery
(SIC 35)—that serve national and international markets. For
these industries, it appears that an establishment’s local
environment matters most.2
This paper employs geographically refined data from Dun &
Bradstreet together with geographic information systems (GIS)
software to study the spatial pattern of entrepreneurship in
New York City for a broad set of industry groups. The key
aspects of our analysis involve regressions of the number of

The authors gratefully acknowledge the financial support of the Kauffman
Foundation, the Center for Policy Research at Syracuse University, and the
Social Sciences and Humanities Research Council of Canada. They also thank
Robert Inman for helpful comments. Excellent research assistance was
provided by Yong Chen and Michael Eriksen. The views expressed are those of
the authors and do not necessarily reflect the position of the Federal Reserve
Bank of New York or the Federal Reserve System.
FRBNY Economic Policy Review / December 2005

29

births and the amount of new-establishment employment in a
census tract on variables that describe the tract’s local
environment. Two sets of such variables are constructed. The
first characterizes the total employment across all industries
within one mile, between one and five miles, and between five
and ten miles of the tract. These measure the degree of
urbanization of the tract, which Jacobs (1969) and others argue
is associated with productivity. The second set of variables
characterizes the employment in individual two-digit SIC
industries. These allow the identification of localization effects,
where the proximity to own-industry activity adds to
productivity (Marshall 1920).
We take a within-city approach to agglomeration, with the
identification of the determinants of the spatial pattern of
births and new-establishment employment coming from
variation in the data within the New York consolidated
metropolitan statistical area (CMSA). Although such an
approach is rare in the literature—Anderson, Quigley, and
Wilhelmson (2004) and Arzaghi and Henderson (2005) are
exceptions—theoretical work on agglomeration argues
forcefully that the effect should be modeled as decaying with
distance rather than being bounded by political borders.3
In addition to being closer to theories of agglomeration, our
within-city geographic approach has an important
econometric advantage: any effects that are fixed at the city
level are captured by the constant term. One such effect is
regional natural advantage. Recognition of the importance of
this effect goes back to Marshall (1920) at least. More recently,
Glaeser, Kolko, and Saiz (2001) show climate to be a strong
predictor of urban growth. To the extent that this sort of
natural advantage influences entrepreneurship at the regional
level, we control for it, and also for any other regionwide
natural advantage that might exist. Although we cannot fully
rule out the possibility that within-city variation in natural
advantages drives some of our results, we believe that most
natural advantages are regional. If so, then spatial variation in
activity within the New York CMSA will be driven primarily by
agglomeration economies and the spatial differences in
productivity they create. This seems to be especially likely when
analyzing the location of information-oriented industries that
are less sensitive to shipping costs.
Separate regressions are carried out for four one-digit
industry groups: manufacturing (SIC 21-39), wholesale trade
(SIC 50-51), services (SIC 70-89), and finance, insurance, and
real estate (FIRE, SIC 60-67). We also estimate models with
employment from all industries in the economy aggregated
together (eighty-two two-digit industries in all). In all of these
models, we include two-digit SIC-fixed effects to control for

30

The Geography of Entrepreneurship

characteristics common to enterprises throughout a given twodigit category. We also estimate one additional model for just
business services (SIC 73). This industry is considered
separately because of its importance in the local economy. In all
the models, we consider whether urbanization and localization
economies are present. More important, our geographically
refined data also allow us to consider whether these effects
attenuate geographically.
Our results are as follows. First, we document the extensive
variation within the New York CMSA in the types of business
activity that take place, including entrepreneurship. Second, in
our analysis of the sources of entrepreneurship, the density of
local employment (urbanization) and the amount of local
employment in an entrepreneur’s own industry (localization)
are both shown to affect entrepreneurship. The influence of
localization is always positive, while the effect of urbanization
is much smaller in magnitude at the margin. For some
industries, it is negative. Third, all of these agglomeration
economies are shown to attenuate with distance. Typically, the
effects of the environment beyond one mile are an order of
magnitude smaller than the effects of the more immediate
environment.
In the next section, we present evidence on the location of
economic activity within New York. Section 3 offers a simple
model of new-establishment formation and discusses the
agglomeration variables used in our estimation. The estimation
results are presented in Section 4.

2. Metropolis 2001: Location Patterns
in the New York Region
2.1 Overview
Nearly fifty years ago, the Graduate School of Public
Administration at Harvard University was asked to carry out a
comprehensive study of the New York region. This mammoth
effort resulted in nine monographs and a summary volume
(Vernon 1960). The New York Metropolitan Region Project
covered nearly every aspect of New York’s economy, including
its labor markets, housing markets, and industrial
organization. Geography was central to all of this analysis.
What goods and services were produced in New York and not
in other places because of New York’s preeminent and peculiar
place in the system of cities? Within New York, where were

different goods produced? Although the study of agglomeration economies was far from mature during the project, the
idea of external increasing returns played a central role in the
answers offered to these questions.
Our goals in this paper are obviously much more modest,
but they are related. We are interested in characterizing where
various activities take place within New York and how
agglomeration economies impact New York’s perpetual
reinvention of itself. This section concerns the first of these
goals. As will become apparent, our analysis departs from the
New York Metropolitan Region Project in at least one
important way: we analyze at a much more refined level of
geography.

2.2 Data
We are able to conduct our analysis at a more refined level of
geography by employing data from Dun & Bradstreet
Marketplace. This database provides a wealth of information
on establishments throughout the New York CMSA. We
employ data from 2001:2 to describe New York’s economic
environment. The data characterize an establishment’s activity
(using the primary standard industrial classification), its
employment, and its U.S. postal ZIP code location. We then
match ZIP codes to the census ZIP code tabulation area
(ZCTA) geography, as well as to the year 2000 census-tract
geography. This procedure enables us to convert all of the
employment data to census-tract geography, which we use as
our standard geographic unit of analysis.4 In future work, the
procedure will facilitate analysis of the relationship between
local employment and residential patterns. However, as noted
earlier, our focus in this paper is on employment and
entrepreneurial activity in manufacturing, wholesale trade,
FIRE, and services. We will address how the data are employed
in our estimation later in the discussion.

center of Manhattan, and is considerably less dense, with 350
people per square mile. Across the rest of the New York CMSA,
population density varies between these two extremes. This
intracity variation is one of the main reasons why our study
looks at agglomeration and entrepreneurship using within-city
variation.
The maps in Charts 1-4 depict employment densities
(employment per square mile) at the county level across the
metropolitan area. Right away, it is clear that with regard to
employment as well, Manhattan is different. Despite the
well-known problems of central cities in general and of
New York in particular, and despite the tendencies of
industries and households to decentralize, the high density
of activity in Manhattan remains unique in the New York
metropolitan area. This pattern holds for manufacturing
(SIC 20-39, Chart 1), wholesale trade (SIC 50-51, Chart 2),
services (SIC 70-89, Chart 3), and FIRE (SIC 60-67,
Chart 4). This result is somewhat surprising. Much popular
urbanism (such as Garreau [1991]) argues that the really
important parts of America’s cities are their peripheries. It is
certainly true that the changes taking place at the urban
fringe are significant. However, it is also true that their

Chart 1

Manufacturing Employment Density (Workers per
Square Mile)
County Level, 2001:2

2.3 County-Level Patterns
Before turning to our more geographically refined
characterization of economic activity in New York, we will
begin by painting a larger but somewhat less detailed picture at
the county level. The New York CMSA is made up of thirty-one
counties. They differ substantially. New York County, which is
essentially equivalent to Manhattan, is extremely dense, with
66,940 people per square mile (<http://www.factfinder
.census.gov>). Dutchess County is sixty-four miles from the

1,000–20,000 (1)
500–1,000 (4)
100–500 (10)
0–100 (16)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 MarketPlace files.
Note: Figures in parentheses are the number of counties in each category.

FRBNY Economic Policy Review / December 2005

31

Chart 2

Chart 4

Wholesale Trade Employment Density (Workers per
Square Mile)

FIRE Employment Density (Workers per Square Mile)
County Level, 2001:2

County Level, 2001:2

10,000–50,000 (1)
500–1,000 (1)
100–500 (7)
0–100 (22)

1,000–10,000 (1)
500–1,000 (1)
100–500 (9)
0–100 (20)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 MarketPlace files.
Source: Dun & Bradstreet, Inc., Second Quarter 2001 MarketPlace files.
Note: Figures in parentheses are the number of counties in each category.

Chart 3

Services Employment Density (Workers per
Square Mile)
County Level, 2001:2

10,000–50,000 (1)
1,000–10,000 (5)
500–1,000 (5)
100–500 (10)
0–100 (10)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 MarketPlace files.
Note: Figures in parentheses are the number of counties in each category.

32

The Geography of Entrepreneurship

Notes: Figures in parentheses are the number of counties in each category.
FIRE is finance, insurance, and real estate.

status as a fringe implies the existence of a center, and the
center still matters, at least for some cities. Of course, as we
observed, New York is unusually dense. Thus, the picture
from this analysis of New York may not apply to more
sparsely developed cities like Houston.
Not surprisingly, the industries differ in their patterns of
centralization. Comparing Charts 1 and 2 shows that
manufacturing and wholesale trade follow roughly similar
patterns, with the latter being more centralized. Given the
importance of services to all twenty-first-century cities, it is not
surprising that Chart 3 shows service sector employment
exceeding 100 workers per mile in more than half of New York
City’s counties. It is also not surprising that employment in the
FIRE industries is highly concentrated in and near Manhattan.
These are known to be highly agglomerated industries.

2.4 Tract-Level Patterns
One might believe that the centralization of the New York
CMSA is adequately depicted in the county maps (Charts 1-4).
However, the maps in Charts 5-8 reveal that this is not true.
They present employment densities at the census-tract level.
Charts 5-8 show, as the county-level maps do, that Manhattan

is overwhelmingly the center of the city’s employment. In fact,
for each of the four industry groups, the center of employment
is not just Manhattan, but Lower Manhattan, defined as
beginning at the southern end of Central Park. Even within
Lower Manhattan, there are places with greater and smaller
densities for each of the four industry groups. Thus, taken as a
whole, the charts clearly establish that there is micro-level
geographic concentration within the New York metropolitan
area.
We begin with Chart 5, which indicates that manufacturing
is concentrated in Midtown, specifically in the Fashion District

(formerly the more modestly named Garment District). There
exist smaller concentrations in the closest areas of Brooklyn,
Queens, the Bronx, and in New Jersey. Despite the deurbanization of manufacturing activity that took place in the
last half of the twentieth century, the manufacturing sector
remains important for New York City. In light of our earlier
claim that New York has been treated as an urban paradigm, it
is important to note that the persistence of manufacturing
activity is probably greater in New York than in other cities.
Chart 6 depicts wholesale trade employment density. As the
earlier county-level map revealed, the pattern for wholesale

Chart 5

Manufacturing Employment Density (Workers per Square Mile)
Census-Tract Level, 2001:2

50,000–500,000 (11)

25,000–50,000 (30)

10,000–25,000 (39)

1,000–10,000 (517)

0–1,000 (4,497)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 MarketPlace files.
Note: Figures in parentheses are the number of tracts in each category for the entire New York consolidated metropolitan statistical area.

FRBNY Economic Policy Review / December 2005

33

trade is very similar to the pattern for manufacturing. Both
industry groups reach their highest employment densities in
Midtown.
Chart 7 shows starkly just how much New York has become
a “service city.” For manufacturing, there are only eleven tracts
where employment density is greater than 50,000 workers per
square mile. For services, there are ninety-four tracts that reach
an employment density of at least 50,000. There are smaller
concentrations of manufacturing in the outer boroughs. The

parallel for services is that most of Brooklyn, Queens, and the
Bronx reach at least moderately concentrated levels of service
employment density. It is worth reiterating that although
service sector employment is present everywhere, it is especially
present in Lower Manhattan.
Chart 8 illustrates employment density for the FIRE
industry group. The chart reveals a somewhat different pattern.
Employment continues to reach its greatest densities in Lower
Manhattan, as with the other industries. Unlike the other

Chart 6

Wholesale Trade Employment Density (Workers per Square Mile)
Census-Tract Level, 2001:2

50,000–500,000 (8)

25,000–50,000 (12)

10,000–25,000 (28)

1,000–10,000 (285)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 MarketPlace files.
Note: Figures in parentheses are the number of tracts in each category for the entire New York consolidated metropolitan statistical area.

34

The Geography of Entrepreneurship

0–1,000 (4,761)

industries, though, for FIRE there are two centers. They are
located Downtown (at the lower tip of Manhattan) and in
Midtown. Also, relative to the other industry groups, there is
really very little high-density employment in FIRE outside
(both upper and lower) Manhattan.
Taken together, the maps in Charts 1-4 and 5-8 paint a
picture of a centralized city, both at the macro (county) and
micro (census-tract) levels. The pattern varies by industry, with
service employment reaching high densities across much of
Manhattan and at least moderate densities in the adjacent
areas. Other industries are concentrated more narrowly.

Manufacturing and wholesale trade are still important for
New York City; they are concentrated in Midtown. FIRE is also
concentrated there, but another concentration also exists
Downtown.
These maps describe the local business environment that
confronts an entrepreneur making the decisions of whether to
start up a new establishment, where to put it, and at what scale
to operate it. These will essentially be the regressors in our
models. The dependent variables are births of new
establishments and new-establishment employment.

Chart 7

Services Employment Density (Workers per Square Mile)
Census-Tract Level, 2001:2

50,000–500,000 (94)

25,000–50,000 (55)

10,000–25,000 (133)

1,000–10,000 (2,570)

0–1,000 (2,242)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 MarketPlace files.
Note: Figures in parentheses are the number of tracts in each category for the entire New York consolidated metropolitan statistical area.

FRBNY Economic Policy Review / December 2005

35

Chart 8

FIRE Employment Density (Workers per Square Mile)
Census-Tract Level, 2001:2

500,000–1,000,000
(2)

50,000–500,000
(39)

25,000–50,000
(22)

10,000–25,000
(28)

1,000–10,000
(261)

0–1,000
(4,742)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 MarketPlace files.
Notes: Figures in parentheses are the number of tracts in each category for the entire New York consolidated metropolitan statistical area. FIRE is finance,
insurance, and real estate.

2.5 Entrepreneurial Density
The maps in Charts 9-12 illustrate the density of newestablishment employment at the tract level. Specifically,
they describe geographic patterns of employment of
establishments in 2004:2 that are less than three years old. It
is well-known that many establishments have very short life
spans (see the references in Caves [1998]). Our births
variable thus understates the true amount of new-

36

The Geography of Entrepreneurship

establishment creation that took place over the period
because we do not take into account those companies that
were created after 2001:2 but closed before 2004:2. Having
said that, it is not obvious that using a shorter horizon
would have been preferable. In this case, our initial period
was chosen to characterize New York City before the
destruction and disruptions associated with September 11.
We chose to look at births over a longer horizon in part to
allow some of the effects of September 11 to work through

the system. Of course, adjustment remains incomplete as of
this writing, but some terminal date needed to be set.
It is immediately clear from Charts 9-12 that
entrepreneurial activity is highly concentrated. Furthermore,
new-establishment employment is greatest near the locations
identified in Charts 5-8 as having the most employment in the

various industry groups. These maps suggest the presence of
geographically attenuating agglomeration economies in
entrepreneurship where the effect is at least partly associated
with own-sector activity (localization).
In sum, the maps in this section paint a picture of the
New York CMSA as remarkably centralized, both at the macro

Chart 9

Manufacturing Employment Density (Workers per Square Mile) at Establishments Three Years of Age or Less
Census-Tract Level, 2004:2

5,000–50,000 (9)

1,000–5,000 (33)

100–1,000 (168)

25–100 (728)

0–25 (4,156)

Source: Dun & Bradstreet, Inc., Second Quarter 2004 MarketPlace files.
Note: Figures in parentheses are the number of tracts in each category for the entire New York consolidated metropolitan statistical area.

FRBNY Economic Policy Review / December 2005

37

and micro levels. Both the number of new establishments and
the employment they bring are also centralized. Entrepreneurial activity appears to be attracted to locations with large
amounts of activity in the same sector. This is as far as simple

descriptive devices like maps can take us. The next section sets
out a model that forms the basis for our estimation of the
relationship between the spatial allocation of business activities
and entrepreneurship.

Chart 10

Wholesale Trade Employment Density (Workers per Square Mile) at Establishments Three Years of Age or Less
Census-Tract Level, 2004:2

5,000–50,000 (5)

1,000–5,000 (13)

100–1,000 (127)

25–100 (504)

0–25 (4,445)

Source: Dun & Bradstreet, Inc., Second Quarter 2004 MarketPlace files.
Note: Figures in parentheses are the number of tracts in each category for the entire New York consolidated metropolitan statistical area.

38

The Geography of Entrepreneurship

Chart 11

Services Employment Density (Workers per Square Mile) at Establishments Three Years of Age or Less
Census-Tract Level, 2004:2

5,000–50,000 (20)

1,000–5,000 (85)

100–1,000 (661)

25–100 (2,014)

0–25 (2,314)

Source: Dun & Bradstreet, Inc., Second Quarter 2004 MarketPlace files.
Note: Figures in parentheses are the number of tracts in each category for the entire New York consolidated metropolitan statistical area.

FRBNY Economic Policy Review / December 2005

39

Chart 12

FIRE Employment Density (Workers per Square Mile) at Establishments Three Years of Age or Less
Census-Tract Level, 2004:2

5,000–50,000 (29)

1,000–5,000 (28)

100–1,000 (146)

25–100 (370)

0–25 (4,521)

Source: Dun & Bradstreet, Inc., Second Quarter 2004 MarketPlace files.
Notes: Figures in parentheses are the number of tracts in each category for the entire New York consolidated metropolitan statistical area. FIRE is finance,
insurance, and real estate.

40

The Geography of Entrepreneurship

3. Model and Estimation Strategy
3.1 Model
The heart of the model is agglomeration economies.
If agglomeration economies exist, then productivity
will vary spatially. This, in turn, implies that births of new
establishments will take place near existing concentrations of
employment, all else equal. However, all else may not be equal.
If there were a local source of natural advantage, firms would
agglomerate even though they had no external effect on each
other. For example, as discussed in Rosenthal and Strange
(forthcoming), the wine industry is concentrated in California
because of favorable climate and other natural features that
facilitate the growing of grapes. As we observed earlier, our
within-city approach controls for natural advantages that
operate at a regional level. To take that idea a step further, we
also include two-digit SIC-fixed effects in all of the models.
This allows the influence of regionwide natural advantages to
differ across two-digit industry subgroups by stripping away all
factors common to enterprises belonging to a given subgroup.
Even with these fixed effects, we cannot rule out the possibility
that local variation in natural advantages may still account for
a portion of the estimated attraction of new economic activity
to existing concentrations of employment. However, for two
reasons, which we elaborate on later, we believe that our results
largely reflect the influence of external economies of scale
rather than natural advantages. To anticipate, the first reason is
that some of our industry groups seem to be quite footloose,
such as services and FIRE. In addition, the attenuation patterns
we document implicitly suggest the presence of factors whose
influence dissipates rapidly, a feature that seems to better fit
local variation in agglomeration than natural advantages.
We begin with a model adapted from Rosenthal and Strange
(2003). Suppose that the price of output is normalized to 1.
In this case, an establishment generates profit equal to
π ( y) = a ( y) f ( x ) – c ( x ), where a ( y ) shifts the production
function f ( x ) , y is a vector of local characteristics (the
components of which will be clarified below), and x is a vector
of factor inputs that cost c ( x ). Input quantities will be chosen
to maximize profits by satisfying the usual first-order
conditions. Employment ( n ), for example, is chosen such that
a ( y) ∂ f ( x ) ⁄ ∂ n – ∂c ( x ) ⁄ ∂ n = 0 .
Establishment births occur if a firm can earn positive
profits, with all inputs chosen at their profit-maximizing levels.
Establishments are heterogeneous in their potential
profitability. This feature is captured by rewriting the profit

function as π ( y, ε ) = max x a ( y) f ( x )( 1 + ε ) – c ( x ). We suppose
that ε is independent and identically distributed across
establishments according to the cumulative distribution
function Φ ( ε ) . For any y , there is a critical level ε ∗ ( y ) such
that π ( y, ε ∗ ( y) ) = 0 and π ( y, ε ) >(<) 0 as ε >(<) ε ∗ ( y ). In
this case, the probability that an establishment is created is
Φ (ε∗ (y)) .
We assume that new establishments are opened at locations
chosen from among all of the census tracts in the New York
CMSA, j = 1 , …, J . We also assume that location and
employment decisions are made taking the prior economic
environment (2001:2) as given. Let the vector y j describe the
local characteristics of each tract. Aggregating over
establishments in a given tract gives the number of births (B)
and total new-establishment employment (N) in industry i
and tract j . We express these as follows:
(1)

Bi j = byi j + b m + b i + ε b, i j ,

(2)

Ni j = nyi j + n m + n i + ε n, i j ,

where ε b and ε n are error terms, b and n are vectors of
coefficients, b m and n m are metrowide constant terms, and bi
and n i are industry-fixed effects. The bm and n m terms capture
any characteristics that impact entrepreneurship that are
common across all industries in the New York metropolitan
area. The industry-specific fixed effects capture any attributes
that are common to entrepreneurship throughout that
industry in the New York area. Together, the metrowide
constant and the industry-fixed effects control for a range of
natural advantages, as we observed earlier.
In addition, these terms are also likely to capture a number
of other unobserved determinants of entrepreneurship that
might vary geographically.5 For example, Blanchflower,
Oswald, and Stutzer (2001) report that “latent entrepreneurship,” the unfulfilled desire for self-employment, varies
substantially across countries. It is reasonable to suspect that it
might also vary between cities. Black, de Meza, and Jeffries
(1996) show the availability of collateral to be an important
determinant of new-enterprise creation in the United
Kingdom. The entrepreneur’s own housing is shown to be the
single most important source of such collateral. Since housing
markets in larger cities are different than in smaller cities, this
may be another metrowide effect captured in the model-fixed
effects. Furthermore, there is a well-documented correlation
between entry and failure. See Caves (1998) for a review of this
literature. This correlation implies that resources that can be
used by new establishments may be more plentiful where there
has previously been activity of a similar sort. Carlton (1983)

FRBNY Economic Policy Review / December 2005

41

includes this in his concept of the “birth potential” of an area.
This is clearly an important issue in estimation where
identification is based on intercity variation in the data. In our
case, however, the identification comes from intracity
variation. As long as firms that fail are free to choose any
location within the CMSA, this effect will be captured by the
fixed effects.
As discussed above, local variation in agglomeration that
influences productivity will affect births and employment at
the new establishments. Thus, the vector yi j will characterize
the spatial distribution of employment as perceived by
industry i in tract j . Specifically, yi j includes the level of
employment within and outside industry i (for i = 1, …, I )
within various distances of the geographic centroid of tract j .
These variables define the level of agglomeration associated
with a given tract and can be measured with our data. We now
explain how.

3.2 Concentric Ring Variables
As discussed above, we employ data from Dun & Bradstreet in
our analysis. Our goal is to assess the relationship between a
census tract’s local business environment and establishment
births and birth employment. To do this, we characterize the
environment of each tract in our sample according to the
2001:2 level of employment. The first step is to compute for
each tract both the total level of employment and the level of
employment in each two-digit industry. It is worth
emphasizing that in our estimation, our employment variables
will then measure activity at the two-digit industry level, and
not at the more general one-digit-level industry group.
The next step is to create a set of concentric ring variables for
both own-industry and aggregate employment. These variables
will allow the measurement of the geographic extent of
agglomerative externalities. They are calculated as follows.
First, employment in a given tract is treated as being uniformly
distributed throughout the tract. Then, using mapping
software, we draw circles of radius ri , i = 1, 5, and 10 miles
around the geographic centroid of each census tract in the
New York CMSA. The level of own-industry employment
contained within a given circle is then calculated by
constructing a proportional (weighted) summation of the
own-industry employment for those portions of the tracts
intersected by the circle. For example, if a circle includes all of
tract 1 and 10 percent of the area of tract 2, then employment
in the circle is set equal to the employment in tract 1 plus
10 percent of the employment in tract 2. The same procedure
is used to calculate the level of other-industry employment

42

The Geography of Entrepreneurship

within each circle. Differencing employment levels for adjacent
circles (by employment type) yields estimates of the levels of
own- and other-industry employment within a given
concentric ring. Thus, the 5-mile ring ( r 5 ) reflects employment
between the 1- and 5-mile circles, and so on out to 100 miles.
Table 1 describes our data, including the rings.6

3.3 Tobit Estimation
We estimate (1) and (2) using a Tobit specification to account
for the censoring of both kinds of entrepreneurial activity at
zero. An alternative would have been to estimate the number
of new establishments in a count model, while estimating
new-establishment employment by Tobit. We chose to
estimate both by Tobit in order to treat both aspects of
entrepreneurship symmetrically. This raises an econometric
issue because noisy estimates of the fixed effects in nonlinear
models typically lead to inconsistent estimates of the slope
coefficients (see, for example, Chamberlain [1980, 1984] and
Hsiao [1986]). Also, Tobit models are known to be more
sensitive to distributional assumptions than are linear
regressions. Our primary response to this issue is that bias
resulting from noisy estimates of fixed effects in nonlinear
models tends to go toward zero as the number of observations
per fixed effect becomes arbitrarily large. Since our sample
has 5,211 tracts per fixed effect (the number of tracts in the
New York CMSA), inconsistency arising from noisy estimates
of the fixed effects is hoped to be small.7

4. The Geography of Entrepreneurship
4.1 Births
This section presents estimates of models relating
entrepreneurship to the local business environment as defined
by the concentric ring variables described above. We begin with
estimates of (1), the new-establishment births model. All
estimation is carried out at the census-tract level.
Table 2 presents two models: Model 1 deals only with
urbanization, the scale of aggregate activity; Model 2 adds
variables capturing localization, the scale of activity in an
establishment’s own industry. In all models, we include
variables capturing activity in an establishment’s immediate
vicinity (within one mile), nearby (between one and five
miles), and further away (within ten miles).

Table 1

Variable Means per Two-Digit Industry and Census Tract by County: All Industries
Existing Own-Industry Employment

State

County

CT
CT
CT
CT
NJ
NJ
NJ
NJ
NJ
NJ
NJ
NJ
NJ
NJ
NJ
NJ
NJ
NJ
NY
NY
NY
NY
NY
NY
NY
NY
NY
NY
NY
NY
PA

Fairfield
Litchfield
Middlesex
New Haven
Bergen
Essex
Hudson
Hunterdon
Mercer
Middlesex
Monmouth
Morris
Ocean
Passaic
Somerset
Sussex
Union
Warren
Bronx
Dutchess
Kings
Nassau
New York
Orange
Putnam
Queens
Richmond
Rockland
Suffolk
Westchester
Pike

Total

New
Census-Tract
County Own-Industry
FIPS Code Establishments
9001
9005
9007
9009
34003
34013
34017
34019
34021
34023
34025
34027
34029
34031
34035
34037
34039
34041
36005
36027
36047
36059
36061
36071
36079
36081
36085
36087
36103
36119
42103

New
Census-Tract
Own-Industry
Establishment
Employment

Within
One Mile

Within One to Within Five to
Five Miles
Ten Miles

Existing All-Industry Employment

Within
One Mile

Within One
to Five Miles

Within Five
to Ten Miles

0.21
0.14
0.16
0.13
0.29
0.14
0.12
0.24
0.18
0.19
0.20
0.25
0.19
0.24
0.25
0.14
0.19
0.16
0.05
0.11
0.06
0.15
0.36
0.17
0.16
0.05
0.07
0.16
0.14
0.13
0.15

1.25
0.64
0.66
0.80
1.57
2.39
0.82
1.65
1.68
1.04
1.13
2.43
0.58
1.28
2.45
0.48
0.99
0.52
0.23
0.51
0.25
0.91
4.21
0.81
0.47
0.25
0.24
0.63
0.74
0.79
0.58

72
7
17
61
128
200
277
9
167
75
33
46
18
153
40
5
114
7
255
17
327
108
3,460
10
5
247
70
37
41
93
1

976
157
314
889
3,949
3,590
22,067
182
1,454
1,547
605
1,085
356
2,638
933
122
2,610
158
5,454
250
11,182
2,313
25,347
199
162
8,984
1,684
870
926
1,694
23

1,806
439
962
1,748
18,220
14,174
26,047
689
2,081
4,004
1,453
3,073
795
7,713
3,017
438
7,319
464
27,965
478
28,917
5,898
21,184
490
666
25,563
13,967
2,533
2,349
4,923
83

5,807
564
1,344
4,959
10,334
16,240
22,428
708
13,521
6,081
2,662
3,717
1,471
12,410
3,264
442
9,223
581
20,622
1,350
26,514
8,736
280,283
811
394
19,979
5,669
3,032
3,341
7,551
72

79,052
12,709
25,469
71,989
319,865
290,762
1,787,452
14,771
117,810
125,333
49,032
87,850
28,865
213,670
75,579
9,856
211,406
12,825
441,752
20,259
905,770
187,393
2,053,141
16,148
13,153
727,692
136,435
70,450
75,021
137,237
1,843

146,311
35,570
77,939
141,613
1,475,853
1,148,106
2,109,836
55,775
168,560
324,359
117,726
248,917
64,362
624,716
244,397
35,474
592,868
37,622
2,265,155
38,752
2,342,297
477,736
1,715,933
39,704
53,913
2,070,562
1,131,321
205,175
190,269
398,743
6,713

0.14

0.98

348

6,193

14,429

28,151

501,593

1,168,765

Source: Dun & Bradstreet, Inc., Second Quarter 2001 and Second Quarter 2004 MarketPlace files.
Notes: Eighty-two industries are represented (standard industrial classifications codes 1-97). “New” refers to establishments three years of age or less.
New-establishment and new-employment counts are from 2004:2; existing employment counts are from 2001:2. FIPS is federal information processing standards.

FRBNY Economic Policy Review / December 2005

43

Table 2

Number of Establishments Three Years of Age or Less in 2004:2
All Industries

Manufacturing

Wholesale Trade

FIRE

Services

Business Services

1.56E-03
(101.60)

6.70E-04
(45.40)

5.67E-03
(45.70)

1.79E-03
(54.66)

2.73E-03
(62.50)

1.44E-02
(37.64)

One to five miles

2.36E-06
(1.71)

2.37E-05
(9.96)

-1.10E-04
(-6.42)

-3.08E-05
(-6.15)

3.53E-06
(0.56)

-1.59E-04
(-3.03)

Five to ten miles

-9.64E-05
(-66.74)

-5.22E-05
(-33.31)

-5.58E-05
(-5.49)

-7.11E-05
(-23.55)

-1.34E-04
(-35.53)

-5.69E-04
(-18.43)

Model 1
All workers (1,000)
Zero to one mile

Memo:
SIC-fixed effects
Censored observations
Uncensored observations
Log-likelihood
Pseudo R2
Model 2
Own SIC workers (1,000)
Zero to one mile

82
235,198
186,893
-275,426.87
0.27

20
76,421
27,799
-34,760.02
0.21

2
830
9,592
-14808.08
0.07

7
16,793
19,684
-22,357.75
0.20

15
20,092
58,073
-92536.19
0.14

8.32E-02
(137.09)

5.52E-02
(50.78)

2.81E-01
(40.72)

3.85E-02
(37.00)

9.78E-02
(89.26)

2.86E-01
(15.55)

One to five miles

-6.17E-04
(-7.04)

1.19E-04
(0.61)

3.84E-03
(2.31)

-2.20E-04
(-1.22)

-7.50E-04
(-4.35)

6.46E-02
(12.85)

Five to ten miles

-2.39E-03
(-36.96)

1.13E-03
(8.30)

4.35E-03
(3.84)

-1.65E-04
(-1.30)

-3.66E-03
(-34.61)

2.04E-02
(8.18)

2.79E-04
(11.69)

3.21E-04
(20.08)

-3.86E-03
(-14.83)

6.04E-04
(13.54)

1.30E-07
(-1.35)

-1.89E-02
(-8.43)

One to five miles

5.82E-06
(2.94)

2.24E-05
(8.49)

-1.66E-04
(-3.04)

-1.70E-05
(-2.41)

1.64E-05
(2.10)

-6.74E-03
(-12.11)

Five to ten miles

-5.27E-05
(-30.56)

-5.68E-05
(-31.96)

-1.80E-04
(-5.13)

-6.31E-05
(-14.40)

-2.68E-05
(-4.26)

-2.00E-03
(-7.67)

All workers (1,000)
Zero to one mile

Memo:
SIC-fixed effects
Censored observations
Uncensored observations
Log-likelihood
Pseudo R2

82
235,198
186,893
-263,299.55
0.31

20
76,421
27,799
-33,372.00
0.24

2
830
9,592
-14035.75
0.12

7
16,793
19,684
-21,624.79
0.23

15
20,092
58,073
-87,534.67
0.19

Source: Dun & Bradstreet, Inc., Second Quarter 2001 and Second Quarter 2004 MarketPlace files.
Notes: t-ratios are in parentheses. SIC is standard industrial classification (code); FIRE is finance, insurance, and real estate.

44

–
22
5,189
-11,720.36
0.07

The Geography of Entrepreneurship

–
22
5,189
-11,523.95
0.08

The first result to notice from Model 1 is that the
urbanization of the immediate environment has a positive
effect on births for all four industry groups. Overall, the effect
is that adding 1,000 workers is associated with .0016 newestablishment births. For manufacturing, adding an additional
1,000 workers within one mile adds .0006 births. For wholesale
trade, the marginal effect of 1,000 workers within one mile is
.0057 births. For services, the effect is .0027 births. For FIRE, it
is .0018 births. For business services, the effect is the largest,
.0144. The effect is significant for all four industry groups.
The effects are also economically meaningful. As we noted
earlier, the mean population density is much greater in
Manhattan than in Dutchess County at the edge of the city
(66,940 per square mile compared with 350 per square mile).
Commuting patterns within the metropolitan area cause
differences in employment density to be even greater: for the
one-mile ring, the mean level of employment is 280,283 in
Manhattan and 3,717 in Dutchess County (Table 1). Changing
only the one-mile employment level in Dutchess County to the
Manhattan level would result in .43 additional new
establishments per tract. By comparison, the mean number of
new establishments in a tract in Dutchess County is .25.
The next result to notice in Table 2 is that the effect
attenuates fairly rapidly. For each industry group, the
coefficient for employment in the one-to-five-mile ring is at
least an order of magnitude smaller than the coefficient in the
one-mile ring. This attenuation is very clear in Chart 13. The
decay is especially pronounced in business services. The
attenuation of the effect of the local business environment is a
result that persists through nearly every specification in this
paper. The result suggests that urban interactions are highly
local in nature. In other words, a business’s neighborhood
matters.
Model 2 considers urbanization and localization together. It
is immediately apparent that controlling for activity in a firm’s
own industry impacts the estimates of the effect of employment
in all industries. For wholesale trade, services, and business
services, the effect of additional total employment within one
mile is either no longer significant or is negative. It is significant
for all industries, FIRE, and manufacturing, but the effect is
reduced by an order of magnitude in the first two cases by half
for the last.
In contrast, the effects of localization are positive and
significant in every case. For all industries, adding 1,000

workers in a firm’s own industry (two-digit SIC) within one
mile is associated with .0832 additional new-establishment
births. For manufacturing, an increase of 1,000 of ownindustry employment within one mile produces an additional
.0552 births. It is important to reiterate: this is the effect of
1,000 additional workers in the establishment’s own two-digit
SIC code. It is not the effect of 1,000 additional workers in the
entire manufacturing industry group. For wholesale trade, the
effect is even larger, at .2810 births; in services, the effect is
.0978 births. In FIRE and business services, respectively, the
effects are .0385 and .2860. These effects are all significant. To
sum up, it appears that some of the urbanization effects present
in Model 1 are instead really localization effects.
One result that Model 2 shares with Model 1 is that if
agglomeration effects exist, they attenuate. The top panel of
Chart 14 presents the urbanization coefficients. As we
discussed, many are negative or are insignificant. The rest are
small. Nevertheless, these coefficients attenuate. The picture in
the bottom panel of Chart 14 is much clearer. Localization
coefficients attenuate in much the same way that urbanization
coefficients do in the urbanization-only Model 1. In this case,
attenuation is most sharp for business services and wholesale
trade.

Chart 13

Model 1: Urbanization Effects
Dependent Variable: Number of Establishments
Three Years of Age or Less in 2004: 2
Impact of 1,000 more workers in 2001: 2
1.60E-02
1.40E-02
All industries
FIRE
Manufacturing
Business services
Services
Wholesale trade

1.20E-02
1.00E-02
8.00E-03
6.00E-03
4.00E-03
2.00E-03
0.00E+00
-2.00E-03

5 to 10
1 to 5
0 to 1
Distance to census-tract centroid (miles)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 and
Second Quarter 2004 MarketPlace files.
Note: FIRE is finance, insurance, and real estate.

FRBNY Economic Policy Review / December 2005

45

4.2 Birth Employment

Chart 14

Model 2: Urbanization and Localization Effects
Dependent Variable: Number of Establishments
Three Years of Age or Less in 2004: 2
Impact of 1,000 more workers in 2001: 2
5.00E-03

Urbanization Effects

0.00E+00
-5.00E-03
All industries
FIRE
Manufacturing
Business services
Services
Wholesale trade

-1.00E-02
-1.50E-02
-2.00E-02
3.00E-01

Localization Effects

2.50E-01
2.00E-01
1.50E-01
1.00E-01
5.00E-02
0.00E+00
-5.00E-02
5 to 10
0 to 1
1 to 5
Distance to census-tract centroid (miles)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 and
Second Quarter 2004 MarketPlace files.
Note: FIRE is finance, insurance, and real estate.

The discussion thus far has focused on the number of newestablishment births taking place in a census tract. This is one
natural measure of the amount of entrepreneurial activity
taking place there. Yet it misses one particularly important
aspect of entrepreneurship: the scale of entry. We now estimate
a model that addresses this aspect.

46

The Geography of Entrepreneurship

The results reported in Table 3 are estimates of (2), the model
of employment at new establishments. As we observed, these
are firms created between 2001:3 and 2004:4. As before, we
begin with a model including only urbanization coefficients,
Model 1. The evidence of urbanization effects here is similar to
the evidence in Table 2 (Model 1). For all industries, the
presence of an additional 1,000 workers within one mile is
associated with .0375 more workers at new establishments. For
all industry groups, total employment within one mile also has
a significant effect on birth employment. The presence of 1,000
additional employees within one mile of a census tract
increases new-establishment employment by .0368 in
manufacturing, by .0510 in wholesale trade, by .1270 in FIRE,
by .0296 in services, and by .1420 in business services. All are
highly significant.
As with the new-establishment births model in Table 2, the
attenuation of the urbanization effects is striking. Chart 15
depicts these effects. For all employment and for each of the
individual industry groups, the effect attenuates by an order of
magnitude between the one- and five-mile rings. As with the
urbanization effects in the births model (Chart 13), business
services exhibits the largest one-mile ring coefficient and the
sharpest attenuation.
Table 3 also presents a model that includes both localization
and urbanization variables in a regression of newestablishment employment. As in Table 2’s births model,
including localization variables impacts the estimates of
urbanization effects. In this case, wholesale trade takes on a
negative sign for the one-mile ring (see the top panel of
Chart 16), as do all of the ring coefficients for business services.
The other three industry groups and all employment have
positive and significant coefficients. Although these
coefficients are smaller than they are in Model 1, they are not as
reduced in size as they are when moving between the
urbanization-only and urbanization-and-localization models
for births.

Table 3

Employment at Establishments Three Years of Age or Less in 2004:2
All Industries

Manufacturing

Wholesale Trade

FIRE

Services

Business Services

3.75E-02
(49.08)

3.68E-02
(25.91)

5.10E-02
(36.41)

1.27E-01
(19.88)

2.96E-02
(49.57)

1.42E-01
(30.87)

One to five miles

4.56E-04
(4.71)

2.35E-03
(10.45)

-8.89E-04
(-4.60)

-2.71E-03
(-2.76)

-2.37E-05
(-0.28)

-1.40E-03
(-2.21)

Five to ten miles

-1.90E-03
(-27.66)

-3.63E-03
(-24.51)

-5.64E-04
(-4.91)

-3.31E-03
(-5.64)

-1.01E-03
(-19.67)

-3.39E-03
(-9.10)

Model 1
All workers (1,000)
Zero to one mile

Memo:
SIC-fixed effects
Censored observations
Uncensored observations
Log-likelihood
Pseudo R2
Model 2
Own SIC workers (1,000)
Zero to one mile

82
235,198
186,893
-973,247.04
0.05

20
76,421
27,799
-152914.36
0.04

2
830
9,592
-38023.11
0.02

7
16,793
19,684
-123836.86
0.01

15
20,092
58,073
-241323.35
0.03

–
22
5,189
-24641.01
0.02

1.37E+00
(41.68)

3.31E+00
(31.47)

2.30E+00
(28.28)

2.72E+00
(12.98)

8.87E-01
(55.40)

2.20E+00
(9.66)

One to five miles

-3.86E-02
(-7.88)

-2.32E-02
(-1.23)

2.08E-02
(1.07)

4.89E-02
(1.34)

-1.27E-02
(-5.09)

4.56E-01
(7.32)

Five to ten miles

9.88E-03
(4.64)

1.58E-01
(11.87)

3.01E-02
(2.25)

-4.86E-03
(-0.19)

-2.57E-02
(-16.67)

1.18E-01
(3.82)

1.57E-02
(17.39)

1.44E-02
(9.01)

-2.68E-02
(-8.77)

4.38E-02
(4.86)

4.77E-03
(6.67)

-1.15E-01
(-4.13)

One to five miles

1.05E-03
(8.02)

2.64E-03
(10.21)

-1.01E-03
(-1.58)

-3.51E-03
(-2.46)

2.68E-04
(2.25)

-4.82E-02
(-6.99)

Five to ten miles

-2.06E-03
(-24.99)

-4.61E-03
(-26.35)

-1.42E-03
(-3.44)

-2.81E-03
(-3.20)

-2.67E-04
(-3.45)

-1.07E-02
(-3.32)

All workers (1,000)
Zero to one mile

Memo:
SIC-fixed effects
Censored observations
Uncensored observations
Log-likelihood
Pseudo R2

82
235,198
186,893
-972,094.22
0.05

20
76,421
27,799
-152333.34
0.04

2
830
9,592
-37636.15
0.03

7
16,793
19,684
-123735.40
0.01

15
20,092
58,073
-239348.99
0.04

–
22
5,189
-24571.90
0.02

Source: Dun & Bradstreet, Inc., Second Quarter 2001 and Second Quarter 2004 MarketPlace files.
Notes: t-ratios are in parentheses. SIC is standard industrial classification (code); FIRE is finance, insurance, and real estate.

FRBNY Economic Policy Review / December 2005

47

Chart 15

Model 1: Urbanization Effects
Dependent Variable: Employment at Establishments
Three Years of Age or Less in 2004: 2
Impact of 1,000 more workers in 2001: 2
1.60E-01
1.40E-01
All industries
FIRE
Manufacturing
Business services
Services
Wholesale trade

1.20E-01
1.00E-01
8.00E-02
6.00E-02
4.00E-02
2.00E-02
0.00E+00
-2.00E-02

Table 3 and the bottom panel of Chart 16 clearly show that
localization has a positive and significant effect on newestablishment employment for all industries and for the
various individual industry groups. The one-mile coefficient is
greatest for manufacturing. It implies that an increase in the
number of own-industry workers within one mile is associated
with an increase in new-establishment employment of 3.3100
workers. The effects are of the same order of magnitude for
(in order of size) FIRE, wholesale trade, and business services.
They are positive and significant, if somewhat smaller, for all
industries and services. Once again, for each industry
regression, the effects attenuate sharply with distance.

5 to 10
1 to 5
0 to 1
Distance to census-tract centroid (miles)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 and
Second Quarter 2004 MarketPlace files.
Note: FIRE is finance, insurance, and real estate.

Chart 16

Model 2: Urbanization and Localization Effects
Dependent Variable: Employment at Establishments
Three Years of Age or Less in 2004: 2
Impact of 1,000 more workers in 2001:2
6.00E-02
Urbanization Effects
4.00E-02
2.00E-02
0.00E+00
-2.00E-02
-4.00E-02

All industries
FIRE
Manufacturing
Business services
Services
Wholesale trade

-6.00E-02
-8.00E-02
-1.00E-01
-1.20E-01
-1.40E-01
3.50E+00

Localization Effects

3.00E+00
2.50E+00
2.00E+00
1.50E+00
1.00E+00
5.00E-01
0.00E+00
-5.00E-01
5 to 10
0 to 1
1 to 5
Distance to census-tract centroid (miles)

Source: Dun & Bradstreet, Inc., Second Quarter 2001 and
Second Quarter 2004 MarketPlace files.
Note: FIRE is finance, insurance, and real estate.

48

The Geography of Entrepreneurship

4.3 The Sources of Agglomeration Economies
We have thus far shown that both urbanization and
localization are related to two aspects of entrepreneurial
activity: the births of new establishments and the total
employment of new establishments. These results relate most
closely to the findings of Rosenthal and Strange (2003), who
also estimate models of births and birth employment. One very
important difference is that the authors look at six select
manufacturing industries (including a computer software
aggregate), chosen in part because each receives large numbers
of births and each exports nationally and internationally.
A large number of births reduces the number of censored
observations in the Tobit models, while marketing abroad
likely increases the degree to which a company’s location is
influenced by local variation in agglomeration economies as
opposed to within-city variation in natural advantages. This
paper extends Rosenthal and Strange (2003) by focusing on
broad one-digit industry groups, using fixed effects to control
for two-digit industry subgroups. This procedure restricts the
slope coefficients to being alike across industry subgroups, but
grouping industries at the one-digit level reduces the number
of censored observations. Despite the difference in
specification, the results in this paper are consistent with those
in Rosenthal and Strange (2003) in terms of showing that rapid
attenuation is the norm.
The result that attenuation is rapid is also consistent with
the finding in the few other studies that consider the decay of
agglomeration economies. Anderson, Quigley, and
Wilhelmson (2004) consider the local impacts of a shift in the
organization of higher education in Sweden. The policy
change—a significant decentralization—is a kind of natural
experiment. The key finding is that the effects are highly
localized. Arzaghi and Henderson (2005) show that external
economies in advertising are also highly localized.8

An important issue touched on earlier is the ability of the
estimation to separate agglomeration effects from natural
advantages or other potential reasons why entrepreneurs
should be attracted to locations with high levels of existing
activity. This would not be a problem for any natural advantage
that affected the entire metropolitan area. There are, however,
natural advantages that are more local. For instance, a port
location may be more productive for a firm engaged in
wholesale trade. In this situation, natural advantages will lead
to high levels of employment, so the coefficients on
employment levels may reflect both natural advantages and
agglomeration effects. Our results show that the effect of
existing activity attenuates rapidly. For this to be explained by
a natural advantage, it would have to be one that attenuated
rapidly as well. This does not seem to describe a port, since
shipping costs are relatively low, especially for informationoriented industries such as FIRE and services.
If the influence of within-city variation in natural
advantages is at most weak, this naturally leads to the question
of what agglomeration economies might be present locally that
are so much weaker at larger distances. This is a particular
aspect of the more general question of what the sources of
agglomeration economies might be. This larger question has
proven very difficult to address. Many plausible sources of
agglomeration economies have been proposed. Marshall’s
(1920) list involves labor market pooling, input sharing, and
knowledge spillovers. Other explanations involve the
availability of consumption externalities (Glaeser, Kolko, and
Saiz 2001) and the management of uncertainty (Strange,
Hejazi, and Tang 2004). There are many other possibilities, as
set out in the survey by Duranton and Puga (2004).
Unfortunately, in many respects, the implications for births,
wages, and productivity of these possible sources are fairly
similar. This makes it difficult to identify particular forces that
give rise to agglomeration economies.
This paper’s key result regarding microfoundations is that
agglomeration economies attenuate rapidly. This does seem to
favor some sources of agglomeration economies over others. In
a sense, agglomeration economies are a transportation cost
issue. Glaeser (1998) suggests the following way to think about
this issue: There are costs of moving goods, costs of moving
people, and costs of moving ideas. The first set of costs is not
especially important for the modern business because the costs
of moving goods have shrunk dramatically over the past 100
years. People are more costly to move, with urban commuting
being a particularly salient example. Although information can
easily be transported electronically, ideas and knowledge are
almost certainly costly to transport. The type of unexpected
synergies that Jacobs (1969) sees as being responsible for the

creation of new work depend on random interactions. These
are much more likely to occur if the interacting parties are quite
close to each other.
All of this suggests that our attenuation result is more
consistent with the high costs of moving ideas than with the
other sources of an agglomeration economy. To the extent that
this interpretation is correct, the ideas being transported must
be Marshallian knowledge spillovers or some other type of
social interaction. In either case, high transportation costs
would be associated with rapid decay. Of course, it is important
to recognize that this interpretation of the observed patterns
has been quite casual. Future research is required to disentangle
more precisely the many agglomerative forces at work.

5. Conclusion
This paper analyzes the spatial pattern of entrepreneurial
activity in the New York consolidated metropolitan statistical
area. Since entrepreneurship takes place against a backdrop of
current activity, we begin by looking at the geography of
activity in four industry groups: manufacturing, wholesale
trade, services, and FIRE. All are shown to be centralized
around Manhattan and the nearer boroughs, with FIRE being
the most centralized. Entrepreneurial activity is also
centralized, with the pattern being quite similar to the pattern
for levels of activity. This suggests that some force is leading
entrepreneurs to agglomerate. There are many candidates that
are consistent with the data, including natural advantages and
Marshallian external economies.
In order to understand the relationship better, we estimate
models of new-establishment births and new-establishment
employment as functions of the local business environment. In
a model that includes only one agglomeration variable—
urbanization, total nearby employment—urbanization is
shown to be positively related to both births and birth
employment. If instead an additional agglomeration variable is
also included—localization, employment in an establishment’s
own industry—then the results change. For all of the industry
groups, localization is shown to be positively associated with
both measures of entrepreneurship. For most of the industry
groups, the influence of urbanization is greatly reduced,
sometimes negative, and no longer significant after controlling
for localization.
In our analysis of entrepreneurship, we take a geographic
approach to agglomeration rather than a political one.
Specifically, we estimate the effects of activity taking place very
close to a census tract (within one mile), fairly close (between

FRBNY Economic Policy Review / December 2005

49

one and five miles), and further away (between five and ten
miles). For nearly all of our many models, the effects of a tract’s
business environment are shown to attenuate sharply. The
effect at five miles is typically at least one order of magnitude
smaller than the effect within one mile. This result speaks to the
question, what is a city? The answer seems to be that many of
the spatial interactions that are central to cities are quite local.
When entrepreneurs must decide on the best location to open
an establishment, they choose one that is close to existing
activity, especially in their own industry. It should be
recognized, however, that by estimating these effects within
one city, we hold constant those factors that are common to
businesses throughout the New York CMSA. Thus, the fact that
we identify a local effect does not preclude the existence of
other effects that operate across cities and regions.

50

The Geography of Entrepreneurship

There are many forces that can explain our paper’s
agglomeration results. Unfortunately, the estimation does not
enable us to identify specific agglomerative forces that are at
work. Whatever the forces may be, however, they appear to
operate at a narrow level of geography. If there are Marshallian
agglomeration economies, then the economies must attenuate
rapidly. This observation suggests—but of course does not
prove—that the effect might be some type of social spillover,
since ideas and learning are costly to transport and allegiances
are costly to maintain over a great distance. If there are also, or
are instead, natural advantages that favor particular locations,
then these too must attenuate rapidly. This could reflect access
to particular neighborhood amenities, for example. In either
case, the important result is rapid attenuation.

Endnotes

1. See Rosenthal and Strange (2004) for a more complete survey.
2. Aharonson, Baum, and Feldman (2004) show the importance of
the local environment for biotechnology.
3. For example, see O’Hara (1977), Ogawa and Fujita (1980),
Imai (1982), Helsley (1990), or Krugman (1993).
4. U.S. Postal Service ZIP code boundaries are established “at the
convenience of the U.S. Postal Service” (<http://www.census.gov/
epcd/www/zipstats.html>). They are based on postal logistics rather
than on a geographic or socioeconomic concept of a neighborhood, in
contrast to census-block or -tract geography. In response, the U.S.
Census Bureau has created a boundary file that approximates the
geographic region associated with each U.S. postal ZIP code based on
the associated year 2000 census blocks found in that ZIP code. The
resulting geographic polygons correspond to an agglomeration of
block-level geography and provide a close approximation of the U.S.
postal ZIP code boundaries. The resulting boundary file is referred
to as the ZCTA file on the Census Bureau’s website and is available
for download. Using that file, we matched the ZIP code IDs from
Dun & Bradstreet to geocode the data. This procedure worked for the
great majority but not all of the ZIP codes in the New York CMSA
(and the United States overall). To identify further the location of the
remaining postal ZIP codes, we augmented the ZCTA file with a 1999
file available on the Census Bureau’s website that reports the latitude
and longitude of the U.S. postal ZIP codes in the United States in 1999.

After merging those coordinates into the year 2000 ZCTA file, we were
able to geocode all but a very small number of the year 2001 ZIP codes
obtained from Dun & Bradstreet. Using that augmented ZCTA
boundary file and the year 2000 census-tract boundary file (also
available from the Census Bureau’s website), we calculated the
correspondence between ZCTA geographic units and census tracts.
Those correspondence weights were used to calculate the number of
establishments and employees present in each census tract given the
original U.S. postal ZIP-code-level data from Dun & Bradstreet.
5. See the review by Shane and Venkataraman (2000).
6. See the Syracuse University Economics Department working paper
version of this paper for a more extensive set of descriptive statistics
(<http://www.maxwell.syr.edu/econ/>).
7. Although for most of the industry regressions to follow there are a
large number of tracts with zero arrivals of new enterprises (and their
associated employment), it should also be noted that for each industry
regression, a large fraction of tracts do receive arrivals. This is clear in
Tables 2 and 3.
8. It is important to emphasize that the attenuation of agglomeration
economies does not mean that separate parts of a city are completely
unrelated. See Haughwout and Inman (2002) for a full study of this
issue.

FRBNY Economic Policy Review / December 2005

51

References

Aharonson, B. S., J. A. C. Baum, and M. P. Feldman. 2004. “Industrial
Clustering and the Returns to Inventive Activity: Canadian
Biotechnology Firms, 1991-2000.” University of Toronto
working paper.
Anderson, R., J. M. Quigley, and M. Wilhelmson. 2004. “University
Decentralization as Regional Policy: The Swedish Experiment.”
Journal of Economic Geography 4, no. 4 (August): 371-88.
Arzaghi, M., and J. V. Henderson. 2005. “Networking Off Madison
Avenue.” Brown University working paper, March. Available at
<http://www.econ.brown.edu/faculty/henderson/madison.pdf>.
Black, J., D. de Meza, and D. Jeffries. 1996. “House Prices, the Supply
of Collateral, and the Enterprise Economy.” Economic
Journal 106, no. 434 (January): 60-75.
Blanchflower, D. G., A. Oswald, and A. Stutzer. 2001. “Latent
Entrepreneurship across Nations.” European Economic
Review 45, no. 4-6 (May): 680-91.

Garreau, J. 1991. Edge Cities: Life on the New Frontier.
New York: Doubleday.
Glaeser, E. L. 1998. “Are Cities Dying?” Journal of Economic
Perspectives 12, no. 2 (spring): 139-60.
Glaeser, E. L, H. D. Kallal, J. A. Scheinkman, and A. Shleifer. 1992.
“Growth in Cities.” Journal of Political Economy 100, no. 6
(December): 1126-52.
Glaeser, E. L., J. Kolko, and A. Saiz. 2001. “Consumer City.” Journal
of Economic Geography 1, no. 1 (January): 27-50.
Glaeser, E. L., and D. C. Mare. 2001. “Cities and Skills.” Journal
of Labor Economics 19, no. 2 (April): 316-42.
Haughwout, A. F., and R. P. Inman. 2002. “Should Suburbs Help
Their Central City?” Brookings-Wharton Papers on Urban
Affairs, 45-94.
Helsley, R. W. 1990. “Knowledge and Production in the CBD.”
Journal of Urban Economics 28, no. 3 (November): 391-403.

Carlton, D. W. 1983. “The Location and Employment Choices of New
Firms: An Econometric Model with Discrete and Continuous
Endogenous Variables.” Review of Economics and
Statistics 65, no. 3 (August): 440-9.

Henderson, J. V. 2003. “Marshall’s Scale Economies.” Journal
of Urban Economics 53, no. 1 (January): 1-28.

Caves, R. 1998. “Industrial Organization and New Findings on the
Turnover and Mobility of Firms.” Journal of Economic
Literature 36, no. 4 (December): 1947-82.

Henderson, J. V., A. Kuncoro, and M. Turner. 1995. “Industrial
Development in Cities.” Journal of Political Economy 103,
no. 5 (October): 1067-90.

Chamberlain, G. 1980. “Analysis of Covariance with Qualitative Data.”
Review of Economic Studies 47, no. 1 (January): 225-38.

Hoover, E., and R. Vernon. 1959. Anatomy of a Metropolis.
Cambridge: Harvard University Press.

———. 1984. “Panel Data.” In Z. Griliches and M. Intriligator, eds.,
Handbook of Econometrics, vol. 2, 1248-1318. New York:
Elsevier.

Hsiao, C. 1986. Analysis of Panel Data. New York: Cambridge
University Press.

Chinitz, B. J. 1961. “Contrasts in Agglomeration: New York and
Pittsburgh.” American Economic Review 51, no. 2 (May):
279-89.
Duranton, G., and D. Puga. 2004. “Micro-Foundations of Urban
Agglomeration Economies.” In J. V. Henderson and J.-F. Thisse,
eds., Handbook of Urban and Regional Economics, vol. 4,
2063-2118. New York: Elsevier.

52

The Geography of Entrepreneurship

Imai, H. 1982. “CBD Hypothesis and Economies of Agglomeration.”
Journal of Economic Theory 28, no. 2 (December): 275-99.
Jacobs, J. 1969. The Economy of Cities. New York: Vintage.
Krugman, P. 1993. “On the Number and Location of Cities.”
European Economic Review 37, no. 2-3 (April): 293-8.

References (Continued)

Marshall, A. 1920. Principles of Economics. London: MacMillan.
Ogawa, H., and M. Fujita. 1980. “Equilibrium Land Use Patterns in a
Nonmonocentric City.” Journal of Regional Science 20, no. 4
(November): 455-75.
O’Hara, D. J. 1977. “Location of Firms within a Square Central
Business District.” Journal of Political Economy 85, no. 6
(December): 1189-1207.
Rosenthal, S. S., and W. C. Strange. 2003. “Geography, Industrial
Organization, and Agglomeration.” Review of Economics and
Statistics 85, no. 2 (May): 377-93.

———. Forthcoming. “The Micro-Empirics of Agglomeration
Economies.” In R. Arnott and D. McMillen, eds., Blackwell
Companion to Urban Economics. Malden, Mass.: Blackwell.
Shane, S., and S. Venkataraman. 2000. “The Promise of
Entrepreneurship as a Field of Research.” Academy of
Management Review 25, no. 1: 217-26.
Strange, W. C., W. Hejazi, and J. Tang. 2004. “The Uncertain City:
Competitive Instability, Skills, and the Strategy of Agglomeration.”
University of Toronto working paper.
Vernon, R. 1960. Metropolis 1985. Cambridge: Harvard
University Press.

———. 2004. “Evidence on the Nature and Sources of Agglomeration
Economies.” In J. V. Henderson and J.-F. Thisse, eds., Handbook
of Urban and Regional Economics, vol. 4, 2119-72. New York:
Elsevier.

The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2005

53

Robert Inman

Commentary

T

here is no economic issue more likely to make or break the
political career of a large-city mayor than the city’s job
growth or decline. Understanding why firms locate where they
do and why they expand or contract has now become an
important part of any mayor’s first course in good governance.
The paper by Stuart S. Rosenthal and William C. Strange
belongs on the syllabus—it is careful research with an
important message. Using a truly extraordinary sample of
business locations by census tract for the New York
metropolitan area, the paper reaches three conclusions. First,
firms are attracted to locations populated by other firms,
particularly in their own industry. The authors conjecture that
this attraction is caused by a production spillover that
economists call agglomeration economies. Second, the
observable reach of these agglomeration economies is strongly
bounded geographically, probably not much further than one
mile from the center of current firm locations. Third, at present
levels of employment density—remember, the New York
metropolitan area and New York City in particular are already
very dense locations—adding a new firm does not appear to
have a very strong further effect on local employment; the
multiplier effect is modest at best, perhaps no more than 25
to 50 new jobs for every 1,000 additional jobs located at an
employment center. These conclusions are valuable, perhaps
provocative, and deserve a close look.
I should note at the outset that I am a great admirer of this
line of research by Rosenthal and Strange. A companion piece

Robert Inman is the Richard K. Mellon Professor of Finance and Economics
at the University of Pennsylvania’s Wharton School.
<inman@wharton.upenn.edu>

to their study, recently published in the Review of Economics
and Statistics, was the first to adopt the authors’ unique
empirical approach to the analysis of business location.1 In that
study, the authors use a national sample of firm locations
organized by ZIP code and reach much the same conclusions,
but only for six narrow, but still interesting, industry
classifications: software, food products, clothing, printing and
publishing, fabricated metals, and machinery. This study
follows their original methodology, but here the authors
examine new firm locations within one metropolitan area, use
a finer geographical grid (census tracts are much smaller areas
than ZIP codes), and search for effects more broadly: first, for
“all industries” and then within the major employment
categories of manufacturing; wholesale trade; finance,
insurance, and real estate (FIRE); and services. The authors
emphasize business services in particular.
The methodology used in both studies is straightforward.
New firms will locate in a census tract if they can make a profit,
where profits are defined by:
(1)

π ( x, A) = p ( A ) ⋅ Q ( x, A) –

∑w ( A )⋅ x ,

where π (x , A) are the profits (appropriately discounted) earned
by the firm by locating in the census tract with a vector of
location attributes A ; x is the vector of inputs the firm must
buy to produce output Q at that location using a locationspecific production function Q ( x, A); p ( A ) is the price the firm
can charge for its output Q , where the price also may be

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

55

location-specific; and w (A ) is the price the firm must pay for
each input, where again, prices may be location-specific.
Location attributes A include measures of local demand
conditions when the firm produces a locally traded product
(for example, restaurants), local supply conditions when the
firm hires locally produced inputs (for instance, labor and most
importantly land), and finally, any local resources that make
the firm more or less productive (such as public infrastructure,
harbors, clean rivers).
Also included in A , and central to the Rosenthal-Strange
analysis, is the density of other-firm employment at a location.
As first stated by Alfred Marshall, having many firms from the
same industry close at hand enables each firm to attract and
encourage specialty inputs, save on the transit costs of needed
natural resources, and perhaps share in the development of
industry-specific innovations. As first noted by Jane Jacobs,
productive synergies may also exist between proximate firms in
different industries. Restaurants thrive near theaters, software
firms stimulate innovations by hardware firms, and hospitals
encourage medical research and development. The presence of
these Marshallian and Jacobian agglomeration economies,
proxied here by existing employment in a firm’s own and
related industries, promises higher total-factor productivity,
greater profits, and, all else equal, new firm arrivals at the
location. In fact, when deciding where to locate, firms are
concerned only with the elements of A. As profit-maximizers,
firms adjust their use of inputs to accommodate local prices,
local resource availability, and local agglomeration economies.
Thus, x = χ ( A ) ; therefore, π ( χ( A ), A ) = Π( A ). If profits
conditional on location attributes are positive, then firms will
locate in the census tract; if not, they will stay away. As any New
Yorker will say: “It’s location, baby!”
Finally, new employment at a location depends upon the
number of new firms—“births” (B ) in Rosenthal and
Strange—and the number of jobs that arrive with these new
firms (N ). Since new firms only arrive if Π( A ) > 0, predicting B
and N entails estimating a pair of regressions of the general
form:
(2)

B = b (A ) and N = n (A )

across a sample of census tracts, each with different values of A,
where B and N represent new establishments and new
employment in the tract, respectively. Rosenthal and Strange
do so, both here and in their earlier national study, except in
this case, the key variables in A are own-industry employment
and other-industry employment at the location. In both
studies, the authors are careful to allow for the fact that some
census tracts—often more than half of those in their sample
here—may actually have no new firms or new employment.

56

Commentary

The authors do not study the effects of location on the loss of
firms and jobs, although this too would be a useful exercise.2
The “structural” profit relationship in equation 1 helps us
understand what might lay beneath Rosenthal and Strange’s
“reduced form” estimates of equation 2 and in particular the
effect of current census-tract employment on the arrival of
new establishments and new employment over the next three
years. Current employment affects firms’ profits in three ways.
First, current employment in a census tract might influence
the price that new firms charge for their products, p ( A ). More
current employment in an industry means more market
competition for locally traded goods and services, causing a
fall in product prices and firm profits; this effect discourages
new firm entry and new jobs. Second, more current
employment in a firm’s own industry raises the price of locally
supplied specialty inputs (for example, skilled labor), while
more current employment in all industries raises the price of
local inputs generally (such as unskilled labor and land).
Higher factor prices lower firm profits so that again new firm
location and employment are discouraged. These two adverse
effects of higher current employment are offset by the
potential gains in total-factor productivity from Marshallian
agglomeration economies with more “own workers”
employment and from Jacobian agglomeration economies
with more “all workers” employment. Whether the two
adverse price effects of more current employment are offset by
the positive effects of current employment’s agglomeration
effects is an empirical issue.3 Positive coefficients for current
employment—the key A variable in this study—in the
estimated new establishment and new employment equations
suggest that positive agglomeration effects offset adverse price
effects; negative coefficients suggest that the negative price
effects dominate (Rosenthal and Strange’s Tables 2 and 3,
respectively).
What do Rosenthal and Strange find? That positive
coefficients, and thus agglomeration economies, seem to
dominate; and when statistically significant negative
coefficients do appear, they usually obtain for “all workers” and
not for workers in the firm’s narrower own industry. (See the
results in the aforementioned Tables 2 and 3 for Model 2). This
outcome makes sense. Negative price effects are most likely to
arise from high factor prices—most likely the price of land and
office space—in this metropolitan area’s very dense, highemployment centers. The results for wholesale trade, FIRE, and
business services are particularly instructive on the point.
Before we embrace the agglomeration explanation,
however, we need to think a bit more critically about exactly
what has been estimated in the authors’ Tables 2 and 3. The
results show a statistically significant positive correlation

between old jobs and new jobs in a firm’s own industry; but
correlations do not signify causation. For example, there may
be a very attractive attribute within current (2001) highemployment tracts—for example, a good highway location or
harbor, low taxes, or easy public transportation—that leads
these tracts to have high new (2002-04) employment as well. If
so, we cannot conclude that current employment is causing
new employment; rather, the cause of both is good
infrastructure, low taxes, or a natural-resource advantage. If
important location attributes are omitted from the RosenthalStrange regressions but they cause both old and new
employment to be jointly larger (or smaller), then the
regression coefficients in Tables 2 and 3 will not be valid
measures of causation. The estimated coefficients will be
upwardly biased (overly large) estimates of the true causal link
from old to new employment. Rosenthal and Strange are aware
of this statistical problem. Their solution is to use industry SICfixed effects as a proxy variable for omitted location attributes;
but unless a firm’s SIC code is strongly correlated with omitted
attributes, this control will be weak.4 Still, we cannot rule out a
causal connection from existing jobs to new jobs. When one
keeps this qualification in mind, the estimates in Tables 2 and 3
stand as plausible upper bounds for a true causal impact of old
jobs on new employment in a tract.
The study’s second conclusion, the one rightly underscored
by Rosenthal and Strange, is in many ways the most important
one. Whether causation or correlation, the connection between
current jobs and new jobs is very local. Almost all of the effect
of current jobs on future jobs is exhausted within one mile of
the center of the census tract. If the connection is causal and
arises from agglomeration economies, then spatially small
governments will be sufficient to recognize, and thus fully
internalize, all the benefits arising from productive firm-tofirm interdependencies. If the observed connection measures
an important omitted public policy—for example,
infrastructure, local tax breaks, or better neighborhood
services—then again the benefits can easily be internalized by a
small local government. Indeed, large but still privately owned
and managed industrial parks might be sufficient to do the job.
This narrow spatial reach for firm or policy interdependencies
means that economic development strategies can be locally
designed, and most importantly, fully funded from locally
raised revenues. Business improvement districts, as small
governments designed to internalize firm and policy spatial
interdependencies, make good sense in light of the RosenthalStrange results. Countywide, citywide, or statewide funding
should be limited only to those development policies with
significant multicommunity benefits—for example, sharing
the fixed costs of large transportation and telecommunication

networks. Beyond that, economic development decision-making
and financing should be kept very local.
Third, and again whether correlation or causation, the
second-order—or multiplier—effects estimated here of adding
new jobs to any location are very small, perhaps no more than
25 to 50 extra jobs for every 1,000 initial jobs brought into a
location.5 In the New York metropolitan area, retaining or
attracting a large employer, such as a financial institution’s call
center, will add those jobs to the location; but there will be a
very modest multiplier effect of at most .05 jobs for every new
job created. The reason for this modest effect is surely the
current density of employment in the New York area. Most
tracts are likely to have sufficient supply capacity to meet the
needs of any new employers brought into the tract. More
important, if the land area needed to accommodate new
employment is scarce, then 1,000 new jobs will simply drive
up rents and thereby discourage additional firm location.
Remarkably, agglomeration economies seem sufficient to
compensate fully for the rent increases imposed by the initial
1,000 jobs—that is, the multiplier is even slightly positive. For
economic development proponents and critics too, however,
the lesson here is clear: In the New York metropolitan area,
multiplier arguments used to justify economic development
policies should be ignored.6
There is a final benefit of Rosenthal and Strange’s work for
those of us who study urban economies. We have an important
new fact against which to calibrate our structural analysis of
firm location in dense urban areas. It is impractical to think
that we will ever be able to disentangle statistically household
utility and firm production functions from the myriad product
and factor market interdependencies that define how real
urban economies perform. What we can do statistically,
however, is identify a set of carefully constructed “reduced
form” facts that any well-specified structural model of an urban
economy must replicate. A failure to “predict” these facts
means that the structural model is likely to have been
misspecified—that is, something is missing. The authors’ work
here, and in their companion national study, gives us one such
fact—I am willing to elevate it now to the status of a “robust
fact”—that our structural models must reproduce. Whatever
policy or technology shock that generates firm demand for X
new jobs in an urban economy, X , and maybe a bit more of
those potential new jobs, must actually locate in the city. In the
end, the model’s beneficial agglomeration effects must
dominate the adverse price effects, but not by too much.
Models that cannot match this benchmark are probably not
appropriate for the study of economic policies in dense cities.
On both the policy and research fronts, the paper by Rosenthal
and Strange makes a valuable contribution.

FRBNY Economic Policy Review / December 2005

57

Endnotes

1. Rosenthal and Strange (2003).
2. I cannot resist mentioning my own work with colleagues on the
adverse effects of inefficient taxation on job location in four cities, one
of which is New York City; see Haughwout et al. (2004).
3. Rosenthal and Strange (2003) provide a cleaner estimate of the
effects of agglomeration on firm location. In that study, they attempt
through sample design to remove the effects of current employment
on p (A ) and w (A ) . First, they examine narrower industry categories
producing goods primarily intended for export from the production
site to national or world markets; thus, p ( A) = p , the “world price.”
Second, they use ZIP code areas as the unit of analysis. Because ZIP
code areas are often very large—sometimes as big as a county—it is
more likely that there will be an elastic supply of labor and land
available to firms. If so, factor prices will be independent of demand
shocks from more employment; thus, w (A ) = w . Assuming that these
identification assumptions hold, the only remaining effects of current
employment on new firm location are due to positive agglomeration
economies.
4. Consider this test: Do all census tracts with many investment
bankers have nearly identical public transportation and low income
taxes? Do all census tracts with many machine shops have equally easy
access to the turnpike? Are all warehousing centers near harbors or
centrally located train yards? The answer is surely no; thus, omitted
attributes will be imperfectly correlated with industry classification.
The issue is how imperfectly correlated they will be.
5. This estimate is computed from Rosenthal and Strange’s Table 3,
Model 2 estimate of the effect of 1(000) additional “all workers”
within the one-mile ring of employment in a given SIC code: .0157

58

Commentary

new workers in each SIC code in each census tract within one mile of
the 1(000) additional workers. There are eighty-one industry SIC
codes within the “all industries” category and roughly ten census tracts
within a one-mile radius. Thus, the total new jobs will be 12.7 jobs =
.0157·(81 SIC industries/tract)·(10 tracts/1-mile radius). In addition
to the “all workers” effect, there will be an “own workers” effect.
Assume that the 1,000 additional workers are spread evenly across the
eighty-one SIC industries—the linearity of the model makes this an
inconsequential assumption—and that the “own workers” effect is
1.37 new jobs per 1(000) current SIC jobs, as estimated in Table 3,
Model 2 for “all industries.” Then the “own workers” effect of the
1(000) current jobs will be an additional 13.7 new jobs within the onemile radius: 13.7 jobs = 1.37[(1/81)·(1)]·(81 SIC industries/tract)·
(10 tracts/1-mile radius). The total new jobs created from 1(000)
additional current jobs is therefore 12.7 + 13.7 = 26.4 new jobs.
I appreciate the authors’ assistance with this calculation. This is only a
partial equilibrium effect, however, measuring the impact in the first
three years after the “arrival” of 1,000 additional jobs and ignoring
any feedback from these 26.4 new jobs back onto the original 2001
economy. I concede the conceptual point but suspect that any
additional effects are small. In conversation, the authors are more
optimistic; they felt that doubling the 26.4 new jobs to 52.8 new jobs
might be a better general equilibrium estimate. Either way, the total
effect of adding 1,000 new jobs is modest.
6. For additional evidence that the multiplier effect of new location on
own- or other-industry employment may be small, even in less dense
counties, see Greenstone and Moretti (2003). The fact that the authors
of that study find that land values rise with own-tax-financed subsidies
to attract firms suggests that efficiency gains and agglomeration
economies are at work. Such a result is consistent with the analysis
here, but again it lacks a sizable multiplier.

References

Greenstone, M., and E. Moretti. 2003. “Bidding for Industrial Plants:
Does Winning a ‘Million Dollar Plant’ Increase Welfare?” NBER
Working Paper no. 9844, July.

Rosenthal, S. S., and W. C. Strange. 2003. “Geography, Industrial
Organization, and Agglomeration.” Review of Economics
and Statistics 85, no. 2 (May): 377-93.

Haughwout, A. F., R. Inman, S. Craig, and T. Luce. 2004. “Local
Revenue Hills: Evidence from Four U.S. Cities.” Review of
Economics and Statistics 86, no. 2 (May): 570-85.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2005

59

Andrew F. Haughwout and Bess Rabin

Exogenous Shocks and the
Dynamics of City Growth:
Evidence from New York
1. Introduction

T

shock it sustained played an important role in the pattern of
the city’s recovery. We argue that several explanations could
account for this economic resilience. One is that based on
previous events, private actors had already reacted to the threat
of terrorism, and that the events of 9/11 were, in a meaningful
sense, anticipated. A second possibility is that a repeat of the
9/11 attacks was regarded as very unlikely. A third possibility
is that the destruction of the World Trade Center, while
unanticipated, came amid a disequilibrium in the city’s real
estate markets and, by chance, happened to reinforce
preexisting trends. Finally, it is possible that public
pronouncements, regulation, and planning played a
substantial role in the economic recovery. Perhaps most
surprising is this fourth possible conclusion—that government
could have a positive effect in such a setting. Yet recent work on
New York City’s real estate markets concludes that regulation
plays an important role in economic development more
generally (Glaeser, Gyourko, and Saks 2004). Such signals are
perhaps particularly effective when an economy is out of
equilibrium, as New York City’s may have been in early 2001.

he response of cities and regions to shocks plays a central
role in our understanding of the spatial organization of
firms and households, which has been shown to have
important implications for economic outcomes ranging from
air pollution to productivity growth. Yet because exogenous,
unanticipated shocks are rarely observed, efforts to identify
their effects are often hampered.
This paper empirically examines the spatial and temporal
responses of the New York City economy to a large, but
spatially concentrated, exogenous shock to its capital stock:
the terrorist attacks of September 11, 2001. Our focus on the
city’s response allows us to draw inferences about how city
economies work, rather than to explore the effects of terrorism
on New York or other cities. We utilize data before and after
9/11 to study the response because we believe that the size,
location, and timing of the shock were unanticipated, and
because the shock was large enough to create substantial
dislocations in the city’s economy. While the actual financial
losses produced by the attacks were not large relative to the size
of the city’s economy, a major element of the shock was the
perception that the city would be in danger of future attacks.
Our analysis reveals that New York City’s economy was
surprisingly resilient to the 9/11 attacks and the damage they
caused, but the shock was associated with significant changes,
particularly in the spatial distribution of activities. Furthermore, the particular character of the city’s economy and the

In the late 1990s, New York City was experiencing
extraordinarily strong growth for such a mature economy.

Andrew F. Haughwout is a research officer at the Federal Reserve Bank
of New York; Bess Rabin, formerly a research associate at the Bank, is an
analyst at Watson Wyatt Worldwide.
<andrew.haughwout@ny.frb.org>

The authors are grateful to Jan Brueckner and Howard Chernick for helpful
comments on earlier versions of this paper. The views expressed are those of
the authors and do not necessarily reflect the position of the Federal Reserve
Bank of New York or the Federal Reserve System.

2. The Effect of 9/11 on New York
City’s Economy

FRBNY Economic Policy Review / December 2005

61

Between 1996 and 2000, private sector employment in the city
grew at a 2.6 percent annual rate, the strongest four-year run in
more than four decades. In each of those years, the rate of city
job growth exceeded that of the nation. Private sector wage and
salary growth also exceeded the national average over this
period, rising 7 percent per year in real terms (Bram 2003).
This economic strength was reflected in broader measures of
activity as well. In January 2000, the New York City index of
coincident economic indicators (CEI), a measure of the shortrun dynamics of economic activity, reached its highest level
since the series began in 1965.1 City housing values were also at
very high levels in both absolute terms and relative to the
nation (Bram, Haughwout, and Orr 2002). Real revenues from
the city’s four largest taxes reached an all-time high, despite
rate reductions, in fiscal year 2000-01 (Edgerton, Haughwout,
and Rosen 2004).
In the subsequent two years, the city experienced a sharp
economic downturn. Private sector jobs reversed their strong
growth and, for the 2001-03 period, fell at a 2.1 percent annual
rate. By November 2003, the CEI had retreated nearly
10 percent from its peak value. Revenues from the city’s four
major taxes declined sharply in real terms during fiscal year
2002, and they had yet to recover their 1999 level by fiscal year
2003.
The sources of this reversal in the city’s fortunes are not
controversial: the 9/11 attacks on the World Trade Center, the
decline in the stock market, and the national recession all
clearly played important roles in the slowing of aggregate city
economic activity.

As of this writing, the World Trade Center site remains
essentially vacant, although the reopened PATH station—the
Lower Manhattan terminus of the Port Authority’s light-rail
system—occupies a small portion of the area. This persistent
loss of productive capital and the potential ongoing threat of
future loss of life and property caused many commentators to
voice concerns about the future of the city as a highly desirable
location for businesses and households.
The attacks occurred as a recession was already under way in
the nation and the city. Employment in New York peaked in
December 2000 and had declined by 60,000 jobs by August.
Another 100,000 jobs were lost between August and October
2001 (Chart 1). The New York City CEI began falling as the
local recession commenced in January 2001 and declined
nearly 0.95 percent in September 2001 alone (Chart 2). This
was the fourth-largest monthly decline in the history of the
index. While the CEI continued to decline until August 2003,
the total peak-to-trough decline totaled 8.9 percent, which was
significantly less deep than those registered during the city
downturns that began in 1969 and 1989. In addition, the rates
of decline before September 2001 and after are approximately
the same, suggesting that the ongoing national recession was an
important factor in the adverse outcomes experienced by the
city economy. For this reason, isolating the effect of the cityspecific shock that struck New York on September 11 requires
controlling, to the extent possible, for the effects of the ongoing
national recession. In the analysis that follows, we accomplish
this by normalizing our results by changes in the national
economy. We thus seek to isolate differential New York City
effects from changes in the national economy as a whole,
whether attributable to 9/11 or to other factors.

2.1 Isolating the City-Specific Component
of the Shock
The destruction of the World Trade Center had several
potential effects on the economy of New York. First, and most
horrific, the attacks took nearly 2,800 lives. In economic terms,
this means that the human capital stock for the entire
metropolitan region was reduced, at least in the short run.
Despite the tragic consequences for the individuals and their
families, the direct impact on the supply of human capital in
New York City—an open economy with more than 3.5 million
jobs and 8 million residents—was small.
The sixteen acres of the World Trade Center site housed
approximately 13.4 million square feet of class A office space,
nearly 30 percent of the Downtown total. This complex was
destroyed on September 11, and several surrounding buildings
were damaged when the towers fell. While some residential
space was affected as well, it was reoccupied relatively quickly.

62

Exogenous Shocks and the Dynamics of City Growth

Chart 1

Employment in New York City
Millions of employees
4.0
3.8
3.6
3.4
3.2
3.0
1965

70

75

80

85

90

95

00

03

Source: U.S. Department of Labor, Bureau of Labor Statistics.
Notes: Data are seasonally adjusted. The bands indicate local recessions.

Chart 2

New York City Index of Coincident
Economic Indicators

September 2001
WTC attack
September 2000
stock market peak

160
150
140
130
120
110
100
90
80
70
1965

70

75

80

85

90

95

00

03

Source: Federal Reserve Bank of New York.
Note: The bands indicate local recessions.

2.2 The City’s Real Estate Markets
The series depicted in Chart 3 is the quarterly Office of Federal
Housing Enterprise Oversight (OFHEO) single-family home
price index for the New York metropolitan area, divided by the
national index. Both indexes, and the resulting series, are
indexed to 100 in 1976:2, when the New York series began.
There is little evidence here that the September 11 attacks on
the World Trade Center reduced the demand for residential
locations in the New York metropolitan area. The chart shows

Chart 3

New York City Area House Prices
Relative to U.S. Average
Index: 1976=100
225
200
175
150
125
100
September 2001

75
1976

80

85

90

95

00

03

Sources: Office of Federal Housing Enterprise Oversight;
Federal Reserve Bank of New York calculations.
Note: The index is based on the ratio of the repeat-sales price measure
for existing single-family homes in the New York City metro area to
that of the United States overall.

the date of the attacks, which occurred during 2001:3. Repeatsale house prices in the metropolitan area were rising faster
than they were in the rest of the nation both before and after the
attacks, as depicted by the steady rise in the index on both sides
of the September 11 point. That is to say, the New York area’s
residential housing market gained ground on the rest of the
nation immediately after the attacks. (Statistical tests fail to
reject the null hypothesis that the trend in the series is the same
before and after 2001:3.) Only after two years had passed, in
late 2003, was there any sign that housing prices in New York
had faltered relative to the nation. Since that period, data not
plotted here suggest that the New York metropolitan area
housing price premium has resumed its rise. Thus, the relative
demand for residential locations in the New York area market
has remained strong since the attacks.
The OFHEO data cover only single-family homes, which are
presumably located primarily in the suburbs. Increased
demand for single-family houses may reflect reduced demand
for Manhattan locations and a decentralization of population
from New York City proper. Such a result, for example, is
consistent with the ideas presented in Mills’ (2002) early
reflections on the implications of urban terrorism. To address
this issue of urban form, we turn to a detailed examination of
the New York City housing market before and after the attacks.

2.3 Neighborhood-Level Microdata
on the City’s Real Estate Markets
Our second housing market analysis is more restrictive in the
sense that it focuses solely on housing units in the city of New
York. However, our data source for this analysis, the New York
City Housing and Vacancy Survey (HVS), allows consideration
of a much broader range of housing types, from rental
apartments to condominiums to single-family homes, with the
mix reflecting the actual housing consumption patterns of city
households.
The HVS is conducted about every three years (the coverage
here is 1991, 1993, 1996, 1999, and 2002). Each survey collects
information on the structural and locational characteristics of
about 18,000 housing units in the city. The structural
characteristics include detailed items such as the number of
bedrooms, the presence of complete kitchen facilities, and the
condition of exterior walls.2 For the purposes of the survey,
New York is divided into fifty-five sub-boroughs, and the subborough location of each unit is identified in the public data.
The HVS data, like the OFHEO data, provide a limited view
of changes in housing demand. In particular, the HVS
complements the OFHEO index in the sense that it allows for a
detailed look at those parts of the city itself expected to have

FRBNY Economic Policy Review / December 2005

63

been affected most by the terrorist attacks and the fear of future
attacks.
To discern the effects of September 11 on the demand for
housing in New York City, we estimate a set of regression
equations of the form V = V ( t, N , H ) , where V is a measure
of unit value (expected sales price for owner-occupied units or
gross rent for rental apartments), t indexes time, N indexes
neighborhood, and H is a vector of housing capital measures.
We interact the fifty-five sub-borough measures with a set
of five survey (year) dummies. Our test consists of looking for
significant negative effects on the 2002 dummies in the city as
a whole or in those sub-boroughs expected to have been
affected most by the attacks.3 Our specification estimates
average trait prices and looks for temporal variation in the
relative value of particular neighborhoods. If variations in traits
whose prices are changing are correlated with neighborhood,
then we may obtain biased estimates of neighborhood effects.
We leave research on this topic to future work, but note that if
components of housing capital that experienced rising prices
are concentrated in Lower Manhattan, then we will understate
the relative depreciation (or overstate the relative appreciation)
of a Lower Manhattan location per se.
We experimented with several specifications of the basic
relationships, including estimating the equation in level and
semi-log forms, eliminating the top and bottom 5 percent of
observations based on value, eliminating top-coded units, and
augmenting the equation with information about financial
arrangements and move-in or lease dates. Each of these
specifications leads to the same qualitative conclusions.

Sub-Boroughs of Lower Manhattan
and Northwest Brooklyn

Results
Table 1 reports the results of two sets of regressions designed to
identify the effects of the September 11 terrorist attacks on the
demand for residential locations in New York City. The figures
are the regression coefficients on year 2002 dummies either on
their own (column 1) or interacted with dummies for a
particular borough (column 2), sub-borough (column 5), or
group of sub-boroughs (columns 3 and 4). If the attacks were
to have broken the trend of absolute price and rental growth in
the city, we would expect negative coefficients to predominate
in the table. Analyzing the evidence on the city’s appreciation
relative to that of the rest of the nation requires another step,
described below.
The first column of the table reports the overall citywide
trends in prices and rents, controlling (as do all specifications
reported here) for the units’ structural characteristics. In
addition, for owner-occupied units, we control for the year in

64

which the owner acquired the unit or, for rental units, the year
the occupant moved in. The requirement that we have
information for all of these variables reduces the sample size to
the approximately 51,000 reported in the table. We present
results from both the level and semi-log specifications.
The results suggest that city residential prices and rents in
2002 were both higher than they were in 1999, the year of the
previous survey. But when we subtract the national increase in
shelter costs, 11.1 percent, only the price increase is statistically
different from zero; rental increases were slightly slower in
New York City than they were in the nation as a whole.4 Note,
however, that we can reject the hypotheses that absolute rents
and prices in New York fell on average; all four estimates in
column 1 are positive and more than twice their standard errors.
The second column of Table 1 reports the price changes in
Manhattan in 2002 relative to 1999, controlling for citywide
time effects. These results reveal a pattern similar to that in the
citywide estimates. Although the point estimate of 12 percent
rental appreciation in Manhattan slightly exceeds the national
average, the standard error of the estimated coefficient does not
allow for rejection of the hypothesis that the New York increase
was the same as the nation’s. Manhattan prices, meanwhile,
grew much more rapidly than did the shelter component of the
national CPIU.
Column 3 reports results for the two Lower Manhattan
sub-boroughs and three Northwest Brooklyn sub-boroughs
(see map). All of these areas benefit from direct accessibility to

Exogenous Shocks and the Dynamics of City Growth

Lower Manhattan
Northwest Brooklyn

Source: U.S. Census Bureau.

the Lower Manhattan central business district, with housing
units typically within a thirty-minute commute on public
transportation.5 We might thus expect residential markets in
these areas to be negatively affected by the attacks. Again, the
data provide little evidence to support this conjecture, although
rental increases are statistically indistinguishable from zero for
these areas as a whole.
Since the attacks occurred in Lower Manhattan, there is the
potential that the area would endure significant reductions in
demand. Columns 4 and 5 address this issue, using two
definitions. In column 4, we include the area that extends as far
north and east as Chinatown, while the column 5 results are
limited to the Financial District and Greenwich Village. Once
again, the evidence suggests price increases relative to the
nation in all these areas as well as significant rent increases in
the area most proximate to the World Trade Center.

Our tests indicate that demand for rental properties in New
York was no stronger than demand in the nation, and in some
areas it may have been weaker. Yet in Lower Manhattan, the
area most affected by the attacks, rents grew strongly. The
apparent divergence between the residential rental market in
Lower Manhattan and that in the rest of the city may be
partially attributable to incentives for residents to locate
Downtown, part of the package of aid that the city received in
the wake of the crisis. Under these programs, residents willing
to make a two-year residential commitment to areas of Lower
Manhattan close to the site of the attacks were eligible to receive
up to $12,000 in grants. Our estimated 1999-2002 rental
increase in Lower Manhattan (Table 1, column 5) less the
increase in the city as a whole is about $325 per month, or
approximately $7,800 over a two-year period. Unfortunately,
we cannot identify which units receive the subsidy, so a direct

Table 1

2002 Price and Rent Effects in New York and Selected Subcity Areas

Prices
Dollars

ln

Monthly rents
Dollars

ln

Citywide
(1)

Manhattan
(2)

Lower
Manhattan,
Lower East Side,
Northwest
Brooklyn
(3)

68,714
(3,732)

151,883
(7,244)

102,709
(11,153)

57,771
(16,742)

113,733
(23,465)

-940
(23,560)

130,467
(14,585)

0.77
(0.03)

1.3
(0.07)

1.03
(0.1)

1.23
(0.15)

2.01
(0.22)

0.38
(0.22)

0.8
(0.13)

39.6
(5.8)

169.1
(8)

91.1
(12.2)

161
(16.8)

365.4
(25.3)

1.85
(21.9)

14.08
(16.8)

0.05
(0.01)

0.12
(0.01)

0.02
(0.02)

0.12
(0.03)

0.37
(0.04)

-0.06
(0.04)

-0.07
(0.02)

Lower
Manhattan,
Lower East Side
(4)

Lower
Manhattan
(5)

Lower East Side
(6)

Northwest
Brooklyn
(7)

Source: Authors’ calculations, using data from the New York City Housing and Vacancy Survey.
Notes: The figures in bold represent increases that are significantly greater than national average increases in the shelter component of the CPIU between
1999 and June 2003 (11.1 percent). The total number of observations for prices is 16,672; the total number for monthly rents is 34,586. All regressions
include controls for structural traits, survey year, rent control status, whether the unit is a condominium or cooperative (price regressions), whether the
owner lives in the building (rent regressions), and year acquired (price regressions) or year the current occupant moved in (rent regressions). Rows labeled
“dollars” are estimated in levels; results reported in rows labeled “ln” are from models in which the dependent variable is a natural logarithm.
For column 1, the coefficient and standard error estimates are on a dummy variable for 2002 prices, relative to 1999 prices. For column 2, the coefficient
and standard error estimates are on a dummy variable for 2002 Manhattan prices, relative to 1999 Manhattan prices. For column 3, the coefficient and
standard error estimates are on a dummy variable for 2002 prices in Lower Manhattan, Chinatown and the Lower East Side, and Northwest Brooklyn,
relative to 1999 prices in the same areas. For column 4, the coefficient and standard error estimates are on a dummy variable for 2002 prices in Lower
Manhattan and in Chinatown and the Lower East Side, relative to 1999 prices in the same areas. For column 5, the coefficient and standard error estimates
are on a dummy variable for 2002 Lower Manhattan prices, relative to 1999 Lower Manhattan prices. For column 6, the coefficient and standard error
estimates are on a dummy variable for 2002 Lower East Side and Chinatown prices, relative to 1999 Lower East Side and Chinatown prices. For column 7,
the coefficient and standard error estimates are on a dummy variable for 2002 Northwest Brooklyn prices, relative to 1999 Northwest Brooklyn prices.

FRBNY Economic Policy Review / December 2005

65

comparison of rent with the value of the subsidy is not possible.
However, since the majority of the units in Lower Manhattan
as we define it are eligible for smaller (or no) subsidies, it seems
most likely that our estimate of the rental increase in the area
incorporates demand effects above and beyond those
stimulated by the subsidy.
Of course, the price of any good, including housing, is
determined by both supply and demand. One potential
explanation for increased rents (prices) in Lower Manhattan is
reductions in the current (expected future) supply of units.
Evidence of the direct effect of the attacks on the housing
supply is hard to uncover. Table 2 displays the number of new
housing units added to the Downtown stock from 1995 to
2004. In Downtown Manhattan, with its paucity of vacant land,
office building conversions are an important source of new
residences, as indicated in the table. Also important is a city
tax-incentive program, adopted in 1995, that offers property
tax abatements for residential conversions Downtown.
The data are difficult to interpret, as the peak year for new
units was 2001—the year of the 9/11 attacks. Since the process
of adding units to the stock takes time, it is reasonable to
suppose that the vast majority of the 2,578 units that came on
line in 2001 were planned before the attacks. Nonetheless,
despite the national recession, the 2002-04 total of 4,167 units
slightly exceeds the 1999-2001 total of 4,098, indicating little
effect on the trajectory of the housing supply after 9/11. In
addition, the 2004 total is the second highest of any year since
1995. The data, then, do not suggest a significant effect on the
supply of Downtown residential units. Given that the supply of
Downtown housing appears to have been changed little by the

Table 2

Downtown Residential Development, 1995-2004
Date Open

Conversions

New Developments

Total by Year

1995
1996
1997
1998
1999
2000
2001
2002
2003
2004

8
0
46
1,454
102
811
2,139
1,366
545
867

0
0
0
152
398
209
439
25
449
915

8
0
46
1,606
500
1,020
2,578
1,391
994
1,782

Totals

7,338

2,587

9,925

Sources: Alliance for Downtown New York; New York City Department
of Housing Preservation and Development.

66

Exogenous Shocks and the Dynamics of City Growth

attacks, we interpret our results as strong evidence that the
demand for residential locations in Lower Manhattan
remained very robust in the wake of 9/11.
For the other areas potentially affected by the attacks, the
signals are less clear. Rents in Northwest Brooklyn were
essentially flat in nominal terms, and thus lagged the national
average in the immediate aftermath of the attacks. Prices,
however, remained strong, growing at a pace significantly
faster than the national average. Meanwhile, on the Lower East
Side, both prices and rents fell relative to the national average.
This last finding complements earlier evidence that businesses
in Chinatown, which is in the Lower East Side neighborhood,
were affected negatively by 9/11-related disturbances in
transportation and telecommunications infrastructure (Asian
American Federation of New York 2002). Yet given that these
were expected to be temporary phenomena—and indeed have
largely been rectified in the years since 2001—the residential
price effects we observe are a puzzle. Of course, long-run
divergences between rents and prices may signal differences
in current conditions and expectations of future conditions.
The 2005 HVS, which will be released in 2006, may help answer
some of these questions.

Some Caveats
We begin by noting that our analysis of the 2002 data is based
on a comparison with 1999, the previous survey year. Because
the 2002 survey was based on results from the 2000 decennial
census, while the 1999 survey relied on the 1990 census,
variations in the under- or overcount of housing units in the
census could affect the results. This will only lead to biased
estimates of the neighborhood effects if changes in the housing
characteristics of miscounted units are correlated with
neighborhood. Such a bias would likely appear as a significant
change in results when sampling weights, which adjusts the
sample data to match the census population characteristics.
The results we describe above obtain whether the regression is
estimated with or without the sampling weights, ameliorating
this concern to some extent.
It is also possible that the prices and rents we observe in
2002, while higher than those in 1999, are lower than they were
immediately before the attacks, a period for which less data are
available. Analysis of actual transactions for which we have
prices provides modest support for the contention that real
prices in Manhattan were stronger in 2002 than in 2001, but the
number of units in the HVS sample that sold in those two years
is too small to allow any strong conclusions to be drawn from
the data. We take some comfort from the fact that the analysis

of annual metropolitan statistical area trends produced
conclusions broadly consistent with those advanced here.
Finally, the 2002 survey was conducted during the first half
of the year, or immediately in the aftermath of the terrorist
attacks of late 2001. Since very little time elapsed between the
attacks and the beginning of the survey, there is potential bias in
the survey responses. This bias could be in either direction:
respondents might not have had time to internalize fully the
negative effect of the attacks on their property values, and thus
might have provided an overly optimistic view of value.
However, Lower Manhattan in the first six months of 2002 was
still very much in the throes of the turmoil created by the
destruction of the World Trade Center and a substantial
amount of city infrastructure (such as roads and subways).
Indeed, the fires ignited by the attacks were extinguished only in
late December 2001, and the cleanup of the site continued until
late May 2002. In these circumstances, the idea that property
owners would be overly optimistic about the value of their
homes seems unlikely. Nonetheless, it is impossible to know for
certain. Again, we take comfort from the fact that the results
here are consistent with the analysis of the OFHEO price index.

2.4 Office Markets
We now examine trends in the market for office space in New
York’s two central business districts—Downtown and
Midtown—using data from the National Real Estate Index.6
These data are collected for class A office space in sixty markets

across the nation. We focus on the two New York markets and,
to control for prevailing national conditions, calculate indexes
measuring appreciation in these markets relative to the nation.
These indexes, which are based in 1985:4, are shown in
Charts 4 and 5.
Note in these charts the trend deterioration of Downtown
office prices and rents relative to Midtown. In rents, this
pattern is evident immediately following the commencement
of the data (Chart 4), although it is most pronounced in the
price data after 1993 (Chart 5). This reduction in the relative
premium for Downtown office locations is part of the longterm trend described by Glaeser and Shapiro (2002).
The September 11 attacks destroyed or rendered
temporarily or permanently unusable nearly 28 million square
feet of class A office space, 13.4 million of which was in the
World Trade Center complex itself. If the demand for Lower
Manhattan locations remained stable, we might expect to see a
strong increase in office rents for the remaining Downtown
office space. There is little evidence of this in Chart 4. Indeed,
nominal class A office rents declined nearly 9 percent between
2001:3 and 2002:3, suggesting that demand fell at the same time
as supply. A decline in demand is consistent with Glaeser and
Shapiro’s view that the attacks hastened the decline of Lower
Manhattan as a principal site for New York City office
locations. Yet this decline was matched by an 8.5 percent
decline in class A rents nationwide, with the result being that
both the Downtown and Midtown indexes depicted in Chart 4
remained essentially flat, with perhaps a modest downward
trend.

Chart 4

Chart 5

Office Rent Indexes

Office Price Indexes

Class A Space, Manhattan Markets Relative to National Average

Class A Space, Manhattan Markets Relative to National Average

Index: 1985:4=100

Index: 1985:4=100

130

160
150

September 2001

120

Midtown Manhattan

September 2001

Midtown Manhattan

140
130

110

120
100

110
100

90
Downtown Manhattan
80
1985

87

89

91

93

95

97

99

01

03

90
80
1985

Downtown Manhattan
87

89

91

93

95

97

99

01

03

Sources: Global Real Analytics, National Real Estate Index;
Federal Reserve Bank of New York calculations.

Sources: Global Real Analytics, National Real Estate Index;
Federal Reserve Bank of New York calculations.

Note: The indexes are based on the ratio of office rents in Manhattan
to that of the United States.

Note: The indexes are based on the ratio of office prices in Manhattan
to that of the United States.

FRBNY Economic Policy Review / December 2005

67

Prices reveal an interesting pattern both before and after
September 11, 2001 (Chart 5). Between 1985:4 and 2003:3,
Downtown office building prices essentially held steady relative
to the nation, while they fell relative to Midtown. Note,
however, that Downtown prices reached a trough in 1998:1
(at which point, Downtown had fallen more than 10 percent
relative to the nation since the end of 1985). From 1998:2 to
2001:2, the Downtown market rallied, and the relative price
index stood at 111.5 on the eve of September 2001. By the close
of 2001, the Downtown market had given back all its gains
relative to the nation, and the index reached a low of 96.8
in 2002:3. There is modest evidence here of a rally in the
Downtown market since that point, as the index rose back
above the break-even point (101.6) by 2003:3.
The fact that the relative Downtown office prices remain
below the peak they reached immediately prior to the
September 11 attacks might be taken as evidence that the
attacks themselves had a very substantial effect on office prices.
There are several points to make here. First, the 2001:2 peak of
the office index (111.5) was anomalous in the sense that it
represented a sharply higher level than it did in the previous
quarter (103.7). Second, the pre-9/11 rise in the index as we
measure it was the result of a modest decline in the national
index and a sharp uptick in the Downtown index.7 That is, the
chart shows a sharp increase in part because of the national
office market downturn. Third, the fact that the Downtown
office market stabilized in the subsequent two years provides
some indication that demanders continue to find locations
there attractive. By the end of the period, the relative
Downtown price index was about 3 percent higher than it had
been three years earlier. However, there is some evidence, as
suggested by Glaeser and Shapiro (2002), of a post-attack shift
in demand to Midtown, where prices have rallied strongly
relative both to the nation and to Downtown since mid-2001.
Statistical tests indicate that both the level and the growth rate
of the ratio of Midtown to Downtown prices per square foot of
office space increased significantly after 2001:3.
Overall, the evidence from the office market suggests a postattack weakening of demand in Lower Manhattan relative to
the rest of the nation, especially in light of the decline in the
supply of space that accompanied the destruction of the World
Trade Center. The most dramatic effects are seen in prices
(Chart 5), although an unusual spike just prior to the attacks
makes the data difficult to interpret. Nonetheless, it is clear that
the dramatic increase in prices that occurred in Midtown has
not been experienced Downtown. In rental markets, there is
some sign of weakening in both Downtown and Midtown,
although there was modest evidence of stabilization in both
areas by the end of 2003.

68

Exogenous Shocks and the Dynamics of City Growth

These data are consistent with a fairly benign view of the
attacks’ effect on the demand for New York locations. As
suggested by Glaeser and Shapiro (2002), it would appear that
Downtown’s appeal to businesses has declined relative to that
of Midtown. However, Downtown demand has held up
reasonably well relative to demand in the nation, especially
given the temporary dislocations associated with the cleanup
and redesign of the World Trade Center and surrounding
areas.
We can calculate the weighted average price increase for all
of Manhattan by applying the Downtown and Midtown shares
of class A space as weights to the relevant price increases. That
calculation yields a 12.6 percent increase in office prices across
Manhattan between 2001 and 2003.

2.5 Summary
Our evidence suggests several interesting features of the 9/11
shock on the New York City economy:
• It destroyed a very significant share of the Downtown
class A office stock.
• The shock exacerbated the effects of the ongoing
recession, and almost certainly contributed to a sharp
loss of city jobs in late 2001.
• Long-run demand for city locations relative to the rest of
the nation appears to have been affected very little;
modest evidence from aggregate real estate prices
suggests that it may have continued to strengthen.
• Long-run demand for residential space in Lower
Manhattan strengthened significantly, but demand in
the short run was weaker.
• Both long- and short-run demand for office space in
Lower Manhattan weakened relative to the rest of the
nation, while demand for Midtown offices rose sharply.

3. Interpreting the Data
What can economic models tell us about what happens to cities
over time when they experience significant shocks? Previous
work on the dynamics of city economies in light of factor
mobility is surprisingly limited. Wildasin (2003) describes a
model in which at least one factor of production is imperfectly
mobile in the short run, and explores the dynamic implications
for tax competition. A key conclusion is that the effect of
shocks depends on whether agents are surprised by them;

anticipated shocks have little or no effects. Glaeser and
Gyourko (forthcoming) examine the implications of capital
durability for paths of urban growth and decline. Both papers
indicate that dynamics are very important to the behavior of
actors and to the interpretation of empirical results.
A few papers provide models that explicitly incorporate
shocks of the sort we examine here. Harrigan and Martin
(2002) study simple equilibrium theoretical models of urban
growth in the face of terrorism. In both models presented, a
large shock is sufficient to reduce the long-run equilibrium size
of the city, but the authors argue that large shocks of this type
are unlikely to occur as a result of terrorism. They conclude
that the transport cost and labor pooling advantages of urban
density are likely to be broad and durable enough to absorb
plausible terrorism shocks in the long run. The models that
these authors adopt are not designed to examine intracity
spatial or temporal dynamics, but their results are broadly
consistent with the evidence from New York.
In a noneconomic approach to the effects of 9/11, Beunza
and Stark (2003) report the results of an ethnographic study of
a financial services firm before and after the 2001 attacks. They
conclude that the organization’s ability to recreate itself was the
result of a complex interaction of human and technological
capital. One theme that clearly emerges is the primacy of
networks across firms and information sharing within the firm.
These findings suggest that spatial concentration of activities is
an enduring feature of advanced service economies, even in
light of sophisticated technologies for transferring and storing
information. These conclusions support those of Harrigan and
Martin while adding some empirical detail to the advantages
conferred by density. One relevant feature of Beunza and
Stark’s study is that it does not presume that the spatial
organization of activity on September 10, 2001, was an
equilibrium allocation, which implies that the dynamics of
recovery will depend on the expected future configuration as
well as the particular character of the shock.
The aggregate effect of shocks on the New York City
economy has been empirically documented by several authors.
Two kinds of shocks have drawn special attention: the 9/11
terrorism shock (Haughwout 2005) and changes in city fiscal
policies (Haughwout et al. 2004). One remarkable feature of
these studies is the very different responses that the city
economy exhibits in response to these different kinds of shocks.
Haughwout et al. find that small changes in tax rates have
substantial effects on city tax bases, which are themselves
determined by city economic activity, including employment.
However, as we indicate, the arguably very large shock caused
by the attacks of September 11 resulted in very little aggregate
effect on the city economy, but it seems to have been associated

with changes in the equilibrium distribution of activities over
space.
Rossi-Hansberg (2004) provides a dynamic general
equilibrium analysis of the effect of a terrorist attack on a city
economy. The paper reaches several conclusions. First, the
long-run effect of a terrorist attack on the overall size of a city
is expected to be substantial, with a benchmark simulation
suggesting that a modestly sized attack would produce city
output declines of between 12 and 21 percent, depending on
commuting costs. Second, the new equilibrium spatial
configuration features no uniform effects on business land
rents, but uniformly higher residential land rents.
In Rossi-Hansberg’s model, the long-run effect of a terrorist
attack is determined by what the attack implies about ongoing
risks of future destruction and the distribution of that threat
over areas of the city, or what the author refers to as the
“terrorism tax.” Policy interventions such as subsidies to
development in areas that are (incorrectly) perceived to be at
elevated risk of future attacks will improve welfare only to the
extent that the public sector has special (correct) information
about the probability of future attacks that it cannot credibly
convey to private actors.
Glaeser, Gyourko, and Saks (2004) emphasize the
importance of land use regulations in influencing the level and
distribution of economic activity in New York. Government’s
role in providing information that affects development may
have been an important factor in the case of New York as well,
although in a different way than those highlighted by RossiHansberg and Glaeser, Gyourko, and Saks. Because
government plays an important role in determining the
equilibrium spatial configuration of activity in New York City,
clear pronouncements about the future equilibrium
configuration provided market players with information in the
face of uncertainty. This information appears to have been
valuable enough to more than offset the terrorism tax that 9/11
imposed on the city, allowing a relatively smooth transition
toward the new equilibrium.

3.1 Understanding New York’s
Response to 9/11
New York’s relatively rapid recovery after 9/11 is a puzzle. How
could such a large shock result in so little aggregate change in
the economy after just two years? One possible explanation is
that while the general public did not anticipate a terrorist attack
of such magnitude, relevant market actors like property
developers and their insurers understood that it was a real
possibility. An example of evidence supporting this argument

FRBNY Economic Policy Review / December 2005

69

is that this was not the first terrorist attack on the World Trade
Center, which had survived an attempt to topple the towers in
1993. Another possibility is that relevant market actors
expected that the shock would never be repeated, or that the
ongoing terrorism tax was very low. Yet neither notion is
supported by evidence from insurance markets. In the
immediate aftermath of the attacks, property insurance prices
soared (Lakdawalla and Zanjani 2005), suggesting that the
shock was unanticipated and that the perceived probability of
further attacks had risen.
As we observe, Lower Manhattan on the eve of the 9/11
terrorist shock was already changing from a primary location
of the financial services and banking industries, centered on
Wall Street. As indicated in Charts 4 and 5, Manhattan office
rents and prices had lagged those in Midtown for at least fifteen
years. Indeed, public construction of the World Trade Center
itself in the 1960s was an effort to resuscitate a lagging
Downtown office market (Glaeser and Shapiro 2002).
Meanwhile, throughout the 1990s, demand for Manhattan
residential locations, including Downtown, was strong.
Prior to 9/11, the movement of office employment to
Midtown was gradual, in part because of a shortage of
accessible, developable land in Midtown; existing stocks of
office capital Downtown; and heavy government regulation in
both markets. Given that only the last of these can be altered in
the short run, it is useful to think about the spatial allocation
of activities in Manhattan prior to 9/11 as a disequilibrium.
A critical feature of this disequilibrium is the central role
played by government in affecting the distribution of activities
in New York. Industries and occupations that place high value
on spatially defined networks dominated employment in pre9/11 Lower Manhattan. For these firms, the geographic
characteristics of places are less important than their economic
and social characteristics. That is, the agglomeration of
financial services firms that exists in Lower Manhattan could
potentially be located anywhere within the greater New York
commuting area, as long as the relevant actors are located
together. As a preexisting agglomeration begins to come apart,
firms lack a means of coordinating their new locations so as to
remain near each other. When the public sector has important
effects on location patterns, government regulators have the
tools at their disposal to serve this coordination function.
In this context, the behavior of public officials in the wake of
the 9/11 terrorist attacks had the potential to be a crucial
determinant of the future level and distribution of activity.
How did officials respond? The federal government
immediately pledged $20 billion in aid to reconstruct the city,
signaling that it was committed to maintaining New York as
the nation’s primary center of economic activity. City officials
responded in several ways. In addition to proposing detailed

70

Exogenous Shocks and the Dynamics of City Growth

plans for the use of the federal money, they made strong and
repeated announcements about the future of Downtown
Manhattan as a 24/7 mixed-use community. In addition, city
officials sought to divert some of the federal resources intended
for Downtown businesses to businesses located elsewhere in
the city.8 Finally, Mayor Michael Bloomberg’s administration
accelerated the process of developing the far West Side of
Manhattan, adjacent to Midtown, as a new premium office
location complete with a new football stadium.
All of these actions served to signal that the city intended
to accommodate the transformation of Downtown into
a residential location. This transformation included the
relocation of financial services jobs from Downtown to
Midtown. All of these actions, whether intentionally or not,
provided valuable information to market participants in the
wake of 9/11. The change from Downtown as a business
location to Downtown as a residential location proceeded
slowly, in part because of the existence of large amounts of
sector-specific capital. The 9/11 attacks destroyed a large
portion of this durable capital in a short period of time. In the
market uncertainty that followed, consistent government
behavior was interpreted as a clear signal that the future
location for business was Midtown.
This view of the evidence is, we believe, consistent with
much of the previous literature on city economies. It places
appropriate weight on the importance of networks and
spillovers, as emphasized by Beunza and Stark (2003). It also
stresses the importance of government activities in general
(Rossi-Hansberg 2004) and in New York (Glaeser, Gyourko,
and Saks 2004). Finally, it provides a potential explanation for
the difference between the findings in Haughwout et al. (2004)
on tax shocks and the relatively small effect of the terrorism tax.
What distinguishes the two is that in the latter case, government is attempting to offset an exogenous shock, while in the
former, government itself is generating a “surprise,” to use
Wildasin’s (2003) language. Combined, these results suggest
that the actions of New York City government are perceived to
be highly credible, both when they signal preferred patterns of
land use and when they signal a redistribution of resources.

4. Conclusion
The resilience of cities to powerful shocks has been
documented by many authors. In this paper, we present and
interpret data on the effects of the September 11 attacks on
New York City. The New York experience is consistent with a
significant role for government in resolving uncertainty in the
immediate aftermath of the attacks. Our results suggest that

cities’ responses will depend on the size of the original shock
and its expected ongoing cost (in this case, the terrorism tax),
whether the preshock spatial configuration was an equilibrium,
and the importance and effectiveness of public sector actors as
coordinating agents.
If this conjecture is valid, then a negative shock to capital
stocks in a city that is in a stable equilibrium will likely reduce
activity in the short run, but absent a long-run cost, long-run
levels and the spatial distribution of activity will return to the
previous equilibrium. But when a city’s spatial configuration is
far from equilibrium, the shock will potentially exert a stronger
effect on the spatial distribution of activity in the long run. In

the case of New York, the fact that the city was not in
equilibrium, as evidenced by the long-term trends away from
Downtown as a business location, and that a very influential
local government provided clear information led to marked
increases in the Midtown premium for business locations and
the Downtown residential premium.
In addition to emphasizing the importance of government
behavior, these results suggest that analysts who study the
effect of shocks on urban economies take into account the
potential effects of disequilibrium on the shock’s effects.
The results also suggest the usefulness of modeling both
the temporal and the spatial dimensions of the shock.

FRBNY Economic Policy Review / December 2005

71

Endnotes

1. The New York City CEI is a broad-based, dynamic single-factor
measure of economic activity, constructed according to the
methodology of Stock and Watson (1989). The index is calculated
from the common movements in four indicators tied to the city’s
labor market: payroll employment, the unemployment rate, average
weekly hours worked in manufacturing, and real earnings. The CEI is
described more fully in Orr, Rich, and Rosen (1999).
2. A complete description of the survey is available at <http://
www.census.gov/hhes/www/housing/nychvs/2002/nychvs02.html>.
3. Because of high correlations among the measures of unit quality,
the specifications reported in Table 1 exclude some variables. These
exclusions have no effect on the coefficients of interest. R2 values for
the regressions range from 0.72 for the price equations to 0.85 for the
rent equations. Detailed results are available upon request.
4. All prices and rents are measured in nominal terms. The shelter
component of the national CPIU increased 11.1 percent between 1999
and June 2002 (Council of Economic Advisers 2005, Table B-61).
Since the rental and owner’s equivalent rent components grew at
similar rates (12.3 percent and 11.1 percent, respectively), we use the

72

Exogenous Shocks and the Dynamics of City Growth

total as our benchmark; disaggregating would not affect our
conclusions. Overall CPIU inflation over this time period was
8.0 percent.
5. Average commutes in New York City outside of Manhattan average
more than forty minutes, placing the four “outer boroughs” sixth,
seventh, eighth, and ninth in the national ranking of longest
commuting times.
6. Global Real Analytics, which produces the index, collects quarterly
information on recently closed office building sales and average rents
for class A office space.
7. The price for a square foot of class A office space in Lower
Manhattan rose from $307 in 2001:1 to an all-time high of $328 in
2001:3, while the national average fell from $215 to $213. Comparing
fourth-quarter prices, we note that Downtown prices were 4.8 percent
higher in 2001 than they were in 2002.
8. See <http://www.lowermanhattan.info/construction/
looking_ahead/residential_growth.asp>.

References

Asian American Federation of New York. 2002. “Chinatown after
September 11th: An Economic Impact Study.” Available at
<http://www.aafny.org/research/911Impact.asp>.

Haughwout, A. F. 2005. “The Long-Term Impact of 9/11 on the
New York City Economy: Evidence from Real Estate Markets.”
Unpublished paper, Federal Reserve Bank of New York.

Beunza, D., and D. Stark. 2003. “The Organization of Responsiveness:
Innovation and Recovery in the Trading Rooms of Lower
Manhattan.” Socio-Economic Review 1, no. 2 (May): 135-64.

Haughwout, A. F., and R. Inman. 2001. “Fiscal Policy in Open Cities
with Firms and Households.” Regional Science and Urban
Economics 31, no. 2-3 (April): 147-80.

Bram, J. 2003. “New York City’s Economy before and after
September 11.” Federal Reserve Bank of New York Current
Issues in Economics and Finance 9, no. 2 (February).

Haughwout, A. F., R. Inman, S. Craig, and T. Luce. 2004. “Local
Revenue Hills: Evidence from Four Cities.” Review of Economics
and Statistics 86, no. 2 (May): 570-85.

Bram, J., A. Haughwout, and J. Orr. 2002. “Has September 11 Affected
New York City’s Growth Potential?” Federal Reserve Bank of
New York Economic Policy Review 8, no. 2 (November): 81-96.

Lakdawalla, D., and G. Zanjani. 2005. “Insurance, Self-Protection
and the Economics of Terrorism.” Journal of Public
Economics 89, no. 9-10 (September): 1891-1905.

Council of Economic Advisers. 2005. Economic Report of the
President 2005. Available at <http://www.gpoaccess.gov/eop/
index.html>.

Mills, E. 2002. “Terrorism and U.S. Real Estate.” Journal of Urban
Economics 51, no. 2 (March): 198-204.

Edgerton, J., A. Haughwout, and R. Rosen. 2004. “Revenue
Implications of New York City’s Tax System.” Federal Reserve
Bank of New York Current Issues in Economics and
Finance 10, no. 4 (April).
Glaeser, E., and J. Gyourko. Forthcoming. “Urban Decline and Durable
Housing.” Journal of Political Economy.
Glaeser, E., J. Gyourko, and R. Saks. 2004. “Why Is Manhattan So
Expensive? Regulation and the Rise in House Prices.” Harvard
University working paper, August. Available at <http://
post.economics.harvard.edu/faculty/glaeser/papers.html>.
Glaeser, E., and J. Shapiro. 2002. “Cities and Warfare: The Impact of
Terrorism on Urban Form.” Journal of Urban Economics 51,
no. 2 (March): 205-24.

Orr, J., R. Rich, and R. Rosen. 1999. “Two New Indexes Offer a Broad
View of Economic Activity in the New York–New Jersey Region.”
Federal Reserve Bank of New York Current Issues in Economics
and Finance 5, no. 14 (October).
Rossi-Hansberg, E. 2004. “Cities under Stress.” Journal of
Monetary Economics 51, no. 5 (July): 903-27.
Rossi-Hansberg, E., and R. Lucas. 2002. “On the Internal Structure of
Cities.” Econometrica 70, no. 4 (July): 1445-76.
Stock, J., and M. Watson. 1989. “New Indexes of Coincident and
Leading Economic Indicators.” NBER Macroeconomics
Annual. Cambridge, Mass.: MIT Press.
Wildasin, D. 2003. “Fiscal Competition in Space and Time.” Journal
of Public Economics 87, no. 11 (October): 2571-88.

Harrigan, J., and P. Martin. 2002. “Terrorism and the Resilience of
Cities.” Federal Reserve Bank of New York Economic Policy
Review 8, no. 2 (November): 97-116.

The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2005

73

Stephen L. Ross

Commentary

1. Introduction

A

Haughwout and Rabin argue that the general stability of
the city’s economy and the surge in Midtown values are
attributable in part to the actions of public officials in the wake
of 9/11. Specifically, New York City officials made strong,
repeated announcements that Downtown Manhattan would be
a mixed-use community while simultaneously accelerating
commercial development in and near Midtown. In this way,
the administration removed uncertainty and facilitated the
private sector response to the dislocations arising from 9/11,
which in turn increased economic stability and raised the longrun value of commercial space in Midtown. In contrast, the
authors cite earlier work by Haughwout et al. (2004) on a fiscal
shock to New York City, which found that small increases in
tax rates led to large, permanent declines in the city’s tax base.

ndrew F. Haughwout and Bess Rabin examine trends in
New York City’s economy and real estate markets prior to
and following the 9/11 terrorist attacks. They analyze trends
both in New York City employment and in an index of
coincident economic indicators (CEI) specific to the city. On
the real estate side, the authors focus on trends in the Office of
Federal Housing Enterprise Oversight’s index of metropolitan
house prices, the New York City Housing and Vacancy Survey,
and the national real estate index (NREI) of class A office space
in Manhattan.
Haughwout and Rabin do not identify any persistent,
negative effects of 9/11 on New York City’s economy or real
estate markets. Employment in the city had already entered a
steep decline prior to 9/11 because of the continuing national
recession, and that decline continued at a comparable pace in
the months that followed. The CEI exhibits a very steady decline
immediately before and after the attacks, and the rate of decline
slows dramatically during the year following 9/11. Both the index
of metropolitan house prices and the Housing and Vacancy
Survey suggest that residential values rose following 9/11, and
the NREI office rent data are fairly flat relative to the national
average during the periods before and after the attacks. The one
exception to this pattern of stability is a dramatic increase in the
NREI of office building prices in Midtown Manhattan and an
associated decline in the index for Downtown.

The value of urban land rises in large part because of some form
of agglomeration economies. If these agglomeration
economies are driven by the efficiencies arising from a large,
diverse labor market, the destruction of commercial office
space during the 9/11 terrorist attacks might have been expected
to increase property values because it created a shortage of
physical capital while leaving the human capital stock in the

Stephen L. Ross is an associate professor of economics at the University
of Connecticut.
<Stephen.L.Ross@uconn.edu>

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

2. Housing Markets, Government
Action, and Price Changes

FRBNY Economic Policy Review / December 2005

75

New York metropolitan area broadly intact. However, if
agglomeration economies arise primarily because of economic
spillovers, the destruction of commercial office space should
have lowered the value of economic spillovers, leading to lower
property values in Downtown and potentially affecting the entire
New York City economy (see Rosenthal and Strange [2005] for
evidence of economic spillovers in the city).
Haughwout and Rabin suggest that the negative effects of lost
economic spillovers were mitigated by government action. In the
wake of any dislocation, firms face uncertainty as to where
private sector activity will locate, and property values are likely to
suffer based on this uncertainty. The authors suggest that New
York City officials solved this coordination problem through the
signals sent in public announcements and actions. As a result,
firms could base location and investment decisions on accurate
expectations concerning the spatial pattern of economic activity
within the city. According to this logic, these government actions
stabilized the city’s economy in general and led to dramatic
increases in the value of Midtown office buildings based on the
expectation of increased economic efficiencies as the Midtown
commercial district continues to grow.
The explanation provided by Haughwout and Rabin seems
reasonable, but I would like to offer an alternative explanation
that appears equally consistent with the data. While the
commercial office space destroyed in the 9/11 attacks
represented 30 percent of the stock of Downtown class A space,
this loss represents a substantially smaller portion of the class A
stock in New York City and an even smaller portion of space
across the entire New York consolidated metropolitan
statistical area (CMSA). Therefore, the effect of 9/11 is that the
shock was relatively small when compared with the economic
size of the entire CMSA. The lost commercial office space may
have been replaced by marginal adjustments across the larger
economy with higher end activities moving to Midtown and
lower end activities moving to space outside Manhattan, which
in turn pushed other activities to office space in the broader
CMSA. Given that a large fraction of workers commute into
Manhattan, such adjustments might only have had a relatively
small effect on the overall labor market.
In this context, the key question is how quickly the real
estate market can adjust to such a large spatial shock to the
relative location of office space supply. Clapp and Ross (2004)
examine the adjustment of the market for owner-occupied
housing in Connecticut in response to economic and
demographic changes. While they find that the relative
demographic composition of towns is affected by such shocks,
they find no evidence of systematic changes in relative town
prices over a two-year time frame. They conclude that
sufficient mobility exists within the owner-occupied housing
markets such that the increased demand in a few towns arising

76

Commentary

from migration is spread across the entire metropolitan
housing market. In a related analysis, Clapp et al. (2005)
examine both the short-run effect (yearly changes) of town
demographic changes on prices and the long-run effect (fouryear change). They obtain very similar results in the two
analyses, suggesting that prices adjust quite rapidly across
housing submarkets. One might expect the market for
commercial office space to adjust relatively quickly when
compared with the market for owner-occupied housing.
Only Haughwout and Rabin’s finding of declining
Downtown and rapidly increasing Midtown office building
values cannot be explained by a simple view that the real estate
market is characterized by the actions of efficient and flexible
actors. These results, however, must be put into context. The
office building price indexes exhibit a high degree of variability,
with a spike in Midtown prices occurring during the fourth
quarter of 1997 that was just as sharp and large as the spike
following the 9/11 attacks. The Midtown and Downtown series
also appear to be negatively correlated for most of the 1990s—not
just the period following 9/11. Finally, office prices in both
Downtown and Midtown began moving off their extremes by the
second quarter following 9/11, and while they have not returned
to their previous levels, the indexes appear to have returned to
levels that are consistent with the trends established after 1995.

3. Conclusion
Haughwout and Rabin provide a very detailed picture of New
York City’s economy and real estate markets leading up to and
following the 9/11 terrorist attacks. Their view that 9/11 had a
relatively minor effect on the city’s economy is quite
convincing. However, the underlying reason behind this
benign effect is unknown. The authors suggest that
government action allowed private commercial activity to
coordinate in Midtown Manhattan, which mitigated the
negative effects of the dislocations caused by 9/11.
In this commentary, I offer a different view—that the shock
was actually quite small relative to the total stock of
commercial office space in the region, and that over a short
amount of time marginal adjustments by individual firms
absorbed the large shock to class A space in Downtown
Manhattan with only relatively minor effects on prices. Unlike
Haughwout and Rabin, I view their post–9/11 findings as
consistent with earlier work suggesting a large impact from a
fiscal shock. In their study, the focus was on a large supply
response, which was probably necessary to keep the after-tax
price of commercial and residential property relatively
unchanged following the shock.

References

Clapp, J. M., A. Nanda, S. L. Ross, and C. Briggs. 2005. “Which School
Attributes Matter? The Influence of School District Performance
and Demographic Composition on Property Values.” University
of Connecticut Department of Economics Working Paper
no. 2005-26.
Clapp, J. M., and S. L. Ross. 2004. “Schools and Housing Markets:
An Examination of School Segregation and Performance in
Connecticut.” Economic Journal 114, no. 499 (November):
425-40.

Haughwout, A. F., R. Inman, S. Craig, and T. Luce. 2004. “Local
Revenue Hills: Evidence from Four Cities.” Review of Economics
and Statistics 86, no. 2 (May): 570-85.
Rosenthal, S. S., and W. C. Strange. 2005. “The Geography of
Entrepreneurship in the New York Metropolitan Area.” Federal
Reserve Bank of New York Economic Policy Review 12, no. 2
(December): 29-53.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2005

77

Keynote Address
Kenneth T. Jackson

The Promised City:
Openness and Immigration
in the Making of a World
Metropolis
1. Introduction

A

t least since the Great Depression, urban specialists have
spent much of their time searching for patterns common
to all cities, thinking about the similarities among crowded
human settlements, and devising new terms—such as central
business district, strip mall, gentrification, and edge city—to
describe phenomena that occur in most metropolitan regions.
All cities, for example, must somehow deal with water supply,
sewage and garbage disposal, fire prevention, criminal justice,
public health, affordable housing, and adequate open space,
and all have to establish governmental structures to cope with
those issues.
Indeed, the Chicago School of Sociology, founded in the
1930s by Ernest W. Burgess, Louis Wirth, and Robert E. Park,
became famous for developing a model of the spatial structure
of the modern industrial metropolis. Using the Windy City
itself as the prototype, the Chicago School shaped the
dominant theoretical and methodological assumptions about
urban development for more than half a century. Even after the
Chicago School came under attack from scholars like Milton
Gordon, Nathan Glazer, Daniel Patrick Moynihan, Nancy
Foner, Herbert Gans, and many others, it continued to be the
paradigm against which other models were measured.1

Kenneth T. Jackson is the Jacques Barzun Professor of History
and Social Sciences at Columbia University.
<ktj1@columbia.edu>

The focus of my remarks is something else entirely. My
purpose is threefold: first, to make the case that the study of
history is essential to understanding the present and future of
any urban area; second, to suggest that in terms of age, size,
density, and demographic patterns, New York has been
different from, rather than typical of, American cities; and third,
to argue that Gotham has been unusually successful for almost
four centuries because of its heterogeneity, not in spite of it;
because of its openness, not in spite of it; and because of its
immigrants, not in spite of them. Certainly, the Hudson River
metropolis has not won many accolades for being gracious or
charming. As John Steinbeck noted decades ago: “It [New
York] is an ugly city, a dirty city. Its climate is a scandal. Its
politics are used to frighten children. Its traffic is madness. Its
competition is murderous. But there is one thing about it.
Once you have lived in New York and it has become your
home, no other place is good enough.”
The little settlement that began at the southern tip of
Manhattan has, however, been welcoming in a more important
sense—it has provided a haven and opportunity for a larger
and more diverse population over more centuries than any
other city in human history.

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

81

2. The Founding of New York
By American standards, New York is old. Founded as Fort
Amsterdam by the Dutch in 1625, it predates Boston (1630),
New Haven (1636), Newark (1666), Charleston (1670),
Philadelphia (1682), Colonial Williamsburg (1699), and a
hundred other places that we generally regard as more historic
than Gotham. St. Augustine (1565) is assuredly older than New
York, but for three centuries and more it consisted simply of a
fort, a couple of chapels, a school, and a few hundred
unremarkable human and animal inhabitants. St. Augustine
was not a city by any reasonable definition and it gained
prominence only in the twentieth century, when it became a
tourist destination because of its age, not its prominence.
Similarly, Jamestown (1607), the first English settlement, never
found its niche and ultimately disappeared into the muck of the
James River, where anthropologists continue in the twentyfirst century to search for what little remains of the town. The
same is true for Plymouth, the Pilgrim village in Massachusetts
that was founded in 1620. It never grew beyond a few small
buildings, fell quickly into ruin, and found new life only in the
twentieth century, when it was reborn and reconstructed as a
kind of historical theme park. Meanwhile, thousands of miles
to the west, Santa Fe began in 1610 as a Spanish colonial
administrative center. But it remained a wide place on a dusty
road until the twentieth century, and not until after World War
II did it find success as an art and cultural center.
New York does not seem “historic” to most people because
it has been so successful for so long that its population has
exploded, its real estate prices have risen dramatically, and its
building lots have seen repeated development. Quite simply,
because it was important in history, it does not have many
buildings that testify to its age—the structures having been torn
down repeatedly by successive generations of developers eager
to cash in on rising real estate values. Charleston, South
Carolina, by contrast, has much of its historic value within its
boundaries precisely because little of historic importance
happened there. Charleston went into long-term decline after
1820 and grew only slightly over the next half-century.
Property values remained low, change was glacial, and old
antebellum houses continued to stand along the waterfront
into the twenty-first century. Such an outcome would be
impossible to conceive in Manhattan, where turbulence,
congestion, and constant building—not to mention fires in
1776, 1778, and 1835—contrived to destroy virtually
everything of the city’s important colonial past.
Of course, other parts of the world boast great cities that are
centuries older than New York, whose age is unimpressive

82

The Promised City

when compared with Athens, Rome, Beijing, Tokyo, London,
Paris, or a thousand other cities. What was Manhattan when
Aristotle and Plato were musing in ancient Greece or when
Caesar conquered Gaul? Of what did the Empire City consist
when the Ming Dynasty moved its capital in 1421 from
Nanking to Peking? And Istanbul, the exotic meeting place
between east and west, was already 900 years old in 1492, when
Christopher Columbus first set sail for a new route to the
Indies.

3. Size
If New York is not old as a settlement by world standards, it
is nevertheless old as a big city by world standards. Indeed,
it was a major metropolis by 1860, when (including
Brooklyn) it had 1 million inhabitants and was larger than
any city on the European continent except Paris. By the end
of the century, Gotham had 3.4 million citizens and was,
after London, the second-largest city on earth and the
richest metropolis anywhere. In 1900, for example,
approximately half of all the millionaires in the United
States, and perhaps a third of those in the entire world, lived
in the New York metropolitan region.
In 2005, Gotham remains the only American municipality
ever to exceed 4 million residents, and each of its five boroughs
would rank as an important city in its own right. Brooklyn
alone was almost as big as Chicago; Queens was larger than
Philadelphia; the Bronx was bigger than Detroit and Cleveland
combined; and Staten Island was more populous than
Pittsburgh, St. Louis, or Atlanta.
Figures for the New York metropolitan region have been
even more impressive over the past century. In 1930, New York
became the first urbanized area in the world to exceed 10
million residents; in 1970, it became the first to exceed 15
million. Although its current thirty-one-county metropolitan
region of 22 million people is exceeded by Tokyo and possibly
by São Paulo and Mexico City, the Hudson River metropolis
remains a human agglomeration of almost unimaginable size.
These statistics remind us that New York has a significance
in history unrelated to the date of its establishment as a Dutch
trading post. Its size and wealth over the past 150 years has
meant that Gotham has had to deal with issues of public health,
public transportation, public safety, fire prevention, water
supply, and a hundred others before they were addressed in a
modern way by Athens, Rome, Moscow, or Istanbul—all of
which were smaller and poorer than New York a century ago.

4. Density and Demographic
Patterns
Why should anyone care whether any city is particularly old?
What does history have to do with our present circumstances?
Demographers have long regarded the spatial arrangement
of the United States as so outside the mainstream that they have
settled on a term, “the North American pattern,” to describe it.
Quite simply, the model of urban settlement in this nation is a
donut, meaning that all the life, energy, and vitality of the
American metropolis is on the edges—in shopping malls,
corporate office parks, and residential subdivisions. In the
older, urban neighborhoods, one finds pathologies of every
description—poverty, public housing, decrepit schools,
graffiti-infested playgrounds, racial minorities, prostitution,
heavy drug use, and visible homeless problems. While the
central business district may feature a few high-end restaurants
and glittering skyscrapers, perhaps even a sports arena, Main
Street is essentially deserted after dark. Indeed, this pattern is so
ingrained in our culture that Americans have devised special
ways of discussing it that are understood by the general
population. When we mention “inner-city problems,” for
example, it is not necessary to spell out what we mean.
New York differs from the North American pattern in three
fundamental ways: 1) the socioeconomic distribution of the
population, 2) the population density of the inner city and the
outer suburbs, and 3) the change in gross density over the past
half-century. Let us consider each of these demographic
patterns in turn.
First, the Hudson River metropolis in some ways follows the
North American pattern. Gotham has more than its share of
famous and expensive suburbs—from Scarsdale, Chappaqua,
Bronxville, and Bedford to the north; to Greenwich, Darien,
and New Canaan to the northeast; to Saddle River, Metuchen,
and Short Hills to the west; and to the Five Towns and Great
Neck to the east. Similarly, the five boroughs include many
desperately poor neighborhoods as well as a disproportionate
share of the region’s public housing and homeless population.
But so it is with all American cities. What makes New York
unusual is that the greatest concentration of wealth on earth is
in the middle of Manhattan, the wealthiest ZIP code address is
10021, and the most expensive real estate is along Park Avenue,
Fifth Avenue, and Central Park West. Moreover, of the 3,137
counties in the United States, the poorest in 2000 was in
western Nebraska, with a per capita income of less than $3,000.
By that measure, the wealthiest single county in the entire
nation was New York County, otherwise known as Manhattan,
with a per capita income in excess of $70,000 in 2000.
This statistic is astonishing, if only because Manhattan has
long been the locus of so much concentrated poverty. After all,

Manhattan contains the nation’s largest Dominican
population, which is mostly poor, as well as Harlem, the
nation’s most famous black community. It includes tens of
thousands of newly arrived Chinatown residents who are
working for below-minimum-wage rates as well as thousands
of unemployed and underemployed actors and actresses. And
the Manhattan total excludes many wealthy families who own
apartments near Central Park but who go to great lengths to
prove that their official residence is somewhere else, the better
to avoid Gotham income taxes. Yet despite all that, Manhattan
comes out as the richest county in the United States, a place not
on the edges but at the center.
Second, New York is assuredly not a donut in terms of
population density or activity. Its central business district far
overshadows any shopping mall or corporate office park, and
no one would argue that the city is deserted after dark or quiet
at night. And no teenager growing up in Fairfield County or
Westchester County or Morris County would likely argue that
the Stamford Mall or the Galleria or the Paramus Mall is where
the action is or is representative of a lifestyle they want to
emulate. They know that the shopping opportunities, sports
arenas, concert halls, restaurants, and nightclubs of Manhattan
easily eclipse anything they will ever find in White Plains,
Garden City, or Saddle River.
But this demographic characteristic goes well beyond the
preferences of young adults. As even a casual examination would
reveal, the United States is a low-density civilization, and its
metropolitan regions spread over larger spaces than those of any
other advanced nations. Rare is the American city (Chicago,
Philadelphia, Boston, San Francisco) with a population density
of more than 10,000 per square mile (a number that would be
typical of cities in Europe or Asia). Many municipalities (San
Jose, Denver, Portland, Houston, Seattle) have densities of fewer
than 5,000 per square mile and some American cities (Memphis,
Jacksonville, Oklahoma City, Kansas City) have densities of
fewer than 2,000 per square mile, or about as many as who live
in completely rural parts of India or Bangladesh. New York, of
course, is quite different. Its population density in 2000 was
more than 25,000 per square mile for the entire city, and many
times that number in most of Manhattan.
Third, Gotham’s density is also unusual in that it is not
declining. In the United States as a whole, especially since 1950,
metropolitan regions have been hollowed out even as the
fringes have developed at a rapid pace. The American city could
be described as a balloon in the twentieth century that was
squeezed in the middle, thus forcing expansion on the edges. In
cities that did not expand their boundaries in the twentieth
century, the total population declined. Thus, Cleveland went
from 915,000 inhabitants in 1950 to 478,000 in 2000; Detroit
went from 1,850,000 to 951,000; Philadelphia went from

FRBNY Economic Policy Review / December 2005

83

2,072,000 to 1,518,000; Pittsburgh from 677,000 to 335,000;
and Buffalo from 580,000 to 292,000. St. Louis is perhaps the
most dramatic case, as it declined from 857,000 in 1950 to
348,000 in 2000.
The same phenomenon is true as well in the exploding cities
of the south and west that expanded their boundaries over the
past 100 years. So that even though Houston, Dallas, San
Antonio, San Diego, Phoenix, and Memphis have grown since
1950 in total population, their densities have declined,
meaning that their area has increased even faster than new
families have moved in.
Only two American cities had population densities that were
higher in 2000 than they were in 1950: New York and San
Francisco. Thus, what is unusual about Gotham is not that
millions of its citizens left for Westchester County or Florida.
Rather, what makes New York City unusual is that somebody
took their place.
And contrary to what has often been the popular perception
in the United States, the density and diversity of New York have
made the city safer than other large American agglomerations.
For example, even in 1992, when the murder toll in Gotham
reached its horrendous peak of 2,245 in a single year, the city
ranked no higher than tenth in the nation in its homicide rate.
In the next thirteen years, the number of murders in New York
plummeted so far (to fewer than 600 per year between 2002 and
2005) that the city no longer ranks among the country’s 150
most dangerous places.

5. Immigration and Diversity
New York has other unique characteristics, among them its
heavy reliance on public transportation, its twenty-four-hour
orientation, and its diverse cultural offerings. Indeed, it would
be easy to argue that taken as a whole, the numerous opera
houses, symphonic opportunities, rock concerts, jazz choices,
dance performances, legitimate theaters, and art museums in
New York provide residents with a cultural richness that Paris,
London, Vienna, Berlin, Tokyo, Milan, Moscow, and Los
Angeles cannot challenge.
The most important characteristic of New York City,
however, has been its openness to newcomers. Essentially,
Gotham has never had a majority culture. It was founded by the
Dutch to trade and to do business, and for that reason the
ruling elite of the small colony were not particularly concerned
about religious, racial, or ethnic differences. Even in the 1640s,
for example, more than eighteen languages were being spoken
on New Amsterdam’s streets—and the town had fewer than
1,000 total residents at the time.

84

The Promised City

The early history of New York contrasted sharply with that
of Boston, where the Puritan’s “city on a hill” worked mightily
to prevent religious dissent and to enforce a kind of theocracy
on the inhabitants. When one strong-willed resident, Anne
Hutchinson, dissented from the ruling orthodoxy, she was put
on trial for heresy and banished from Boston and the
Massachusetts Bay Colony.
Such an action would have been inconceivable in the Dutch
settlement at the mouth of the Hudson River. Following the
traditions of the Netherlands, then the most liberal and
tolerant nation in Europe, the city fathers of New Amsterdam
followed a kind of “live and let live” policy. They did not
particularly care whether one went to church or believed in any
god at all, regarding such issues as matters of personal
preference.
When the English took the city in 1664 and renamed it New
York, they retained much of its Dutch flavor and its tradition
of openness.
After the thirteen colonies won their independence and
transformed themselves into the United States, Gotham
continued to be unusual in the heterogeneity of its citizenry. In
1900, for example, New York had more Irish than Dublin,
more Italians than Naples, and more Germans than Hamburg.
Indeed, the kleindeutschland neighborhood below Fourteenth
Street in Lower Manhattan would have ranked as the thirdlargest city in the Kaiser’s German Empire. The almost
unbelievable diversity on the streets was captured in print by
the young radical John Reed, who gained fame by joining the
Russian Revolution in 1917 and writing about his experiences
in Ten Days That Shook the World. Before he died of
tuberculosis in his adopted land, however, he wrote about his
early life in Gotham:
New York was an enchanted city to me. I wandered about
the streets, from the soaring imperial towers of
downtown, along the East River docks, smelling of spices
and the clipper ships of the past, through the swarming
East Side, alien towns within alien towns, where the
smoky glare of miles of clamorous pushcarts made a
splendor of shabby streets. I knew Chinatown and Little
Italy, Sharkey’s and McSorley’s saloons, the Bowery
lodging houses and the places where the tramps gathered
in winter, the Haymarket, the German village and the
dives of the Terderloin. The girls that walked the streets
were friends of mine, and the drunken sailors off ships
from the world’s end. I knew how to get dope, where to go
to hire a man to kill an enemy. Within a block of my house
was the adventure of the world. Within a mile was every
foreign country.

Even in 2005, many global cities remain largely
homogenous. In Tokyo, for example, ethnic homogeneity is so
ingrained in the culture that Koreans who have lived in Japan
for their entire life are derisively called Zainichi, which means
to stay in Japan. In school, boys and girls shun them as
playmates; as adults, they are considered inferior and are not
eligible for important or prestigious government jobs.
Similarly, in Shanghai, Beijing, Seoul, Moscow, Hong Kong,
and São Paulo, one or two ethnic groups make up more than
90 percent of the total population. Other cities have become
heterogeneous only since World War II—one thinks of Paris,
Vancouver, Toronto, Sydney, Melbourne, and Berlin. London,
as always, is a leader among cities. Leo Benedictus, for example,
noted in 2005 that 300 languages were being spoken by the
people of London, that 2.2 million people in the city had been
born outside England, and that the city had at least fifty
nonindigenous communities with populations of 10,000 or
more. As he wrote, “Virtually every race, nation, culture, and
religion in the world can claim at least a handful of Londoners.”
But New York remains in a class by itself, as it has been since
the middle of the seventeenth century. According to the 2000
census, 2.93 million foreign-born persons, up from 2.18 million
in 1990, lived in the five boroughs, and unlike the British, who
count persons from Wales and Scotland as foreign born,
Americans do not classify persons from California or Texas or
Mississippi as foreign born, although they have to travel farther
than someone from Northern Ireland to get to the cultural and
financial capital. Significantly, the largest group of foreignborn persons in Gotham—those from the Dominican
Republic—account for only 14 percent of the newcomer total.
Quite simply, New York is the immigrant metropolis, and it
has a more diverse population than any other city in the
history of man. Queens alone is the most polyglot place on
earth, with 1,028,339 “official” foreign-born persons in
2000, or 46 percent of the total.

6. The Jewish Experience
New York has transformed many ethnic and racial groups—
the Dutch, the English, the Irish, the Germans, the Italians,
African-Americans, the Greeks, for example—who in turn
have transformed the metropolis. No other group, however,
reveals the peculiar history and challenges of New York better
than the Jews.
Quite simply, the major events in New York’s Jewish history
reflect the larger history of the metropolis. The first small band
of Jews to reach New Amsterdam arrived on September 1, 1654,
from Portuguese Brazil, where they had been forced to leave.

Their initial reception in Manhattan was not much better
because Peter Stuyvesant, the last of the four Dutch governors of
the town, had no use for the newcomers and wanted to send
them on their way. But his superiors in Amsterdam learned of
the controversy and reminded Stuyvesant that the purpose of the
colony was to encourage trade and to welcome opportunities for
business growth, not to encourage some sort of Christian
conformity. Properly chastened, the governor allowed the Jews
to remain, and even to hold religious services in their homes. By
the time the English captured the city in 1664, the Jews were
already holding public services. Called Shearith Israel, the
congregation rented quarters on Beaver Street and had about 100
members by the end of the seventeenth century.
The second major shift in Jewish New York came between
1825 and 1875, when a large number of German, Austrian,
Bohemian, and Hungarian Jews came, largely after the
revolution of 1848. This group, which later formed the core of
what Stephen Birmingham would call “Our Crowd,”
exemplified the theme of aspiration.
The third major moment in New York Jewish history lasted
from about 1881, when the Russian pogroms began in earnest,
until 1924, when restrictive immigration laws at least
temporarily cut off the flow of newcomers from eastern
Europe. These were the peak years of immigration, captured in
prose by Emma Lazarus’s famous poem The New Colossus and
in physical form by the Statue of Liberty. And while life on the
Lower East Side was never easy, those years and those streets
exemplified the theme of hope.
The fourth major moment came in the 1930s, when German
refugees fleeing Hitler congregated in Washington Heights and
when second-generation Jews from the Lower East Side
became, as Deborah Dash Moore has argued, “at home in
America,” moving away from Rivington and Essex and
Delancey and Orchard Streets to places like East New York in
Brooklyn and the Grand Concourse in the Bronx.
Since World War II, there has been an exodus of the Jewish
population from the five boroughs to places like Scarsdale and
Great Neck or to Florida and the Sunbelt more generally. At the
same time, the growth of the Orthodox and Hassidic
populations in Crown Heights, Williamsburg, and Borough
Park has meant that the Jewish proportion of the city’s
population has stabilized.

7. The Decline of Industrial
and Port Employment
So what? Are there larger lessons to take from the New York
experience in terms of tolerance and openness to newcomers?

FRBNY Economic Policy Review / December 2005

85

The suggestion of my remarks is an emphatic yes. New York
has not only been the Promised City for the Jews, but also for a
succession of other immigrant groups. Both the city and the
immigrants themselves benefited from the exchange, whether
successful entrepreneurs like Andrew Carnegie and Alexander
T. Stewart or penniless newcomers who only dreamed of
economic success, political opportunity, and religious
freedom. Taken as a group, they transformed what in 1775 was
a second-tier city in the British Empire into what by 1950 was
variously considered the Capital of the Twentieth Century, the
Capital of Capitalism, or, as the late Pope John Paul II famously
said, the Capital of the World.
The constant infusion of new energy and ideas into the
metropolis over the years enabled New York to meet economic
and technological challenges that destroyed the prospects of
competing cities. Consider how the engines of Gotham’s
prosperity have changed over the past half-century. In 1955,
the twin underpinnings of the metropolitan economy were
manufacturing and the port. Indeed, at midcentury, Gotham
was the most important industrial city in the world. German
and Japanese competitors had of course been blasted into
ruins, and other European cities were still recovering from the
conflict. Chicago and Pittsburgh were of course dominated by
factories of every description, but their populations were so
much smaller than that of New York that the value added by
manufacturing and the total employment in production was
less than half that of Gotham. The same was true of Detroit
with its automotive plants or Los Angeles with its aircraft
construction. What made New York unusual was the absence
of heavy industry and instead the presence of thousands of little
factories where operatives were sewing buttons onto overcoats,
building and repairing warships, making razor blades and file
cabinets, producing chewing gum and caskets, bottling milk
and brewing beer, printing checks and magazines, and turning
out hats, blouses, and skirts by the millions—usually in
businesses with fewer than 1,000 employees.
What happened to New York’s industries? In the past halfcentury, more than three-quarters of them have disappeared as
manufacturing employment in the city declined from more
than 1 million in 1950 to fewer than 200,000 at the turn of the
century. Brewing is perhaps typical. In 1900, Gotham was
home to more than ninety breweries, mostly concentrated in
Greenpoint and Williamsburg in Brooklyn; as late as 1960,
New York produced more beer than Milwaukee and St. Louis
combined. By 1975, however, the industry was dead in the city,
and in 2005, not a single brewery, other than a micro-pub,
remains in the five boroughs.
The harbor has followed a similar trajectory. A half-century
ago, the Port of New York was the busiest and most important in
the world, and it had held that position for more than a century.

86

The Promised City

During the second half of the nineteenth century, there were
many years when the volume of trade passing through the
Manhattan, Brooklyn, and Staten Island docks was greater than
that of every other harbor in the United States combined. It was
not just a world port, it was the world port. During World War I,
freight trains backed up all the way to Pennsylvania and beyond
awaiting their turn to unload cargo destined for France and the
Western Front. The pattern was similar during the Second
World War, when Gotham was again the major point of transshipment for men and material heading for North Africa, Italy,
and England, and through Normandy and France to the German
heartland. Practically every tank, gun, soldier, and uniform
involved in the invasion of Europe passed through the New York
docks on their way overseas.
The 1954 motion picture classic, On the Waterfront, starring
Marlon Brando, illustrated the powerful role of the harbor in
the economy, as it depicted the tens of thousands of stevedores
who showed up every morning and afternoon in the hope of
getting the chance to unload boxes or bags from a ship.
Recreational boating and swimming were rare because the East
and Hudson Rivers were so crowded with tugboats and
commercial shipping.
What happened to the Port of New York? In the past halfcentury, it has been eclipsed by Rotterdam and Hong Kong and
Los Angeles and Long Beach. More important, its thousands of
jobs were rendered unnecessary because of the switch to
containers. These rectangular metal boxes, now forty feet in
length and longer, are stacked and unstacked on great
container ships that ply to waterways of the world. But they no
longer require gangs of stevedores; instead, one man in the cab
of a hoist, another who places a hook onto a container, and
another who guides it to the ground (or onto the rear of a
tractor-trailer truck) are able to accomplish the entire process
in less time and with less pilferage and loss than a hundred men
could have done a half-century earlier.
Thus, manufacturing and the port have both essentially
disappeared from the economy of New York. But unlike
Detroit or Cleveland or Newark or Buffalo or Pittsburgh,
Gotham reinvented itself as a different kind of city, a place on
the leading edge of the service and white-collar economies. As
a result, New York City has more and better jobs in 2005 than
it did in 1905 or 1955.

8. Openness, Tolerance, and Change
Change, openness, and tolerance are at the heart of what New
York is and what New York represents. For more than three
centuries, it has been more diverse and more open than any

other important city. Because of its history and its diversity,
Gotham has long been a haven for dissent. It is no accident that
the NAACP traces its origins to Manhattan and not to
Mississippi, or that the Communist Party made New York its
headquarters for the entire twentieth century, or that the Gay
Rights Movement reportedly began in the Stonewall bar in
Greenwich Village in 1969. New Yorkers as individuals are
probably no more tolerant than residents of South Carolina or
Oregon, as racial and ethnic confrontations too numerous to
mention in the city’s boroughs (fatal incidents in Howard
Beach, Crown Heights, and Bay Ridge, are just a few examples)
remind us. But the density, diversity, and size of New York have
made public dissent possible by granting anonymity to almost
anyone who wants it. A troublemaker in Mississippi could
easily be identified, located, and punished. But New York is far
too big and complex for its residents to concern themselves
with the politics, religion, or ethnicity of strangers.

No one has done a better job than E. B. White of
describing this essential characteristic of the great American
metropolis. “New York,” he wrote in 1949, “blends the gift
of privacy with the excitement of participation, and better
than most dense communities New York succeeds in
insulating the individual against all enormous and violent
and wonderful events that are taking place every minute.”
He continued with what remains the most succinct sentence
yet written about the big and gritty city: “New York is
peculiarly constructed to absorb almost anything that comes
along, whether a thousand-foot line out of the East or a
twenty-thousand man convention out of the West, without
inflicting the event on its inhabitants, so that every event is
in a sense optional, and the inhabitant is in the happy
position of being able to choose his spectacle and so
conserve his soul.”

FRBNY Economic Policy Review / December 2005

87

Endnotes

1. On the use of theory by urban historians, see Gilfoyle (2001).

References

Burrows, E. G., and M. Wallace. 1999. Gotham: A History of
New York City to 1898. New York: Oxford University Press.

Jackson, K. T., and D. Dunbar, eds. 2002. Empire City: New York
through the Centuries. New York: Columbia University Press.

Gilfoyle, T. 2001. “United States Urban History: Theoretical Graveyard
or Interpretative Paradise?” In H. Krabbendam, M. Roholl, and
T. de Vries, eds., The American Metropolis: Image and
Inspiration. Amsterdam: VU University Press.

Kaiser, C. 1997. The Gay Metropolis, 1940-1996. Boston:
Houghton-Mifflin.

Heilbrun, M., ed. 2000. Inventing the Skyline: The Architecture
of Cass Gilbert. New York: Columbia University Press.
Heinze, A. R. 1990. Adapting to Abundance: Jewish Immigrants,
Mass Consumption, and the Search for American Identity.
New York: Columbia University Press.

Kessner, T. 1977. The Golden Door: Italian and Jewish
Immigrant Mobility in New York City, 1880-1915.
New York: Oxford University Press.
Mollenkopf, J., and M. Castells, eds. 1991. Dual City: Restructuring
New York. New York: Russell Sage Foundation.
Moore, D. D. 1981. At Home in America: Second Generation
New York Jews. New York: Columbia University Press.

Hood, C. 1993. 722 Miles: The Building of the Subways
and How They Transformed New York. New York:
Simon and Schuster.

Rosenwaike, I. 1972. Population History of New York City.
Syracuse, N.Y.: Syracuse University Press.

Howe, I. 1976. World of Our Fathers: The Journey of the East
European Jews to America and the Life They Found and
Made. New York: Harcourt, Brace, Jovanovich.

Stansell, C. 2000. American Moderns: Bohemian New York and
the Creation of a New Century. New York: Henry Holt and
Company.

Jackson, K. T., ed. 1995. The Encyclopedia of New York City.
New Haven: Yale University Press.

Waldinger, R. 1996. Still the Promised City? African-Americans
and New Immigrants in Postindustrial New York.
Cambridge, Mass.: Harvard University Press.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
88

The Promised City

George J. Borjas

Immigration Trends
in the New York
Metropolitan Area
1. Introduction

T

here has been a resurgence of large-scale immigration in
the United States and in many other countries in recent
decades. Not surprisingly, the impact of immigration on
economic conditions in the receiving country is often a topic of
contentious policy debate. In the U.S. context, this concern has
motivated a great deal of research that attempts to document
how the U.S. labor market has adjusted to the large-scale
immigration in the past few decades. Much of this research has
focused on analyzing the determinants of the skill composition
of the foreign-born workforce (see the survey in Borjas [1994]).
This analytical focus can be easily justified by the fact that the
skill composition of the immigrant population is perhaps the
key determinant of the social and economic consequences of
immigration.
For example, the connection between the skill composition
of the immigrant population and the fiscal impact of
immigration is self-evident. The many programs that make up
the welfare state tend to redistribute resources from highincome workers to persons with less economic potential.
Skilled workers, regardless of where they were born, typically
pay higher taxes and receive fewer social services.
Skilled immigrants may also assimilate quickly. They might
be more adept at learning the tools and “tricks of the trade”
that can increase the chances of economic success in the
United States, such as the language and culture of the

George J. Borjas is the Robert W. Scrivner Professor of Economics and Social
Policy at Harvard University’s John F. Kennedy School of Government and
a research associate at the National Bureau of Economic Research.
<gborjas@harvard.edu>

American workplace. Moreover, the structure of the American
economy changed drastically in the 1980s and 1990s, and now
favors workers who have valuable skills to offer (Katz and
Murphy 1992). It seems, therefore, as if high-skill immigrants
would have a head start in the race for economic assimilation.
The skill mix of immigrants also determines which native
workers are most affected by immigration. Low-skill
immigrants will typically harm the economic opportunities of
low-skill natives, while high-skill immigrants will typically have
a similar effect on high-skill natives.
Finally, the skills of immigrants determine the economic
benefits achieved from immigration. The United States
benefits from international trade because it can import goods
that are not available or are too expensive to produce in the
domestic market. Similarly, a country can benefit from
immigration because it can import workers with scarce
qualifications and abilities.
In addition to measuring the relative skill endowment of
immigrants, the existing literature also stresses the economic
consequences that arise from the fact that immigrants cluster in
a small number of geographic areas (Friedberg and Hunt 1995;
Card 2001). It is well known that New York City and its
environs have been an important immigrant gateway for more
than a century. Although the geographic gravity of modern
immigration has shifted to other parts of the United States,
such as California, Texas, and Florida, the New York
metropolitan area remains an important receiving site. In 2000,

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

91

15.7 percent of all foreign-born workers resided in the
New York metropolitan area—down from 24.5 percent in
1970, prior to the resurgence of immigration.
This paper documents the impact of recent changes in
immigration settlement patterns on the skill endowment of
immigrants in the New York metropolitan area. The empirical
analysis uses the available U.S. census microdata between 1970
and 2000 to examine two related questions that inevitably lie at
the core of any study of immigration’s economic impact in the
New York area:
• Which types of immigrants choose to settle in
New York?
• How do these immigrants compare with the native-born
population of the New York region and with the
immigrants who choose to settle elsewhere?

2. Basic Trends
Our analysis uses data drawn from the 1970-2000 Integrated
Public Use Microdata Series (IPUMS) of the U.S. census.1 The
data contain information on the skills and labor market
outcomes of millions of workers in the United States.
Throughout this study, persons who are not citizens or who are
naturalized citizens are classified as immigrants; all other
persons are classified as natives.2 To examine the contribution
of immigration to the workforces of particular geographic
areas, we focus on the sample of workers aged twenty-five to
sixty-four who are not in the military and who are not enrolled
in school.
The growth of the foreign-born workforce in the New York
metropolitan area in the past two decades has corresponded
roughly with the growth of the foreign-born workforce in other
parts of the country. Chart 1, for example, illustrates trends in
the immigrant share—that is, the percentage of the workforce
that is foreign born—in the New York metropolitan area and
in the rest of the country (labeled “outside New York” in the
chart). In 1970, 15.8 percent of the workforce in the New York
metropolitan area was foreign born. The figure grew to
25.1 percent by 1990 and to 34.0 percent by 2000. This growth
rate is significantly faster than the growth rate in the immigrant
share outside the New York area, where the immigrant share
grew from 4.5 percent in 1970 to 11.9 percent in 2000.
Chart 1 also shows, however, that the immigrant share grew
even faster in some other metropolitan areas. In particular,
the chart summarizes the experience of three other large

92

Immigration Trends in the New York Metropolitan Area

metropolitan areas that are important gateways for
immigrants: Los Angeles, Miami, and Chicago. Both the
Los Angeles and Miami metropolitan areas have a substantially
larger immigrant share, and their immigrant share grew at a
much faster rate during the 1970-2000 period. In 1970, for
example, the New York metropolitan area had a slightly higher
immigrant share than did the Los Angeles metropolitan area
(15.8 percent and 12.6 percent, respectively). By 2000,
however, the immigrant share in the Los Angeles metropolitan
area had risen to 44.1 percent, a 10 percentage point difference
over the share in the New York metropolitan area. In Miami,
the immigrant share rose from 28.5 percent to 63.6 percent
over the same period.
One important difference between immigration to the
New York metropolitan area and to other parts of the country
lies in the national origin mix of the immigrant population.
It is well known that partly as a result of the policy changes
introduced by the repeal of the national origins quota system in
the 1965 Amendments to the Immigration and Nationality Act,
the national origin mix of immigrants shifted from Europe and
Canada to countries in Latin America and Asia beginning in the
1970s. Table 1 shows the difference in the national origin mix
of immigrants in the various U.S. regions as of 2000. The data
reveal that there is a great deal more diversity in the national
origin mix of the immigrant population in the New York

Chart 1

Trends in the Immigrant Share of the Workforce
By Area
Percent
70
60

Miami metro area

50

Los Angeles
metro area

40
30

New York
metro area

20

Outside New York

10
Chicago metro area

0
1970

80

90

00

Source: Author’s calculations, based on U.S. Census Bureau’s
1970-2000 Integrated Public Use Microdata Series.
Notes: The workforce is defined as the group of persons aged
twenty-five to sixty-four who are not enrolled in school and who
worked in the civilian sector at least one week in the year prior to
each decennial census. The immigrant share is the fraction of the
workforce that is foreign born.

Table 1

National Origin Mix of Immigrants, 2000
Percentage of Immigrant Stock Originating
in a Particular Country

Country
All immigrants
Canada
Mexico
Central
America
Cuba
West Indies
Europe
China
Korea
Philippines
Vietnam
India

New
York
Metro
Area

Outside
New
York

Los
Angeles
Metro
Area

Miami
Metro
Area

Chicago
Metro
Area

0.8
4.2

2.6
35.0

1.1
45.0

0.5
1.9

1.0
42.4

6.3
1.6
22.9
19.7
7.2
2.7
3.1
0.5
6.9

7.4
2.8
3.9
13.3
4.4
2.5
5.4
3.9
4.5

13.1
0.6
0.4
6.1
4.9
4.1
6.2
4.8
2.0

15.4
43.9
14.2
3.5
0.5
0.1
0.5
0.2
0.5

3.0
0.7
1.3
23.8
3.1
2.3
5.1
1.1
7.1

Source: Author’s calculations, based on U.S. Census Bureau’s 2000
Integrated Public Use Microdata Series.
Notes: Figures are calculated using the sample of persons aged twentyfive to sixty-four who are not enrolled in school and who worked in the
civilian sector at least one week in the year prior to each decennial census. The “outside New York” region is composed of the sample of persons residing outside the New York metro area.

metropolitan area than there is outside the New York area or in
other selected metropolitan areas.
Not surprisingly, outside the New York metropolitan area,
immigration is dominated by the Mexican origin population:
35.0 percent of immigrants and 40.0 percent of newly arrived
immigrants (that is, immigrants who have been in the United
States fewer than five years) outside the New York area are of
Mexican origin. In contrast, only about 4.2 percent and
8.9 percent of the immigrant and newly arrived immigrant
workforces in New York, respectively, are of Mexican origin.
In fact, the largest immigrant group in the New York
metropolitan area comprises those who originate in the West
Indies (which includes Jamaica and the Dominican Republic).
In 2000, 22.9 percent of immigrants in New York originated
in the West Indies. Outside the New York area, however,
immigration from the West Indies accounted for only
3.9 percent of the immigrant workforce. Equally interesting,
the second largest group of immigrants in the New York area is
formed by European immigrants; they make up 19.7 percent of
the immigrant workforce.

In contrast to the national origin mix of immigrants in
New York, consider the composition of the immigrant
workforce in the three other metropolitan areas (Table 1).
Between 40 percent and 50 percent of the immigrants in each
of these metropolitan areas belong to a single national origin
mix. In Los Angeles, 45.0 percent are of Mexican origin; in
Miami, 43.9 percent are of Cuban origin; and in Chicago,
42.4 percent are of Mexican origin.
It is well known that there are substantial differences in
socioeconomic outcomes among the various national origin
groups that make up the immigrant population and that
Mexican immigrants, in particular, tend to have relatively low
educational attainment and wages. As a result of these national
origin differentials, Table 1 suggests that the economic impact
of immigration on the New York area will likely differ
substantially from the impact on other metropolitan areas—
even if those other regions have roughly similar levels of
immigration.
We conclude this section by describing the occupational
distribution of immigrant men in New York and of immigrant
men outside New York.3 The first two columns of Table 2
present the basic distributions. The data indicate that a
relatively large fraction of immigrant men in the New York area
tend to be employed in management occupations and in sales.
These two occupations alone, in fact, employ a quarter of
immigrant men in the New York metropolitan area. The
concentration of immigrants in these occupations, of course,
could reflect the fact that the New York occupational structure
may be heavily weighted toward those types of jobs. To adjust
for the fact that the occupational distribution of immigrant
men in a particular region is affected by the occupational
structure of the local labor market, we report in the last two
columns of Table 2 the statistic given by the ratio of the
percentage of immigrants employed in a particular occupation
to the percentage of natives employed in the same occupation
in a particular region. A value of 1 for this statistic would imply
that immigrant and native men have the same proportional
representation in the particular occupation in the local labor
market. In the New York metropolitan area, immigrant men
tend to be underrepresented in such occupations as
management, business operations, legal, and protective service,
and are overrepresented in health care support, production,
and transportation and material moving. Remarkably, a
comparison of the last two columns of the table suggests that,
with only a few exceptions, there is a great deal of similarity in
the degree of immigrant penetration in particular occupations
in New York and outside New York.

FRBNY Economic Policy Review / December 2005

93

Table 2

Occupational Distribution of Immigrant Men, 2000

Percentage of Immigrants Employed
in Occupation
Occupation
All immigrant men
Management occupations
Business operations specialists
Financial specialists
Computer and mathematical occupations
Architecture and engineering
Life, physical, and social science
Community and social service
Legal
Education, training, and library
Arts, design, entertainment, sports
Health care practitioners and technical
Health care support
Protective service
Food preparation and serving
Building and grounds cleaning and maintenance
Personal care and service
Sales
Office and administrative support
Farming, fishing, and forestry
Construction trades
Extraction workers
Installation, maintenance, and repair workers
Production
Transportation and material moving

Percentage of Immigrants Employed
in Occupation Relative to Percentage of
Natives Employed in Occupation

New York Metro Area

Outside New York

New York Metro Area

Outside New York

13.9
2.4
3.6
3.8
2.4
0.9
1.1
2.9
3.4
3.9
2.7
0.5
5.4
1.9
3.4
1.1
11.7
8.5
0.1
7.4
0.0
5.8
5.2
7.9

12.3
2.0
2.0
3.0
3.6
1.0
1.1
1.2
2.7
1.8
2.3
0.4
3.2
1.7
3.2
0.9
10.3
6.4
0.8
10.5
0.2
7.9
11.5
10.1

0.6
0.6
0.6
1.2
0.9
1.1
0.6
0.2
0.4
0.5
1.1
2.0
0.4
3.7
1.6
1.4
0.8
0.8
2.0
1.3
0.3
1.0
2.1
1.5

0.7
0.6
0.6
1.4
1.1
1.3
0.6
0.3
0.6
0.8
1.2
1.2
0.3
3.6
1.9
1.2
0.7
0.8
3.9
1.2
0.4
0.8
1.3
0.9

Source: Author’s calculations, based on U.S. Census Bureau’s 2000 Integrated Public Use Microdata Series.
Notes: Figures are calculated using the sample of persons aged twenty-five to sixty-four who are not enrolled in school and who worked in the civilian sector
at least one week in the year prior to each decennial census. The “outside New York” region is composed of the sample of persons residing outside the
New York metro area.

3. The Skills and Earnings
of Immigrants
The skill composition of the immigrant population is the key
determinant of the economic impact of immigration. This
section examines how the skills and economic performance of
immigrants in the New York area compare with those of native
workers in the region as well as with those of foreign-born
workers in other regions of the country. In addition, we
document the extent to which regional differentials in
immigrant skills and economic performance have changed
over time.
Table 3 presents the trend in the distribution of educational
attainment for male native and immigrant workers. Because of
the rising level of educational attainment among native

94

Immigration Trends in the New York Metropolitan Area

workers, the table shows a significant decline in the fraction of
native working men who are high-school dropouts in all
geographic areas between 1970 and 2000. Outside the New
York metropolitan area, for example, the fraction of native
workers who are high-school dropouts fell from 40.0 percent to
8.0 percent between 1970 and 2000. In New York, the decline
was equally steep, from 37.2 percent to 5.7 percent.
The New York metropolitan area, however, witnessed a
much more rapid increase in the fraction of natives who are
college graduates. In the New York area, the fraction of male
workers with at least sixteen years of schooling rose from
20.1 percent to 41.5 percent between 1970 and 2000, or an
increase of 21.4 percentage points. Outside the New York area,
the fraction rose from 15.2 percent to 28.8 percent, or an
increase of 13.6 percentage points. This dramatic improvement

in the relative educational attainment of the native-born
workforce in the New York area will play an important role in
our discussion of regional differences in the relative economic
performance of the foreign-born workforce.
As it did among the native-born workforce, the fraction of
immigrants who are high-school dropouts fell between 1970
and 2000, with the decrease being steeper in the New York
metropolitan area. In New York, the fraction of immigrants
who are high-school dropouts fell from 52.3 percent to
21.5 percent, or a decrease of 30.8 percentage points. This
decline contrasts strikingly with the much more modest
15.8 percentage point drop that occurred outside the New York
metropolitan area, from 48.6 percent to 32.8 percent. Similarly,

Table 3

Distribution of Educational Attainment for Male
Workforce
Natives

Immigrants

1970

2000

1970

2000

New York metro area
High-school dropouts
High-school graduates
Some college
College graduates

37.2
31.5
11.3
20.1

5.7
27.2
25.6
41.5

52.3
22.5
9.7
15.5

21.5
30.7
18.2
29.7

Outside New York
High-school dropouts
High-school graduates
Some college
College graduates

40.0
33.2
11.6
15.2

8.0
33.1
30.2
28.8

48.6
21.8
11.1
18.4

32.8
23.5
17.2
26.6

Los Angeles metro area
High-school dropouts
High-school graduates
Some college
College graduates

27.4
32.5
20.7
19.5

4.7
21.5
34.8
39.0

45.0
22.7
14.9
17.3

39.4
22.6
16.8
21.2

Miami metro area
High-school dropouts
High-school graduates
Some college
College graduates

36.2
31.3
13.2
19.4

8.2
26.9
29.4
35.6

51.7
21.6
12.1
14.6

22.2
32.3
23.3
22.2

Chicago metro area
High-school dropouts
High-school graduates
Some college
College graduates

36.7
32.7
13.7
17.0

5.4
26.9
30.2
37.6

54.1
18.5
11.2
16.2

31.5
26.4
15.7
26.4

Source: Author’s calculations, based on U.S. Census Bureau’s 1970-2000
Integrated Public Use Microdata Series.
Notes: Figures are calculated using the sample of persons aged twenty-five
to sixty-four who are not enrolled in school and who worked in the
civilian sector at least one week in the year prior to each decennial census.
The “outside New York” region is composed of the sample of persons
residing outside the New York metro area.

there was a more rapid increase in the relative number of
foreign-born workers who are college graduates in New York
than there was elsewhere. In New York, the fraction of the
foreign-born workforce with a college degree rose from
15.5 percent to 29.7 percent, or an increase of 14.2 percentage
points. In contrast, the share of foreign-born college graduates
outside the New York area rose only from 18.4 percent to
26.6 percent, or an increase of 8.2 percentage points.
In sum, relative to the rest of the country, the New York
metropolitan area experienced a dramatic improvement in the
educational attainment level of its workforce between 1970 and
2000—for both native-born and foreign-born workers. The
New York area’s advantage is even more dramatic when the
trends in educational attainment are compared with the trends
experienced by other immigrant-receiving metropolitan areas.
In Los Angeles, for example, the share of immigrant men who
are high-school dropouts fell by only 5.6 percentage points
over the period, from 45.0 percent to 39.4 percent, while the
share who are college graduates rose by only 3.9 percentage
points, from 17.3 percent to 21.2 percent. Similarly in Miami,
the fraction of immigrants who are college graduates rose from
14.6 percent to 22.2 percent, or a 7.6 percentage point increase.
Note, however, that the improvement in the educational
attainment of the immigrant workforce in the New York
metropolitan area—although steep relative to that of the
immigrant workforce elsewhere—occurred concurrently with
an even faster improvement in the educational attainment of
New York’s native-born workforce. As a result, it will be
instructive to determine the trends in economic performance
of immigrants in New York not only relative to the native-born
population in the New York area, but also relative to the
foreign-born workforce that chooses to settle elsewhere.
Consider the trend in the wage differential between
immigrant and native workers within a certain geographic
region. Chart 2 summarizes the 1970-2000 trend in the log
weekly wage differential between male immigrant and native
workers in a particular region. Contrast initially the log wage
gap between immigrants and natives in the New York
metropolitan area with that found outside the New York area.
The chart reveals two interesting facts. First, immigrants living
outside the New York metropolitan area have a higher wage
relative to natives than do immigrants living in the New York
area. In other words, relative to the native workforce in the
specific region, immigrants are somewhat more skilled outside
the New York area. In 2000, for example, the log wage gap
between immigrants and natives stood at -.41 in New York and
-.22 outside New York, implying approximately a 34 percent
wage gap between immigrants and natives in New York and a
20 percent wage gap outside New York.4 Second, both in
New York and outside New York, the wage disadvantage of

FRBNY Economic Policy Review / December 2005

95

Chart 2

Trends in the Log Weekly Wage of Immigrant Men
Relative to the Wage of Native Men
By Area
Log wage gap
0
-0.1

Outside New York

-0.2

Chicago
metro area

-0.3
Miami
metro area

-0.4
-0.5

New York
metro area

Los Angeles metro area

-0.6
1970

80

90

00

Source: Author’s calculations, based on U.S. Census Bureau’s
1970-2000 Integrated Public Use Microdata Series.
Note: Figures are calculated using the sample of persons aged
twenty-five to sixty-four who are not enrolled in school and who
worked in the civilian sector at least one week in the year prior to
each decennial census.

immigrants relative to that of natives grew steadily between
1970 and 2000, and the rate of decline was approximately the
same in both regions.
Chart 2 also shows how the relative wage disadvantage of
immigrants differs across the main immigrant-receiving
metropolitan areas. Most striking is the experience of
Los Angeles, where the wage disadvantage grew dramatically
between 1970 and 2000. By 2000, immigrants in Los Angeles
earned approximately 41 percent less than native-born
workers.
As noted above, the trend in the log wage gap between
immigrants and natives in a particular geographic region does
not provide a complete picture of what is happening to
immigrant skills because native skills have been changing over
time as well—and the dramatic improvement in native
educational attainment in the New York area may account
for a large part of the increasing relative disadvantage of
immigrants in that area. In other words, the tracking provided
in Chart 2 isolates the trend in the relative economic standing
of immigrants in a particular geographic region—but it may
provide a very misleading picture about whether a certain
region is attracting a more skilled immigrant workforce than
are other regions.
To isolate what is happening to immigrant skills in
New York as compared with immigrant skills elsewhere, we
contrast the wage of immigrants in New York with the wage
of immigrants in other parts of the country. One important
difficulty with this type of comparison is the presence of

96

Immigration Trends in the New York Metropolitan Area

differences in wage levels across metropolitan areas that reflect
cost-of-living differences.5 To adjust for these cost-of-living
differentials, we use the respective Bureau of Labor Statistics
cost-of-living index for each particular metropolitan area to
deflate the wage data reported in the various censuses.
Chart 3 illustrates the change in the (deflated) log weekly
wage of immigrants in the New York area relative to
immigrants in other areas. Compare initially the trend in the
real wage of immigrants in New York with that of immigrants
in the rest of the country.6 In 1970, the typical New York area
immigrant earned slightly less than the typical immigrant
residing outside New York (the log wage gap was -.01), and the
immigrant position worsened slightly between 1970 and 1980
(the log wage gap in 1980 stood at -.03). Although the data are
somewhat noisy, the chart reveals that there was a general
improvement in the real wage of immigrants in New York
relative to that of immigrants elsewhere between 1980 and 2000,
so that by 2000 the log wage gap stood at .037. In short, at the
same time that the wage of immigrants in New York was falling
relative to that of natives in New York, it was improving relative
to that of immigrants employed outside the New York area.
The comparison between immigrants employed in
New York and in some of the other immigrant-receiving
metropolitan areas indicates that immigrants in New York are
substantially more skilled than the immigrants who settle in
Los Angeles or Miami. The difference between Los Angeles and
New York is particularly striking. In 2000, the log wage gap of
.126 between the two groups of immigrants implied that

Chart 3

Log Weekly Wage of Immigrant Men in the New York
Metro Area Relative to the Wage of Native Men
By Area
Log wage gap
0.3
New York versus
Miami

0.2

New York versus
Los Angeles

0.1
0
-0.1

New York versus
rest of country

New York versus
Chicago

-0.2
1970

80

90

00

Source: Author’s calculations, based on U.S. Census Bureau’s
1970-2000 Integrated Public Use Microdata Series.
Note: Figures are calculated using the sample of persons aged
twenty-five to sixty-four who are not enrolled in school and who
worked in the civilian sector at least one week in the year prior to
each decennial census.

New York immigrants earned about 14 percent more than
their counterparts in Los Angeles.
The difference in the results between Charts 2 and 3 implies
that a systematic evaluation of the economic impact of
immigration in the New York area will inevitably have to
confront the fact that, while New York immigrants are
relatively more skilled than immigrants elsewhere, they are
relatively less skilled than native workers in New York—and
that while the skill advantage of New York’s immigrants
relative to immigrants elsewhere is growing over time, the skill
disadvantage of New York’s immigrants relative to New York’s
natives is also growing. In an important sense, the New York
area is doing quite well competing for skilled immigrants in
the “immigration market,” but the skill level of the native
New York workforce is increasing even more rapidly, so that
even the relatively skilled immigrants attracted by New York’s
labor market are at an increasing disadvantage in the local
economy.
Many studies in the modern literature on the economics of
immigration focus on analyzing how the earnings potential of
immigrant workers adapts to the host country’s labor market.7
In the past two decades, this literature has concentrated on
measuring both the “assimilation” and “cohort” effects that
jointly determine the evolution of the relative wage of
immigrants over time (Chiswick 1978; Borjas 1985, 1995). The
assimilation effect arises because immigrants acquire relatively
more human capital than do native workers as they accumulate
experience in the U.S. labor market. As a result, the human
capital stock of immigrants grows relative to that of natives,
and immigrants experience faster wage growth. Cohort effects
arise because there may be permanent differences in skills
among immigrant waves. For example, the immigrants who
arrived in the late 1990s may be different (as reflected, for
example, by the entry wage) than the immigrants who arrived
in the late 1970s, who, in turn, might differ from those who
arrived in the late 1950s.8
Chart 4 summarizes the evidence on interregional
differences in cohort effects over the past thirty years by
looking at the trend in the log wage gap between native workers
and immigrants who belong to the cohort of newly arrived
immigrants at each census date (that is, immigrants who have
been in the United States fewer than five years as of the census
date) in a particular geographic region. Consider initially the
cohort effect for the immigrants who are residing outside the
New York metropolitan area shortly after their arrival in the
United States. The trend in their relative wage clearly indicates
that the relative wage of consecutive immigrant cohorts
declined between 1970 and 1990, from a 20 percent wage
disadvantage in 1970 to 35 percent in 1990. Interestingly, this
trend was reversed in the 1990s. By 2000, the wage disadvantage

of newly arrived immigrants living outside the New York
metropolitan area rose to 31 percent.
The comparison of the trend for cohort effects among
immigrants living outside the New York area with the cohort
effects for immigrants residing in the New York area yields two
interesting findings. First, newly arrived immigrants in the
New York area tend to do systematically worse than newly
arrived immigrants elsewhere in the country—relative, of
course, to natives in each of the respective geographic regions.
In 1990, for example, the relative wage disadvantage of newly
arrived immigrants living in the New York area was 41 percent,
as compared with a disadvantage of 35 percent for newly
arrived immigrants living outside New York. Second, the
“uptick” in the relative skills of new immigrants arriving
between 1990 and 2000 is not found among newly arrived
immigrants settling in the New York area.
Borjas and Friedberg (2004) have recently shown that the
uptick in cohort quality for immigrants who arrived in the late
1990s (at the national level) can be explained in terms of a
simple example that has significant policy relevance. In
particular, the entire uptick disappears when the relatively
small number of immigrants who are employed as computer
scientists and engineers is excluded from the analysis. Although
the census does not provide information on the type of visa that
immigrants use to enter the country, it is probably not a

Chart 4

Log Weekly Wage of Newly Arrived Immigrant Men
Relative to the Wage of Native Men
By Area
Log wage gap
-0.2
-0.3

Outside New York

-0.4
New York
metro area

-0.5
-0.6

Miami
metro area

-0.7
-0.8

Chicago
metro area

Los Angeles metro area

-0.9
1970

80

90

00

Source: Author’s calculations, based on U.S. Census Bureau’s
1970-2000 Integrated Public Use Microdata Series.
Notes: Figures are calculated using the sample of persons aged
twenty-five to sixty-four who are not enrolled in school and who
worked in the civilian sector at least one week in the year prior to each
decennial census. The sample of newly arrived immigrants includes
foreign-born persons who have been in the United States for fewer than
five years as of the census date.

FRBNY Economic Policy Review / December 2005

97

coincidence that the increase in the relative number of hightech immigrants occurred at the same time that the size of the
H-1B visa program grew substantially. This program allows
employers to sponsor the entry of temporary workers in
“specialty occupations.” Most of the workers entering the
country with an H-1B visa are employed either in computerrelated occupations or in engineering (70 percent in 2000).9
Between 1990 and 1994, the number of H-1B visas hovered
around 100,000 annually. This number increased to 144,548 in
1996, to 240,947 in 1998, and to 302,326 in 1999.10
It turns out that the growth in high-tech employment for
native workers was roughly similar in New York and outside
New York, but the growth in high-tech employment for newly
arrived immigrants lagged slightly in the New York area. In
1990, for example, about 3.5 percent of native workers were
employed in computer-related occupations or engineering.
In 2000, the fraction of natives employed in these high-tech
occupations stood at 5 percent both in New York and outside
New York. Among immigrants, however, the fraction
employed in high-tech occupations increased by 4.5 percentage
points, from 3.0 percent to 7.5 percent, in New York, but by
5.3 percentage points, from 3.6 percent to 8.9 percent, outside
New York. It would be of great interest to explore whether the
relatively slow growth of foreign-born high-tech employment
in the New York metropolitan area (due, perhaps, to the
concentration of H-1B employment on the West Coast) could
explain the differential cohort effects revealed by the data.

As noted earlier, the changing log wage gap between
immigrant and native workers in each metropolitan area could
also reflect a region-specific changing mix of skills in the
native-born workforce. To isolate the status of the newly
arrived immigrant population in New York relative to that of
newly arrived immigrants residing elsewhere in the country, we
calculate the (real) wage of immigrants in the New York
metropolitan area relative to the real wage of immigrants in
other parts of the country. Chart 5 summarizes the trends in
this adjusted real wage. Although the trends are noisy, the data
clearly indicate that newly arrived immigrants in the New York
area typically earn substantially more than newly arrived
immigrants in other parts of the country.
Finally, the 1970-2000 census data can also be used to
measure the extent of “economic assimilation,” the
improvement in the relative wage of a specific immigrant
cohort over time. Chart 6 uses a simple methodology to

Chart 6

Economic Assimilation of Immigrant Men (Relative
Wage of Immigrants Who Entered the Country at
Ages Twenty-Five to Thirty-Four)
By Area
Log wage gap
0.1

Outside New York

Arrived in
1965-69

0

Log Weekly Wage of Newly Arrived Immigrant Men
in the New York Metro Area Relative to the Wage
of Newly Arrived Immigrant Men in Other Areas
Log wage gap

Arrived in
1985-89

-0.2

-0.3

0.4

New York versus
Los Angeles

New York versus
rest of country

0.3

Arrived in
1975-79

-0.1

Chart 5

0.2

New York versus
Miami

0.1

-0.1

New York Metro Area
Arrived in
1965-69

-0.2

Arrived in
1975-79

-0.3
New York versus
Chicago

0

-0.4
Arrived in
1985-89

-0.1
1970

80

90

00

Source: Author’s calculations, based on U.S. Census Bureau’s
1970-2000 Integrated Public Use Microdata Series.
Notes: Figures are calculated using the sample of persons aged
twenty-five to sixty-four who are not enrolled in school and who
worked in the civilian sector at least one week in the year prior to
each decennial census. The sample of newly arrived immigrants
includes foreign-born persons who have been in the United States
for fewer than five years as of the census date.

98

Immigration Trends in the New York Metropolitan Area

-0.5
1970

80

90

00

Source: Author’s calculations, based on U.S. Census Bureau’s
1970-2000 Integrated Public Use Microdata Series.
Note: Figures are calculated using the sample of persons aged
twenty-five to sixty-four who are not enrolled in school and who
worked in the civilian sector at least one week in the year prior to
each decennial census.

calculate rates of economic assimilation within specific regions
of the country. Consider first the group of immigrant men
living outside the New York area who arrived in the late 1960s
when they were twenty-five to thirty-four years old. The top
panel of Chart 6 shows that these immigrants earned about
11 percent less than comparably aged native workers at the
time of entry (as observed in the 1970 census). Move forward
ten years to 1980, when both the immigrants and the natives
were thirty-five to forty-four years old. The wage gap between
the two groups has essentially disappeared. Move forward
again ten years to 1990, when the workers are now forty-five to
fifty-four years old. The data indicate that immigrants now
earn about 2.8 percent more than native workers. Overall, the
process of economic assimilation exhibited by this cohort
reduced the initial wage disadvantage of these immigrants by
about 14 percentage points over a thirty-year period—with
most of the growth occurring in the first ten years after
immigration.
Contrast this pattern with the rate of economic assimilation
measured for immigrants who arrived when they were twentyfive to thirty-four years old in 1970 and resided in the New York
metropolitan area at the time of each census observation
(Chart 6, bottom panel). They entered the country with a
22.5 percent wage disadvantage. Unlike their counterparts
who lived outside New York, the wage gap between these
immigrants and native workers in New York remained
relatively constant over the next thirty years. By 2000, the wage
disadvantage between these workers still stood at 22.9 percent.
Although it may be tempting to conclude from these
calculations that immigrants in the New York metropolitan
area do not experience much economic assimilation, it is
unlikely that this interpretation is correct. For example, there is
a great deal of interregional internal migration between
New York and other parts of the country in both the foreignborn and native-born workforces. Suppose, for instance, that
these internal migration flows lead to a large number of low-

skill immigrants moving into the New York metropolitan area
after their initial settlement elsewhere, or lead to the outmigration of high-skill immigrants who initially settled in the
New York area. These internal migration flows could easily
generate the perverse assimilation paths illustrated in the
bottom panel of Chart 6. As a result, the intriguing differences
in the synthetic assimilation profiles generated by the tracking
of specific cohorts across various census data sets suggest that
the differential internal migration decisions of immigrant and
native workers in the New York metropolitan area remain an
important topic for future research.

4. Summary
This paper uses data drawn from the 1970-2000 Integrated
Public Use Microdata Samples of the U.S. census to analyze the
trends in the educational attainment and earnings of
immigrants in the New York metropolitan area. Although the
growth of immigration in California, Texas, and Florida in
recent decades has shifted the geographic gravity of
immigration in the United States, the New York metropolitan
area remains an important receiving site. In 2000, 15.7 percent
of all foreign-born workers resided in the New York
metropolitan area.
The empirical analysis presented here documents the
observation that although the immigrants who settle in the
New York area tend to be more skilled than the immigrants
who settle elsewhere, they tend to be less skilled than nativeborn workers in the New York area. Moreover, because of the
dramatic improvement in the educational attainment of
native-born workers in New York in recent decades, the
(relative) economic disadvantage experienced by immigrants
in New York has widened.

FRBNY Economic Policy Review / December 2005

99

Endnotes

1. These data are available at the University of Minnesota’s IPUMS
website (<http://www.ipums.umn.edu/usa/index.html>). The data
contain a 1 percent sample of the U.S. population in 1970 and a
5 percent sample in 1980-2000.

5. Note that these differences do not play a role in the data
summarized in Chart 2 because these data difference the earnings of
immigrants and natives within a metropolitan area at a particular
point in time.

2. This definition implies that persons born abroad of American
parents or persons born in American territories are classified as
natives. Some of the variables reported in the census, such as annual
earnings, refer to the year prior to the survey. We avoid confusion
by always referring to the data in terms of the census year.

6. To deflate the wage for immigrant workers residing outside the
New York metropolitan area, we simply use the national aggregate of
the consumer price index.

3. The remainder of the analysis focuses on the trends in skills and
earnings of the male workforce. The trends in the relative wage of
immigrant women (and interregional differences in those trends) are
likely to be heavily influenced by the selection issues that characterize
the huge differences in female labor force participation rates both
across groups and across regions.

7. Borjas (1994) and Smith and Edmonston (1997) survey this
extensive literature.
8. The cross-section correlation may also be contaminated by cohort
effects if there is selective out-migration of immigrants, so that the
trend in the earnings of “survivors” over time will not measure the
actual earnings growth experienced by a particular immigrant cohort.
9. U.S. Immigration and Naturalization Service (2002).

4. The percentage wage gap implied by a specific value of the log wage
gap, x, is given by ex – 1.

100

Immigration Trends in the New York Metropolitan Area

10. U.S. Immigration and Naturalization Service (various years).

References

Borjas, G. J. 1985. “Assimilation, Changes in Cohort Quality, and the
Earnings of Immigrants.” Journal of Labor Economics 3, no. 4
(October): 463-89.

Katz, L. F., and K. M. Murphy. 1992. “Changes in Relative Wages,
1963-87: Supply and Demand Factors.” Quarterly Journal
of Economics 107, no. 1 (February): 35-78.

———. 1994. “The Economics of Immigration.” Journal
of Economic Literature 32, no. 4 (December): 1667-1717.

LaLonde, R. J., and R. H. Topel. 1992. “The Assimilation of Immigrants
in the U.S. Labor Market.” In G. J. Borjas and R. B. Freeman, eds.,
Immigration and the Work Force: Economic Consequences
for the United States and Source Areas, 67-92. Chicago:
University of Chicago Press.

———. 1995. “Assimilation and Changes in Cohort Quality
Revisited: What Happened to Immigrant Earnings in the 1980s?”
Journal of Labor Economics 13, no. 2 (April): 201-45.
Borjas, G. J., and R. M. Friedberg. 2004. “What Happened to
Immigrant Earnings in the 1990s?” Unpublished paper,
Harvard University, March.
Card, D. 2001. “Immigrant Inflows, Native Outflows, and the Local
Labor Market Impacts of Higher Immigration.” Journal
of Labor Economics 19, no. 1 (January): 22-64.
Chiswick, B. R. 1978. “The Effect of Americanization on the Earnings
of Foreign-Born Men.” Journal of Political Economy 86, no. 5
(October): 897-921.
Friedberg, R. M., and J. Hunt. 1995. “The Impact of Immigration on
Host Country Wages, Employment, and Growth.” Journal
of Economic Perspectives 9, no. 2 (spring): 23-44.

Ramos, F. 1992. “Out-Migration and Return Migration of Puerto
Ricans.” In G. J. Borjas and R. B. Freeman, eds., Immigration and
the Work Force: Economic Consequences for the United
States and Source Areas, 49-66. Chicago: University of Chicago
Press.
Smith, J. P., and B. Edmonston, eds. 1997. The New Americans:
Economic, Demographic, and Fiscal Effects of
Immigration. Washington, D.C.: National Academy Press.
U.S. Immigration and Naturalization Service. 2002. “Report on
Characteristics of Specialty Occupation Workers (H-1B):
Fiscal Year 2000.” April. Washington, D.C.
———. Various years. Statistical Yearbook of the Immigration
and Naturalization Service. Washington, D.C.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2005

101

Stephen J. Trejo

Commentary

eorge J. Borjas’ paper provides a very clear and
convincing analysis of the labor market skills and
earnings of immigrant workers in the New York metropolitan
area. The author compares New York immigrants with
U.S. natives residing in the same metropolitan area and with
immigrants residing elsewhere in the United States, including
other large cities such as Los Angeles, Chicago, and Miami.
Using decennial census data, Borjas tracks these comparisons
over the 1970-2000 period. A key finding is that New York
workers, immigrants as well as natives, are more skilled
than workers in the rest of the country. Interestingly, the
skill advantage of New York immigrants relative to other
immigrants has widened over the past thirty years, but so has
the skill disadvantage of New York immigrants relative to
New York natives.
The empirical analysis is transparent, sensible, and
compelling. Initially, I worried about Borjas’ decision to group
island-born Puerto Ricans with U.S. natives rather than with
immigrants. Although Puerto Ricans are U.S. citizens and
therefore not subject to restrictions on their migration to the
mainland, those who do migrate face some of the same
adjustment issues as other foreign-born workers. Moreover,
I was concerned that the exclusion of relatively low-skilled
Puerto Ricans from his immigrant sample was driving Borjas’
finding that New York immigrants are more skilled than
immigrants living elsewhere in the United States. It turns out,
however, that the author’s findings are not sensitive to whether

G

island-born Puerto Ricans are grouped with immigrants or
natives. For example, in the 2000 census data, redefining
island-born Puerto Ricans as immigrants would increase the
size of the New York metropolitan area immigrant sample by
less than 10 percent and would have a negligible impact on
estimates of the average education or earnings of either
immigrants or natives in the area. As Borjas shows, the national
origins of immigration flows to New York are much more
diverse than those to other U.S. gateway cities; thus, the overall
pattern of immigration flows into New York is not dominated
by the characteristics of immigrants from any one source
country. Indeed, over the last couple of decades, substantial
inflows of Mexicans and Central and South Americans have
joined the sizable Puerto Rican and Dominican populations
that had already been established, making the New York
metropolitan area perhaps the only place in the United States
with significant numbers of Latin American immigrants from
virtually all of the major Hispanic national origin groups.
I do not doubt Borjas’ basic empirical findings about
New York immigrants, but I do question how we should
interpret these findings. For example, how much of the skill
advantage of New York metropolitan area immigrants relative
to other U.S. immigrants derives from differences in national
origins, especially when we consider the fact that New York
receives comparatively few low-skilled immigrants from
Mexico? This question could be answered with a simple
decomposition analysis, similar to what Borjas has done in

Stephen J. Trejo is an associate professor of economics at the University
of Texas at Austin.
<trejo@eco.utexas.edu>

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

103

previous work on immigration. The answer is of interest
because it would reveal the extent to which New York can
attract more skilled immigrants from a given source country.
As another example, consider Borjas’ finding that, outside
the New York metropolitan area, the trend of declining skills
for new immigrant arrivals reverses in 2000, but this reversal
does not occur in the New York area. Citing his recent work
with Rachel Friedberg, Borjas attributes the uptick in
immigrant skills observed in nationwide data for 2000 to the
large number of high-tech H-1B immigrants who arrived in
the late 1990s, and he speculates that the absence of such an
uptick in New York may reflect a smaller influx of H-1B
immigrants there.
For two reasons, however, I doubt that the H-1B visa
program is the entire story here. First, Borjas shows that
between 1990 and 2000, the share of new immigrants employed
in high-tech occupations grew only slightly less in New York
(from 3.0 to 7.5 percent) than it did outside New York (from
3.6 to 8.9 percent). I am skeptical that this small difference
accounts for the fact that immigrant skills were falling in
New York over this period while they were rising in the rest
of the country. Certainly, it would be a simple matter for Borjas
to replicate for New York the analysis that he and Friedberg
conducted at the national level and, in that way, evaluate the
accuracy of his speculation. Second, I believe that even at the
national level, more is going on than just the effects of the
H-1B program. Borjas and Friedberg show that, when they
exclude immigrants who work in high-tech occupations, the
average skills of new immigrants are similar in 1990 and 2000.
Therefore, the influx of high-tech immigrants in the late 1990s
(many of whom are presumably H-1B admissions) might
explain the rise in immigrant skills between 1990 and 2000,
but it cannot explain why the downward trend, observed from
1970 to 1990, halted in 2000. Even after one excludes high-tech
workers, immigrant skills leveled off between 1990 and 2000,

rather than declined, as the preceding twenty-year trend led
us to expect.
Finally, as other researchers do, Borjas argues that the
skill level of immigrant workers is an important issue for
U.S. policy, but he provides only a cursory discussion of what
the optimal skill mix of U.S. immigrants might look like.
The underlying tone of the paper suggests that Borjas views
skilled immigrants as better for the United States than unskilled
immigrants, but a more explicit discussion of this topic would
have been enlightening. As Borjas notes, skilled immigrants
probably have a more favorable effect on government budgets
because they tend to pay more taxes and receive less public
assistance. From an international trade perspective, however,
the United States might be thought of as having a relative
abundance of skilled labor; therefore, it would make sense to
import unskilled labor via both trade and immigration. As
such, unskilled immigration and the unskilled labor embodied
in imported goods might be two sides of the same coin. In
this context, it is interesting to note that unskilled U.S.
immigrants seem to concentrate in sectors that produce
nontraded goods and services (for example, construction,
restaurants, hotels, and domestic service). Perhaps unskilled
immigrants are a viable substitute for imports in these sectors.
At any rate, a bit more discussion of optimal immigration
policy could have provided a nice framework for interpreting
the provocative empirical findings that Borjas so deftly reveals.
For instance, what should we make of the widening skill
gap between New York natives and immigrants? Is this a
“problem”? Evidently, New York is doing quite well in the
competition with other U.S. cities to attract skilled immigrants,
and it is doing even better in the competition to attract skilled
natives. Is this a good thing for New York or for the United
States as a whole? Answers to questions like these will help us
to understand the policy consequences of Borjas’ findings.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
104

Commentary

John Mollenkopf

Trajectories for the
Immigrant Second Generation
in New York City
1. Introduction

I

t has become a truism to say that immigration has
transformed American society since 1965. Beginning with
“gateway” cities like New York and Los Angeles, the effect of
new immigrants now extends to small pork- or chickenprocessing towns in Iowa or North Carolina. Indeed, the
March 2004 annual demographic supplement to the Current
Population Survey (CPS) indicates that almost 12 percent of
America’s residents were born abroad, doubtless an
underestimate. In places where first-generation immigrants
concentrate, like New York City, immigrants now make up
almost half the adult population—and in the case of Miami,
more than three-fifths. This outcome has led scholars to
undertake many studies of the new immigrants, for example,
using individual traits to model individual earnings or looking
at the school performance or health conditions of the children
of immigrants.
One leading researcher, George Borjas, has warned that the
relatively low skill levels of recent immigrants bode poorly for
their lifetime earnings and chances for upward mobility (Borjas
1990, 1999). Incorporating new immigrant ethnic groups also
poses many other challenges, such as heightened tensions
among ethnic and racial groups (Gerstle and Mollenkopf 2001).
Despite problematic aspects of the effect of immigration,

John Mollenkopf is executive director of the Center for Urban Research
of the Graduate Center of the City University of New York.
<jmollenkopf@gc.cuny.edu>

however, many observers, including this one, think that the
new immigrants constitute a clear net plus for American
society. Immigrants are “positively selected” from their
populations of origin (Feliciano 2005). They pass a difficult test
by resettling themselves and their families in the United States.
They often take jobs natives do not want to perform, work hard
for long hours, contribute a great deal of entrepreneurial
creativity, and bring valuable cultural capital—qualities that
their wages or other standards may not reflect immediately.
While competition from immigrants may put some low-skilled
natives, often members of minority groups, at a disadvantage
in the labor market—and indeed highly skilled immigrants
may compete against highly skilled natives—it seems to me
that the strong work effort, relatively low labor cost, and varied
talents of immigrants expand the overall economy and benefit
most native-born people. Certainly, the official New York City
position is that immigrants have prevented the city from
becoming smaller, poorer, and more like Philadelphia (Lobo
and Salvo 2004, p. xiv). Regardless of how many books scholars
write on this topic, however, they are not likely to resolve
anytime soon the question of whether new immigrants are
good or bad for America.
That may not be the most important question, however.
Instead, the fates of their children—the new second
generation—will likely shape how we evaluate the current
Much of the data and most of the ideas presented here have been developed in
collaboration with Philip Kasinitz, Mary Waters, and Jennifer Holdaway, my
partners in the Study of the Immigrant Second Generation in Metropolitan
New York. They provided valuable criticism of the first draft but do not
necessarily agree with all of my conclusions. They inspired the good qualities
of this study, and any remaining errors are mine alone.
The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.
FRBNY Economic Policy Review / December 2005

105

epoch of immigration. If the children of immigrants continue
on their parents’ upward path, the judgment is likely to be
positive. After all, we judge the last great era of immigration,
the 1880s to the 1920s, to have been a success because
subsequent generations advanced, on average, beyond the
previous ones (DiNardo and Estes 2000; Card 2005). As more
and more descendants of post-1965 immigrants come of age
today, scholars have begun to focus on what is happening to
them. In addition to studies of individual outcomes, studies of
this group, which includes native-born children of immigrants,
have considered their family and neighborhood contexts
(Kasinitz, Mollenkopf, and Waters 2004). To paraphrase Max
Frisch, “we asked for workers, but families came.”
The children of immigrants are numerous. The March
2004 CPS indicates that 10.6 percent of America’s residents
are native-born individuals with at least one immigrant
parent (who might, following Rumbaut [2003], be termed
2.0- or 2.5-generation immigrants). If we subtract the
1.5-generation youngsters (defined as those who arrived by
age twelve and then grew up here) from the immigrant total
and add them to the native children with at least one
immigrant parent, then adult immigrants over seventeen
make up about 9.4 percent of the national population,
while their 1.5-, 2.0-, and 2.5-generation children make up
12.9 percent. According to the March 2004 CPS, more than
half the youngsters under eighteen in New York and almost
two-thirds of those in Los Angeles County have at least one
immigrant parent. Clearly, the fates of these youngsters are
vital to the future of such cities.
The decennial census provides a way to take a more detailed
look at young people growing up in immigrant households
than is possible from the Current Population Survey. Unlike
the CPS, the census no longer asks where one’s parents were
born. But if we look at young people still living in their parents’
homes, we can use the U.S. Census Bureau’s Public Use
Microdata Sample (PUMS) to identify the nativity of parents.
The 2000 PUMS indicates that 1.62 million biological children,
adopted children, or stepchildren under the age of eighteen
lived in families headed by their parent or parents in New York
City in 2000.1 (As they age past eighteen, children are
increasingly likely to leave their parents’ households,
preventing us from knowing from the census the nativity of
their parents. Almost all of those younger than eighteen,
however, live in their families of origin, so we can analyze them
from census data.) About 1 million of these youngsters lived in
families with a household head and spouse, while 619,000 lived
in families with a householder, typically the mother, and no
spouse present. (Such families often did, however, include
other adults, such as an unmarried partner or a grandparent.)
Table 1 shows that 513,000 (50.8 percent) of those living in
two-parent families had two immigrant parents, while another

106

Trajectories for the Immigrant Second Generation

Table 1

Families with Related Children under Age Eighteen
and Number of Related Children under Age
Eighteen by Nativity of Family Head and Spouse
and Family Type

Household
Type
NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

Households

Percentage
of
Households

Related
Children

Percentage
of Related
Children

211,472
259,959
67,743
299,504
183,441
1,022,119

20.7
25.4
6.6
29.3
17.9
100.0

373,410
370,227
122,763
512,537
249,047
1,627,984

22.9
22.7
7.5
31.5
15.3
100.0

Source: U.S. Census Bureau, 2000 5 Percent Public Use Microdata
Sample.
Notes: The sample is all New York City households in 2000 with related
children under age eighteen. NB is native born, FB is foreign born, 2PF
is two-parent family, 1PF is single-parent family.

123,000 (12.2 percent) had one immigrant and one native
parent. Almost two-thirds of those growing up in two-parent
households therefore had at least one immigrant parent.
Among children living in single-parent families, two-fifths had
a foreign parent. Taking both types of families together, we
note that children with at least one immigrant parent thus
made up 54 percent of the young people in New York City
families in 2000. If something differentially bad is happening to
them, or even a large subset of them, it would not be good for
the city’s future.
There is reason to worry about the future of this second
generation. While New York City can be tough on any young
person, regardless of where their parents were born, the
children of immigrants face extra difficulties. First, only a third
of New York City’s 3 million households are families with
related children under eighteen. (In other words, two-thirds of
the households do not face the burdens of rearing children.)
Within that group of families with children, those headed by
immigrant parents are much less likely to speak English at
home (only 19 percent do, as opposed to 60 percent of those
headed by native parents) and they may not even understand
English (about a quarter, as opposed to only 4 percent of native
parents).2 Only half the parents in immigrant families are
citizens, compared with 100 percent of native parents, giving
them far less political influence than native parents have.3
Most crucially, immigrant parents are less likely to be well
educated than native parents: a third lack a high-school degree,
compared with one-fifth of native parents; while only a fifth
have college degrees, compared with a quarter of native
parents. As a consequence, they have less income. Immigrant

parents had a mean household income of $54,404 in 1999,
compared with $73,983 for native parents. Although white
immigrants move to New York, only 18 percent of immigrant
parents classify themselves as non-Hispanic whites, compared
with 41.5 percent of native parents. Immigrant parents often
live in neighborhoods surrounded by families with similar
characteristics, potentially reinforcing their disadvantages.
While living among fellow immigrants may also convey some
advantages—for example, through employment opportunities
available via ethnic networks—it would not seem logical that
they outweigh the challenges of immigrant life. In short, kids
growing up in immigrant families have parents with less
English facility, less education, less political clout, and less
income than those growing up in native families. It would not
be surprising if these factors constituted barriers to their
progress.
Scholars speculating about second-generation trajectories
have also worried that the larger social patterns of racial
inequality and discrimination will force those children of
immigrants who are not classified as white into the ranks of
persistently poor native minorities. Gans (1992), for example,
was concerned that being black would trump the aspirations
for upward mobility of dark-skinned children of immigrants,
and his hypothesis received support from Waters’ (2001)
ethnography of Afro-Caribbeans in New York City. Building
on this concern, Portes and his colleagues developed the
“segmented assimilation” model of second-generation
trajectories (Portes and Zhou 1993; Portes 1995; Zhou 1997;
Portes and Rumbaut 2001a, pp. 44-69, 280-6; 2001b,
pp. 303-12).
While the nuances and subtleties of this formulation allow
for a wide variety of individual outcomes, its core idea is that
whether they like it or not, groups of immigrants are forced
to face the fundamental American condition of racial
stratification and discrimination. Depending on immigrants’
national origins, group socioeconomic characteristics, and the
particular conditions of the places where they end up settling,
the segmented assimilation model posits three general
trajectories that groups might follow. A positive reception from
the white middle-class majority would enable light-skinned
immigrants from relatively high-income countries to
assimilate relatively easily into the mainstream. Racial
inequality, however, would force dark-skinned immigrants
from poorer countries to assimilate downwardly into a native
minority lower class. Groups that cannot easily be classified
into white and black categories, however, might try to retain
their cultural distinctiveness in service of economic
achievement, especially when a group has developed a strong
ethnic economy.

While this model has been subject to theoretical and
substantive criticism (Waldinger and Feliciano 2004; Alba and
Nee 2003), the notion that the dynamics of racial inequality in
host societies will force major parts of the second generation
toward downward mobility and socioeconomic exclusion has
motivated a growing and intense debate in the United States
and Europe. While Europe lacks an exact analog to AfricanAmericans as an historically subordinated domestic racial
group in the United States, many European nations must
contend with difficult colonial legacies (European Commission
2003). In both places, some second-generation immigrant
groups occupy particularly problematic positions. Most firstgeneration immigrants who entered bad situations in the
receiving countries ultimately had higher earnings or income
over time than they would have had in their old countries
(otherwise, they would have gone home). In the United States,
a striking number moved well beyond their low starting points.
As a result, some degree of upward mobility seems practically
built into the first-generation immigrant experience, even if
earnings remain low compared with those of natives. We can
make no such assumption about the second generation. In fact,
first-generation achievements may soften the secondgeneration desire for mobility, even as the new second
generation remains less well positioned than its native peers to
make the transition to adulthood (Mollenkopf et al. 2004).
What, then, do the data tell us about how the passage of the
children of immigrants through adolescence to young
adulthood compares with that of the children of native parents
in New York City? How do the characteristics of the parents, or
the choices they and their children make, or the experiences
they accumulate, shape such important outcomes as
educational attainment, entry into the labor market, and family
formation? Does the impact on children in immigrant families
differ from the impact on youngsters with native-born parents?
And how do racial differences affect the answers to these
questions?
Until now, researchers have had only limited data to explore
the trajectories of the second generation. Although the CPS in
1994 began to ask about a parent’s place of birth, this relatively
small random sample of the national population is designed to
gather labor market information on adults, not detailed
demographic and life-course information on specific immigrant
groups in specific locales. (The CPS sample included 2,564
individuals in New York City in 2004.) One can combine CPS
samples from different years, but this does not overcome limits
on the kinds of questions the CPS asks or on the structure of its
sample. The PUMS sample is not subject to this problem
because it is 100 times larger than the CPS sample, but it does
not identify parents’ nativity once a youngster moves out of the

FRBNY Economic Policy Review / December 2005

107

family of origin. The PUMS also reports only the answers to the
twenty-nine questions on the census long form.
To address these data shortcomings, the Russell Sage
Foundation initiated a research project that enabled the author
and his colleagues to gather data on representative samples of
young adults aged eighteen to thirty-two from five immigrant
group backgrounds (Dominican, Colombian/Ecuadoran/
Peruvian, Anglophone Afro-Caribbean, Chinese, and Russian)
and three native-born racial and ethnic groups (white, AfricanAmerican, and Puerto Rican) living in metropolitan New York.
The project is the Immigrant Second Generation in
Metropolitan New York (ISGMNY) study.4 This paper uses the
2000 PUMS data on youngsters under eighteen in New York
City to paint a broad, descriptive picture of the earlier years and
uses the ISGMNY data to examine the details for specific
groups as they enter adulthood.

2. The Parental Context
We have noted that immigrant parents tend to have less
English language ability, education, and income than nativeborn parents. When comparing the two groups, however, it is
useful to distinguish both their racial and ethnic backgrounds
and family forms so we can analyze similar groups. Table 2
shows the distribution of families by nativity, race, and form.5
Three patterns emerge. First, the different racial groups tend to

Table 2

Families by Type and Race of Family Head
Percentage of Households with Related Children
under Age Eighteen
Race of Family Head
Family Type

Hispanic

NH
Black

NH
Asian

NH
White

Total

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

4.4
8.6
2.1
8.9
8.4
32.4

3.8
12.6
1.1
5.1
6.0
28.6

0.2
0.1
0.4
7.9
1.4
9.9

11.9
3.6
2.8
5.6
1.3
25.2

20.7
25.4
6.6
29.3
17.9
100.0

Source: U.S. Census Bureau, 2000 5 Percent Public Use Microdata
Sample.
Notes: The sample is all New York City households in 2000 with related
children under age eighteen. NB is native born, FB is foreign born, 2PF
is two-parent family, 1PF is single-parent family, NH is non-Hispanic.
The family head may be either sex. Native American and NH other-race
households (3.8 percent of total) are not reported.

108

Trajectories for the Immigrant Second Generation

have strikingly different family forms. Overall, 57 percent of all
families with children under eighteen have a household head
and spouse; however, this is true of more than four-fifths of
white and Asian households, less than half of Hispanic
households, and only a third of black households. Second,
within these broad racial groups, the native families are more
likely to be single-parent families than are the immigrant
families. Finally, these broad racial categories have different
mixes of native and immigrant families. Black and Hispanic
families are roughly evenly split between native and immigrant
parents, but white households are predominantly native and
Asian households are predominantly immigrant. These
patterns have a number of implications.
How does controlling for a family’s race affect the previously
noted differences in English-language use, education, and
income between native and immigrant parents? It turns out
that the native-immigrant parental language gap is greatest
among whites and large among Asians, but far less wide among
blacks and Hispanics. This is because most black immigrants
come from English-speaking countries in the Caribbean, so
most speak English at home—just like the native born do.
Similarly, most Hispanic immigrant families speak Spanish at
home, but so do almost all native Hispanic families. To the
extent that differences in household language from the native
racial and ethnic comparison group impede the transition to
adulthood, the differences should have the greatest impact on
whites and Asians, less of an impact on blacks (although it is
still an issue for Haitians), and the smallest impact on Hispanic
immigrant families.
Controls for race and family form also attenuate the
educational gap between immigrant parents and their native
counterparts. Table 3 shows parental levels of education across
native and immigrant families, controlling for race and family
form. In general, all three factors—race, nativity, and family
form—seem to have a stronger relationship to educational
outcomes. In general, the rates of college education are much
greater for white (42 percent) and Asian (35 percent) family
heads than for black (14 percent) and Hispanic (8 percent)
family heads. (White two-parent families are also much more
likely to have a college-educated spouse.) Within each of these
racial groups, heads of two-parent families are always more
likely to have college educations than are heads of one-parent
families.
After controlling for race and family form, however, we note
that the pattern between native and immigrant family heads
and spouses is less clear. For whites and Asians, the native-born
parents are substantially more likely to be college educated
than are the immigrant parents in both one- and two-parent
families; this is also true, to a narrower extent, for Hispanic
families. Blacks, however, constitute an exception: the

immigrant parents are more likely to be college educated than
are the native parents. Note that although the racial groups
differ greatly in terms of parental levels of education, and
whites and Asians have higher levels than blacks and Hispanics,
blacks are not the group with the lowest levels.

These controls also shed light on the overall patterns of
employment, workers in the family, and household income
(Table 4). Once more, racial differences are strong, with white
and Asian parents having substantially higher rates of
employment than black and especially Hispanic parents. As
might be deduced from the high levels of education among

Table 3

Education of Family Head and Spouse by Race
of Family Head and by Family Form and Nativity

Table 4

Employment of Family Head and Spouse and Median
Household Income by Race of Family Head and
Family Form

Percentage of Households with Related Children
under Age Eighteen

Race of
Family
Head
Hispanic

NH black

NH Asian

NH white

Total

Family
Form and
Nativity

Family
Head
Lacks
HighSchool
Diploma

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

Households with Related Children under Age Eighteen

Family
Head Has
B.A.

Spouse
Lacks
HighSchool
Diploma

Spouse
Has B.A.

31
43
29
50
50
44

10
06
13
09
07
08

32
—
32
51
—
43

11
—
15
08
—
10

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

22
28
14
26
28
26

16
10
26
20
14
14

22
—
15
28
—
24

15
—
24
16
—
17

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

07
18
14
29
31
28

53
22
55
35
30
35

07
—
18
34
—
33

48
—
46
28
—
29

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

07
14
13
18
19
12

47
32
44
37
36
42

07
—
09
19
—
11

45
—
41
34
—
41

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

15
31
18
33
38
29

33
12
31
24
14
21

15
—
18
35
—
26

32
—
29
20
—
25

Family
Mean
Household
Head
Spouse
Workers
Type and Employed Employed in Family
Nativity
(Percent) (Percent) (Percent)

Median
1999
Household
Income

Hispanic

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

65
42
68
61
48
54

50
—
53
41
—
45

2.62
1.82
2.68
2.78
2.16
2.33

$47,000
$16,100
$43,000
$36,900
$20,900
$28,400

NH black

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

65
50
75
75
66
61

62
—
66
64
—
63

2.72
1.90
2.81
2.88
2.25
2.29

$54,000
$21,100
$56,000
$55,000
$30,000
$33,000

NH Asian

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

74
58
81
77
67
75

63
—
59
49
—
50

2.60
2.20
2.84
2.73
2.37
2.68

$64,000
$30,500
$67,000
$40,750
$33,900
$40,900

NH white

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

85
67
82
76
61
79

60
—
57
48
—
56

2.71
2.05
2.66
2.61
2.07
2.55

$83,100
$42,000
$71,000
$49,000
$27,300
$64,300

Total

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

77
50
76
71
57
65

58
—
58
49
—
53

2.69
1.90
2.70
2.74
2.20
2.42

$66,600
$21,610
$57,220
$43,000
$26,000
$38,000

Race of
Family
Head

Source: U.S. Census Bureau, 2000 5 Percent Public Use Microdata
Sample.

Source: U.S. Census Bureau, 2000 5 Percent Public Use Microdata Sample.

Notes: The sample is all New York City households in 2000 with related
children under age eighteen. NB is native born, FB is foreign born, 2PF
is two-parent family, 1PF is single-parent family, NH is non-Hispanic.
The family head may be either sex. Native American and NH other-race
households (3.8 percent of total) are not reported.

Notes: The sample is all New York City households in 2000 with related
children under age eighteen. NB is native born, FB is foreign born, 2PF
is two-parent family, 1PF is single-parent family, NH is non-Hispanic.
The family head may be either sex. Native American and NH other-race
households (3.8 percent of total) are not reported.

FRBNY Economic Policy Review / December 2005

109

income as a measure of social achievement. Immigrant
household incomes compare well with those of their native
counterparts, given the disadvantages they face. Note also that
white and Asian immigrant household incomes lag those of
their native counterparts, partly because native whites are the
best-positioned group and native Asians are relatively few. The
incomes of Hispanic immigrants lag those of their native
counterparts the least, partly because both groups are having
the hardest time. Remarkably, black immigrant household
incomes are doing the best compared with incomes of their
native counterparts, despite the fact that this group is
theoretically most at risk of downward assimilation.
The ISGMNY gives more detail on the family backgrounds
of immigrant second-generation and native young adults aged
eighteen to thirty-two who grew up in New York City and still
live there. Some of the major dimensions are given in Table 5.
As hinted at in the PUMS data, the type of family situation in
which young people grow up and enter adulthood is an
important factor differentiating blacks and Hispanics from
whites and Asians, and to a lesser degree native parents from
immigrant parents. Table 5 shows how fragile family life has
been for many young New Yorkers, especially members of
native minority groups. More than half of African-Americans
and large minorities of West Indians, Puerto Ricans, and
Dominicans grew up without ever knowing a parent, usually
the father. Even a third of the native white children grew up
without one biological parent. Of those who did grow up with
two parents, in many cases those parents had split up by the
time the child reached young adulthood, so that significantly

white parents, their income levels are even higher than their
employment rates compared with other groups. Family form
also has a strong effect on employment rates and income, with
two-parent families by definition being much more likely to
have an employed spouse, more workers in the family, and
higher incomes than single-parent families.
Finally, nativity counts too, but not in a consistent way.
Among Hispanics and whites, immigrant parents are
somewhat less likely to work than their native-born
counterparts; among blacks and Asians, however, they are
more likely to be working. Immigrant single parents are also
more likely to work than their native-born counterparts in
every group but whites. (This is probably related to the fact that
the black and Hispanic native-born single parents are
substantially more likely to have had public assistance income.)
Finally, the fact that immigrant families consistently have a
higher mean number of workers than their native-born
counterparts is also significant. This combined work effort
helps to bring the median household incomes of the immigrant
families closer to, and in some cases actually above, those of
their native counterparts, despite their parental gaps in
education and English-language proficiency. In particular, it is
noteworthy that the median household income of the
immigrant black, Hispanic, and Asian single-parent families
exceeds that of their native counterparts, given the relative
prevalence of this family form among blacks and Hispanics.
Beyond the ways in which two-parent families have obvious
material advantages over single-parent families, work conveys
moral authority in our society, and the mainstream often takes

Table 5

Family Background: Children of Immigrants and Native Born
Percent

Group
CEP
DR
PR
WI
NB
CHI
RJ
NW

Grew Up with
Both Parents

Parents Still
Together

More Than
Two Parental
Figures

68.1
58.9
55.0
52.4
43.0
88.9
82.0
68.5

51.2
40.1
34.9
32.0
21.0
79.8
73.0
47.5

16.6
14.2
12.2
20.4
9.8
25.7
28.8
11.7

Mean Number
of Siblings
Growing Up
1.98
2.35
2.16
2.23
2.69
1.55
1.00
1.65

Father Lacks
High-School
Diploma

Father Has B.A.
or Higher

Mother Lacks
High-School
Diploma

Mother Has
B.A. or Higher

26.8
44.4
41.0
14.7
22.2
38.1
5.4
11.2

18.1
15.5
10.0
24.4
17.1
19.3
58.7
35.8

33.2
48.8
37.9
10.3
16.1
42.9
4.6
11.7

12.3
7.1
11.8
25.6
19.8
14.8
68.2
39.2

Source: Immigrant Second Generation in Metropolitan New York study.
Notes: The sample is people aged eighteen to thirty-two who grew up and still live in New York City. CEP is parents are from Colombia, Ecuador, or Peru,
DR is parents are from the Dominican Republic, PR is parents are native Puerto Rican, WI is parents are from Anglophone West Indies, NB is parents are
native black, CHI is parents are Chinese born abroad, RJ is parents are Jews from former Soviet Union, NW is parents are native white.

110

Trajectories for the Immigrant Second Generation

fewer than half have an intact family of origin for many of the
groups we studied. Among our native black respondents, only
one in five has such a situation. (In every comparison, the
situation is more dire for the native groups.) Conversely, the
immigrant groups often had additional adult figures beyond
their parents in their household, such as a grandmother or
uncle. Meanwhile, the groups that had relatively few parent
figures to care for them also had larger mean numbers of
siblings, with the native black families being the largest. This
points toward what might be called differing “family strategies
of intergenerational mobility” across the groups being
analyzed—with some groups having significantly higher ratios
of adults caring for children and working to receive income
relative to the number of children to be cared for.
Finally, Table 5 makes it clear that most of the minority and
immigrant young people we interviewed have parents with
relatively low levels of education; even the native whites who
grew up in New York City did not come from particularly welleducated families. Only the Russian parents stand out as highly
educated. (If we include native whites who grew up outside
New York, educational attainment for white parents would be
substantially higher.) Within this overall pattern of relatively
low rates of parental education, several striking differences
emerge across the groups. The Dominican and Puerto Rican
parents are the least educated, followed by the Chinese, the
black groups and the South Americans are in the middle, and
the West Indian parents are the best educated, while the two
white groups have the highest levels of education. In each case,
the immigrant parents are somewhat better educated than their
native counterparts, with the Russian Jewish parents enjoying a
particular advantage over the parents of native white New
Yorkers. To the extent that parental education is a dominant
factor in explaining children’s educational attainment, and
therefore their lifetime earnings, we might expect the outcomes
for the children to follow the same general pattern (Sewell et al.
2001, pp. 20, 27).

3. Second-Generation Outcomes
The census PUMS data provide only very limited information
for assessing the educational outcomes of the new second
generation—whether school-age children are enrolled in
grades appropriate for their age and whether they have
completed those grades in a timely manner. (PUMS also tells us
whether enrollment is in a public or private institution.)
However limited this measure is, it is still an important
yardstick. Since PUMS provides the most complete coverage,
we begin with this source. To explore enrollment in an age-

appropriate grade, we calculate measures to determine whether
a child was enrolled in fifth grade or higher by age twelve or was
enrolled in ninth grade by age sixteen. (Since children typically
enter the first grade at age six, they have definitely fallen behind
if they are not enrolled in the fifth grade six years later or in the
ninth grade ten years later.) Table 6 presents the results for
young New Yorkers categorized by their family’s nativity and
form and the race of the head of the household.
Looking first at the 526,000 youngsters aged twelve to
seventeen, we note that about 2.5 percent overall have failed to

Table 6

Enrollment in Appropriate Grade and Private High
School, Related Children under Age Eighteen
by Household Type and Nativity and by Race
of Householder
Percent

Household
Type and
Nativity

Not
Enrolled in
Fifth Grade
by Age
Twelve

Not
Enrolled in
Ninth
Grade by
Age Sixteen

Enrolled in
Private
High
School

Hispanic

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

3.8
3.3
2.1
2.6
2.9
3.0

6.2
8.0
3.1
7.7
6.6
7.0

19.6
10.3
12.3
12.4
7.7
11.7

NH black

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

2.6
2.8
2.4
2.3
1.4
2.4

6.4
7.5
6.4
5.7
4.7
6.3

13.4
7.8
14.0
13.3
10.4
10.6

NH Asian

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

1.4
0.0
7.1
2.5
1.4
2.5

4.4
0.0
13.9
4.8
4.3
4.8

10.9
0.0
8.2
8.2
5.8
7.9

NH white

NB 2PF
NB 1PF
Mixed 2PF
FB 2PF
FB 1PF
Total

1.7
1.9
1.1
1.5
4.0
1.7

3.6
4.0
3.7
4.2
10.7
4.2

54.0
35.3
61.7
31.4
31.8
45.7

Race of
Family Head

Source: U.S. Census Bureau, 2000 5 Percent Public Use Microdata
Sample.
Notes: NB is native born, FB is foreign born, 2PF is two-parent family,
1PF is single-parent family, NH is non-Hispanic. The family head may be
either sex. Native American and NH other-race households (3.8 percent
of total) are not reported.

FRBNY Economic Policy Review / December 2005

111

enroll in the fifth grade. Table 6 suggests that this trend does
not vary greatly across racial groups, although whites are doing
best and Hispanics worst, with blacks and Asians in between
and blacks actually doing better than Asians. For Hispanics and
blacks, the children in immigrant households are doing better
than those in the comparable native-born households, but the
opposite is true in white and Asian families. Family form does
not seem to have a consistent or marked impact, which may be
good news. Table 6 shows similar patterns for the 170,000
youngsters aged sixteen or seventeen. Whites continue to be
the least likely not to have achieved the appropriate grade for
their age, while Hispanic children are the most likely to be
lagging. Blacks have now moved in front of Asians to be the
second most likely group to be lagging. Children in native-born
single-parent families are now more at risk than those in twoparent families across all racial groups, but unexpectedly,
children in immigrant single-parent families are less likely to be
behind than children in native-born single-parent families,
except for white immigrant single-parent families, which seem
to be having large and increasing difficulties over time
compared with the other racial groups. As before, the largest
consistent differences seem driven by race. Family form and
nativity count, but not as expected. Strikingly, the children in
Hispanic and black immigrant single-parent families are less
likely to be lagging their native counterparts, but children in
Hispanic and black immigrant two-parent families are more
likely to be lagging.
The racial differences in age-appropriate grade enrollment
are accentuated by the fact that white families are more than
four times as likely to send their children to private high
schools compared with the other racial groups. Hispanic and
black native two-parent families are also more likely than other
groups to send their youngsters to private high schools; singleparent families, with less means, are less likely to do so.
Ironically, the group that shows the highest levels of
educational attainment in relation to their parents’ low levels of
education—the children growing up in Asian immigrant
families—are the most likely to stick with the public high
schools. As the work of the ISGMNY has shown, the Asian
second generation is the most able to navigate the New York
City public school system to find the best schools, while the
black and Hispanic groups are the least able (Mollenkopf et al.
2001). Since the age limit of seventeen for the PUMS data
prevents us from computing high-school graduation rates, the
ISGMNY data, presented in Table 7 in a form comparable to
that of the prior PUMS data, confirm these patterns.
Table 7 shows the strong differences in outcomes according
to the race, family form growing up, and nativity of the families
of our respondents. The two native minority groups, AfricanAmericans and particularly Puerto Ricans, are most likely to

112

Trajectories for the Immigrant Second Generation

Table 7

Educational Attainment by Group and Family Form
Growing Up
Percent

TwoParent
Family

No HighSchool
Diploma

HighSchool
Diploma,
No B.A.

B.A./
Enrolled

Males,
No HighSchool
Diploma

CEP

Yes
No
Total

14.4
16.3
15.0

45.5
48.1
46.3

40.1
35.6
38.7

13.8
23.1
16.5

DR

Yes
No
Total

19.4
23.9
21.2

47.1
50.3
48.4

33.5
25.8
30.3

23.8
24.8
24.0

PR

Yes
No
Total

23.0
39.3
30.4

48.5
44.6
46.8

28.4
16.1
22.8

25.7
31.9
28.4

WI

Yes
No
Total

15.8
20.9
18.2

50.0
49.4
49.7

34.2
29.7
32.0

16.4
23.9
19.6

NB

Yes
No
Total

19.9
26.6
23.7

56.7
54.3
55.3

23.4
19.1
21.0

22.5
31.4
27.8

CHI

Yes
No
Total

8.1
15.5
8.9

22.3
25.9
22.7

69.6
58.6
68.4

8.3
20.0
10.0

RJ

Yes
No
Total

7.5
8.0
7.6

18.5
32.0
20.9

74.0
60.0
71.5

10.7
14.3
11.4

NW

Yes
No
Total

16.1
13.7
15.3

33.9
45.1
37.4

50.0
41.2
47.2

6.4
3.8
5.8

Group

Source: Immigrant Second Generation in Metropolitan New York study.
Notes: The sample is people aged eighteen to thirty-two who grew up
and still live in New York City. CEP is parents are from Colombia,
Ecuador, or Peru, DR is parents are from the Dominican Republic, PR
is parents are native Puerto Rican, WI is parents are from Anglophone
West Indies, NB is parents are native black, CHI is parents are Chinese
born abroad, RJ is parents are Jews from former Soviet Union, NW is
parents are native white.

lack a high-school diploma and least likely to have a B.A. (or to
be seeking one). Failure to obtain a high-school degree ranges
23 percentage points, from a low of 7.6 percent among Russian
Jews to a high of 30.4 percent among Puerto Ricans. (The
spread on college achievement is greater, 50 percentage points,
from 21 percent among African-Americans to 71.5 percent
among Russians.) The spread across family types is smaller, but
still marked, generally on the order of 5 to 7 percentage points,
depending on the group. As the last column of Table 7 suggests,

the men in each group are doing less well than the women in
both types of families. In particular, except for native whites,
males growing up in families headed by their mothers seem
particularly vulnerable—the rate at which they fail to get a
high-school diploma ranges from only 3.8 percent among
native whites to almost 33 percent among Puerto Ricans and
African-Americans. This result is worthy of a paper all its own;
suffice it to say that young men are more exposed to the
vicissitudes of the street and negative encounters with
authority while also being surrounded by a peer culture that
values toughness and boldness, while young women receive
more encouragement for academic achievement and are more
sheltered from the street by their families. (These patterns hold
even when looking at all respondents who grew up in the
metropolitan area, so they are not simply the product of the
out-migration of the more successful members of less
successful groups.)
Much about these outcomes jibes with the standard status
attainment model. Young adults from groups characterized by
two-parent families, better educated parents, parents with jobs,
and fewer siblings did the best. Those who grew up in the
opposite contexts generally had the hardest time getting an
education. Still, multivariate analysis that regresses educational
outcomes on family and parental characteristics shows that
significant group differences remain even after applying these
family controls (for elaboration on this point, see Kasinitz et al.
[forthcoming]). As one can sense from Table 7, the Chinese are
doing extraordinarily well given their modest family origins—
indeed, they are far outperforming what family backgrounds
alone would predict—while Puerto Ricans and AfricanAmericans are achieving significantly less education than
family background alone would predict. That the secondgeneration youngsters are getting consistently although not
hugely more education than their native counterparts even
after controlling for family background says as much about
how bad things are for native minorities as it does for how well
the children of immigrants are doing. Nevertheless, it is
noteworthy that after family background is controlled for, the
educational attainment of second-generation South
Americans, Dominicans, and West Indians is not statistically
significantly different from that of New York–bred native
whites. (Of course, because these second-generation groups
have different family backgrounds than do whites, they are not
getting as much education as whites in absolute terms.)
One important fork in the road faced by young New Yorkers
is where to go to high school. While the literature on
educational attainment has found that school characteristics
do not have much effect on educational attainment net of
family background, that seems not to be the case in New York
City. Some high schools had high graduation and college

attendance rates, while our respondents told us that others
lacked discipline or had teachers who they felt disrespected
their students. These characteristics were clearly associated
with post-secondary enrollment net of family characteristics
(Mollenkopf et al. 2001). Faced with bad public schools, many
families sought private alternatives for their children, mostly
parochial schools (or Jewish yeshivas in the case of Russian
youngsters).
Table 7 shows that native whites were most likely to exit the
public school system, followed by Russians and South
Americans. The pattern across family types shows that, except
for Chinese and Russians, where there were no differences, the
two-parent families were consistently more likely to send their
children to private high schools, largely because their incomes
were higher and more could afford to do so. Interestingly, two
groups with quite different educational attainment profiles,
native blacks and Chinese, were the most likely to attend public
high schools, followed by native blacks, West Indians,
Dominicans, and Puerto Ricans. One reason why the Chinese,
unlike the other second-generation groups, were highly likely
to stay in the public schools is that they tended to live in less
segregated neighborhoods near whites that had better primary
schools that fed into better high schools. Whites, Russians, and
Chinese were least likely to go to public high schools in the
bottom quintile of school performance rankings. Indeed,
almost one-fifth of Chinese went to one of New York City’s
famed selective high schools, such as Brooklyn Tech or
Townsend Harris in Queens, as did one out of ten Russians.
Meanwhile, a third of those from the poorer Hispanic
groups—Dominicans and Puerto Ricans—went to badly
performing public high schools, as did a quarter of native
blacks and a fifth of West Indians. These high schools drew
from the poorest neighborhoods of the city, had
overwhelmingly minority student bodies, and often had many
students from Spanish-speaking families. The table shows that
many two-parent families, even from these relatively low
income groups, sacrificed to take their children out of the
public system.
These different kinds of high schools tracked directly into
the disparate experiences with post-secondary education
already outlined above. Using the U.S. News and World Report
ranking system, with National I being the highest rating and
Regional IV the lowest rating, Table 8 shows the percentage of
those attending college whose institution falls into the lowest
category. While the pattern overall is similar to that for highschool quality, several departures stand out. West Indians, who
had been less likely than African-Americans to attend the
lowest performing high schools, were about as likely to attend
the lowest ranked colleges and universities. In addition, the
Russian second generation, which had almost entirely avoided

FRBNY Economic Policy Review / December 2005

113

Table 8

Type of High School and College Attended
and Educational Attainment by Group and Family
Form Growing Up
Percent

Attended
Regional
IV
College

Aged
TwentyFive and
Older
with B.A.

TwoParent
Family

Public
High
School

Lowest
HighSchool
Quintile

CEP

Yes
No
Total

80.5
89.2
83.2

12.1
10.0
11.4

10.0
25.0
15.6

24.5
21.9
23.8

DR

Yes
No
Total

84.6
92.3
87.7

29.7
36.4
32.4

25.0
46.7
35.5

26.1
20.8
24.1

PR

Yes
No
Total

82.0
92.2
86.6

33.3
39.8
36.5

29.4
40.0
35.1

14.1
11.3
12.9

WI

Yes
No
Total

85.6
94.7
89.9

14.4
24.6
19.4

38.9
44.4
41.7

27.6
13.6
21.5

NB

Yes
No
Total

92.8
94.1
93.5

22.3
25.5
24.2

44.4
45.5
45.1

14.3
9.0
11.3

CHI

Yes
No
Total

95.3
91.2
94.9

7.6
7.3
7.6

3.1
0.0
2.9

60.0
18.2
56.7

RJ

Yes
No
Total

82.6
82.0
82.5

0.0
0.0
0.0

42.9
33.3
40.0

45.3
22.2
39.4

NW

Yes
No
Total

58.3
62.0
59.5

12.0
4.0
9.3

0.0
14.3
7.7

22.6
19.0
21.7

Group

Source: Immigrant Second Generation in Metropolitan New York study.
Notes: The sample is people aged eighteen to thirty-two who grew up
and still live in New York City. CEP is parents are from Colombia,
Ecuador, or Peru, DR is parents are from the Dominican Republic, PR
is parents are native Puerto Rican, WI is parents are from Anglophone
West Indies, NB is parents are native black, CHI is parents are Chinese
born abroad, RJ is parents are Jews from former Soviet Union, NW is
parents are native white.

the low-performing public high schools, also often found itself
in the lowest ranked post-secondary institutions. Meanwhile,
the Chinese almost entirely escaped them and were among the
most prevalent of any group in higher ranked institutions. The
last column of Table 8 looks only at those young people who
grew up and still live in New York who are aged twenty-five to
thirty-two and who have had more time to complete a college
degree. Two second-generation groups, Chinese and Russians,

114

Trajectories for the Immigrant Second Generation

substantially outperformed all the others in attaining a B.A.
and in performance, followed by Dominicans, native whites,
West Indians, and South Americans—all bunched around one
in five. Puerto Ricans and native blacks achieved only half that
rate. For every group, children growing up in two-parent
families were more likely to have gotten their degrees.
Outcomes other than education are also of considerable
interest, particularly labor force status and the balance between
working and parenting. These are summarized in Table 9. The
majority of every group of our respondents found a job by age
twenty-three, in most cases the great majority. South
Americans, Chinese, Russians, and West Indians all had
employment rates that exceeded that of whites. Once again,
however, the two native-born minority groups, AfricanAmericans and Puerto Ricans, were least likely to be working.
Reciprocally, a third of African-American and a quarter of
Puerto Rican young adults were neither at work nor attending
school. (Subtracting the first two columns of data in Table 9
from 100 yields the percentage of those in each group who are
attending school but not working.) Growing up in a one- or
two-parent family did not seem to have a great direct effect on
participation in the labor force, although those from twoparent families were consistently somewhat more likely to have
a job. Only among Chinese, Russians, and whites, where
growing up in a single-parent family was comparatively rare,
did this seem to have a big effect on people neither having a job
nor going to school at age twenty-three or older. Having an
arrest record probably was related to labor market status: the
males among our respondents were twice as likely as the
females to have been arrested. Table 9 shows that a good many
males in every group except Chinese and Russians were likely
to have gotten into trouble with the police, rising to one-third
among African-Americans. Except for Dominicans, males
growing up in single-parent families were more likely, and in
some cases substantially more likely, to have been arrested.
Needless to say, this can have a deleterious effect on one’s job
prospects, although the damage is likely greater for minority
young people than for whites (Pager 2003).
Similarly, most of our respondents remain unmarried and
are not cohabiting with a partner. Only among Dominicans are
a majority married or cohabiting. Chinese are far and away the
least likely to be forming relationships, just as they are among
the more likely to be working or going to school. Interestingly,
those who grew up in two-parent families are consistently less
likely to have formed a serious relationship, while those who
grew up in single-parent families are more likely to have exited
their parent’s household and formed a new relationship of their
own. More troubling are the continuing patterns of forming
single-parent households among African-Americans and
Puerto Ricans and to a lesser extent West Indians and

Table 9

Labor Force Participation, Male Arrest, and Family Formation by Group and Family Form Growing Up

Two-Parent
Family

Aged
Twenty-Three
and Older,
Working
(Percent)

Aged
Twenty-Three
and Older, Not
Working and
Not in School
(Percent)

Males Aged
Eighteen to
Thirty-Two,
Ever Arrested
(Percent)

Aged
Twenty-Three
and Older,
Not Married
or Cohabiting
(Percent)

Females Aged
Eighteen to
Thirty-Two with
Children but
No Partner
(Percent)

Mean 1999
Household
Income

CEP

Yes
No
Total

79.6
85.2
81.2

13.9
13.0
13.6

15.5
28.0
19.3

58.4
51.9
56.5

7.8
3.8
5.8

$46,200
$28,600
$40,400

DR

Yes
No
Total

78.6
71.0
75.9

15.9
26.0
19.5

21.3
19.7
20.6

43.7
31.9
39.5

9.1
16.0
12.2

$34,900
$21,400
$29,400

PR

Yes
No
Total

70.9
69.0
70.1

26.0
28.7
27.1

24.1
27.7
25.7

56.7
40.2
50.0

17.7
25.5
21.2

$33,300
$24,600
$29,400

WI

Yes
No
Total

81.0
78.3
79.8

13.0
12.0
12.6

21.4
30.4
25.1

60.0
54.2
57.4

18.5
19.0
18.8

$50,900
$30,700
$41,600

NB

Yes
No
Total

63.1
63.8
63.5

31.7
33.9
33.4

25.9
41.9
35.2

62.1
52.8
57.0

45.3
36.3
40.2

$27,700
$24,800
$26,100

CHI

Yes
No
Total

81.1
70.0
80.1

12.0
20.0
12.6

6.7
25.0
9.0

80.8
61.9
79.0

0.9
0.0
0.8

$43,300
$29,200
$41,700

RJ

Yes
No
Total

84.3
75.0
82.0

7.2
21.4
10.8

10.5
15.4
11.4

53.0
40.7
50.0

1.8
0.0
1.4

$50,100
$54,500
$57,900

NW

Yes
No
Total

81.1
68.0
77.8

14.9
20.0
16.2

14.0
40.0
20.7

53.4
58.3
54.6

6.0
8.3
6.7

$42,300
$29,000
$37,700

Group

Source: Immigrant Second Generation in Metropolitan New York study.
Notes: The sample is people aged eighteen to thirty-two who grew up and still live in New York City. CEP is parents are from Colombia, Ecuador, or Peru,
DR is parents are from the Dominican Republic, PR is parents are native Puerto Rican, WI is parents are from Anglophone West Indies, NB is parents are
native black, CHI is parents are Chinese born abroad, RJ is parents are Jews from former Soviet Union, NW is parents are native white.

Dominicans, many of whom grew up in such households.
Table 9 shows that about twice as many African-American
women—two out of five—have had children but are neither
cohabiting nor married. This is also true for about one out of
five Puerto Rican and West Indian women.
Given the high level of risk among the native minority
groups—African-Americans and Puerto Ricans, followed at
some distance by West Indians and Dominicans—it is perhaps
not surprising that these groups have lower rates of labor force
participation and educational attainment and the lowest mean

household incomes. Across the board, those who grew up (and
often still live in) singe-parent families have lower mean
household incomes. By contrast, Chinese and Russians are
more likely to grow up in two-parent families and attend better
schools; the men are less likely to face arrest and the women are
much less likely to have had children on their own. (Chinese, in
particular, are also highly unlikely even to get married in their
twenties.) They have the highest mean family incomes, indeed
higher than that of native whites who grew up and still live in
New York City.

FRBNY Economic Policy Review / December 2005

115

4. Conclusion: How Race, Nativity,
Family Form, and Gender Affect
Young People in New York City
Massey (2005) correctly observes that the “segmented
assimilation” model should not be portrayed as holding that
race by itself will trump ethnicity, family background, gender,
and other factors in determining the trajectories of the second
generation. Indeed, its central insight is just the reverse—
that under the right circumstances, ethnicity and family
background can allay the impact of racial discrimination. At
the same time, the work of Portes and Rumbaut consistently
presents African-Americans as the archetypical group for
whom family and ethnic resources have failed to save them
from being pushed to the bottom. The data presented here do
not support that argument in several respects. First, Table 4
points out that neither African-American nor Afro-Caribbean
households have the lowest mean household incomes in New
York City—instead, native Hispanic households, largely
Puerto Rican, occupy that position—and they do not generally
classify themselves as black. (Most native Hispanic heads of
households with children in New York City chose “other race”
or “white” in the 2000 census; only about 10.8 percent gave
their race as “black.”) Similarly, members of Dominican
immigrant households also suffer more on many measures
than do African-American households, and they too generally
do not say they are black (12.8 percent gave “black” as one of
their races). Clearly, the fact that African-Americans and West
Indians speak English at home, while Puerto Ricans and
Dominicans generally speak Spanish at home, gives them one
advantage over Hispanics. In any case, these data suggest that,
however strong the force of racial discrimination may be in
New York, black families appear more capable of negotiating it
than Hispanic families.
Portes and Rumbaut’s formulation emphasizes that the
selectivity of immigration, the human and social capital of
immigrant families and communities, and the varying context
of reception will affect group trajectories (2001a, pp. 44-69).
Yet they note that the first barrier facing the children of
nonwhite immigrants is “the persistent practice of
discrimination based on [physical differences], especially
against black persons” (pp. 55-6). The authors posit that this
interacts with two other closely related factors—the hourglass
central-city economy wrought by deindustrialization and
suburbanization, and the “emergence of an adversarial outlook
and deviant lifestyles in American inner cities”—to keep
“second- and third-generation offspring of ‘colored’ minorities
bottled up in the inner city while simultaneously preventing
them from taking advantage of emerging opportunities in the
new postindustrial economy” (pp. 58-9). The result, in their

116

Trajectories for the Immigrant Second Generation

view, is the “‘hyperghetto’—veritable human warehouses
where the disappearance of work and the everyday reality of
marginalization led directly to a web of social pathologies”
(pp. 59-60).
From this description, it is hard to avoid the conclusion that
because African-Americans often live in poor neighborhoods
plagued by joblessness, broken families, and adversarial
attitudes and behaviors, they constitute a negative example that
the children of immigrants should avoid, if at all possible. In
Portes and Rumbaut’s analysis, the dark-skinned, relatively
poor immigrant groups at risk of being located in such places
should use any family resources and strategies they can to
escape. Otherwise, they will be prone to downward
assimilation. It is no exaggeration to say that the segmented
assimilation model portrays native blacks as having the worst
outcomes.
It is therefore theoretically interesting that the data clearly
show that African-Americans in New York are not at the
bottom and that black immigrants, largely from the
Anglophone Caribbean, are doing even better than native
blacks. If the causal mechanisms underlying the segmented
assimilation model are at work, then these groups must have
more family and community resources to resist and overcome
racial discrimination than that model suggests. This should
prompt us to rethink whether black communities do indeed
constitute such a negative model. In the ISGMNY, West
Indians are getting more education than African-Americans,
even after taking their somewhat higher parental levels of
education and employment into account. So being
phenotypically black and living near African-Americans may
not be as much of a barrier as the segmented assimilation
model seems to posit. Indeed, the substantial levels of
education and income achieved by many African-Americans in
New York may provide a positive model, not a negative one.
The data presented here should also lead us to reflect on why
Hispanic groups, not black groups, seem the most adversely
affected by the mechanisms of racial and economic inequality
in New York City. As Massey (2005) notes, Hispanic groups
occupy an ambiguous position in America’s black-white
hierarchy and come from societies that have different ways of
categorizing African ancestry (Itzigsohn 2004; Itzigsohn,
Giorguli, and Vazquez 2005). Dominicans, Puerto Ricans, and
other Hispanic groups in New York City are clearly not
comfortable placing themselves along a black-white axis and
choose “other” on the census race question. It is also clear that
the Dominican Republic and other sending societies have
complicated racial classification systems of their own that differ
from that of the United States. Race cannot be dismissed as a
factor, but it needs to be understood in light of how African
ancestry may interact with growing up in a Spanish-speaking

environment to produce even more challenges than simply
being black or simply speaking Spanish. The fact that the
census data and the ISGMNY show that Puerto Ricans and
Dominicans are experiencing the most difficulties should
prompt more analysis of this question.
Second, we need to dissect more minutely why young adult
children growing up in South American, Dominican, or West
Indian immigrant families are going to somewhat better
schools, achieving somewhat more education, and doing better
at avoiding arrest and single parenthood than those growing up
in very similar native Puerto Rican and African-American
families. For example, West Indians growing up in singleparent families are half as likely as African-Americans to have
earned a B.A. at age twenty-five or older, while those growing
up in two-parent families are twice as likely (Table 8). The
children of Chinese immigrants, though nonwhite, have
managed to make extraordinary educational progress despite
their parents’ low level of education. The segmented
assimilation model suggests that these patterns reflect the
immigrant parents’ ability to avoid the poorest, most
segregated native minority neighborhoods characterized by
street crime and poor schools. But there may also be other
factors at work, and we need to specify what they are.
Third, one way forward suggested by this analysis is to focus
on what we might call multigenerational strategies for
accumulating capital and transferring it across generations.
The most successful children come from groups in which
families often have two parents—as well as other adults—
earning wages and caring for relatively few children. The
Chinese excel with respect to the ratio of working adults to
children. While it is true that Chinese parents relentlessly
expect their children to perform well in school, they also
provide them with higher household incomes, live in
neighborhoods with better schools, keep them out of the labor
force while they study, and find the bureaucratic pathways to
the best schools in the New York City public school system.
Children growing up in African-American and Puerto Rican
families also have parents with relatively low levels of
education, but they often live in single-income families that

cannot afford to move out of the poorest neighborhoods with
the worst performing schools and the highest exposure to
crime and arrest.
Finally, the Russian and Chinese second generation has
outdistanced the native white young people who grew up and
remain in New York City, especially when parental education
and income are taken into account. Russian parents had very
high levels of education, but few were able to translate their
credentials into professional careers, and many spent time on
public assistance. Though some Chinese parents, such as those
from Taiwan or Hong Kong, had professional degrees, the
great majority had low levels of education and little ability to
speak English. The fact that they have done so well should
remind us that our native white New Yorkers—often from
Irish, Italian, or even Jewish working- and lower-middle-class
backgrounds—faced a good number of obstacles growing up as
well. Our image of successful young white New Yorkers is
shaped by how many of them—a third or more—grew up and
were educated elsewhere and came to New York as young
adults to make a professional career.
Despite the success of many members of native minority
groups, the data here present a distressing picture of outcomes
for many Puerto Ricans and African-Americans in New York
City. The high levels of poverty and single-parent families
among the adults show signs of being reproduced in the next
generation. (Given how many African-Americans grew up in
single-parent families in segregated settings, their
accomplishments are all the more remarkable.) Even when
native white New Yorkers grow up in single-parent families or
attend poorly performing schools, they have significant
advantages over their African-American and Puerto Rican
peers. They are far less likely to have neighbors in the same
position and far more likely to own their homes or have
relatives who can tie them into job opportunities. Because it
encapsulates a complex dynamic of scarce family resources,
high obstacles to success, and a risky environment, race still
counts very much in New York City. Just because some
children of immigrant minority parents can avoid its worst
effects, that does not lessen the sting on those who cannot.

FRBNY Economic Policy Review / December 2005

117

Endnotes

1. This analysis covers own and related children in families composed
of householders and their spouses, if any. However, about 8.4 percent
of the residents of New York City live in subfamilies, that is, the own
or related children of the household head or spouse have children of
their own. We do not analyze the experience of these children—the
grandchildren of the householder—who make up about 2.6 percent
of New York City’s residents. They would also qualify as members of
the second generation if their parents—the children of the
householder—were foreign born.
2. The data are from the 2000 census 5 Percent PUMS for New York
City and include the individual records on household head; spouse, if
any; and children in households with one or more own or related
children.
3. Although my daughter’s experience with the New York City public
schools highlighted the importance of having parents capable of
engaging the bureaucracy for me, Philip Kasinitz has emphasized the
degree to which noncitizenship poses a problem for the children of

118

Trajectories for the Immigrant Second Generation

noncitizens. Only half of all immigrant parents become citizens, and
they are less likely to vote than are native-born parents.
4. Support for the project was provided by the Russell Sage
Foundation, the Andrew W. Mellon Foundation, the Rockefeller
Foundation, the Ford Foundation, the UJA-Federation, and the
National Institute for Child Health and Human Development. Survey
data on 4,000 individuals were collected in 1998 and 1999; follow-up
in-person, in-depth interviews were conducted with a subsample of
346 individuals in 2000, with 152 reinterviewed in 2002. The Russell
Sage Foundation funded a counterpart study that gathered data in
2004: Immigrant Integration in Metropolitan Los Angeles, directed by
Rubén Rumbaut and Frank Bean of the University of California,
Irvine; Min Zhou, of the University of California, Los Angeles; and a
number of their colleagues.
5. These racial categories consolidate as a distinct group Hispanics
from all races.

References

Alba, R., and V. Nee. 2003. Remaking the American Mainstream:
Assimilation and Contemporary Immigration. Cambridge,
Mass.: Harvard University Press.

Kasinitz, P., J. Mollenkopf, and M. C. Waters, eds. 2004. Becoming
New Yorkers: Ethnographies of the New Second
Generation. New York: Russell Sage Foundation.

Borjas, G. J. 1990. Friends or Strangers: The Impact of
Immigrants on the U.S. Economy. New York: Basic Books.

Kasinitz, P., J. Mollenkopf, M. C. Waters, and J. Holdaway.
Forthcoming. Inheriting the City: Immigrant Origins and
American Dreams. New York: Russell Sage Foundation.

———. 1999. Heaven’s Door: Immigration Policy and the
American Economy. Princeton, N.J.: Princeton University Press.
Card, D. 2005. “Is the New Immigration Really So Bad?” Remarks
delivered at the conference on Immigration in the U.S.: Economic
Effects on the Nation and Its Cities. Federal Reserve Bank of
Philadelphia, April 28-29 (available at <http://www.phil.frb.org/
econ/conf/immigration/card.pdf>).
Card, D., J. E. DiNardo, and E. Estes. 2000. “The More Things Change:
Immigrants and the Children of Immigrants in the 1940s, the
1970s, and the 1990s.” In G. Borjas, ed., Issues in the Economics
of Immigration. Chicago: University of Chicago Press for NBER.
European Commission, Directorate General for Research. 2003.
“Migration and Social Integration of Migrants.” January. Brussels.
Feliciano, C. 2005. “Educational Selectivity in U.S. Immigration:
How Do Immigrants Compare to Those Left Behind?”
Demography 42, no. 1 (February): 131-52.
Gans, H. 1992. “Second Generation Decline: Scenarios for the
Economic and Ethnic Futures of the Post-1965 American
Immigrants.” Ethnic and Racial Studies 15, no. 2: 173-93.
Gerstle, G., and J. Mollenkopf, eds. 2001. E Pluribus Unum?
Contemporary and Historical Perspectives on Immigrant
Political Incorporation. New York: Russell Sage Foundation.
Itzigsohn, J. 2004. “The Formation of Latino and Latina Panethnic
Identities.” In N. Foner and G. M. Frederickson, eds., Not Just
Black and White, 197-218. New York: Russell Sage Foundation.
Itzigsohn, J., S. Giorguli, and O. Vazquez. 2005. “Immigrant
Incorporation and Racial Identity: Racial Self-Identification
among Dominican Immigrants.” Ethnic and Racial Studies 28,
no. 1 (January): 50-78.

Lobo, A. P., and J. Salvo. 2004. “The Newest New Yorkers 2000.”
October. New York City Department of City Planning.
Massey, D. S. 2005. Commentary on “Trajectories
for the Immigrant Second Generation in New York City,”
by J. Mollenkopf. Federal Reserve Bank of New York
Economic Policy Review 12, no. 2 (December): 121-3.
Mollenkopf, J., M. C. Waters, J. Holdaway, and P. Kasinitz. 2004. “The
Ever-Winding Path: Ethnic and Racial Diversity in the Transition
to Adulthood.” In R. A. Settersten, Jr., F. F. Furstenberg, Jr., and
R. G. Rumbaut, eds., On the Frontier of Adulthood: Theory,
Research, and Public Policy. Chicago: University
of Chicago Press.
Mollenkopf, J., A. Zeltzer-Zubida, J. Holdaway, P. Kasinitz, and
M. C. Waters. 2001. “Chutes and Ladders: Educational Attainment
among Young Second Generation and Native New Yorkers.” Paper
prepared for the project on New Immigrants in New York City.
International Center for Migration, Ethnicity, and Citizenship of
the New School University.
Pager, D. 2003. “The Mark of a Criminal Record.” American Journal
of Sociology 108, no. 5: 937-75.
Portes, A. 1995. “Children of Immigrants: Segmented Assimilation
and Its Determinants.” In A. Portes, ed., The Economic
Sociology of Immigration: Essays on Networks, Ethnicity,
and Entrepreneurship, 248-79. New York: Russell Sage
Foundation.
Portes, A., and R. G. Rumbaut. 2001a. Legacies: The Story of the
Immigrant Second Generation. Berkeley: University of
California Press and New York: Russell Sage Foundation.

FRBNY Economic Policy Review / December 2005

119

References (Continued)

———, eds. 2001b. Ethnicities: Children of Immigrants in
America. Berkeley: University of California Press and New York:
Russell Sage Foundation.
Portes, A., and M. Zhou. 1993. “The New Second Generation:
Segmented Assimilation and Its Variants.” The Annals of the
American Academy of Political and Social Science 530:
74-97.
Rumbaut, R. G. 2003. “Conceptual Issues, Methodological Problems,
and New Empirical Findings in the Comparative Study of the
‘Immigrant Second Generation’ in the United States.” Paper
presented at the conference on The Second Generation in North
America and Europe. Rockefeller Conference Center, Bellagio,
Italy, June 18-23.

Waldinger, R. D., and C. Feliciano. 2004. “Will the New Second
Generation Experience ‘Downward Assimilation’? Segmented
Assimilation Reassessed.” Ethnic and Racial Studies 27, no. 3
(May): 376-402.
Waters, M. C. 2000. Black Identities: West Indian Dreams
and American Realities. Cambridge, Mass.: Harvard
University Press.
Zhou, M. 1997. “Segmented Assimilation: Issues, Controversies, and
Recent Research on the New Second Generation.” International
Migration Review 31, no. 4 (winter): 975-1008.

Sewell, W. H., R. M. Hauser, K. W. Springer, and T. S. Hauser. 2001.
“As We Age: The Wisconsin Life Study, 1957-2000.” University
of Wisconsin-Madison Center for Demography and Ecology
Working Paper no. 2001-09, November.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
120

Trajectories for the Immigrant Second Generation

Douglas S. Massey

Commentary

ohn Mollenkopf ’s paper uses data from two sources to
consider patterns of assimilation among second-generation
immigrants in New York City. From the 2000 Public Use
Microdata Sample for New York, he compiles data on
household type by generation, race, and ethnicity, and shows
how household income and school enrollment are conditioned
by variation in these variables. He then turns to the Immigrant
Second Generation in Metropolitan New York study, on which
he is a principal investigator, to extend his analysis beyond
what can be accomplished using census data alone.
By relating race and ethnicity to family structure, income,
and education, Mollenkopf seeks to challenge the hypothesis of
segmented assimilation formulated by Portes and Zhou (1993)
and elaborated by Portes and Rumbaut (2001). He finds that
African racial origin does not necessarily trump class, family
background, gender, and other factors in determining
socioeconomic outcomes, and on this basis concludes that
segmented assimilation is unsupported as a theoretical
explanation. From the data presented in the paper, however,
I do not believe that he is justified in reaching this conclusion,
for two major reasons.
First, by reducing the hypothesis of segmented assimilation
to the simple idea that race trumps other factors in determining
trajectories among the second generation, Mollenkopf
transforms what is very broad and subtle theory into a stylized
caricature of itself. In fact, the model of segmented assimilation
posits that immigrant adaptation and integration are

J

“structured” by specific elements of an immigrant group’s
auspices of departure and context of reception. Race and racial
discrimination are just one of several structuring factors
mentioned by Portes and his colleagues. The auspices of
departure revolve around the original motivation for
international migration. Whether people are leaving their
homeland to flee political persecution, escape a natural
disaster, maximize returns to human capital, or overcome
missing or failed markets will determine much about the
configuration of human, social, and cultural capital that
immigrants bring with them and the strategies they then
employ to advance their interests in American society. The
ability of different groups to advance their interests, whatever
they may be, is also conditioned by the context of reception,
which includes government policies that determine an
immigrant’s legal status (such as temporary worker, asylee,
refugee, undocumented immigrant, or permanent resident
alien), the point of insertion into the labor market (primary,
secondary, or enclave), residential location (size of community,
kind of neighborhood), and patterns and levels of racial and
ethnic discrimination (in various markets). All of these factors
must be considered when testing the concept of segmented
assimilation, not just race and racial discrimination.
My second reservation is that the analysis too quickly
dismisses race as a structuring factor in the experience of
second-generation immigrants. Mollenkopf notes that
households headed by neither native-born nor immigrant

Douglas S. Massey is a professor of sociology and public affairs
at Princeton University.
<dmassey@princeton.edu>

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

121

blacks have the lowest mean household incomes, and that
Hispanics—primarily Puerto Ricans, who are not generally
black—occupy that position. Moreover, he observes that
households headed by Dominicans also suffer as much or more
on many measures than those headed by African-Americans,
and they too generally say they are not black. I do not believe,
however, that these results by themselves justify the conclusion
that race is not a salient, perhaps even a predominant, factor in
determining the experience of second-generation immigrants
in New York City.
For one thing, the fact that immigrant blacks are better off
than Puerto Ricans and Dominicans does not negate the
hypothesis that immigrants are stratified along racial lines,
because the tabulations presented in the paper do not control
for the selectivity of the original migration or the structuring
elements in the context of reception. Whereas Puerto Rican
migration to the U.S. mainland was overwhelmingly working
and lower class, black Caribbean immigrants were generally
selected from the lower professional and middle classes.
Moreover, although Dominicans tend to have higher class
origins than do Puerto Ricans, they are nonetheless generally
less selected than black Caribbeans, and a larger share of
families in this population are undocumented. In order to
conclude that race is not a major factor influencing outcomes
such as income and school enrollment in the second
generation, we really need more sophisticated regression
models that control for the human, social, and cultural capital
possessed by different immigrant groups. Even then, there is
always the possibility that unobserved heterogeneity arising
from variation in the auspices of departure could bias estimates
of racial effects. Given the analysis conducted in Mollenkopf’s
paper, we are not really in a very good position to judge the
relative importance of race as a factor in the experience of
immigrants and their children in New York City.
I also question the wisdom of pointing to poor outcomes
among Puerto Ricans and Dominicans as evidence to challenge
the hypothesis of racial hegemony. This strategy is problematic

122

Commentary

because both populations contain large numbers of people who
are descended from forebears of African origin. Even though
relatively few respondents in either group may identify
themselves as “black,” that does not mean that native white
Americans would not put them in this racial category and treat
them accordingly, subjecting them to higher levels of
discrimination than other immigrants. The fact that most
Puerto Ricans and Dominicans identify themselves as “other
race” reflects the Caribbean conceptualization of race as a
continuum from white to black rather than the dichotomous
conceptualization that historically has prevailed in the United
States; it does not mean that they have no African ancestry. In
fact, when one compares socioeconomic outcomes among
Caribbean Hispanics who identify themselves as white, other,
or black, one generally finds that those in the “other” category
lie much closer in status to blacks than to whites, suggesting the
operation of distinctly racialized processes (see Massey and
Bitterman [1983] and Denton and Massey [1989]).
What Mollenkopf’s paper ultimately presents are some
interesting tabulations that document differentials in income
and education by generational status, race, ethnicity, and
family background in New York. However, these data are
insufficient by themselves to test the model of segmented
assimilation, which incorporates many other structuring
elements besides race into its explanatory model. Simple crossclassifications are also insufficient to judge the relative
importance of race itself as a stratifying agent without the
introduction of controls into much more complicated
statistical models. Segmented assimilation theory may or may
not ultimately hold up when subject to systematic scrutiny
using data from the Immigrant Second Generation in
Metropolitan New York study, but the tabulations presented
represent only the very first steps in a much longer journey to
examine that theory.

References

Denton, N. A., and D. S. Massey. 1989. “Racial Identity among
Caribbean Hispanics: The Effect of Double Minority Status on
Residential Segregation.” American Sociological Review 54,
no. 5 (October): 790-808.
Massey, D. S., and B. Bitterman. 1983. “Explaining the Paradox of
Puerto Rican Segregation.” Social Forces 64, no. 2 (December):
306-31.

Portes, A., and R. G. Rumbaut. 2001. Legacies: The Story of the
Immigrant Second Generation. Berkeley: University of
California Press and New York: Russell Sage Foundation.
Portes, A., and M. Zhou. 1993. “The New Second Generation:
Segmented Assimilation and Its Variants.” The Annals of the
American Academy of Political and Social Sciences 530:
74-97.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2005

123

Guillermina Jasso, Douglas S. Massey, Mark R. Rosenzweig, and James P. Smith

Immigration, Health, and
New York City: Early Results
Based on the U.S. New
Immigrant Cohort of 2003
1. Introduction

E

very year, several hundred thousand persons become legal
permanent residents (LPRs) of the United States,1
averaging 781,848 in the 1991-95 period, 771,307 in the 19962000 period, and 944,884 in the 2001-04 period.2 They include
new arrivals to the United States (some coming for the very first
time) as well as persons already living in the United States,
having come earlier on a temporary visa or without documents
and now achieving the coveted LPR status. Mingled with their
hopes and dreams are the personal characteristics that
propelled the move—the peculiar migrant energy—and the
myriad faculties, experiences, attributes, and skills that will
shape the immigrant trajectory.
Immigrants settle in one point within the vast U.S.
geography. Classically, there are four great reception areas:
the two coasts, Chicago, and the southern border. New York
City was the gateway for the great migrations of the turn of
the twentieth century, and it remains a major destination for
new immigrants.3 Repeatedly, the city has been shaped and
reshaped by the distinctive characteristics of successive waves

Guillermina Jasso is a professor of sociology at New York University;
Douglas S. Massey is a professor of sociology and public affairs at Princeton
University; Mark R. Rosenzweig is a professor of economics at Yale University;
James P. Smith is the RAND Corporation Chair in Labor Markets and
Demographic Studies.
<gj1@nyu.edu>
<dmassey@princeton.edu>
<mark.rosenzweig@yale.edu>
<smith@rand.org>

of new immigrants; new immigrants, in turn, like their
native-born counterparts who arrive from Seattle and Iowa
City and Laredo, have found in New York City both haven
and spur.
Among the things immigrants bring with them to the
United States is their health set: the combination of health
levels and health behaviors. This paper has the twofold
objective of exploring immigrant health and doing so with an
emphasis on New York City. We make use of a new data source,
the New Immigrant Survey (NIS)—the first longitudinal
survey of a nationally representative sample of new legal
immigrants to the United States—drawing information from
Round 1 of its fiscal year 2003 cohort, known as NIS-2003.
(At this writing, the data from Round 1 are being prepared for
initial public release in 2005, and plans are under way for
fielding Round 2.) An important additional objective of this
paper is to make known the availability of this new data source,
which will enable researchers to address a wide variety of
topics, from language acquisition and identity formation to
religion dynamics, not to mention the staples of studies of
immigration, such as selectivity, emigration, and naturalization.

This is a revised version of the paper presented. The New Immigrant Survey
project is supported by the National Institutes of Health (grant HD33843), the
National Science Foundation (grants SRS-9907421 and SES-0096867), the
U.S. Citizenship and Immigration Services, the Office of the Assistant
Secretary for Planning and Evaluation, and the Pew Charitable Trusts; the
authors gratefully acknowledge their intellectual and financial support. Many
colleagues have provided valuable insight, especially Jennifer A. Martin and
Adriana Lleras-Muney. The views expressed are those of the authors and do
not necessarily reflect the position of the Federal Reserve Bank of New York or
the Federal Reserve System.
FRBNY Economic Policy Review / December 2005

127

1.
2.

Two questions dominate the study of immigrant health:
What is the health status of a new immigrant?
What is the immigrant’s health trajectory over the life
course?

The first question, the selection question, encompasses all
factors and mechanisms in both origin and destination
countries that influence who migrates—including, for
example, origin-country skill prices and destination-country
visa allocation regimes—some of which are, directly or
indirectly, attentive to matters of health. The second question,
variously called the assimilation or incorporation question,
focuses on the health-relevant aspects of the receiving country
environment and the immigrant’s resources and behaviors in
the new country.
At first blush, the immigrant health problem considers
health at arrival and examines subsequent health. For example,
a popular story in recent years has been that of a healthy person
immigrating to the United States and subsequently acquiring
some of the bad eating habits associated with American fast
food, leading to health decline.
Migration is complicated, however, and we argue that a
more faithful approach would incorporate the health effects of
the migration process itself, which may begin long before
“arrival” and may differ for immigrants facing different
migration-relevant environments, such as different visa
regimes (Kasl and Berkman 1983; Vega and Amaro 1994;
Jasso 2003; Jasso et al. 2004). For example, navigating the visa
application process may be quite stressful, illegal immigrants
are constantly in fear of discovery and deportation, some legal
immigrants have “conditional” visas for two years after
admission to legal permanent residence, and immigrants may
face prejudice.
Prolonged exposure to stressful circumstances has been
shown to have powerful negative effects on a variety of bodily
systems (McEwan and Lasley 2002). One important set of
effects is cardiovascular. Chronically elevated levels of
adrenaline increase blood pressure associated with the human
stress response and raise the risk of hypertension. At the same
time, elevated fibrogen levels increase the likelihood of blood
clots and thrombosis while the build-up of “sticky” white blood
cells causes the formation of arterial plaques that contribute to
atherosclerosis. Excessive stress also causes the production of
excess glycogen and fat, raising the risk of obesity; and the
suppression of insulin during periods of stress leads to
excessive blood sugar and a greater risk of Type II diabetes
(McEwan and Lasley).
Chronic stress also compromises the human immune
system, suppressing the human immune response and
increasing susceptibility to illness and infection (McEwan and

128

Immigration, Health, and New York City

Lasley 2002). Under some circumstances, it may also overstimulate the immune system, causing it to attack targets
within the body that normally do not pose a threat, leading to
the expression of inflammatory diseases such as asthma and
autoimmune diseases such as multiple sclerosis, arthritis, and
Type I diabetes (McEwan and Lasley).
Attentiveness to the migration process suggests that if the
migration process is stressful, then the appropriate time for
assessing health selectivity is at the time of the migration
decision—rather than at the time of actual migration—and,
further, that assessment of health change subsequent to
immigration should take into account heterogeneity in the
sources of health change and their timing.
Accordingly, and building on the health and immigration
literatures, we formulate a model that distinguishes between
the permanent and transitory components of health and that
identifies three distinct sources of change in the transitory
component of immigrant health: 1) visa stress, defined as the set
of stresses related to the process of obtaining legal permanent
residence; 2) migration stress, defined as the set of stresses
related to the process of moving from one country to another,
net of the visa application process; and 3) U.S. exposure,
conceptualized as dietary and environmental factors.
Each of the three sources of health effects has a distinctive
temporal span and affects distinctive subpopulations. For
example, U.S. exposure affects everyone, not only immigrants;
migration stress affects all international movers, whether or not
they have to go through the visa process, including, to
illustrate, persons born in Puerto Rico or American Samoa and
persons who, though born in the United States, were raised
abroad by their foreign-born parents, possibly since infancy;
and visa stress affects only those who must obtain legal
permanent residence. With respect to the time dimension, visa
stress presumably ends with admission to LPR (or, as will be
seen, somewhat earlier for refugees and somewhat later for
conditional immigrants); migration stress probably ends at
some point after inception of U.S. residence; and U.S. exposure
effects do not end, although positive effects may be accentuated
and negative effects mitigated by discerning choices and
behaviors.
Accordingly, to assess health selectivity, it is important to
measure health before the onset of visa stress, migration stress,
and U.S. exposure, or to control for their operation in the
estimating equations. And assessing health changes requires
isolating the separate effects of the three sources of health
change.
Overall, the contributions of this paper include: 1) a sharp
distinction between health at the time of the migration decision
and health at admission—the former being the variable of
interest in exploring health selectivity; 2) a distinction between

three sources of health change among immigrants (and
concomitantly among others); 3) a description of key healthrelevant features of the U.S. immigration system and of NIS
data, which will enable substantial new work among
immigration and health researchers; 4) an NIS-based
description of recent legal immigrants both to the United States
in general and to New York City in particular; and 5) a
preliminary NIS-based estimation of health selectivity, health
change, visa depression, and body-mass index (BMI).

2. Immigration and Health
2.1. A Brief Overview of U.S. Legal
Immigration
An immigrant visa is a scarce commodity, as more persons
would like to immigrate to the United States than current or
foreseeable law permits.4 In the face of high demand for
immigrant visas, the United States allocates visas by means of
a system that includes family reunification and employment
criteria, as well as humanitarian and diversity considerations.
In brief, the system of visa allocation in the period since 1921
may be characterized by three features. First, the United States
restricts the number of immigrants (restricting since 1921 the
number from the Eastern Hemisphere, and since 1968 the
number from the Western Hemisphere as well). Second,
immediate relatives of adult U.S. citizens—defined as spouses,
minor children, and parents—are exempt from numerical
restriction.5 Third, numerically limited visas are allocated via
two sets of preference categories: one for family-sponsored
immigrants, the other for employment-based immigrants.
Over the years, the United States has altered both the definition
of immediate relatives of U.S. citizens (for example, in 1952 by
extending to U.S. citizen women the right, already held by men,
to sponsor the immigration of an alien spouse outside the
numerical limitations) and the system for granting numerically
limited visas (for instance, by establishing a structure of
preference categories in 1965 but not placing the Western
Hemisphere under that structure until 1977, and subsequently
revising the preference categories in the Immigration Act of
1990). Under current law, the number of visas available
annually in the family preference categories is at least 226,000,
but may be larger (though never larger than 480,000)
depending on the previous year’s volume of numerically
unrestricted immigration; in the employment-based categories,
the annual number of visas available is at least 140,000, but may
be larger if there are unused family preference visas.6

Additionally, U.S. immigration law provides legal
permanent resident visas on humanitarian and diversity
grounds. On humanitarian grounds, persons admitted to the
United States with refugee visas or given asylee status (both
refugee and asylee visas are nonimmigrant temporary visas)
may adjust to legal permanent residence after residing in the
United States for one year. There is no ceiling on refugee
adjustments to permanent residence, and the number has
ranged in recent years from a low of 39,495 in fiscal year 1999
to 118,528 in fiscal year 1996; in contrast, asylee adjustments
are constrained to 10,000 per year. On diversity grounds, the
United States grants 50,000 visas annually to nationals of
countries from which the number of numerically limited
immigrants is less than 50,000 in the preceding five years.
Eligibility requirements include a high-school degree or
equivalent, or two years’ work experience (within the
preceding five years) in an occupation requiring two years of
training or experience; selection is by lottery.7
Finally, U.S. immigration law provides for the legalization
of certain persons illegally in the United States, through the
registry provisions or via cancellation of removal.8 Of course,
illegal persons may also acquire LPR via all the other immigrant
visa categories.
Among family-based and employment-based immigrants,
a key actor in the migration process is the visa sponsor (also
known as the “petitioner”)—the individual (or firm, in the case
of some employment-based immigrants) who, as relative or
employer of the prospective immigrant, establishes the latter’s
eligibility for an immigrant visa.9 The visa sponsor initiates the
paperwork. For all family-sponsored immigrants and for a
subset of employment immigrants, the visa sponsor must also
become the main support sponsor, assuming responsibility
for the immigrant’s support, should the immigrant require
assistance, and signing an affidavit of support contract.10
Additionally, the prospective immigrant must pass a
medical examination to ensure that he or she is not
inadmissible on medical grounds. The medical grounds for
inadmissibility are grouped into four categories: 1)
communicable disease of public health significance (such as
tuberculosis or syphilis), 2) lack of required vaccinations (for
example, for polio and hepatitis B), 3) physical or mental
disorders with harmful behavior, and 4) drug abuse or
addiction. Thus, U.S. immigration law plays a part in shaping
the immigrant’s health status at admission to legal permanent
residence.
In most visa categories except those for immediate relatives
of U.S. citizens (spouse, parent, minor child), visas are awarded
not only to the individual qualifying for an immigrant visa
but also to his or her spouse and minor children who are
“accompanying, or following to join” the immigrant principal.

FRBNY Economic Policy Review / December 2005

129

2.2. Health Implications of the U.S. Visa
Allocation System

Health Selection
The U.S. visa allocation system has several implications for
immigrant health at the time of the initial migration decision.
A priori, the spouses of U.S. citizens—approximately a third
of adult immigrants—would be expected to be healthy; the
marital tastes of U.S. citizens, assortative mating mechanisms,
and the energies and attributes required for participation in the
international marriage market would militate to produce
healthy spouses. Employment-based immigrants would also
be expected to be in superior health, again in view of their
participation in international labor markets. Similarly, the
children of U.S. citizens would be expected to be healthy,
especially given their youth. On the other side of the ledger, less
healthy immigrants may include refugees (who may have
suffered many privations) and parents of U.S. citizens (who
may be of advanced age).

Health Trajectory—Visa Stress
The visa allocation system also has implications for the health
trajectory during the visa application process. While all visa
classes require assembling documents—such as birth
certificate, marriage certificate, police record, military
record—and filling out forms, they differ on the requirements
for a sponsor and for an affidavit of support.
Numerically limited and numerically unlimited visas differ
in the time required to obtain them. The overall waiting period
has two phases. The first phase, applicable only to numerically
limited visas, involves waiting for availability of a visa. Visa
waiting times vary by both class of admission and country of
origin; for example, in April 2005, there was no delay for some
employment-based visas, but the delay for family-based visas
ranged from four years in the first family category (unmarried
sons and daughters of U.S. citizens) for natives of all countries
except Mexico and the Philippines to more than twenty-two
years in the fourth family category (siblings of U.S. citizens) for
persons from the Philippines (see U.S. Department of State
[various years]).
The second phase of the waiting period consists of
application processing. Of course, for prospective migrants
who qualify for a numerically unlimited visa, this phase is
coterminous with the entire waiting period. The length of this

130

Immigration, Health, and New York City

phase varies with administrative factors, such as the number of
personnel assigned to immigrant visa processing and whether
changes in immigration law make necessary the design of new
forms and/or retraining of personnel.
As would be expected, qualifying for an immigrant visa is an
overriding concern for prospective immigrants to the United
States, and visa allocation law is a critical component of the
environment faced by prospective immigrants. Accordingly,
the time waiting for a visa may be a time of accumulating visa
stress.
In some situations, all or some of the waiting period is spent
in the United States. For example, persons with legal temporary
nonimmigrant visas—as foreign students, say, or H-1B
specialty workers—may be applying for legal permanent
residence under family or employment provisions of the law.
Some persons do not qualify for a legal permanent visa under
any provision of the law. They may enter the United States with
a legal temporary visa and then lapse into illegality. Or they
may enter the United States illegally (that is, “without
inspection”).
For most persons admitted to LPR, visa stress ends on the
day of admission. The date of admission to permanent
residence is a milestone in an immigrant’s life. The new
immigrant, who may be arriving from abroad at a U.S. port
of entry (a “new arrival”) or may be adjusting to permanent
residence from a legal temporary visa in the United States (an
“adjustment of status”), acquires a set of privileges, including
that of sponsoring the immigration of certain kin. The passport
is stamped to indicate admission to legal permanent residence,
the “green card”—the paper evidence of legal permanent
residence—is ordered, and the clock starts on the residency
requirement for naturalization.
For some categories of immigrants, visa stress may end
earlier or later than admission to LPR. The main category of
immigrants for whom visa stress may end prior to admission to
LPR is that of refugees, who gain permanent admission when
they are admitted with a (nonimmigrant temporary) refugee
visa. Arguably, for refugees, the stressful part of the application
process ends with arrival in the United States. Refugees may,
but need not, adjust to legal permanent residence; they are
eligible to do so after one year. Asylees also may, but need not,
adjust to legal permanent residence, and they are eligible to do
so after one year; however, in contrast to refugees, there is an
annual ceiling of 10,000 on their adjustment. We may surmise
that the ceiling generates stress, and thus for asylees visa stress
would definitely continue until admission to permanent
residence.
Meanwhile, for a subset of immigrants, visa stress does not
end on the date of admission to LPR. These are the conditional

immigrants—chiefly spouses of U.S. citizens and of LPRs, in
marriages of less than two years’ duration, and employmentbased investor immigrants—whose visas are conditional for
two years and who must apply for removal of the conditionality
restrictions.

take up residence in the United States; this situation, in which
U.S. permanent residence operates as insurance, has come to
light in the course of NIS fieldwork. The first two cases,
involving children, may be more useful for empirical
identification of the operation of visa stress and migration
stress, given that the situation is exogenous, the choices and
decisions made by the parents and not by the children.

2.3. The Distinction between Visa Stress
and Migration Stress
Individuals may be subject to visa stress and not migration
stress, or, conversely, to migration stress but not visa stress.
This distinction paves the way for future research in identifying
the separate effects of these two potential sources of health
change.11

Migration Stress without Visa Stress
Not all persons who move permanently to the United States
from a foreign country require a visa, and thus such persons
would be vulnerable to migration stress but not to visa stress.
Two important subpopulations may be considered; they may
be regarded as “natural” comparison groups in migration
research: 1) U.S. citizens who are natives of territories of the
United States, such as Puerto Rico, American Samoa, and the
Northern Marianas, and 2) U.S. citizens who were born in the
United States to foreign-born parents and raised abroad, such
as the young children of foreign students. These groups may
experience all the migration stress associated with an international move, but none of the visa stress. Future research
might undertake a sharp examination of the two distinct kinds
of stresses by studying one or more of these groups together
with new immigrants. Here we focus on new legal immigrants,
most of whom experience both visa stress and migration
stress.12

3. Theoretical and Empirical
Framework
3.1. Modeling Immigrant Health

Health Selection
Consider an adult residing in a foreign country and contemplating a permanent move to the United States. At the time of
the migration decision—roughly when the first steps are taken
to obtain legal permanent residence in the United States—he
or she has a certain level of healthiness. The distribution of
healthiness among all prospective immigrants to the United
States around the world at this stage of the immigrant career is
determined by selectivity forces, including U.S. immigration
criteria. Of course, the intensity of self-selection on healthiness
may vary; for example, refugees may be less self-selected on
health than are employment immigrants. The healthiness
distribution may be a composite distribution, consisting of
several distinct subdistributions corresponding to distinct
migration flows.
We conceptualize overall healthiness H as having two
p
components—a permanent component, denoted h , and a
t
transitory one, denoted h :
(1)

Visa Stress without Migration Stress
The opposite may also arise—persons who experience visa
stress but not migration stress. Three cases come to mind. The
first two pertain to children raised in the United States who
might either be born in the United States to diplomat parents
and thus not citizens at birth or foreign born and raised in the
United States by illegal parents. Such children are often fully
“American” in sensibility but must undergo the visa process.
The third case pertains to persons who acquire LPR but never

p

t

H = h +h .

Following the standard model, pioneered by Grossman (1972),
health is an important form of human capital, and includes
both a persistent time-invariant component and a timevarying component (Strauss and Thomas 1998).
We assume that immigrants make their initial migration
decision based on the permanent component of their
healthiness. If the transitory component of health does not
change between the initial migration decision and the actual
migration, then health selectivity can be inferred from
observed healthiness at migration. If, however, the transitory

FRBNY Economic Policy Review / December 2005

131

component changes, then observed healthiness at migration
would provide a biased estimate of the persistent component,
and hence of the selectivity forces. As sketched above and as we
will discuss, there is reason to believe that the transitory
component changes nonrandomly. Accordingly, understanding health selectivity in migration requires attentiveness
to the permanent component and thus, in empirical analysis,
attentiveness to observed healthiness at the time of the initial
migration decision, rather than at immigration.
The selectivity forces on health differ for different migration
streams. In general, the decision to migrate can be thought of
as a balance between the gains and costs of migrating—or, as
the Romans put it, ubi bene, ibi patria: Where one is well-off,
there is one’s country. To the extent that economic
considerations play a part—as they no doubt do for most
immigrants who will join the labor force—we can begin with a
model of migration in which the individual migrates if the
economic gains from migrating exceed the costs (as set forth in
Jasso et al. [2004]). Incorporating wages, skill prices, and skill
transferability, as well as costs of migrating, yields the
implication that the higher the skill prices in a country of origin
and the greater the country’s distance from the United States,
the higher the skill levels of its emigrants to the United States.
If skill levels are higher among healthier people, then the gains
from migrating will be greater for healthier individuals and
migrants will be positively self-selected on health. Thus, ceteris
paribus, the higher a country of origin’s skill prices and the
greater its geographic or cultural distance from the United
States, the greater the health selectivity of U.S. immigrants
from that country.
Labor market considerations may be less important or not
important at all for older immigrants and immigrants who do
not plan to work, as well as for refugees who are fleeing for their
lives. Accordingly, such immigrants may be less positively
selected on health. Of course, individuals who become refugees
in the United States are the survivors of extreme situations, and
thus may possess higher levels of health.
Moreover, migration to the United States may be fueled by
the freedoms and other aspects of the American social and
political climate, independent of economic considerations, and
it is not obvious how health selection would operate. For
example, a young person may want to live in a society where
parental permission to marry is not required or where a baby
may be given any name one chooses or where one can stop
going to church without fear. These “freedom gains” would
not necessarily be greater for healthier individuals. Thus,
immigrants primarily seeking freedom gains would not be
positively selected on health.

132

Immigration, Health, and New York City

Health Trajectory—Visa Stress
The initial migration decision is followed by the process of
applying for permanent residence. As discussed above, this
process can be highly stressful, and the transitory component
of health declines in response to visa stress. Similarly, living in
the United States illegally is highly stressful, and the transitory
component declines.13
The decline in the transitory component of health can be
characterized by its magnitude, by the length of time during
which the decline occurs, and by the shape of the decline (such
as its steepness). These aspects of the decline may vary by
migration stream. For example, visa stress may be greater for
immigrants requiring an affidavit of support (all family
immigrants and a subset of employment immigrants) than for
other immigrants, and therefore the magnitude of the decline
may be greater for these immigrants; visa stress may also be
greater for illegals.
Among applicants for legal immigrant visas, permanent
residence is eventually obtained. At that point, visa stress ends,
and we may conjecture that observed healthiness—more
precisely, the transitory component of health—begins an
upward trajectory. The incline, like the decline, may be
characterized by its magnitude, by the length of the recovery
period, and by its shape. And, as with the decline, aspects of the
recovery period may also vary by immigrant stream. Except for
normal aging, one might imagine that following the recovery
period, the immigrant returns to the original level of observed
healthiness, so that the magnitude of the decline would equal
the magnitude of the incline, unless, of course, the stresses have
been so severe or prolonged that the body’s physiology is
altered (Seeman et al. 1997; Smith 1999).14
This model raises several new empirical questions,
including: 1) whether the steepness of the decline and the
steepness of the recovery are related, 2) whether the duration
of the application process affects the duration of the recovery
period, and 3) whether, within the application and recovery
periods, steepness, total decline/recovery, and duration are
related.

Health Trajectory—Migration Stress
and U.S. Exposure
Additionally, as we discussed, there are two other effects that
must be incorporated into the model. The first is the migration
stress associated with adjusting to life in a new country. It
includes stress due to different language, different customs,

and so on. As with visa stress, migration stress may end, and its
health effect may be characterized by decline and recovery,
with attention similarly paid to magnitude, duration, and
steepness.
The second, U.S. exposure, involves the possibly deleterious
effect of the U.S. environment. It has been conjectured that the
combination of a possibly less healthy diet and environmental
agents may induce a deterioration of the immigrant’s health
(Frisbie, Cho, and Hummer 2001; Rumbaut and Weeks 1996).
Of course, an opposite conjecture is also plausible, given that:
1) health-relevant conditions are more favorable in the United
States than in many origin countries; 2) immigrants experience
large gains in earnings, on average, after immigration;15 and
3) immigrants, whose propensity to invest in themselves is
visible in their migration behavior, are likely to invest in their
health, taking advantage of their earnings gains and new
opportunities in the United States.16

Health Trajectory—Disentangling Visa Stress,
Migration Stress, and U.S. Exposure
It is illuminating to contrast these three sets of effects on
immigrant health, and we do so along two dimensions: first, by
noting their spatio-temporal character; second, by highlighting
comparison groups.
Visa stress is tightly linked to the visa process. It begins with
the first filing, proceeds differentially by visa class, and ends
with admission to LPR, or, for conditional immigrants, at
removal of the conditionality restrictions.17 Moreover, visa
applicants are subject to visa stress, regardless of where they are
located, whether in the origin country or in the United States.
In contrast, migration stress and U.S. exposure have
different life spans, independent of the visa process and both
beginning with inception of U.S. residence. Moreover, as
discussed above, migration stress and U.S. exposure affect
different subsets of people. U.S. exposure affects all residents,
whether native born or foreign born. Migration stress affects all
movers, whether they go through the visa process or, as
discussed earlier, are already U.S. citizens (such as persons born
in Puerto Rico or the foreign-raised, U.S.-born children of
foreign students). Table 1 provides a brief summary of the
three sources of health change and the subpopulations at risk.
Two examples illustrate. First, consider Pato Pascual. He
came to the United States to study oenology, obtaining a Ph.D.
Halfway through his studies, he fell in love with and married a
U.S. winemaker, who sponsored his immigration as the spouse
of a U.S. citizen. He worries that the immigration authorities
will not believe that he is really in a love marriage; he worries

Table 1

Sources of Health Change, by Subpopulation
at Risk
Visa
Stress

Migration
Stress

U.S.
Exposure

Yes
Yes

No
No

Yes
Yes

Yes

Yes

Yes

Legal immigrants, potentially
in NIS, not residing in United States
Various types

Yes

No

No

Other persons (not immigrants),
in NIS, residing in United States
U.S. citizen sponsors of spouses

No

No

Yes

No

Yes

Yes

No

Yes

Yes

Subpopulation
Legal immigrants, potentially
in NIS, residing in United States
Born under diplomatic status (DS1)
Living in United States since infancy
All other immigrants residing
in United States

Newcomers (not immigrants),
not in NIS, residing in United States
Born in U.S. territories
Born in United States, raised abroad
by foreign-born parents

about obtaining all the documents that are needed; he worries
that the documents will be lost, etc. For him, U.S. exposure and
migration stress began when he started school; he shares U.S.
exposure with everyone who lives in the area (including his
new bride), and he shares migration stress with everyone who
comes from another country, including a golden classmate
with a U.S. passport but little knowledge of English who was
born in Baltimore when her parents were graduate students.
Visa stress, however, began when his wife filed the first
application for his legal permanent residence.
Meanwhile, Caperucita Roja applied for a diversity visa in
her home country of Peru, went through the entire visa process
in Peru, and arrived in Chicago with her visa, receiving the
stamp on her passport in the “secondary” inspection area at
O’Hare. For her, visa stress ended on the day that U.S. exposure
and migration stress began.
This discussion suggests that for assessing both migration
stress and U.S. exposure effects, the point at which inception of
U.S. residence occurs is a critical time. The visible effects, if any,
of migration stress and U.S. exposure will differ depending on
whether inception of U.S. residence occurs before admission to
permanent residence or at admission to permanent
residence—that is, before or during the decline associated with
visa stress or at its end. If the combined migration-U.S.
exposure effect is zero, then both the visa-stress decline and the
post-LPR recovery are unaffected. However, when inception of

FRBNY Economic Policy Review / December 2005

133

U.S. residence occurs prior to admission to legal permanent
residence, a positive net effect of the combined migration-U.S.
exposure would attenuate the visa-stress decline, while a
negative net effect would exacerbate it. Moreover, the
combined migration-U.S. exposure net effect would also alter
the recovery incline, exaggerating it if positive, attenuating or
even reversing it if negative.18

3.2. Empirical Framework—Data,
Measurement, Estimation
Data are drawn from Round 1 of the New Immigrant Survey’s
first full cohort, a probability sample of new legal immigrants
whose administrative records were compiled by the U.S.
government during a seven-month period in 2003. The NIS2003 drew a sample that undersampled immigrants admitted
as the spouse of a U.S. citizen (who constitute about a third of
adult new legal immigrants) and oversampled employmentvisa principals and diversity-visa principals (two categories that
are smaller but in which there is much interest). In order to
reach sampled individuals as soon as possible after admission
to LPR, the sample was drawn in eight replicates (the first and
last replicates were half-month replicates, the other six were
full-month replicates). Interviews were conducted with the
main sampled immigrant (8,573—achieving a response rate of
69 percent), the spouse of the main sampled immigrant (if he
or she was living in the household—4,336), and with up to two
children aged eight to twelve (1,062). Information was
obtained on virtually every sociobehavioral domain, including
migration history, schooling, employment, as well as earnings
histories, language and religion histories, marital history,
health, health behaviors, and health care. Information was
also obtained on all children under eighteen residing in the
household, and cognitive assessments were carried out on
children aged three to twelve.
To ensure sample coverage and data quality, a basic
principle of the NIS is that all persons are interviewed in the
language of their choice. Accordingly, interviews were
conducted in English, Spanish, Chinese, Russian, and eightytwo other languages, plus sign language. The mean and
median time elapsed between admission to LPR and interview
were seventeen weeks and fourteen weeks, respectively. (For
further detail on the NIS project, the NIS-2003 sampling
design, language design, and questionnaires, see Jasso et al.
[forthcoming].)
Full empirical assessment of the immigrant health model
that we have sketched is quite demanding, requiring health
measures at several carefully chosen points in time: 1) at or just

134

Immigration, Health, and New York City

before the start of the visa application process, 2) at inception
of U.S. residence, 3) at admission to legal permanent residence,
4) at several points between the start of the application process
and admission to legal permanent residence, and 5) at several
points after inception of U.S. residence and after admission to
permanent residence.
Further, measuring health is no simple matter. Here we use
two types of measures: the subjective assessment of overall
health widely used in U.S. data collection and a subjective
measure of health change.
The subjective assessment of overall health asks, “In general,
would you say your health is: ?” and provides five response
categories: excellent, very good, good, fair, and poor. Previous
research suggests that subjective assessment of overall health
accords well with objective measures (Ware and Donald 1978;
Wallace and Herzog 1995). Nonetheless, it is possible that
measured healthiness includes a new component—the
immigrant’s style of reporting, a style that may be understated
or overstated. Moreover, the style of reporting may also have
both a permanent component and a transitory component.
Thus, overall health, subjectively measured (denoted H∗ ),
may contain four distinct components: the two health
components introduced earlier plus two style-of-reporting
components—a permanent component of the style of
p
reporting, denoted s , and a transitory component of the style
t
of reporting, denoted s :
(2)

p
t
p
t
H∗ = h + h + s + s .

The NIS-2003 Round 1 data include three subjective
assessments of health, pertaining to three points in time:
1) during childhood (“when you were growing up, from birth
to age 16”), 2) at the time of the migration decision (“at the
time of that first filing that started the process for the
immigrant visa that you now have”), and 3) at the time of the
interview.
All the measures capture the same permanent health
component and permanent style component. They differ,
however, in the transitory health component and the transitory
style component.
With respect to the transitory health component, the
question on healthiness at the time that the first application
was filed taps healthiness prior to the start of visa stress; the
childhood question does so as well, provided that the sample is
restricted to respondents for whom the first filing occurred
after they were age sixteen. In contrast, the question on current
healthiness taps overall healthiness at a point subsequent to
admission to permanent residence. The precise difference
between the transitory health components in the at-filing and
the current assessment depends on: 1) whether inception of

U.S. residence has occurred prior to the first immigration
application filing, in which case the U.S. exposure effects and
migration stress have started, and 2) whether the immigrant
visa is conditional, in which case visa stress has not ended by
the time of the interview.
With respect to the transitory style component, it is
tempting to assume that because the measures are obtained
at the same time, they contain the same transitory style
component. However, one pertains to the present and the
other two to the past. The measure of current healthiness is
subject to underestimation, to avoid displaying hubris or
jinxing one’s health. The measures of past healthiness are
probably more free of style distortions, although they may be
subject to overestimation, if the past is remembered fondly.

Health Selection Equation
To estimate the health selection equation, we use two
subjective measures of overall healthiness: during childhood
and at the time of the first filing. These measures approximate
a pure measure of the permanent component of health at the
time of the initial self-selection. They are imperfect, however,
because inception of U.S. residence may already have occurred,
and thus migration stress and the effects of U.S. exposure
may already have begun. To correct for this effect, we use
information on whether the new immigrant is adjusting to
LPR while already residing in the United States. Moreover, to
distinguish between effects of legal and illegal prior residence,
we define two binary adjustment variables, one for adjusting
from a legal status and the other for adjusting from an illegal
status.
To control for the transitory style component, we exploit
the language feature of the NIS, including a control for whether
the interview was conducted in English (Jasso 2003). To ensure
that interview language does not operate as a proxy for English
language skill, which could be associated with investments in
health, we also include in the specification the interviewer’s
assessment of the respondent’s fluency in English.
In one version of the health selection equation, we include
binary variables for continent of birth and for the top-ten
origin countries; in the second version, we include skill prices
and distance from the United States, interacted with visa
category, plus origin-country GDP per adult equivalent.19, 20
Note that as NIS survey rounds accumulate, it will be
possible to use individual-specific fixed-effects estimation to
obtain sharper estimates of the permanent component of
health and thus of the health selection equation.

Health Change Equation
To assess the effects of visa stress, migration stress, and
exposure to the U.S. environment, we make use of a question
tapping health change between inception of U.S. residence and
the baseline-round interview. For immigrants whose U.S.
residence started at admission to LPR, visa stress ended at
admission to LPR for all sample members except those with
conditional visas, and thus the health change reflects migration
stress and U.S. exposure, plus the recovery from visa stress. For
immigrants whose U.S. residence started at some point prior to
admission to LPR (which could have been before or after the
first visa filing), the health change also reflects visa stress.
Accordingly, the specifications include the adjustment
variables and a dummy variable for a conditional visa. We
expect adjustees to have greater incidence of health
deterioration and lower incidence of health improvement, due
in part to the visa stress experienced by adjustees and in part to
the greater duration of the period of migration stress and U.S.
exposure. The specifications also include the time elapsed
between admission to LPR and the baseline interview; this
variable targets the joint effects of migration stress and U.S.
exposure after the end of visa stress (or net of visa stress, for
immigrants with conditional visas).

4. Basic Characteristics
of the NIS-2003 Cohort
4.1. General Characteristics
We begin by presenting an overview of the basic characteristics
of the NIS-2003 immigrants—sex ratio and sex-specific
average age and schooling and the proportions adjustee and
fluent in English (Table 2). The table also reports the
proportions in each of the thirteen major visa categories, plus a
residual category, as well as basic characteristics for each of the
visa categories. There is great heterogeneity across migration
streams. For example, average schooling is highest among
employment principals and diversity principals, and, by
mechanisms of assortative mating, among their spouses, and
lowest among parents of U.S. citizens, legalization immigrants,
and spouses of LPRs. Age, of course, differs, as would be
expected when some categories are reserved for parents and
others for offspring under age twenty-one. Overall English
fluency is high, almost 49.4 percent among men and
43.5 percent among women—with higher proportions among

FRBNY Economic Policy Review / December 2005

135

Table 2

Basic Characteristics of New Legal Immigrants Aged Eighteen and Older: NIS-2003 Cohort
Age
Visa Category

Schooling

Percentage Adjustees

English Fluency

Percentage
Female

Men

Women

Men

Women

Men

Women

Men

Women

62.9
82.4
66.1
41.9
51.4
53.4
32.6
77.3
41.2
49.2
41.7
76.0
49.6
51.8
56.4

32.9
43.8
65.5
20.2
48.5
50.3
37.2
40.2
32.3
37.7
40.8
44.5
38.7
35.9
38.7

32.6
40.1
62.7
20.2
48.2
46.3
36.8
35.3
32.8
34.5
38.2
43.2
38.0
36.2
39.1

12.6
8.48
8.75
11.5
11.8
13.0
15.7
14.6
14.5
14.6
12.8
13.3
9.04
12.1
12.3

13.1
7.79
6.93
11.9
11.1
10.9
15.2
15.3
14.5
13.1
11.8
11.0
8.42
11.8
11.6

81.6
51.0
25.5
46.1
8.97
4.03
78.8
57.1
8.47
5.21
100
100
100
24.2
57.9

72.5
63.4
33.5
41.4
12.9
3.94
55.2
76.2
11.4
3.52
100
100
100
23.0
57.0

56.2
24.8
26.6
58.2
41.9
37.7
81.0
72.3
55.3
41.4
46.2
32.9
26.7
44.5
49.4

54.0
19.3
24.4
50.8
25.7
17.8
81.7
79.3
47.4
42.8
41.1
37.4
17.2
36.8
43.5

Spouse of U.S. citizen (34.1%)
Spouse of legal permanent resident (2.44%)
Parent of U.S. citizen (11.9%)
Minor child of U.S. citizen (3.38%)
Sibling of U.S. citizen (3.94%)
Spouse of sibling (2.49%)
Employment principal (6.02%)
Employment spouse (3.63%)
Diversity principal (5.53%)
Diversity spouse (2.58%)
Refugee/asylee/parolee principal (5.35%)
Refugee/asylee/parolee spouse (1.22%)
Legalization (7.98%)
Other (9.36%)
All immigrants
Source: New Immigrant Survey, 2003 Cohort, Round 1.

Notes: The sample size is 8,573. Estimates are based on weighted data. The measure of English fluency requires that either the interview be conducted entirely
in English or that the interviewer give the respondent’s English the highest rating (“very good”). Among the subset coded fluent in English, 89.5 percent
completed the interview entirely in English.

Table 3

Basic Characteristics of New Legal Immigrants in New York City Aged Eighteen and Older: NIS-2003 Cohort

Age
Percentage
Female

Visa Category
Spouse of U.S. citizen (23.8%)
Spouse of legal permanent resident (1.17%)
Parent of U.S. citizen (12.4%)
Minor child of U.S. citizen (5.95%)
Sibling of U.S. citizen (3.79%)
Spouse of sibling (2.84%)
Employment principal (3.94%)
Employment spouse (2.62%)
Diversity principal (9.62%)
Diversity spouse (5.15%)
Refugee/asylee/parolee principal (7.09%)
Refugee/asylee/parolee spouse (2.04%)
Legalization (1.38%)
Other (18.2%)
All immigrants

56.0
—
63.1
32.0
39.1
48.8
41.3
—
42.9
56.4
27.1
—
—
42.3
48.8

Schooling

Percentage
Adjustees

English Fluency

Men Women

Men

Women

Men

Women

Men

Women

35.7
—
64.1
19.5
49.3
52.0
39.2
—
32.4
38.2
42.9
—
—
36.5
39.3

12.1
—
9.29
11.7
11.2
12.2
14.8
—
14.7
14.6
13.6
—
—
12.0
12.3

12.3
—
5.16
11.7
9.06
8.53
14.2
—
15.0
13.6
13.4
—
—
12.0
11.2

70.8
—
6.04
17.4
0
0
84.6
—
7.77
5.93
100
100
100
9.21
36.9

44.6
—
9.03
21.7
6.95
0
74.3
—
7.81
4.07
100
100
100
7.48
30.1

63.5
—
20.5
55.5
62.1
88.0
63.7
—
60.5
24.5
57.1
—
—
51.7
53.7

62.1
—
33.2
38.2
24.1
9.34
83.6
—
45.8
42.9
61.8
—
—
48.3
46.2

34.3
—
61.9
19.5
49.9
49.4
40.0
—
33.0
36.8
47.0
—
—
36.8
40.9

Source: New Immigrant Survey, 2003 Cohort, Round 1.
Notes: The sample size is 866. Estimates are based on weighted data. The measure of English fluency requires that either the interview be conducted entirely
in English or that the interviewer give the respondent’s English the highest rating (“very good”). Among the subset coded fluent in English, 95.5 percent
completed the interview entirely in English. The sample sizes for spouse of legal permanent resident, employment spouse, refugee spouse, and legalization
immigrants are too small to report summary characteristics.

136

Immigration, Health, and New York City

employment principals and the spouses and children of U.S.
citizens.21
Approximately 9.27 percent of the new immigrants declared
New York City to be their initial residence.22 Table 3 summarizes
the basic characteristics for this subset. The New York City–
bound immigrants differ in several important ways from the
larger set. First, the proportion female is lower by almost
8 percentage points (48.8 percent versus 56.4 percent). Second,
and consistent with the sex ratio, the proportion achieving LPR
via marriage to a U.S. citizen is substantially lower—24 percent
versus 34 percent. The New York City group has a smaller
proportion who are employment principals (4 percent versus
6 percent) and a larger proportion who are diversity principals
(9.6 percent versus 5.5 percent), and among employment
principals, a substantially larger proportion who are female
(41 percent versus 33 percent). Third, the proportion adjusting
status is markedly lower in the New York City subset (by
20 percentage points among men and 27 percentage points
among women), reflecting in part the smaller proportion of
marriages to U.S. citizens but also fewer adjustments even
among these couples. Fourth, New York immigrants display
somewhat greater English fluency (53.7 percent versus
49.4 percent among men and 46.2 percent versus 43.5 percent
among women).
The patterns in Tables 2 and 3 suggest differences in the
origin countries of immigrants who settle initially in New York
City and their counterparts who settle elsewhere in the country.
Table 4 displays the top five origin countries for the entire set
of immigrants as well as for the New York City and non-New
York City subsets. As shown in the middle and lower panels,
the two areas share only one country in the top five—China,
which is the second-leading origin country in New York City
and fifth among the non-New York City immigrants. Besides
the largely nonoverlapping sets of top-five countries, the other
important difference concerns the somewhat greater evenness
among the New York City top five, in contrast to the nonNew York City countries, which are dominated by Mexico.
As we observed, a basic principle of the NIS design is that
every respondent is interviewed in his or her preferred
language. Consistent with the greater English fluency among
the New York City subset, 47.9 percent of the New York
immigrants preferred English, compared with 40.6 percent in
the rest of the country. English preference among New York
City immigrants was led by immigrants from Guyana and
Jamaica, virtually all of whom preferred English. In contrast,
among non-New York City immigrants, English preference
was led by immigrants from India and the Philippines, but the
proportions from those two countries preferring English did
not exceed 73 percent.

The NIS included the two questions on race and ethnicity
that are standard in U.S. surveys. Among the New York City
immigrants, the largest racial/ethnic group consisted of nonHispanic Asians, of whom there are 27 percent, followed
closely by non-Hispanic whites (25 percent), non-Hispanic
blacks (17 percent), Hispanic whites (16 percent), Hispanics
who did not provide race (5 percent), and non-Hispanics
who also did not provide race (4 percent). In contrast, among
the non-New York City immigrants, the largest group was
Hispanic whites (30 percent), followed closely by non-Hispanic
Asians (28 percent), non-Hispanic whites (19 percent), nonHispanics who did not provide race (14 percent), non-Hispanic
blacks (10 percent), and Hispanics who did not provide race
(6 percent). The different origin-country distributions help
explain these patterns. For example, the different proportions
of Hispanic whites (16 percent in the New York City subset
versus 30 percent in the non-New York City subset) can be

Table 4

Top Five Countries of Origin among New Legal
Immigrants Aged Eighteen and Older, by Sex
and Initial Residence
Men

Women

All

All immigrants
Mexico
India
El Salvador
China
Philippines
Top five

16.2
7.19
6.82
5.14
4.19
39.5

18.7
7.36
6.49
5.61
5.60
43.8

17.6
7.28
6.13
5.49
5.40
41.9

Immigrants with initial residence
in New York City (n = 866)
Dominican Republic
China
Guyana
Jamaica
Ecuador
Top five

11.9
11.3
7.84
6.5
4.67
42.2

14.3
10.8
5.45
4.63
4.55
39.7

13.1
11.0
6.28
5.05
4.27
39.7

Immigrants with initial residence
not in New York City (n = 7,707)
Mexico
India
El Salvador
Philippines
China
Top five

18.0
7.67
7.55
4.54
4.34
42.1

20.2
7.66
6.95
6.01
5.16
46.0

19.3
7.66
6.67
5.92
4.83
44.4

Source: New Immigrant Survey, 2003 Cohort, Round 1.
Notes: The sample size is 8,573. Estimates are based on weighted data.

FRBNY Economic Policy Review / December 2005

137

attributed in part to different rates of declaring this
combination (Hispanic white) among the top origin
countries—51 percent among the New York City group from
the Dominican Republic and 72 percent in the larger non-New
York City group from Mexico.
Recall the higher proportion who are diversity principals in
the New York City group (9.6 percent versus 5.5 percent). An
important feature of recent immigration is that the diversity
visa program has, as intended, generated new streams of
immigrants from countries that have been underrepresented.
Thus, almost half of diversity principals are from Africa—
44 percent in the NIS-2003 cohort. And the fraction of Africaborn diversity principals who reside in New York City is larger
than the corresponding fraction of other immigrants
(12 percent versus 9 percent).
New York City has a large concentration of foreign-born
persons—currently estimated at 36 percent of the population.
Accordingly, the pool of marriageable persons is likely to be
substantially foreign born, generating a higher-than-average
proportion of foreign born among the U.S. citizen sponsors of
spouses. As expected, while overall 47 percent of the U.S.
citizen sponsors of spouses are native born, in the New York
City immigrant subset, the corresponding figure is less than
half—22 percent.
Finally, we examine home ownership among immigrants in
the NIS-2003 cohort. New York City differs from the rest of the
country in the proportion who own their home, and, indeed,
in the ethos surrounding home ownership. Overall, more than
26 percent of the new immigrants already own their home—as
well as 37 percent of adjustee immigrants, who have had more
time in the United State. Not surprisingly, however, the
corresponding figures for the New York City subset are
7 percent and 13 percent—or roughly 28 to 35 percent of the
nationwide figures.

4.2. Health Characteristics

Health Self-Assessment
Table 5 reports the immigrants’ assessments of their health at
the time of the initial filing, which started the process by which
they became legal permanent residents, reported at the baseline
interview. As shown, overall the new immigrants thought of
themselves as quite healthy at the time of the initial selfselection—almost three-fourths judged themselves to be in
excellent or very good health and only slightly more than

138

Immigration, Health, and New York City

4 percent in fair or poor health. In general, male immigrants
judged themselves to be healthier than did female
immigrants—although the largest difference is in the
“excellent” category, which may reflect mechanisms other than
actual health (such as male brashness or female wish to avoid
hubris). There is a pronounced difference between those with
very little schooling and those with a very high amount of
schooling (53 percent of those with more than sixteen years of
schooling pronouncing themselves to be in excellent health
versus 27 percent among those with less than nine years of
schooling).
Comparable figures (not shown) for the New York City
contingent of immigrants indicate that at each of the three time
points, New York immigrants are substantially healthier than
other immigrants. For example, in the assessment of health
at the time of first filing, 59 percent of the New York City
immigrants judged their health to be excellent versus 41 percent
of the non-New York City immigrants.

Health Change
In Table 6, we present the immigrants’ reported health change
between the last time they came to live in the United States and
the time of the baseline interview, by visa category and
separately for new arrivals and adjustees. As discussed earlier,
for “true” new arrivals, visa stress will have ended at arrival
(except for conditional immigrants) and all effects will be due
to migration stress and U.S. exposure. For adjustees, the period
since last arrival will also include a period of visa stress followed
by the post-LPR recovery phase. Moreover, the length of the
interval is substantially greater for adjustees than for new
arrivals (less than four months for new arrivals and more than
five years for adjustees, on average). As shown in the table, the
results indicate that while similar proportions report improved
health (20 percent of new-arrival immigrants and 22 percent of
adjustee immigrants), a much larger proportion of adjustee
immigrants report deteriorating health (14 percent versus 4
percent). This health decline could be due to the greater
likelihood that for adjustees, arrival occurred before the start of
the decline associated with visa stress or it could be due to the
longer interval during which migration stress and the effects of
U.S. exposure are experienced.
Immigrants who settle in New York City have a smaller
proportion with deteriorating health than immigrants who
settle elsewhere—7.1 percent versus 10.2 percent—a difference
almost completely offset by the larger fraction of New York
City immigrants whose health remained the same.

Table 5

Health Status at Time of First Filing for Immigrant Visa, Self-Reported at Baseline Round:
NIS-2003 Immigrants Aged Eighteen and Older

Health Status
Five-Category Variable (Percent)
Characteristic or Population

Excellent

Very Good

Good

Fair

Poor

Index
(Mean)

42.6
47.4
38.8
26.7
52.9

30.9
29.8
31.8
28.2
32.5

22.5
19.5
24.8
36.9
12.7

3.53
3.07
3.88
7.37
1.93

0.52
0.25
0.74
0.83
0.04

3.11
3.21
3.04
2.73
3.36

45.4
28.1
21.4
54.8
37.8
38.2
52.8
43.2
56.8
50.4
44.3
37.0
37.2
48.4

31.1
37.7
29.3
31.4
35.7
37.9
32.4
38.1
30.3
30.7
28.7
24.4
24.1
31.0

20.8
31.2
37.1
10.7
23.3
22.6
13.7
15.2
12.3
18.2
20.6
30.3
33.0
18.4

2.09
3.00
11.0
2.63
3.21
1.33
0.96
3.49
0.24
0.75
4.93
5.28
5.44
2.16

0.58
0
1.18
0.36
0
0
0.17
0
0.35
0
1.40
3.04
0.29
0.09

3.19
2.91
2.59
3.38
3.08
3.13
3.37
3.21
3.43
3.31
3.10
2.87
2.92
3.25

59.0
39.2
44.7
62.2
39.8
48.6

24.9
35.5
33.3
26.9
27.9
28.7

12.6
21.9
18.6
7.26
27.1
20.1

2.97
2.99
2.50
0
4.75
2.34

0.62
0.35
0.82
3.71
0.56
0.23

3.86
3.10
3.19
3.44
3.02
3.23

33.2
47.7
39.1
42.7
31.4

28.3
29.8
24.7
39.0
42.2

32.5
19.7
30.7
17.1
19.5

5.32
2.64
5.47
1.13
6.90

0.81
0.17
0
0.09
0

2.88
3.22
2.98
3.23
2.98

43.1
42.2

31.5
30.5

21.4
23.3

3.78
3.34

0.28
0.71

3.13
3.10

Selected basic characteristics
All immigrants
Male immigrants
Female immigrants
Schooling less than nine years
Schooling more than sixteen years
Visa category
Spouse of U.S. citizen
Spouse of legal permanent resident
Parent of U.S. citizen
Child of U.S. citizen
Sibling of U.S. citizen
Spouse of sibling
Employment principal
Employment spouse
Diversity principal
Diversity spouse
Refugee/asylee principal
Refugee/asylee spouse
Legalization
Other
Continent of birth
Africa
Asia
Europe
Oceania
North America
South America
Top five countries of birth
Mexico
India
El Salvador
Philippines
China
Adjustment of status
New arrivals
Adjustees
Source: New Immigrant Survey, 2003 Cohort, Round 1.

Notes: The health status variable is coded 0-4, with poor coded 0. Estimates are based on weighted data.

FRBNY Economic Policy Review / December 2005

139

Table 6

Health Change between Most Recent Arrival “to
Live” and First Interview after Admission to Legal
Permanent Residence: NIS-2003 Immigrants
Health Change

Visa Category
New-arrival immigrants
Spouse of U.S. citizen
Spouse of legal permanent
resident
Parent of U.S. citizen
Minor child of U.S. citizen
Sibling of U.S. citizen
Spouse of sibling
Employment principal
Employment spouse
Diversity principal
Diversity spouse
Other
All new-arrival immigrants
Adjustee immigrants
Spouse of U.S. citizen
Spouse of legal permanent
resident
Parent of U.S. citizen
Minor child of U.S. citizen
Sibling of U.S. citizen
Spouse of sibling
Employment principal
Employment spouse
Diversity principal
Diversity spouse
Refugee/asylee/parolee
principal
Refugee/asylee/parolee
spouse
Legalization
Other
All adjustee immigrants

Time since
Arrival
(Years)

Worse

Same

Better

.325

5.72

75.3

19.0

.289
.312
.279
.313
.305
.323
.045
.316
.351
.291
.305

3.05
5.91
2.12
3.39
3.47
5.40
3.74
2.86
2.79
2.18
4.05

88.0
68.4
72.5
82.8
79.7
79.1
74.3
79.7
79.5
77.2
76.2

8.91
25.7
25.4
13.8
16.9
15.5
22.0
17.5
17.7
20.7
19.7

5.20

13.2

67.3

19.5

6.34
6.41
7.24
8.16
—
2.61
2.19
3.67
—

11.7
15.9
12.4
16.7
—
13.2
11.7
5.72
—

63.5
64.3
59.8
60.0
—
67.7
74.5
69.1
—

24.7
19.7
27.9
23.5
—
19.1
13.9
25.2
—

6.89

16.3

57.5

26.2

6.18
11.1
9.64
5.25

24.1
17.2
6.55
14.0

51.7
50.7
68.9
63.6

24.2
32.0
24.6
22.3

Source: New Immigrant Survey, 2003 Cohort, Round 1.
Notes: Estimates are based on weighted data. Missing estimates pertain
to subsets with fewer than twenty observations.

5. Multivariate Results
5.1. Health Selection
Ordered-logit estimates of the health selection equation are
reported in Table 7, with three specifications (sex-specific and
pooled ) for the two health measures: health at first filing and

140

Immigration, Health, and New York City

health during childhood. The objective is to estimate the
selectivities associated with the permanent component of
health. Health at first filing is a good approximation of the
permanent component of health among new-arrival
immigrants but it is not as good for adjustees, whose U.S.
residence may have antedated the first filing so that they may
have already been experiencing migration stress and the effects
of U.S. exposure. Additionally, both new-arrival and adjustee
immigrants may have suffered harm in the origin country prior
to the first filing. We address these possibilities by including
control variables, such as the adjustee variables, and by
estimating the health selection equation with health during
childhood as the dependent variable. Health during childhood,
for example, is likely to be free of the harm effects and free as
well of migration stress and U.S. exposure—unless U.S.
residence started before age sixteen.
We first assess the controls for sources of change in the
transitory component of health. The controls we inserted for
adjustees, as well as the control for having suffered harm,
operate as predicted and most of the estimates are statistically
significant. For example, the two adjustee variables are jointly
highly statistically significant and both are negative, indicating
that the observed health of adjustees is indeed lower than that
of new arrivals, consistent with a negative net effect of the
combined migration stress and U.S. exposure. The negative
effect of adjusting from an illegal status is substantially larger
than that of adjusting from a legal status, consistent with the
operation of visa stress. Similarly, the effect of having suffered
harm in the origin country is negative and statistically
significant in the pooled and male specifications of the firstfiling equation and not significant and of mixed sign in the
childhood equation, indicating that men were more vulnerable
to such harm and that on average it occurred after childhood.23
With respect to the control for style of reporting, the effect
of being interviewed in English was statistically significant in all
specifications except the male childhood one and positive, net
of skill in English, consistent with the hypothesized association
between English and a style of reporting that does not refrain
from declaring high healthiness.
Turning now to our main focus, the health selectivities, we
note that the estimates indicate that men are more highly
positively selected on health than are women and that racial/
ethnic characteristics and area of origin are importantly linked
to health selection. The coefficients on the racial/ethnic
categories indicate that Hispanic whites are the most positively
selected for health, followed by non-Hispanic black men; the
least selected for health are Hispanics who decline to declare a
race.
The visa category variables are jointly significant in the
women’s at-filing equation and in both the men’s and women’s

childhood health equations. Comparison of the coefficients in
the at-filing and childhood equations reveals interesting
patterns. Among men, legalization immigrants are among the
most robust in childhood, but by the time of the first filing they
are less healthy; in contrast, refugee principals are less robust in
childhood but by the time of the first filing they are healthier
than many of their fellow immigrants. Among women,

diversity principals are the most positively selected for health,
followed by employment principals.
The joint tests for the continent and country dummies
indicate high statistical significance in all cases except one—the
continent dummies in the male first-filing equation. The
coefficients (not shown) indicate that immigrants from North
America (which includes Canada, Mexico, and Central

Table 7

Selected Estimates, Ordered-Logit Health Selection Equation: NIS-2003 Immigrants
Aged Eighteen and Older at Time of First Filing for Legal Permanent Residence
Specification
Health at First Filing
Variable
Sex
Age at first filing
Age squared
Age joint test chi2 (2 df)
Suffered harm in origin country
Hispanic, no race
Hispanic, white
Not Hispanic, Asian
Not Hispanic, black
Not Hispanic, white
Race/ethnicity joint test chi2 (5 df)
Spouse of U.S. citizen
Parent of U.S. citizen
Child of U.S. citizen
Employment principal
Diversity principal
Refugee/asylee principal
Legalization
Visa category joint test chi2 (7 df)
Adjustee, not illegal
Adjustee, illegal
Adjustee joint test chi2 (2 df)
Interview in English
English “very good”
Continent dummies joint test chi2 (5 df)
Country dummies joint test chi2 (10 df)
Number of observations
Log pseudolikelihood

Health during Childhood

All

Men

Women

All

Men

Women

-.260
(7.08)
0.0266
-.000597
133.01
-.287
(2.40)
-.469
.199
-.230
-.0893
-.0330
24.6
.0855
-.161
.145
.107
.314
-.0158
-.0943
11.9
-.107
-.440
27.1
.211
(2.78)
.385
(5.18)
27.3
892.1
7,517
-8332.30

—

—

—

—

.0123
-.000426
35.8
-.372
(2.61)
-.270
.0884
-.228
.0747
-.0217
9.76
.147
-.375
.366
-.00945
.219
.136
.0528
12.3
-.142
-.467
11.4
.212
(1.94)
.479
(4.27)
6.57
892.7
3,687
-3904.21

.0340
-.000680
93.3
-.188
(1.23)
-.631
.302
-.219
-.318
-.0142
36.2
.0673
-.0663
-.251
.272
.405
-.250
-.235
27.4
-.0843
-.408
18.1
.216
(2.59)
.295
(3.72)
20.9
563.3
3,830
-4408.25

-.109
(2.73)
.0326
-.000329
11.6
-.0479
(.52)
-.567
.0228
-.144
.144
.325
41.3
.101
-.120
-.227
.0707
.229
.0360
.109
11.7
-.0616
-.389
20.8
.171
(2.00)
.222
(3.15)
37.4
470.0
7,246
-7891.35

.00697
-.000045
.49
-.0985
(.82)
-.514
.00298
-.177
.160
.382
37.0
.104
-.255
-.475
-.0650
.0889
-.0139
.445
23.6
.0718
-.493
26.6
.173
(1.51)
.277
(2.91)
18.1
685.1
3,569
-3804.76

.0496
-.000502
15.5
.0111
(.08)
-.621
.0399
-.0972
.100
.247
14.9
.107
-.110
.0744
.166
.380
.00360
-.218
18.2
-.155
-.295
12.8
.193
(2.22)
.153
(1.82)
25.3
534.7
3,677
-4066.73

Source: New Immigrant Survey, 2003 Cohort, Round 1.
Notes: The dependent variables are coded 0-4, with poor coded 0 and excellent coded 4. Standard errors are corrected for heteroskedasticity due to clustering
by origin country; absolute values of asymptotic t-ratio appear in parentheses under parameter estimates for numeric and binary variables. Joint tests are
reported for multiple-category categorical variables. Cut-points are not shown.

FRBNY Economic Policy Review / December 2005

141

America and the Caribbean) and Africa are the most highly
positively selected for health, while immigrants from Europe
(the omitted category), Asia, and Oceania are the least
positively selected for health. Of course, for any individual
immigrant, these effects have to be combined with the country
effects. For example, the coefficients for India and Mexico
indicate the highest and lowest selectivities, respectively, so that
combining the country and continent effects alters the picture
somewhat.
The area-of-origin effects point to mechanisms involving
country characteristics. Table 8 presents ordered-logit

estimates of the health selection equation based on an
economic model in which selection responds to skill prices and
the origin country’s distance from the United States. Control
variables—the adjustee variables, for example, and the Englishlanguage variables—operate as they do in the previous
equations. However, the new results indicate important
selectivity by origin-country skill prices. The joint test of skill
prices and skill prices interacted with visa category indicates
that these effects are highly statistically significant in all three
at-filing specifications and in the women’s childhood health
specification. In contrast, distance and its interactions with visa

Table 8

Selected Estimates, Ordered-Logit Health Selection Equation, with Skill Prices and Distance:
NIS-2003 Immigrants Aged Eighteen and Older at Time of First Filing for Legal Permanent Residence
Specification
Health at First Filing
Variable
Sex
Age at first filing
Age squared
Age joint test chi2 (2 df)
Suffered harm in origin country
Race/ethnicity joint test chi2 (5 df)
Spouse of U.S. citizen
Parent of U.S. citizen
Child of U.S. citizen
Employment principal
Diversity principal
Refugee/asylee principal
Legalization
Visa category joint test chi2 (7 df)
Adjustee, not illegal
Adjustee, illegal
Adjustee joint test chi2 (2 df)
Interview in English
English “very good”
Skill price interacted with visa joint test chi2 (8 df)
Distance interacted with visa joint test chi2 (8 df)
Real GDP per adult equivalent
Number of observations
Log pseudolikelihood

Health during Childhood

All

Men

Women

All

Men

Women

-.251
(6.05)
0.302
-.000618
97.9
-.234
(1.67)
34.4
-.697
-.556
-.454
.100
-.0800
-.584
1.80
38.9
-.140
-.621
47.4
.263
(3.37)
.376
(4.72)
28.3
13.6
8.34e-08
(.01)
6,449
-7151.94

—

—

—

—

.0115
-.000425
30.4
-.426
(2.64)
18.2
-.121
-.899
.0550
-.0400
-.338
-.237
1.76
24.7
-.165
-.673
16.4
.245
(2.19)
.448
(3.50)
35.4
7.57
-1.75e-06
(.23)
3,196
-3368.99

.0438
-.000741
64.8
.0854
(.48)
30.5
-.394
-.382
-2.24
.219
.110
-1.21
3.10
13.5
-.137
-.573
36.2
.284
(3.12)
.307
(3.82)
29.3
15.8
8.98e-07
(.10)
3,253
-3758.69

-.0961
(2.16)
.0386
-.000400
12.2
-.0828
(.68)
48.8
.286
-.222
-.259
.283
-.222
-.477
.483
17.1
-.113
-.548
31.5
.252
(2.96)
.198
(2.66)
5.33
8.41
2.06e-06
(.29)
6,207
-6777.83

.00645
-.000057
.22
-.229
(1.49)
32.3
.297
-.363
.0128
.289
-.0481
-.471
.678
17.3
.0402
-.644
29.1
.313
(2.55)
.269
(2.52)
7.74
21.2
-1.30e-06
(.14)
3,091
-3302.51

.0650
-.000662
25.3
.164
(.82)
25.3
.343
-.273
-2.24
.242
-.602
-.419
.742
13.7
-.208
-.460
17.5
.209
(2.53)
.112
(1.39)
21.4
10.5
6.88e-06
(.93)
3,116
-3446.55

Source: New Immigrant Survey, 2003 Cohort, Round 1.
Notes: The dependent variables are coded 0-4, with poor coded 0 and excellent coded 4. Standard errors are corrected for heteroskedasticity due to clustering
by origin country; absolute values of asymptotic t-ratio appear in parentheses under parameter estimates for numeric and binary variables. Joint tests are
reported for multiple-category categorical variables. Cut-points are not shown.

142

Immigration, Health, and New York City

category produce mixed results, achieving statistical
significance only among women in the at-filing equation and
among men in the childhood equation. It is possible that
distance is becoming less important as globalization takes root.
Finally, we re-estimated all the specifications in Tables 7
and 8 and include a binary variable for initial residence in
New York City. The estimates are uniformly highly statistically
significant and positive, indicating that at the initial selection,
those immigrants who claim New York City as their first home
after admission to LPR are more highly positively selected on
health than are their fellow immigrants who settle elsewhere.

5.2. Visa Depression
In the health model sketched above, an important factor is the
visa application process itself and the associated visa stress that
may negatively affect health. We turn now to immigrants’
subjective experience of visa stress. A question in the NIS-2003
Round 1 interview asks, “During the past 12 months, have you
ever felt sad, blue, or depressed because of the process of
becoming a permanent resident alien?” For convenience, we
use “visa depression” as shorthand for feeling “sad, blue, or
depressed. . . .” All respondents except for thirty-three achieved
LPR during the twelve months before the interview (the mean
time elapsed between LPR and interview was seventeen weeks;
the median time elapsed was fourteen weeks). Overall,
15.9 percent of the men and 18.5 percent of the women
reported becoming depressed because of the visa process.
There is substantial variation in the experience of visa
depression across visa category and origin country/region.
Furthermore, notwithstanding the greater overall depression
among women, the gender pattern itself varies, with men
reporting higher depression rates among employment and
refugee spouses. Visa depression is larger for adjustees than for
new arrivals (by 2.0 percentage points among women and
4.5 percentage points among men). A question for future
research concerns the possibility that visa depression may be
reduced if visa stress is not experienced jointly with migration
stress.24
The figures for the New York City immigrants indicate that
the incidence of visa depression is lower among them and
substantially so for men (10.7 percent versus 15.9 percent).
Relatedly, the gender differential is substantially larger in
New York City than it is in the larger cohort. Rates of visa
depression are high among the city’s largest immigrant
contingent: those born in the Dominican Republic—the rates
are more than twice those of all New York City immigrants,
among both women and men (21 percent among men and

39 percent among women). At the other extreme, not a single
case of visa depression was reported among China-born
immigrant men in the New York City subset. Like the
immigrants in the larger cohort, New York City adjustees
have higher depression rates than do new arrivals; this is
substantially so among women (20.5 percent versus
15.8 percent).
To explore visa depression in a multivariate context, we
estimate a binary logit specification that includes age, race/
ethnic background, visa category, years of schooling, the two
adjustee variables, and binary variables for continent and
selected country of origin, both for the sample as a whole and
separately for men and women. The specification also includes
a binary variable for having suffered harm in the origin
country. Table 9 reports the results. As one would expect from
the raw figures, women are significantly more likely to report
visa depression. Moreover, the visa depression process differs
importantly by gender, with apparently gender-specific risk
and protective factors.
Having suffered harm in the origin country is a strong
predictor of visa depression among men, but it does not reach
statistical significance among women, although it remains
positive. The visa category variables are jointly significant for
men but not for women. It is no surprise that among men,
legalization principals are more likely to report visa depression
or that having a spouse or parent who is a U.S. citizen confers
some protection against visa depression. What is surprising is
that among women, having a spouse or parent who is a U.S.
citizen appears not to provide substantial protection against
visa depression. Moreover, among men, visa stress may be
more manageable in the origin country than in the United
States. The two adjustee variables are highly statistically
significant among men, positive, and of approximately the
same magnitude, suggesting that the lack of protection against
depression while being in the United States prior to becoming
a legal permanent resident is independent of legal or illegal
status. Among women, however, the two adjustee variables are
far from statistically significant, negative, and of magnitudes
close to zero. Thus, the data hint that the origin-country
environment protects men from visa stress but does not
influence, in either direction, women’s higher propensity for
visa depression.25
The racial/ethnic variables are jointly significant in the
women’s equation but not in the men’s. Of the groups
identified, and net of origin area, non-Hispanic whites have the
strongest likelihood of reporting visa depression.
Schooling does not protect against visa depression, on net,
though the nonsignificant and small coefficients could be
masking the opposite operation of two mechanisms—one

FRBNY Economic Policy Review / December 2005

143

positive, the other negative. For example, high schooling might
indeed make it easier to handle the vicissitudes of the visa
process, while at the same time exacerbating the costs of
waiting for LPR.
Finally, we re-estimated the equations with a binary variable
for New York City. Immigrants who settle there are less likely
to report having experienced visa depression than their
counterparts who settle elsewhere in the country. This effect is

Table 9

Selected Coefficients of Binary Logit Estimate
of Visa Depression Equation: NIS-2003
Specification
Variable
Sex
Age at admission
to legal permanent residence
Age squared
Age joint test chi2 (2 df)
Hispanic, no race
Hispanic, white
Not Hispanic, Asian
Not Hispanic, black
Not Hispanic, white
Race/ethnicity joint test chi2 (5 df)
Schooling (years)
Spouse of U.S. citizen
Parent of U.S. citizen
Child of U.S. citizen
Employment principal
Diversity principal
Refugee/asylee principal
Legalization
Visa category joint test chi2 (7 df)
Adjustee, not illegal
Adjustee, illegal
Adjustee joint test chi2 (2 df)
Suffered harm in origin country
Continent dummies joint test chi2 (5 df)
Country dummies joint test chi2 (10 df)
Intercept
Number of observations
Log pseudolikelihood

All

Men

Women

.177
(2.20)

—

—

.419
-.000579
14.9
.148
-.0413
-.0254
.0243
.161
5.08
.107
(1.03)
.0143
.123
.113
.249
-.169
-.336
.354
28.9
.186
.220
4.40
368
(3.33)
19.6
6284.14
-2.87
(7.47)
8,149
-3660.62

.0316
-.000465
4.54
.335
.129
-.0571
.0482
-.00868
2.44
.00936
(.58)
-.234
.0717
-.302
.138
-.188
-.866
.466
39.1
.512
.543
16.3
.440
(3.11)
7.42
770.99
-2.57
(4.57)
3,951
-1706.83

.0458
-.000627
14.7
-.00346
-.205
.00976
-.0188
.323
13.4
.0110
(.99)
.128
.123
.459
.142
-.215
.107
.212
9.37
-.0524
-.00311
.30
.285
(1.79)
21.8
8276.78
-2.80
(4.64)
4,198
.-1926.80

Source: New Immigrant Survey, 2003 Cohort, Round 1.
Notes: Standard errors are corrected for heteroskedasticity due to clustering by origin country; absolute values of asymptotic t-ratio appear in
parentheses under parameter estimates for numeric and binary variables.
Joint tests are reported for multiple-category categorical variables.

144

Immigration, Health, and New York City

highly statistically significant among men ( β = -.559, absolute
value of asymptotic t-ratio = 2.6) and almost twice as large as
the not-quite-significant coefficient among women ( β = -.312,
absolute value of asymptotic t-ratio = 1.86).

5.3. Body-Mass Index and Time
in the United States
Overweight and obesity have increased in the United States
over the past forty years (Ogden et al. 2004). Accordingly, there
is much interest in the causes and correlates of the increasing
American girth. Immigrants present a useful laboratory for
studying overweight. How do they compare with Americans?
And what happens to their weight as they adjust to life in the
United States?
The New Immigrant Survey asks respondents to provide
their height and weight. Thus, the data enable analysis of three
key characteristics—weight, height, and body-mass index. We
examined BMI (weight in kilograms divided by the square of
height in meters) among the NIS-2003 immigrants and among
their native-born counterparts in the 1999-2002 sample of the
National Health and Nutrition Examination Survey
(NHANES), published in McDowell et al. (2005), focusing on
the mean and selected percentiles, separately by age and sex.
NHANES data are collected by trained health technicians in
mobile examination centers, and thus are no doubt more
accurate than the self-reported data collected in the NIS.
Nonetheless, the contrasts point to some unmistakable results.
In brief, immigrants have lower BMI than do Americans in the
NHANES sample—lower mean, lower median, and, with only
two exceptions, lower percentiles at every age.
A key question pertains to the effects on weight of living in
the United States. Mean BMI is larger for adjustees than for
new arrivals among both men and women and in every age
group except, for both sexes, the sixty to sixty-nine age group.
Of course, increasing BMI may be healthful, if BMI at arrival in
the United States was too low. A BMI below 18.5 is considered
to represent being underweight. Mean BMI in the new-arrival
subsets is never below 18.5. Indeed, the fifth percentiles for the
whole cohort are never below 18.5. Accordingly, it appears that
the increase in BMI associated with time in the United States
does not indicate an increase in health.
To explore in a multivariate context the effect of time on
BMI in the United States, we specify and estimate a model with
sex, age, age squared, visa-fixed effects, the two adjustee
variables (adjusting from a legal status and adjusting from an
illegal status), and continent and country dummies. Table 10
reports the results, estimated for the sample as a whole as well
as separately for men and women. The results indicate that the

two adjustee variables are jointly statistically significant in all
three equations and are both positive—BMI increases with
time in the United States. Their relative effects, however, are
sex-specific. Among women, the effect of time spent illegally is
double the effect of time spent legally, while for men, the two
effects are more similar, though the pattern is the reverse of

Table 10

Selected Estimates, Ordinary Least Squares
Equation of Determinants of Body Mass Index:
NIS-2003
Specification
Variable
Sex
Age at Round 1 interview
Age squared
Age joint test chi2 (2 df)
Hispanic, no race
Hispanic, white
Not Hispanic, Asian
Not Hispanic, black
Not Hispanic, white
Race/ethnicity joint test chi2 (5 df)
Schooling (years)
Spouse of U.S. citizen
Parent of U.S. citizen
Child of U.S. citizen
Employment principal
Diversity principal
Refugee/asylee principal
Legalization
Visa category joint test chi2 (7 df)
Adjustee, not illegal
Adjustee, illegal
Adjustee joint test chi2 (2 df)
Conditional visa
Continent dummies joint test
chi2 (5 df)
Country dummies joint test
chi2 (10 df)
Intercept
Number of observations
R2

All

Men

Women

-.852
(3.59)
.363
-.492
79.2
-.519
-.339
-1.26
.240
-.322
1.89
-.0709
(2.95)
-.348
.770
.159
-.183
-.303
.342
.184
5.58
.570
.677
11.8
-.527
(2.38)

—

—

.370
-.00362
42.9
-.492
.110
-1.08
-.300
-.241
2.33
-.0286
(1.18)
.280
-.141
-.213
-.323
-.322
.313
.464
1.28
.645
.401
8.13
-.148
(.39)

.320
-.00282
61.6
-.583
-.895
-1.37
.709
-.542
1.98
-.0860
(2.42)
-.582
1.15
.458
-.285
-.0248
.504
.126
4.26
.407
.916
11.3
-.523
(1.97)

1.68

2.46

1.22

93.7
18.0
(20.4)
7,802
.124

91.0
17.8
(15.7)
3,884
.100

213.9
17.6
(12.6)
3,918
.158

Source: New Immigrant Survey, 2003 Cohort, Round 1.
Notes: Standard errors are corrected for heteroskedasticity due to
clustering by origin country; absolute values of asymptotic t-ratio appear
in parentheses under parameter estimates for numeric and binary
variables. Joint tests are reported for multiple-category categorical variables.

that found among women, with time spent legally producing
greater girth. This result suggests that among illegals in the
United States, men may be more likely than women to be
employed in high-exertion occupations; stress, too, may be
a factor.
The results suggest other gender-based differences in BMI.
Racial background is statistically significant only for men, and
non-Hispanic Asian men are thinner than other immigrants.
Visa category, in contrast, is significant only for women, with
new immigrants who are sponsored by U.S. citizen spouses
significantly thinner and parents sponsored by U.S. citizen
offspring significantly heavier. As well, immigrant women with
conditional visas are statistically significantly thinner. Given
that 99 percent of the women with conditional visas are spouses
of U.S. citizens married for less than two years, this result
further suggests that, net of time in the United States, female
thinness is not only an asset in the marriage market but also a
further asset in the early years of marriage.
The continent dummies are jointly significant only for men,
but the country dummies are significant for both sexes. If we
rank-order the summed continent and country coefficients
(not shown) for all ten countries, the country with the highestgirth women is Guatemala, followed by El Salvador, Mexico,
Haiti, the Dominican Republic, Colombia, India, the
Philippines, China, and Vietnam. Among men, the rankordering of countries would begin with Mexico, followed by
the Dominican Republic, Guatemala, El Salvador, Colombia,
the Philippines, Haiti, India, China, and Vietnam.
Thus, among both women and men, and net of visa category
and time spent in the United States, immigrants from the
Western Hemisphere have the highest girth and immigrants
from Asia the lowest. This pattern immediately suggests the
possible operation of selection mechanisms; if thinness is
productive in the United States, then immigrants will be more
positively selected on thinness the greater the distance from the
United States. Of course, before exploring this question in
greater depth, it is important to assess BMI in the parent
populations of the origin countries. As well, it is useful to
consider the possible role of such mechanisms as the extent of
regulation in the origin country and the type of civil law, as
discussed by Cutler, Glaeser, and Shapiro (2003). It is
interesting to note that the highest-girth countries in our
sample tend to be countries with a French-origin civil law,
which runs counter to the hypothesis of Cutler, Glaeser, and
Shapiro. Of course, highly regulated countries, besides
producing girth-lowering effects via technology, also may
inhibit development of greater knowledge as well as techniques
for self-control (applying to the BMI context Vives’ [1522-40]
classic argument for gender equality). However, sharp
assessment of the effects of regulation and civil law origin

FRBNY Economic Policy Review / December 2005

145

requires careful characterization of all origin countries
represented in the sample, a task outside the scope of this paper
but an important one for future work.
Schooling achieves statistical significance for women but
not for men. Its effect is to reduce BMI, doing so nontrivially,
by .086 of a point for each year of schooling. Thus, a college
graduate will have BMI .688 lower than an immigrant who did
not go beyond the eighth grade.
Finally, estimation of the regression equations including a
binary variable for New York City does not in any specification
produce a statistically discernible New York City effect. Thus,
it appears that immigrants who settle there are neither thinner
nor fatter than other immigrants.

5.4. Health Change in the United States
Our work thus far includes several results pertinent to health
trajectory and the sources of health change. From the health
selection equation, we already know that among immigrants
already in the United States at the time of the first filing for legal
permanent residence, the combined effects of migration stress
and U.S. exposure are negative. Moreover, the health selection
equation also provides evidence of visa stress, because the effect
of adjusting from an illegal status is in all specifications larger
than the effect of adjusting from a legal status (Tables 7 and 8).
From the visa depression equation, we already know that
adjustee men are more likely to become depressed due to the
visa process than are new-arrival men, suggesting that visa
stress is more manageable in the origin country, at least for men
(Table 9). Finally, from the BMI equation we already know that
time in the United States increases girth (Table 10).
To assess further the sources of health change, we estimate
the determinants of the self-reported health change between the
most recent arrival “to live” in the United States and the baseline
interview. Recall that the vast majority of immigrants reported
no health change—76 percent of new-arrival immigrants and
64 percent of adjustee immigrants—with the proportions
whose health deteriorated registering 4 percent among new
arrivals and 14 percent among adjustees. There are two possible
reasons for the greater health deterioration among adjustees:
1) only the adjustees experienced visa stress in the interval, and
2) either/or both migration stress and U.S. exposure differ
qualitatively for LPRs and non-LPRs (especially LPR applicants
who may be in the United States illegally). To distinguish
among these effects, the health change equation includes not
only the two adjustee variables but also a variable for the time
elapsed between admission to LPR and the baseline interview.
Table 11 reports the results of the ordered-logit
specification. As shown, the two adjustee variables are jointly

146

Immigration, Health, and New York City

highly statistically significant among both women and men.
The coefficients differ, however, in that while the effect of
adjusting from a legal status is about the same for both sexes—
negative and of similar magnitude—the effect of adjusting
from an illegal status is negative for men but positive for
women. Two possible interpretations are that the deleterious
effect of illegal residence is larger for men than for women—
consistent with the effects in the selection equation (Tables 7
and 8) and with the visa depression effects (Table 9)—and that
women recover faster than men.
The effect of having a conditional visa is negative, as
expected, for both women and men, but is not statistically
significant, indicating that the effect is weak.
Finally, the effect of time since admission to LPR is positive,
statistically significant, and of a nontrivial magnitude among
men, but not statistically significant and close to zero among
women. Men’s health appears to increase with each passing day
as an LPR, net of health effects prior to obtaining LPR. Health
benefits from U.S. exposure outweigh the lingering or
dwindling effects of migration stress. Put differently, if
migration stress exerts a negative effect on health, then the pure
effect of U.S. exposure must be positive. However, if the

Table 11

Ordered-Logit Estimates of Determinants of Health
Change between Most Recent Arrival “to Live” and
First Interview after Admission to Legal Permanent
Residence: NIS-2003 Immigrants
Specification
Variable
Sex
Age at Round 1 interview
Age squared
Age joint test chi2 (2 df)
Adjustee, not illegal
Adjustee, illegal
Adjustee joint test chi2 (2 df)
Conditional visa
Time since admission to legal
permanent residence (years)
Number of observations
Log pseudolikelihood

All

Men

Women

-.0717
(1.65)
.0001827
-.492
4.09
-.312
-.0437
22.34
-.0777
(.76)
.207
(1.49)
7,660
-6060.89

—

—

-.0310
.000260
12.0
-.318
-.253
16.3
-.126
(.79)
.414
(2.40)
3,988
-3125.77

-.00617
.0000984
1.57
-.303
.145
18.1
-.0238
(.21)
-.00243
(.01)
4,232
-3365.28

Source: New Immigrant Survey, 2003 Cohort, Round 1.
Notes: Standard errors are corrected for heteroskedasticity due to
clustering by origin country; absolute values of asymptotic t-ratio appear
in parentheses under parameter estimates for numeric and binary
variables. Joint tests are reported for multiple-category categorical
variables. Cut-points are not shown.

migration gains experienced by new LPRs (including the
freedom gains) outweigh migration stress, then the effect of
U.S. exposure could be negative (and outweighed by the net
positive effect of migration “stress”).
We also re-estimated the equations including a binary
variable for the New York City immigrants. The coefficient is
not statistically significant in any specification, though it is
positive in all three.

6. Concluding Note
This paper explores immigrant health, emphasizing New York
City and using for the first time a large database in the final
stages of preparation for public release: Round 1 of the New
Immigrant Survey’s immigrant cohort of 2003. We formulated
a health model based on two related insights: 1) if migration is
stressful, then the appropriate time for assessing health
selectivity is at the time of the migration decision rather than at
the time of the actual migration, and 2) assessment of health
change subsequent to immigration should take into account
heterogeneity in the sources of health change and their timing.
The model distinguishes between the permanent and transitory
components of health and identifies three distinct sources of
change in the transitory component: visa stress, migration
stress, and U.S. exposure. Though not all the data required for
a thorough empirical assessment have become available, we
estimated several key components of the envisioned analyses.
To examine health selectivity, we relied on self-reported
health at the time of the initial filing for an immigrant visa;
we also looked at health during childhood (to guard against
contamination of health at the initial filing by changes in health
already in progress among immigrants residing in the United
States at the time of the migration decision). Our results
indicate that men are more positively selected for health than
women (though we cannot yet rule out differential reporting
styles by sex—future rounds of the longitudinal survey will
enable controlling for the style of reporting via fixed-effects
estimation). Diversity immigrants appear to be among the
most positively selected for health. Among men, legalization
immigrants are the most robust during childhood, but by the
time of the first filing, they rank lower on health than many of
their fellow immigrants. Health selectivity is responsive to skill
prices in the country of origin, but results for the effects of
distance are somewhat mixed.
Women are more likely than men to report experiencing
sadness or depression because of the visa process, and the
pattern of effects appears to differ across the sexes. Men with a
spouse or parent who is a U.S. citizen are less likely to

experience visa depression, but women do not appear to
receive a similar benefit from their kin. Men adjusting to legal
permanent residence in the United States are more likely to
experience visa depression than new-arrival immigrant men, a
finding that suggests that visa stress may be more manageable
in the origin country, but only for men—women’s propensity
to visa depression is not responsive to location.
We also examined body-mass index. Among both women
and men, time in the United States increases girth. It does so
differentially, however, depending on legal status prior to
admission to legal permanent residence. Among women, the
effect of time spent illegally is double the effect of time spent
legally, while for men the two effects are more similar, though
the pattern is the opposite of that found among women, with
time spent legally producing greater girth. This result suggests
that among illegals in the United States, men may be more
likely than women to be employed in high-exertion
occupations. Women admitted to legal permanent residence as
the spouses of U.S. citizens are substantially thinner than other
immigrants, and women married for less than two years are
even thinner, suggesting that female thinness is an asset not
only in the marriage market but also in the early years of
marriage.
The combined effects of migration stress and U.S. exposure
are negative in the time before admission to legal permanent
residence but non-negative afterwards and positive among
men. It thus would appear that the pure effect of U.S. exposure
is positive, at least after legal permanent residence and for men,
but we cannot rule out the possibility that migration gains—
such as freedom gains—are high, outweighing both migration
stress and the possible negative effect of U.S. exposure.
Finally, those immigrants who claim New York City as their
first home after admission to legal permanent residence are
more highly positively selected on health than their fellow
immigrants who settle elsewhere. Moreover, they are less likely
to report having experienced visa depression than other
immigrants. However, they are neither thinner nor fatter than
the rest of the cohort.
These results are obtained from a survey conducted soon
after admission to legal permanent residence. It will be
important to track change in the health of surveyed individuals
with the passage of time. Visa stress, already ended for most of
the cohort, will end for all with the removal of conditionality
restrictions. Migration stress presumably will run its course, if
it has not already done so for some cohort members. The
effects of U.S. exposure—positive or negative—will continue
for those in the cohort who remain in the United States, and it
will be possible to assess whether, and how, growth in U.S.specific skills enables immigrants to extract greater health
benefits and mitigate health hazards.

FRBNY Economic Policy Review / December 2005

147

Endnotes

1. The abbreviation LPR denotes both legal permanent resident and
legal permanent residence. The context should make clear whether
reference is to a person or to a status.
2. Immigration figures refer to the total non-IRCA (Immigration
Reform and Control Act) legalization number of new LPRs (see the
Immigration and Naturalization Service’s 2001 Yearbook, Table 4; its
earlier iterations; and 2004 data posted on the Yearbook of Immigration
Statistics website).
3. New York State is the second-leading state of intended residence for
new LPRs (after California) and the New York metropolitan area is the
second-leading metro area (after the Los Angeles-Long Beach area). At
the turn of the twentieth century, New York was the leading intended
state of residence, followed by Pennsylvania, Illinois, and
Massachusetts. (See the Immigration and Naturalization Service’s/
Office of Immigration Statistics’ Statistical Yearbooks for further detail
and the Dillingham Commission Reports for historical information.)
4. Following official terminology, we use “immigrant” interchangeably with “legal immigrant” and “legal permanent resident
alien.” Legal immigrants have the right to reside permanently in the
United States, to engage in most occupations, to sponsor the
immigration of certain relatives, and, after completing a residency
requirement, to become citizens of the United States. Besides legal
immigrants, there is a large set of legal nonimmigrants who have
temporary residence visas; legal temporary visas provide for legal
residence for a temporary period and for a specific purpose. Examples
of nonimmigrants include foreign students, tourists, and a variety of
workers, including representatives of foreign news media, computer
specialists, athletes, and entertainers. Additionally, there are
individuals in the United States illegally who qualify for neither legal
permanent residence nor legal temporary residence or who have
violated the terms of a legal temporary visa. Both legal temporary
residents and illegal migrants may be desirous of attaining legal
permanent residence.
5. A few other classes of individuals are also exempt from numerical
restriction, some as a permanent feature of U.S. law (such as American
Indians born in Canada and children born abroad to alien residents),
others under temporary provisions (such as the special three-year
program in effect in 1992-94 for spouses of aliens legalized under the
Immigration Reform and Control Act of 1986). Additionally, special
legislation has permitted refugees previously admitted with temporary
documents to adjust to permanent resident status outside the
numerical limitations.

148

Immigration, Health, and New York City

6. For a succinct description of U.S. visa allocation law, see the U.S.
Citizenship and Immigration Services’ and State Department’s
websites, in particular, the Office of Immigration Statistics’ Yearbook
of Immigration Statistics and the State Department’s Visa Bulletin. For
elaboration from a social science perspective, see Jasso, Rosenzweig,
and Smith (2000).
7. The number of persons admitted as refugees is set annually by the
President in consultation with Congress; the ceiling has fluctuated in
the range of 75,000 to 100,000. The diversity lottery program was
begun in fiscal year 1987 on a trial basis and made a part of U.S.
immigration law under provisions of the Immigration Act of 1990.
8. Registry provisions allow for the adjustment to LPR of persons who
have resided continuously in the United States since a given target
date; currently, that date is set at January 1, 1972. Cancellation of
removal, together with the kindred suspension of deportation
provisions in effect before 1997, similarly provide for adjustment
to LPR.
9. A small number of family-sponsored and employment-based
immigrants may self-petition. These include, in the case of family
visas, spouses and children of deceased or abusive U.S. citizens and
legal permanent residents, and, in the case of employment visas,
investors and individuals of great renown. For further detail, see the
requisite forms: Forms I-130, I-140, I-360, and I-526, available on the
U.S. Citizenship and Immigration Services’ website.
10. Additional “joint” sponsors may be brought in if the visa sponsor
cannot fulfill the support requirement alone. For details, see the I-864
affidavit of support package of forms on the U.S. Citizenship and
Immigration Services’ website.
11. Moreover, as we show, individuals subject to both visa stress and
migration stress may experience them at different times. For example,
consider adjustees who have spent many years in the United States as
legal nonimmigrants before applying for LPR; migration stress for
them may have ended long before the onset of visa stress.
12. Notice how such a study will require new vocabulary; the U.S.
citizen “newcomers” are not “immigrants” as that term is almost
universally used.
13. For a discussion of migration and visa stresses, see Kasl and
Berkman (1983) and Vega and Amaro (1994). Illustration of these
stresses is plentiful. For example, the website of an immigration law

Endnotes (Continued)

firm begins with the following description of visa stress: “Immigrating
to the United States is a complicated procedure that can cause
tremendous stress for the individual wishing to immigrate.
MacKenzie-Hughes, LLP is the area’s premier immigration law firm,
and we work hard to smooth the process and minimize the anxiety for
our clients” (<http://www.imm-usa.com>). And the stresses may be
even greater for illegal migrants, who must live partly in the shadows
and face threats of deportation. Other components of visa stress
include the constraints on international travel, which may cause
family hardships (U.S. Immigration and Naturalization Service 1992).
14. As noted above, for some categories of immigrants, the trajectory
would be somewhat different. For refugees, visa stress may end at the
time of the temporary (nonimmigrant) admission, while for
conditional immigrants (spouses of U.S. citizens who have been
married for less than two years, and investors), visa stress may not end
until removal of the conditionality restrictions two years later.
15. Among immigrants in the nationally representative New
Immigrant Survey Pilot who were employed in the United States at the
time of the baseline round and who had worked abroad within the past
ten years, earnings gains were substantial: on average, they were
$10,306 for men (a 68 percent increase) and $6,146 for women
(a 62 percent increase). (Gains are denominated in dollars based on
estimates of the country-specific purchasing power of the currencies
from the Penn International Comparisons Project [Summers and
Heston 1991].)
16. For elaboration of the relationship between income and health,
see Smith (1999).
17. Note that recent changes in the law, as well as the new climate in
the wake of the September 11 attacks, raise the possibility that visa
stress does not end until naturalization. Indeed, even with
naturalization, the immigrant is not completely safe, for unlike nativeborn citizens, an immigrant can be denaturalized and deported (for
cause). Further thought is needed in order to modify the model
presented in this paper to accommodate the possibility of lifelong,
albeit possibly mild, visa stress.

18. Note that among illegal migrants, a net positive combined effect of
migration stress and U.S. exposure would attenuate the decline, while
a net negative effect would exacerbate it.
19. It is not possible to insert a full set of country-specific fixed effects,
because a nontrivial number of countries (26 out of 168) are
represented by a single immigrant. Our solution is to include the
continent dummies plus ten country dummies.
20. Estimates of origin-country skill prices are based on recent work
that uses information on immigrant earnings in the last origincountry job before immigration and in the first U.S. job after
immigration, expressed in PPP-adjusted figures (Summers and
Heston 1991), together with country characteristics such as schooling
levels and school quality (based on Barro and Lee [1993]) and GDP
(Jasso and Rosenzweig 2005).
21. All descriptive statistics are based on weighted data, adjusting for
the over- and undersampling of the design.
22. Initial residence is the address to which new immigrants request
that their green card be mailed.
23. Indeed, the proportion who suffered harm in the origin country
was larger by almost 3 percentage points among men than among
women—8.3 percent versus 5.5 percent.
24. Such a result would echo the findings of sociologists and
psychologists a quarter-century ago on the multiple stresses associated
with both entering puberty and shifting to a new school at the same
time (Simmons and Blyth 1987).
25. It is illuminating to recall that Simmons and Blyth’s (1987) insight
into the effects of reaching puberty and transitioning to middle school
at the same time was also gender-specific.

FRBNY Economic Policy Review / December 2005

149

References

Barro, R., and J.-W. Lee. 1993. “International Comparisons of
Educational Attainment.” Journal of Monetary Economics 32,
no. 3 (December): 363-94.

Kasl, S. V., and L. Berkman. 1983. “Health Consequences of the
Experience of Migration.” Annual Review of Public Health 4:
69-90.

Cutler, D. M., E. L. Glaeser, and J. M. Shapiro. 2003. “Why Have
Americans Become More Obese?” Journal of Economic
Perspectives 17, no. 3 (summer): 93-118.

McDowell, M. A., C. D. Fryar, R. Hirsch, and C. L. Ogden. 2005.
“Anthropometric Reference Data for Children and Adults: U.S.
Population, 1999-2002.” Advance data from Vital and Health
Statistics, no. 361. Hyattsville, MD: National Center for Health
Statistics.

Frisbie, W. P., Y. Cho, and R. Hummer. 2001. “Immigration and the
Health of Asian and Pacific Islander Adults in the United States.”
American Journal of Epidemiology 153, no. 4 (February 15):
372-80.
Grossman, M. 1972. “On the Concept of Health Capital and the
Demand for Health.” Journal of Political Economy 80, no. 2
(March-April): 223-55.
Jasso, G. 2003. “Migration, Human Development, and the Lifecourse.”
In J. T. Mortimer and M. Shanahan, eds., Handbook of the
Lifecourse, 331-64. New York: Kluwer.
Jasso, G., D. S. Massey, M. R. Rosenzweig, and J. P. Smith. 2004.
“Immigrant Health—Selectivity and Acculturation.” In N. B.
Anderson, R. A. Bulatao, and B. Cohen, eds., Critical
Perspectives on Racial and Ethnic Differences in Health
in Late Life, 227-66. Washington, D.C.: National Academy Press.
———. Forthcoming. “The U.S. New Immigrant Survey: Overview
and Preliminary Results Based on the New-Immigrant Cohorts
of 1996 and 2003.” In B. Morgan and B. Nicholson, eds.,
Immigration Research and Statistics Service Workshop on
Longitudinal Surveys and Cross-Cultural Survey Design:
Workshop Proceedings, 29-46. London: Crown Publishing.
Jasso, G., and M. R. Rosenzweig. 2005. “Selection Criteria and the Skill
Composition of Immigrants: A Comparative Analysis of
Australian and U.S. Employment Immigration.” Paper presented
at the conference “Skilled Migration Today: Prospects, Problems,
and Policies,” cosponsored by the Columbia University Program
on Immigration Economics, the Russell Sage Foundation, and the
Council on Foreign Relations, New York City, March.
Jasso, G., M. R. Rosenzweig, and J. P. Smith. 2000. “The Changing Skill
of New Immigrants to the United States: Recent Trends and Their
Determinants.” In G. J. Borjas, ed., Issues in the Economics of
Immigration, 185-225. Chicago: University of Chicago Press.

150

Immigration, Health, and New York City

McEwen, B., and E. N. Lasley. 2002. The End of Stress as
We Know It. Washington, D.C.: The Joseph Henry Press.
Ogden, C. L., C. D. Fryar, M. D. Carroll, and K. M. Flegal. 2004. “Mean
Body Weight, Height, and Body-Mass Index, United States 19602002.” Advance data from Vital and Health Statistics, no. 347.
Hyattsville, MD: National Center for Health Statistics.
Riding, A. 2002. “Unlikely Star Captivates the French.” New York
Times, April 4.
Rumbaut, R. G., and J. Weeks. 1996. “Unraveling a Public Health
Enigma; Why Do Immigrants Experience Superior Prenatal Health
Outcomes?” Research in the Sociology of Health Care 13:
337-91.
Seeman, T. E., B. H. Singer, J. W. Rowe, R. I. Horwitz, and B. S. McEwen.
1997. “Price of Adaptation—Allostatic Load and Its Health
Consequences.” Archives of Internal Medicine 157, no. 19
(October 27): 2259-68.
Simmons, R. G., and D. A. Blyth. 1987. Moving into Adolescence:
The Impact of Pubertal Change and School Context.
New York: Aldine.
Smith, J. P. 1999. “Healthy Bodies and Thick Wallets: The Dual
Relation between Health and Economic Status.” Journal of
Economic Perspectives 13, no. 2 (spring): 145-66.
Strauss, J., and D. Thomas. 1998. “Health, Nutrition, and Economic
Development.” Journal of Economic Literature 36, no. 2
(June): 766-817.
Summers, R., and A. Heston. 1991. “The Penn World Table (Mark 5):
An Expanded Set of International Comparisons, 1950-1988.”
Quarterly Journal of Economics 106, no. 2 (May): 327-68.

References (Continued)

U.S. Department of Homeland Security. Various years. Yearbook of
Immigration Statistics. Washington, D.C.: U.S. Government
Printing Office.

Vega, W., and H. Amaro. 1994. “Latino Outlook: Good Health,
Uncertain Prognosis.” Annual Review of Public Health 15:
39-67.

U.S. Department of State. Various years. Visa Bulletin.

Vives, J. L. 1522-40 (1947). Obras Completas. Two volumes.
Translated from Latin into Spanish, annotated, and with an
introduction by L. Riber. Madrid, Spain: M. Aguilar. Especially
De Officio Mariti of 1528.

U.S. Immigration Commission. 1911. Reports of the Immigration
Commission. Forty-two volumes. Washington, D.C.: U.S.
Government Printing Office. (Also known as the Dillingham
Commission Report.)
U.S. Immigration and Naturalization Service. 1943-78. Annual
Report of the Immigration and Naturalization Service.
Washington, D.C.: U.S. Government Printing Office.
———. 1979-2001. Statistical Yearbook of the Immigration
and Naturalization Service. Washington, D.C.: U.S.
Government Printing Office.

Wallace, R. B., and R. Herzog. 1995. “Overview of the Health Measures
in the Health and Retirement Study.” Journal of Human
Resources 30, supplement: S84-S107.
Ware, J., A. Davies-Avery, and C. Donald. 1978. “General Health
Perceptions.” Publication no. R-1987/5-HEW. Santa Monica,
Calif.: RAND Corporation.

The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2005

151

Adriana Lleras-Muney

Commentary

uillermina Jasso, Douglas S. Massey, Mark R. Rosenzweig,
and James P. Smith use unique, newly collected data to
look at the health of immigrants and how it changes from the
time they decide to immigrate until they are established in the
United States. The authors surveyed a sample of new legal
immigrants in 2003 and collected detailed data on the legal type
of immigration. The new data also contain several health
indicators, including self-reported health status (SRHS) at
various stages of the immigration process. Finally, the authors’
data provide information on health changes during that
process.
A number of very interesting conclusions emerge from the
analysis. I will comment on several aspects of the paper,
starting with issues related to the health measures employed,
then moving on to the interpretation of the results, and ending
with some questions about the broader implications of this
research.
Although the new data improve greatly upon previous data,
it is worth noting that the three health measures used in the
paper—SRHS, body-mass index (BMI), and depression—have
some limitations. For health status, questions are asked both
about levels at various points in time and changes between time
periods. All of these outcomes are self-reported at a single point
in time, shortly after the person has obtained legal entry into
the United States.
Self-reported health status can be problematic because it is a
subjective measure. Even though it correlates well with more

G

“objective” measures of health, it is probably subject to many
cultural biases, which are likely to be important in this study,
given that immigrants come from various countries. SRHS may
be a better predictor of underlying health in some countries
and for some subgroups. For example, in the United States
SRHS is a better predictor of mortality for men than it is for
women (Case and Paxson 2005).
Another issue is that these health questions are asked in the
context of immigration. Several questions specifically ask the
interviewee to rate their health at a given time in the
immigration process. The depression question is asked with
respect to the visa process itself. Immigrants may therefore be
afraid of reporting themselves in poor health. Even if
immigrants are not consciously or directly afraid of answering
the health questions, their answers may be biased because of
the context in which they are asked. For instance, question
“D3” asks individuals whether their health has changed since
coming to live in the United States. Among those who have
recently been admitted to the country, this question is likely to
focus attention on a “happy” event (successful immigration);
thus, they may be more likely to report improvements in their
health. Similar biases have been reported elsewhere, for
example, when measuring well-being more generally
(Kahneman, Diener, and Schwarz 2003, ch. 4). Finally, it is
worth noting that even though the authors collected data on
health at various points in time, this information is
retrospective and thus subject to the usual recollection biases.

Adriana Lleras-Muney is an assistant professor of economics and public policy
at Princeton University.
<alleras@princeton.edu>

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

153

Although the empirical estimation is clear, I question the
authors’ interpretation of the results. The first question of
interest is the so-called health selection issue, namely, the
question of whether immigrants are more or less healthy than
the average person in their country of origin. It is not clear to
me how one can infer the health of immigrants relative to that
of their nonimmigrant counterparts without information on
the health of those who did not immigrate.
For example, the authors conclude that men are more
positively selected for health than are women. All the
estimations compare the health of men and women who
immigrated. What the results show is that immigrant men are
healthier than immigrant women (according to self-reported
health). But this finding does not imply that men are more
positively selected on health than are women. For instance, it is
well known that women are more likely to report themselves in
worse health than men in the United States and elsewhere (see,
for example, Case and Paxson [2005]). If in fact the health of
men is better than that of women in the country of origin
(suppose, for example, that men’s distribution is shifted to the
right), then immigrant women could be more positively
selected than immigrant men and be in worse health than
immigrant men. Similar arguments can be made when
interpreting the results on the health selection of immigrants
by type of visa.
There are additional difficulties in interpreting the findings,
due to the fact that immigrants come from different countries
and it is not possible to include country-fixed effects. To
continue with the example above, we note that it is possible that
men and women come from different countries and thus are
drawn from different health distributions. Without further
assumptions or additional data, it is unclear whether the
findings in the paper can shed light on the health selection
process.
At a broader level, it would be helpful to relate the specific
questions investigated—that is, what is immigrant health? and
how does it change over time?—to larger policy or academic
questions of interest. For example, why is it important to know

154

Commentary

whether immigrants are more or less healthy than their
nonimmigrant countrymen? Would the answer to this
question, for instance, inform immigration policy? If so, how?
There could be many reasons why the selection issue is of
interest, but these are not stated.
Similarly, it would be interesting to know why it is
important to understand the trajectory of immigrant health.
One reason mentioned in the paper is that failure to
understand the trajectory of health during migration may lead
to erroneous conclusions about the health selection process:
because of transitory shocks to health during the immigration
process, measures of immigrant health at a given point in time
may be biased. However, given that the survey collects data on
health prior to immigration and is therefore subject to this bias,
more needs to be said about why the health trajectory itself is of
interest. For example, do we want to provide special health
services to particular immigrants during the immigration
period? Do we want to inform them about how their health
may suffer throughout the process?
An interesting question that this work starts to address is the
assimilation question, namely, does the health of immigrants
improve or decline upon reaching the United States? The
authors report that for all immigrants, BMI increases with time
spent in the United States. But the implications of this finding
are not clear. It is not possible to determine whether BMI is
increasing because of the various changes in an immigrant’s
life, including changes in jobs and earnings (which may have
been similar in the country of origin), or because of the
environment in which the immigrant lives. The environment
(which includes, for example, pollution and eating habits) may
affect immigrants and natives alike. In order to understand
better the mechanisms at work, one has to compare
immigrants with natives.
Jasso et al. use new data to begin answering an ambitious set
of questions associated with immigrant health. Our
understanding of many of these questions will certainly
improve because of the extraordinarily detailed data presented
by the authors.

References

Case, A., and C. Paxson. 2005. “Sex Differences in Morbidity and
Mortality.” Demography 42, no. 2 (May): 189-214.
Kahneman, D., E. Diener, and N. Schwarz, eds. 2003. Well-Being:
Foundations of Hedonic Psychology. New York:
Russell Sage Foundation.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2005

155

Amy Ellen Schwartz and Leanna Stiefel

Public Education in the
Dynamic City: Lessons
from New York City
1. Introduction

T

he plight of urban schools and their failure to educate
students adequately and efficiently have occupied the
national discussion of public schools in America over the past
quarter-century. While there is little doubt that failing schools
exist in rural and suburban locations, the image of city school
systems as underfinanced, inefficient, inequitable, and
burdened by students with overwhelming needs is particularly
well entrenched in the modern American psyche.
As the largest school district in the nation, New York City
attracts particular attention to its problems. To some extent,
this image reflects realities. New York City school children, like
many urban students around the country, are more likely to be
poor, nonwhite, and immigrant, with limited English skills and
greater instability in their schooling. Moreover, the new waves
of immigrants from around the world bring students with a
formidable array of backgrounds, language skills, and special
needs. The resulting changes in the student body pose special
challenges for schools. At the same time, despite a decade of
school finance litigation and reform, New York continues to
have trouble affording the class sizes, highly qualified teachers,
and other resources that its suburban neighbors enjoy. Finally,
there is evidence of continuing segregation and disparities in
performance between students of different races and ethnicities.
Nonetheless, not all the news is bad. As we describe in detail,
our work on New York City’s public schools—which includes
extensive research on immigrant children—and our separate
Amy Ellen Schwartz is a professor of public policy, education, and economics
and Leanna Stiefel a professor of economics at New York University.
<amy.schwartz@nyu.edu>
<leanna.stiefel@nyu.edu>

work on school reform offer several reasons for optimism.
First, immigrant students, who might be viewed as among
those most seriously at risk of failure, are doing quite well. Our
research suggests that although immigrants are somewhat
segregated from the native born, this factor has little impact on
the resources available in the schools they attend. Even more,
immigrants in elementary and middle schools earn higher
scores on average than do the native-born students who are
otherwise similar to them, and the “immigrant advantage”
increases over time, perhaps following the students’
acclimation or acquisition of English language skills.
Second, the school system is changing and not at all static.
Each school year sees new schools open and old ones close,
reorganization and reform of existing schools, and changes in
curriculum, governance, and budgeting procedures, among
other experiments. Whether these changes lead to
improvements in test scores, more efficient use of resources, or
greater equity is not always clear, but any notion that the
system is intransigent and static seems inapt.
Third, advances in methods and the availability of data
combined with increased public pressure for accountability
have led to improvements in the quantity and quality of
evaluations of the various reforms and a new emphasis on
evidence to guide decision making. In some ways, New York
City has been at the forefront of this movement by tracking
expenditures at the school level, which allows for analysis of
cost-effectiveness, and providing student-level data to
researchers working to evaluate reforms in its schools.
The views expressed are those of the authors and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

157

Finally, some reforms and experiments are yielding positive,
if modest, results. For example, evidence suggests that the first
wave of small high schools created in the mid-1990s has been
more successful at getting students to graduation without a
significantly higher cost per graduate. In addition, budgeting
reform introduced by Rudy Crew, the former chancellor of
New York City’s public schools, and other whole-school
reforms also seem to be yielding small positive effects on
student test scores.
Our paper discusses all of these issues in greater detail,
drawing lessons for urban schools in the conclusion.

2. New York City Public School
Children
As the largest school district in the nation, New York City
educates more than 1.1 million students in roughly 1,300
public schools, with a student population that is diverse and
challenging. To illustrate, we present the third-grade class of
2000-01 (Table 1, panel A). This cohort included roughly
72,000 native-born students, more than 33 percent of whom
are black, nearly 40 percent Hispanic, roughly 9 percent Asian,
and 14 percent white. Poverty is alarmingly common. More
than 75 percent of the students are poor (as measured by
eligibility for free lunch) and another 8 percent are near poor
(as measured by eligibility for reduced-price lunch). Further,
more than 33 percent of the students come from homes in
which English is not the primary language and 5 percent have
sufficiently limited English skills to be eligible for English as a
Second Language or bilingual-education services.
At the same time, this cohort includes more than 10,000
students born outside the United States (hereafter referred to as
immigrant or foreign-born students). That is, roughly one out
of every eight third graders was foreign born in 2000-01. (Note,
however, that because many of the native-born students are
themselves children of immigrants, these figures in some sense
understate the impact of immigrants on the public schools.)
Immigrants differ noticeably from the native born in racial
composition: more than 25 percent are Asian, less than
20 percent are black, 36 percent are Hispanic, and 18 percent
are white. An even greater share of the foreign born are poor or
near poor—in fact, only about 10 percent of foreign-born
students are not poor. As one might expect, immigrant
students are far more likely to come from homes in which
English is not the primary language (more than 75 percent) and
to be limited-English-proficient, or LEP (more than 25 percent).
New York City’s immigrant population is extraordinarily

158

Public Education in the Dynamic City

diverse, hailing from more than 200 countries and speaking
more than 160 languages and dialects. While some arrive with
strong academic backgrounds, rich and stable home lives, and
poised for success in American schools, others arrive less well
prepared, needing remediation, supplemental support, and
special attention.

Table 1

Selected Characteristics of Third-Grade Students
by Nativity Status
Native Born

Foreign Born

0.089
0.369
0.399
0.142
0.499
0.363
0.775
0.080
0.056

0.277
0.184
0.359
0.180
0.495
0.778
0.825
0.074
0.269

-0.014
0.917

0.143
0.664

-0.012
0.938
71,931

0.118
0.765
10,428

0.072
0.382
0.374
0.171
0.504
0.371
0.776
0.069
0.099

0.231
0.188
0.389
0.192
0.496
0.739
0.812
0.073
0.354

0.006
0.981

-0.040
0.786

0.006
0.985
62,513

-0.035
0.874
10,845

Panel A: 2000-01
Percentage of students who are
Asian
Black
Hispanic
White
Female
Non-English-speaking at home
Eligible for free lunch
Eligible for reduced-price lunch
Limited-English-proficient
Test data
Reading
Mean score
Percentage taking test
Math
Mean score
Percentage taking test
Number of students
Panel B: 1995-96
Percentage of students who are
Asian
Black
Hispanic
White
Female
Non-English-speaking at home
Eligible for free lunch
Eligible for reduced-price lunch
Limited-English-proficient
Test data
Reading
Mean score
Percentage taking test
Math
Mean score
Percentage taking test
Number of students
Source: Authors’ calculations.

Notice, however, that many foreign-born students do quite
well in school. Panel A of Table 1 reports the mean
performance on standardized tests in reading and math. (For
comparison purposes, these scores have been normalized for all
students in a grade to produce a mean of zero and a standard
deviation of 1.) Foreign-born students with sufficient English
skills to take the standardized tests perform on average at a
higher level than the native born. (Of course, many students do
not take the tests, making it difficult to disentangle causality
here. We return to this issue shortly.)
Finally, note that the student body changes over time,
driven by differences in immigrants as well as in the native
born. Consider the differences between this cohort and a
similar cohort five years earlier. Panel B of Table 1 shows the
characteristics of the third-grade cohort of 1995-96. Notice that
there are considerably fewer native-born students in this
cohort—nearly 9,000—but the number of immigrants is
roughly constant. Thus, immigrants are even more important
in this group. Further, the racial composition is different—
fewer Asians, more Hispanics, more whites. While poverty
rates are roughly similar, limited English proficiency is
significantly more prevalent in both the native- and foreignborn populations in 1995-96. Finally, the proportion of
students taking standardized tests is considerably higher in the
earlier period, and the disparities in performance between the
native- and foreign-born populations are almost zero.
This comparison of cohorts, however, ignores the change
that occurs within a cohort over time, and our analysis suggests
that this intra-cohort change is important. To illustrate, we
consider the change in the third-grade cohort of 1995-96 by
its eighth-grade year, 2000-01. As Table 2 shows, more than
20 percent of the students had left the New York City public
school system (attritted) either to attend other public schools
or private schools, and the attritters are significantly less likely
to be black and more likely to be white. Further, the attritters
are somewhat less likely to be poor, but more likely to be near
poor, and they perform better on both reading and math tests
than do continuing students.
Even more important than those who left are those who
entered. Consider the eighth-grade cohort of 2000-01. Table 3
distinguishes between two groups of students in the cohort—
those who entered in third grade or before and therefore were
part of the third-grade cohort of 1995-96, and those who
entered after third grade.1 All told, nearly 33 percent of the
eighth graders were not attending third grade in any public
school in New York City five years earlier. The fraction entering
after kindergarten is undoubtedly higher. Interestingly, while
the differences between the attritters and continuing students
are relatively modest, the differences between the early and late
entrants are stark. Nearly 45 percent of students entering after
third grade are foreign born, compared with 15 percent of the

Table 2

Mean Characteristics of Third-Grade Students
by Attrition Status, 1995-96
Continuing
Students

Attritting
Students

0.097
0.360
0.376
0.166
0.505
0.424
0.789
0.068
0.135
0.861

0.091
0.326
0.376
0.204
0.498
0.429
0.740
0.076
0.139
0.828

Test data
Mean score in reading
Mean score in math

-0.015
-0.010

0.076
0.053

Number of students
Percentage of all third graders in 1995-96

56,463
77.8

16,142
22.2

Percentage of students who are
Asian
Black
Hispanic
White
Female
Non-English-speaking at home
Eligible for free lunch
Eligible for reduced-price lunch
Limited-English-proficient
Native born

Source: Authors’ calculations.
Notes: Continuing students are those registered in third grade in both
1995-96 and 2000-01. Students need not be continuously enrolled.

Table 3

Selected Means for Eighth-Grade Students
by Entrance Status, 2000-01
Entered by Entered after
Third Grade Third Grade
Percentage of students who are
Asian
Black
Hispanic
White
Female
Non-English-speaking at home
Eligible for free lunch
Eligible for reduced-price lunch
Limited-English-proficient
Native born

0.113
0.342
0.354
0.189
0.523
0.426
0.725
0.097
0.027
0.854

0.142
0.354
0.374
0.126
0.458
0.535
0.821
0.077
0.245
0.553

Test data
Mean score in reading
Mean score in math

0.111
0.119

-0.361
-0.340

46,566
68.202

21,711
31.798

Number of students
Percentage of all eighth graders in 2000-01
Source: Authors’ calculations.

Notes: The table presents 2000-01 means for New Your City public school
students enrolled in the eighth grade in 2000-01 and enrolled in the third
grade in 1995-96. Students need not be continuously enrolled.

FRBNY Economic Policy Review / December 2005

159

early entrants. Nearly 25 percent of the late entrants are
limited-English-proficient in the eighth grade, compared with
only 3 percent of the early entrants. Late entrants are
significantly more likely to be poor and significantly less likely
to be white. Finally, the late entrants score substantially lower
on the standardized tests than do the early entrants.
The implications for policy are real. The success or failure of
the public schools in delivering an eighth-grade class ready for
high school hinges, in no small way, on the performance of
students educated by schools outside the New York City public
school system and, among those, a substantial number of
schools in other countries. Put differently, this implies that
there may be limits to the extent to which improving early
childhood education, for example, can improve the highschool readiness of students at the end of middle school—an
important goal for educators and parents. More generally, this
turnover suggests that the implementation of school
accountability for student performance may have to be done in
a way that recognizes the particular difficulties of educating a
student body that has high levels of turnover.

homogenously black, for example, and others that have very
few blacks. The same can be said for many groups.
Just as the student population is dynamic and changing, so
too are the public schools. To some extent, these changes reflect
policy or economic changes affecting a wide range of schools.
Labor market returns to education are ever-increasing,
heightening the pressure to prepare students for the labor
market and college. There has been an increasing focus on test
scores and accountability across the nation, exemplified by the
terms of the federal No Child Left Behind Act, which requires
the tracking of test scores and gains in various ways. New York
State has imposed its own set of accountability reforms,
including high-stakes tests in fourth grade and eighth grade
and rising standards for Regents high-school diplomas. The
possibility of significant changes in school finance looms, as the
state negotiates the implications of the Campaign for Fiscal
Equity lawsuit, and policymakers and educators consider
where the money will come from and how to spend it.

Table 4

3. New York City Public Schools
Just as New York City’s public school students are diverse, so
too are the city’s schools. To illustrate, we present descriptive
statistics for 865 elementary and middle schools in 2000-01
(Table 4). The average elementary or middle school enrolled
roughly 830 students and spent nearly $11,000 per pupil, about
$6,200 of which was for expenses other than teachers
(including administrators, support staff, books, and materials).
The teacher-pupil ratio averaged .079, or roughly one teacher
for every thirteen students. On average, about 80 percent of
these teachers were licensed and permanently assigned, more
than 70 percent had master’s degrees, nearly 60 percent had
more than two years in their current school, and more than
50 percent had more than five years of experience. At the same
time, the standard deviations on nearly all of these variables
are substantial. While some schools enroll more than 1,000
students, others have only a couple of hundred. In some
schools, virtually all teachers are licensed, while others have
relatively few with licenses. School spending varies widely,
driven by differences in teachers and the needs of students, as
we discuss in greater detail.
Equally important is the variation in the characteristics of
students. While the average school is roughly 16 percent white,
the standard deviation is 23. Similar variability is seen in the
other race groups. New York City public schools run the
spectrum of racial diversity—there are schools that are virtually

160

Public Education in the Dynamic City

Mean Characteristics of Elementary
and Middle Schools, 2000-01
Variable
Total per-pupil expenditures
Nonteacher per-pupil expenditures
Teacher-pupil ratio
Average school enrollment
Percentage of teachers
Licensed and permanently assigned
With master’s degree
With more than two years of experience
With more than five years of experience
Percentage of students in schools
Female
White
Black
Hispanic
Asian and other
Eligible for free lunch
Eligible for reduced-price lunch
Native
Non-English-speaking at home

Mean

Standard
Deviation

$10,907
$6,183
0.079
829.7

$3,169
$2,102
0.020
402.3

80.9
72.6
59.1
51.5

17.8
15.7
19.2
15.1

49.2
16.5
35.8
36.6
11.1
72.3
7.5
86.0
40.4

3.2
23.1
30.7
25.9
15.3
23.9
5.0
10.0
24.5

Source: Authors’ calculations.
Notes: The sample is 865 schools with students in either fifth or eighth
grade (573 have only fifth graders; 194 have only eighth graders; 98 have
both fifth and eighth graders). Schools serving only special-education
students are excluded. Eligibility for free lunch is calculated only for
students with nonmissing data. Native students are those born on
U.S. soil.

The various pressures from within and from outside have
yielded many changes in the New York City schools. For
example, consider recent governance changes. Just a couple of
years ago, New York City Mayor Michael Bloomberg gained
control over the school district, earning the power to appoint
the chancellor and assuming the responsibility for the district’s
performance. Chancellor Joel Klein quickly implemented a
reorganization of the governance of the schools. The thirty-two
community school districts, which had primary responsibility
for elementary and middle schools in the city, were reorganized
into ten considerably larger instructional regions. Curriculum
reform soon followed along with changes in third- and fifthgrade promotion policies. The effort to build new small schools
continued, following the belief that small schools are more
successful, funded in part by the Gates Foundation. These are
just some examples of the many changes affecting public
education in New York City. Others include charter schools,
vouchers, reforms to the high-school articulation process,
teacher certification, and principal training.
Change and reform, however, are not new. As shown in
Table 5, the period between 1996-97 and 2002-03 witnessed
quite a bit of turnover in the schools. Every year in that period
saw a set of schools close and an even larger set of schools open.
By the end of the period, there were roughly 10 percent more
schools than there were six years earlier and, of the 1,160
schools operating in 2002-03, roughly 15 percent had opened
in the past five years. (These statistics exclude adult-education
schools and special-education schools, among others.) Wholeschool reforms and governance reforms were implemented
during the terms of many previous chancellors, including Crew
and Harold Levy.

Table 5

Schools Opening and Closing by Year

1996-97
1997-98
1998-99
1999-2000
2000-01
2001-02
2002-03

Closed

Opened

Operating
Schools

—
9
8
18
9
25
13

—
30
11
59
23
24
43

1,052
1,073
1,076
1,117
1,131
1,130
1,160

Source: Authors’ calculations.
Notes: Closed is defined as no longer operating during that year. Citywide
special-education schools, schools in prisons, adult-education schools,
nonpublic schools, and community-based-organization schools are
excluded from the sample. Only schools with nonzero registration are
included.

In general, the motivation for the various reforms and
changes can be characterized as aiming to improve the
efficiency of resource use and/or the performance either of
students overall or particular groups of students. Of course, not
all changes and reforms are effective, and it is crucial to
consider whether these programs are efficient. Doing so,
however, is far from straightforward.

4. Is Change Good?
Not all change is good, and distinguishing between which
innovations are successful and worthy of replication and which
are not is crucial to improving schools. Unfortunately,
distinguishing between “what works” and “what doesn’t work”
in education is particularly complicated compared with doing
so in other settings and, while there has been woefully little
attention paid to this in the past, there is quite a bit of attention
being paid right now. For example, the U.S. Department of
Education created and funds the What Works Clearinghouse
(WWC) to provide answers and disseminate findings by
reviewing and vetting evaluations based upon their scientific
validity and reliability.2
To the economist, the fundamental criteria for evaluating
reforms center on their effect on equity and efficiency, which
must be carefully defined to be useful. Even then, applying
these criteria requires confronting and resolving a host of
conceptual and practical difficulties. Efficiency requires that
resources be deployed in such a fashion that the greatest
amount of output is produced with the inputs used. Figuring
out what works requires assessing whether a reform or
innovation had an effect on outputs and figuring out what
works best requires an understanding of the impact on cost.
Thus, we need to define and measure carefully changes in
outputs, changes in inputs, and ultimately the relationship
between these—the production function for education. (See
Stiefel et al. [2005] for more on measuring school efficiency.)
In an ideal world, there is broad consensus on the
appropriate measures of efficiency as well as abundant data
tracking these measures across students, schools, and school
districts over time. In addition, new programs and reforms are
best implemented using randomized experiments that allow us
to disentangle easily the causal relationship in the data.
Unfortunately, these conditions are rarely met. Data on school
resources are rarely tracked at the school level. There is only
grudging consent to the use of test scores to measure output
and little consensus on which subjects and what types of scores
to use. (While No Child Left Behind has put the federal
emphasis on a set of tests and statistics, it is not at all clear that

FRBNY Economic Policy Review / December 2005

161

these will be broadly accepted by state education departments
and school districts around the country.) However,
administrative data on individual students that can be used to
track their performance over time are increasingly available,
and there are some jurisdictions in which expenditures and
other variables are measured at the school level. New York City
is one of these.
While few reforms are intentionally adopted in a
randomized fashion, the complexity of the New York City
system has often meant that reforms are not universally
implemented at one moment and there is often some
randomness in timing and/or implementation of the reforms,
creating opportunities to disentangle causality in the impact
estimate.
Another important criterion for assessing reforms revolves
around equity, and again there are both practical and
conceptual issues. If we agree that our concern is the equitable
treatment of students (compared with, say, teachers), then we
need to resolve several issues. First, equitable treatment for
which students—low-performing, poor, black, Hispanic, girls
or boys, disabled, English language learners?—to name just a
few. While the ideal reform affects all equally, it is rarely, if ever,
the case.
Second, how do we measure improvement in equity? What
sort of measure is appropriate? If greater equity is achieved
when a reform reduces disparities in performance between two
groups—say, between blacks and whites—then it is almost
certainly the case that the reform delivers greater improvements in performance for one group than the other. Put
differently, are we looking for equity in levels or in gains?
Third, we need to decide whether to focus attention on the
equity in the distribution of resources (inputs measured in
dollars, teacher counts, teacher qualifications, say) or in the
distribution of outputs (such as test scores or graduation
outcomes), as we have implicitly assumed in our earlier
discussion. Finally, there are the usual difficulties inherent in
distributional analyses—alternative measures are available and
they are not always consistent. (See Berne and Stiefel [1984] for
more on equity measurement in education.)
Despite these difficulties, recent experience indicates that
progress is being made in evaluating school reforms, in
assessing changes in both efficiency and equity. New York City
in many ways is an excellent “laboratory” for studying
schools—the student body is large and diverse; the many
schools vary widely in size, composition, organization, and the
like; and schools change over time. Further, the Department of
Education collects (and has been willing to provide to
researchers) detailed data on students, including test scores,
socio-demographics, language skills, and nativity, along with

162

Public Education in the Dynamic City

comprehensive school-level expenditure data. Thus, we have
been able to explore in some detail the treatment and
experience of immigrant students in the New York City public
schools and to assess the effects of recent reform efforts. We
now turn to a brief discussion of some examples from our
research on New York City’s public school students.

5. The Education of Immigrant
Students
How well immigrant students fare in New York City public
schools reflects, in large part, how well the school system
responds to change. New countries, new languages, and new
challenges are the norm, rather than the exception. One
particular concern regarding immigrants derives from their
propensity to settle in communities with others from their own
country. This strong link between residential location and
elementary school attendance may well mean that immigrant
children will go to segregated schools with few native-born
students, which carries with it concerns about access to social
networks, peers, English language acquisition, and, to the
extent that immigrants are less active politically, about the
prospect of creating school communities that are insufficiently
funded.
Measures of exposure and isolation show that this concern
may be misplaced (Table 6). In fact, immigrants are not very
segregated at all. As of 1998-99, the typical elementary or
middle-school student went to a school in which 76.3 percent
of his or her schoolmates are native born. The isolation index
of .237 is not very high either. To be sure, some specific groups
of foreign born, such as those from the Dominican Republic,
the former Soviet Union, or China, are more highly isolated—
their own-group isolation indexes are 10.5 percent, 17.5 percent,
and 13.4 percent, respectively. However, certainly compared
with the racial segregation of nonwhite (at 90.4 percent) or
free- and reduced-lunch-eligible (also 90.4 percent) students,
this level of segregation is mild.
Of course, the native-born peers with whom immigrant
students attend school may be children of immigrants
themselves, leaving open the possibility that their schools will
be less well supported than other schools. Immigrant
advocates, for example, often do not distinguish between the
foreign born and children of immigrant communities, and
claim that immigrants do not receive their fair share of
spending. At the same time, parents of native-born students
wonder if immigrants are taking resources from their children.
How do resources vary with the representation of immigrants?

Table 6

Exposure of New York City Public Elementary and
Middle-School Students, Immigrant and Native Born,
1998-99
Exposure
to Native
Born

Isolation
Index

Percentage
of Total
Students

Native born

0.854

0.854

0.839

Foreign born
Recent immigrant
Limited English skills

0.763
0.767
0.750

0.237
0.117
0.106

0.161
0.073
0.050

Born in Dominican Republic
Born in Mexico, Central America,
or Spanish South America
Born in other Caribbean
Born in former Soviet Union
Born in South Asia
Born in China, Taiwan, or
Hong Kong

0.803

0.105

0.031

0.758
0.811
0.669
0.723

0.071
0.093
0.175
0.066

0.026
0.024
0.017
0.016

0.696

0.134

0.012

0.841

0.904

0.844

0.836

0.904

0.866

Nonwhite
Eligible for free or
reduced-price lunch
Source: Ellen et al. (2002, Table 4).

Our examination of the distribution of spending suggests
that immigrant students receive the same level of most school
resources that native-born students receive. To be specific, we
estimated school-level expenditure regressions for New York
City elementary and middle schools in the late 1990s. In these
models, we controlled for features of the school population
that traditionally garner more resources for schools—the
percentage of poor, special-education, LEP students, for
example—and found that, ceteris paribus, the percentage of
immigrants in the schools rarely affects the per-pupil amount
devoted to students as a whole. The representation of
immigrants was significant only for nonclassroom
expenditures and the percentage of teachers who have
permanent teaching certification, and these work in opposite
directions. Put differently, immigrants seem to draw resources
in just the same way that native-born students do—because of
their poverty status, English proficiency status, and specialeducation needs. Thus, we conclude that there is no “smoking
gun” suggesting that immigrants are treated inequitably.
Of course, this equity concern about resources is closely tied
to the question of how immigrants perform in the New York
City schools. That immigrants receive resources equal to those
of similar, native-born students may or may not be an efficient
use of resources, depending on how immigrants do in school.

Performance significantly below that of the native born might
suggest that resources would be more efficiently used by
redistributing toward immigrants. Thus, we consider the
academic performance of immigrant students; in brief, our
findings suggest that this concern is unnecessary.
We estimate the nativity gap in performance—the
difference in average test scores of foreign- and native-born
students—for different grades and years in the late 1990s, using
various specifications of a regression model to control for other
differences between immigrant and native-born students.
Table 7 presents representative results for fifth- and eighthgrade reading and math test scores. (As before, test scores have
been normalized to a mean of zero and a standard deviation
of 1.) Column 1 shows the unadjusted mean differences in
performance; column 2 shows the size of the nativity gap once
we control for the previous year’s performance (a value-added
specification); column 3 shows the estimated nativity gap once
we include a full set of control variables. On the whole, the
evidence suggests that foreign-born students outperform
native-born students, ceteris paribus.
Of course, while foreign-born students might do better on
the whole, there may well be significant differences among the
immigrants masked in the overall category. As Table 8 shows,
there are marked differences in the characteristics of students
from different regions of the world. For example, while nearly
all of the Dominican students are poor, poverty is less common
among Europeans. Again, while 25 percent of the Dominican
students are LEP, only 1 percent of Caribbean students require
English remediation. Further, special-education rates differ
significantly across regions. Finally, there are differences in the
length of time students have attended the New York City public
schools. While native-born students have been enrolled for
nearly five years, which is consistent with kindergarten entry,
foreign-born students average more than one year less in the
schools. While students from some regions differ marginally
from the native born, students from other regions are
significantly more recent additions. Do these differences
translate into differences in performance across regions? As
Table 9 illustrates, we find that once we control for differences
in the underlying characteristics of students, there are relatively
few differences across regions, although Russian and Chinese
students perform particularly well. (We present results for
reading tests; similar results are obtained for math.)
Notice, however, that these cross-sectional snapshots may
be misleading. Suarez-Orozco (2001) argues that “among
immigrants today, length of residence in the United States
seems associated with declining health, school achievement,
and aspirations.” This argument is shared by other researchers.
While the hypothesis that the superior academic performance

FRBNY Economic Policy Review / December 2005

163

Table 7

Selected Regression Results for Reading and Math Tests, Foreign-Born Students by Grade and Year
Reading
(1)
Fifth grade, 1997-98
Foreign born
Number of observations
R2
Fifth grade, 2000-01
Foreign born
Number of observations
R2
Eighth grade, 1997-98
Foreign born
Number of observations
R2
Eighth grade, 2000-01
Foreign born
Number of observations
R2
Prior-year test score
Additional variables

(2)

Math
(3)

(4)

(5)

(6)

0.122***
(0.019)
64,971
0.00

0.126***
(0.010)
64,971
0.54

0.070***
(0.010)
64,971
0.57

0.061***
(0.022)
66,629
0.00

0.105***
(0.010)
66,629
0.58

0.050***
(0.009)
66,629
0.60

0.083***
(0.018)
71,141
0.00

0.089***
(0.010)
71,141
0.47

0.020
(0.018)
71,141
0.45

0.115***
(0.021)
72,509
0.00

0.108***
(0.012)
72,509
0.55

0.043***
(0.014)
72,509
0.55

-0.004
(0.024)
57,465
0.00

0.037***
(0.010)
57,465
0.58

0.023*
(0.014)
57,465
0.60

-0.029
(0.028)
59,749
0.00

0.062***
(0.012)
59,749
0.56

0.026*
(0.014)
59,749
0.58

0.014
(0.027)
57,152
0.00

0.058***
(0.013)
57,152
0.54

0.035***
(0.013)
57,152
0.59

0.099***
(0.027)
59,024
0.00

0.148***
(0.013)
59,024
0.59

0.065***
(0.013)
59,024
0.62

No
No

Yes
No

Yes
Yes

No
No

Yes
No

Yes
Yes

Sources: Schwartz and Stiefel (forthcoming, Table 5); authors’ calculations.
Notes: The sample is New York City public school students who took a reading or math test. Robust standard errors are in parentheses. Demographic
characteristics include age and a set of dummies indicating eligibility for free lunch, eligibility for reduced-price lunch, sex, race, and the existence of missing
data. Educational characteristics are language other than English frequently spoken at home, took the language assessment battery (LAB), percentile on the
LAB, scored at or below the 40th percentile on the LAB, part-time special-education participation, prior-year test score, and whether the student took the test
in the prior year. School resources are nonteacher expenditures (in thousands of dollars), teacher-pupil ratio, percentage of teachers with more than five years
of experience, percentage of teachers with more than two years in the school, percentage of teachers licensed and permanently assigned, percentage of
teachers with a master’s degree, enrollment (in hundreds), and dummy variables indicating that teacher characteristic and expenditure data are nonmissing.
Cohort variables are dummies for the number of years in the New York City public schools. The teacher-pupil ratio is instrumented with the prior-year
enrollment and enrollment squared.
***Statistically significant at the 10 percent level.
***Statistically significant at the 5 percent level.
***Statistically significant at the 1 percent level.

164

Public Education in the Dynamic City

Table 8

Characteristics of Fifth- and Eighth-Grade Students by Region, 1997-98

Region

Percentage
Limited-EnglishPercentage
Percentage
Proficient
Special Education
Female

Years in
New York City
Public Schools

Number
of Students

Percentage Eligible
for Free Lunch

Percentage Eligible for
Reduced-Price Lunch

206
1,911
459
1,409
471
329
729
1,296
1,127
638
219
252

77.7
83.5
69.1
94.7
45.2
59.0
84.1
86.1
56.8
71.3
68.0
56.3

7.3
6.4
10.2
2.3
21.0
14.9
8.2
5.6
10.6
10.2
7.3
12.7

6.3
1.0
6.5
25.1
2.8
6.1
0.0
18.5
2.0
5.8
3.2
3.6

4.9
5.5
3.3
5.3
4.0
3.0
6.6
6.9
4.7
5.8
8.2
6.7

51.0
54.0
50.1
49.7
50.5
51.7
52.8
46.6
49.1
45.9
51.1
46.0

2.7
3.1
4.1
4.4
3.9
3.7
3.1
4.4
3.9
4.0
4.3
3.8

9,046
55,925

76.6
73.5

8.1
7.0

8.2
4.3

5.5
9.7

50.2
51.1

3.8
4.9

224
2,890
678
1,667
665
382
956
1,784
1,230
693
257
266

70.1
74.7
66.4
92.2
46.6
61.3
77.2
84.1
49.8
70.9
68.5
59.4

8.9
7.4
12.4
2.1
17.4
12.6
7.6
5.5
13.4
11.4
5.4
10.2

6.3
2.0
10.8
35.5
7.4
6.0
0.5
21.2
2.0
8.1
5.1
4.1

2.2
4.6
3.7
4.4
2.4
3.4
4.0
5.3
2.5
2.0
4.7
6.4

49.6
52.7
49.0
49.2
52.3
52.9
53.2
46.9
48.9
47.0
37.0
55.6

3.4
4.2
6.0
6.0
5.3
5.2
4.3
6.3
4.7
5.5
5.9
5.7

11,692
45,773

72.9
66.8

8.3
7.9

10.9
3.1

4.0
8.6

50.1
50.9

5.2
7.7

Panel A: Fifth grade
Africa
Caribbean
China
Dominican Republic
East Asia
Eastern Europe
Guyana
Latin America
Russia
South Asia
West Asia
Western Europe
All foreign born
All native born
Panel B: Eighth grade
Africa
Caribbean
China
Dominican Republic
East Asia
Eastern Europe
Guyana
Latin America
Russia
South Asia
West Asia
Western Europe
All foreign born
All native born

Source: Schwartz and Stiefel (forthcoming, Table 8).

FRBNY Economic Policy Review / December 2005

165

Table 9

Regional Regression Results for Reading
Education Production Functions, Foreign-Born Students
Fifth Graders

Russia
Eastern Europe
Western Europe
China
East Asia
South Asia
West Asia
Africa
Dominican Republic
Caribbean
Guyana
Latin America
Constant
Number of observations
R2

Eighth Graders

1997-98

2000-01

1997-98

2000-01

0.135***
(0.037)
0.082*
(0.043)
0.123**
(0.048)
0.161***
(0.042)
0.083***
(0.030)
0.045*
(0.027)
0.100**
(0.046)
0.082
(0.051)
0.121***
(0.021)
0.033*
(0.018)
-0.155***
(0.029)
0.106***
(0.019)
0.109
(0.081)
64,971
0.57

-0.116
(0.119)
0.017
(0.055)
0.044
(0.048)
0.143***
(0.044)
0.068
(0.043)
-0.039
(0.033)
0.079
(0.053)
0.190***
(0.053)
0.065***
(0.024)
-0.016
(0.030)
-0.037
(0.035)
0.067***
(0.024)
-0.379
(0.494)
71,141
0.45

0.157***
(0.045)
0.116***
(0.038)
0.058
(0.041)
0.097***
(0.034)
0.090***
(0.028)
-0.023
(0.038)
-0.026
(0.038)
0.043
(0.050)
0.053**
(0.021)
-0.006
(0.020)
-0.135***
(0.043)
0.015
(0.018)
1.124**
(0.445)
57,465
0.60

0.315***
(0.073)
0.151***
(0.057)
0.087*
(0.051)
0.080
(0.051)
-0.042
(0.036)
0.034
(0.039)
-0.084*
(0.045)
0.079
(0.054)
0.071***
(0.020)
-0.057***
(0.019)
-0.102***
(0.038)
-0.004
(0.021)
1.876***
(0.582)
57,152
0.57

Source: Authors’ calculations.
Notes: The model includes controls for free-lunch eligibility, reduced-price-lunch eligibility, gender, age, ethnicity/race, English proficiency, language
assessment battery scores, special-education status, prior-year reading and math scores, teacher-pupil ratio, teacher experience, teacher tenure, teacher
licensing, teacher education, and school enrollment. Cohort dummies control for the number of years in the New York City public schools. Students who
have zero to one year in the New York City public schools entered the system in the 1997-98 school year. Specifically, they entered on or after November 1,
1996. Students who have at least one but less than two years entered between November 1, 1995, and October 31, 1996. Fifth graders with five or more years
in the New York City public schools entered on or before October 31, 1992. The teacher-pupil ratio is instrumented with the prior-year enrollment and
enrollment squared.
***Statistically significant at the 10 percent level.
***Statistically significant at the 5 percent level.
***Statistically significant at the 1 percent level.

166

Public Education in the Dynamic City

of immigrant students “disappears” with time in the United
States (that is, performance converges to the lower
performance of native-born students) has intuitive appeal and
surface validity, there is relatively little statistical evidence to
support it. To address this concern, we investigate the
evolution of performance of a cohort of students attending
New York City schools from third through eighth grades, using
a regression model to control for a range of time-varying
characteristics and student-fixed effects to capture unobserved
time-invariant characteristics. We find that the performance of
immigrants diverges from that of native-born students
(Chart 1). Separate analysis by race group suggests that the
time path differs across groups (Chart 2). White immigrants
diverge the most from their white native-born counterparts,
while Hispanic immigrants show some early divergence but
then begin to converge back in later grades. Overall, we find
little evidence for convergence.
We have examined several dimensions of the treatment of
immigrant students in the New York City public schools—
a group that presents special challenges because of the students’
late entry into the schools, limited English proficiency, and the
like, and that may well be at particular risk because of the
group’s potentially low level of political clout. Our results are

Chart 2

Regression-Adjusted Nativity Gap by Year and Race
Standard Academic Progress Cohort; Calendar Time;
Reading Scores
Difference in standardized test score (FB-NB)
1.0
0.8
0.6

White

0.4

Black
Hispanic

0.2

Asian
0
1995-96

1996-97 1997-98 1998-99 1999-2000 2000-01
School year

Source: Schwartz and Stiefel (2005, Figure 7).
Notes: The standard academic progress cohort includes students
originally enrolled in the third grade in 1995-96 who remained enrolled
every year through the 2000-01 school year and it progresses one grade
each year. The nativity gap is defined as the difference between the
average z-score of foreign-born (FB) and native-born (NB) students.
It is interpreted as the number of standard deviations by which foreignborn students outperform native-born students. The performance of
students tested outside the indicated year and grade is not included.
Models include student-fixed effects.

Chart 1

Regression-Adjusted Nativity Gap by Year
Standard Academic Progress Cohort; Calendar Time

encouraging. Segregation is relatively mild, resource allocation
seems equitable, and, perhaps most importantly, immigrant
student performance is good and trending upward. In the end,
it seems that immigrants may well be good for the New York
City public schools.

Difference in standardized test score (FB-NB)
1.0
0.8
0.6
Reading gap
0.4
Math gap
0.2

6. Evaluating School Reforms in
New York City: Some Examples

0
1995-96

1996-97 1997-98 1998-99 1999-2000 2000-01
School year

Source: Schwartz and Stiefel (2005, Figure 3).
Notes: The standard academic progress cohort includes students
originally enrolled in the third grade in 1995-96 who remained enrolled
every year through the 2000-01 school year and it progresses one grade
each year. The nativity gap is defined as the difference between the
average z-score of foreign-born (FB) and native-born (NB) students.
It is interpreted as the number of standard deviations by which foreignborn students outperform native-born students. The performance of
students tested outside the indicated year and grade is not included.
Models include student-fixed effects.

The dynamic nature of New York City’s public schools
provides a natural laboratory for new educational policies and
reforms. How well do these work? We examine three recent
reforms, using data provided by the New York City
Department of Education. The first, the Performance-Driven
Budgeting (PDB) initiative, changed the way that resources are
allocated within schools. The second, the New York Networks
for School Renewal (NYNSR) project, is an example of wholeschool reform, not unlike others implemented elsewhere, such
as Success for All. The third is the small-schools initiative,

FRBNY Economic Policy Review / December 2005

167

which continues as new small schools are opening each year in
New York City and elsewhere. The methodology is relatively
straightforward and replicable and, because it relies upon
administrative data, it is relatively inexpensive. The implication
is that evaluation is both possible and affordable and needs to
be integral to policymaking. As we observed, our findings are
generally positive. Reforms yield positive, if small, effects on
student outcomes.

6.1 Performance-Driven Budgeting
In 1996, New York City Schools Chancellor Rudy Crew
initiated an effort to move budgeting decisions toward schoollevel decision makers and to tie the new budgeting practice to
school performance. Termed Performance-Driven Budgeting,
the underlying logic was that decision makers closer to the
student are better able to align resources with academic needs.
The centerpiece of our analysis of this reform is a schoollevel production function linking student performance on
fourth- and fifth-grade tests to school inputs (teacher resources
and expenditures). The effect of the PDB reform was identified
as the difference in school performance before and after the
PDB intervention, relative to the schools that did not
implement PDB—in essence, a difference-in-difference design.
As shown in Table 10, the coefficients on the “implemented
PDB” variable indicate a positive, albeit small, effect of around
.06 standard deviations in reading and math (in fourth-grade)
test scores. To put this effect size in context, we note that
educators as a rule-of-thumb aspire to sizes of .25 when
initiating specific curriculum reforms; racial test-score gaps
between white and black or Hispanic students are around .7.
Thus, .06 is indeed small, but it is also positive.

6.2 The New York Networks
for School Renewal
The New York Networks for School Renewal project had a
somewhat different genesis, beginning in 1995-96 with eighty
founding schools. Representing a model of whole-school
reform, which involves voluntary networks and small school
sizes, the project was initiated with a $25 million, five-year
grant from the Annenberg Foundation.3 Our analysis of
NYNSR uses student-level data to estimate the effect of the
reform on students attending fourth, fifth, or sixth grade in
1995-96, as well as an “intent-to-treat design” to disentangle
the effect of the reform from all other changes. Table 11

168

Public Education in the Dynamic City

Table 10

The Effect of PDB Participation on Standardized
Tests
Fourth Grade

Fifth Grade

Dependent variables

Reading

Math

Reading

Math

Participation variable
Implemented PDB in
1997-98
Number of observations
R2

0.0557**
(0.0254)
2,436
0.9234

0.0599**
(0.0269)
2,436
0.9290

0.0568**
(0.0247)
2,436
0.9252

0.0187
(0.0263)
2,436
0.9304

Source: Stiefel et al. (2003, Table 5).
Notes: PBD is the Performance-Driven Budgeting initiative. All
regressions are weighted by enrollment share. All dependent variables are
measured in z-scores. Test scores in all years are from the CTB (reading) or
CAT (math) normal curve equivalents, except for 1998-99 fourth-grade
reading and math scores. Fourth-grade students were given new state
reading and math tests in 1998-99, and the Board of Education reports
their scaled test scores. Regression equations include a set of teacher
characteristics (percentage licensed, with more than five years of
experience, with more than two years of experience, with a master’s
degree; average number of days absent per year) and a set of school
characteristics (percentage students female, Asian and other, black,
Hispanic; average daily attendance; percentage eligible for free lunch,
limited-English-proficient, resource room participant, special-education,
recent immigrant) as well as school- and year-fixed effects and a group of
missing value indicators, the log of expenditures, and enrollment and a
constant term. Standard errors are in parentheses.
***p<.10.
***p<.05.
***p< .01.

illustrates our results, showing two- or three-year (long-term)
changes in reading and math test scores in two differently
specified models. On the whole, the impact estimates are
positive, with many statistically different from zero, and no
evidence exists of any negative effect. In addition, the size of the
effects, when significant and positive, is between .16 and .25,
considerably higher than those found for the PDB reforms.

6.3 Small-Schools Initiative
In the mid-1990s, reformers turned their attention to
improving the performance of American high-school students.
While various initiatives have been attempted—including
offering child care on school sites and imposing graduation test
requirements—one of the most enduring, visible, and wellfunded initiatives is the “small-schools” movement. Headlines
have trumpeted New York City’s (and Chicago’s) efforts to

Table 11

Long-Term-Impact Analysis of NYNSR Participation on Standardized Reading and Math Scores, by Cohort
Fourth-Grade Cohort
1998-99
(Grade 7)
Baseline reading regressions
NYNSR
R2
Including school characteristicsa
NYNSR
R2
Number of observations in all models
Baseline math regressions
NYNSR
R2
Including school characteristicsa
NYNSR
R2
Number of observations in all models

Fifth-Grade Cohort
1998-99
(Grade 8)

Sixth-Grade Cohort
1997-98
(Grade 8)

0.161***
(0.036)
0.627

0.165***
(0.063)
0.634

0.064
(0.044)
0.646

0.155**
(0.063)
0.636
4,947

0.029
(0.065)
0.655
4,842

0.062
(0.043)
0.658
5,981

0.251***
(0.045)
0.666

0.039
(0.048)
0.678

0.047
(0.040)
0.645

0.229***
(0.056)
0.680

-0.113*
(0.062)
0.699

0.001
(0.077)
0.667

4,977

6,153

5,024

Source: Schwartz, Stiefel, and Kim (2004, Table 4).
Notes: NYNSR is the New York Networks for School Renewal project. Test scores are measured in z-scores transformed from normal curve equivalents for
the CTB (reading) or CAT (math) exams, except for the DRP reading test scores in 1994-95 and state reading (ELA) and math test scores for the eighth grade
in 1998-99. Huber’s robust standard errors are reported in parentheses. All regressions include 1994-95 and 1995-96 test scores. Dummies are used for
students who are female; exposed to a language other than English; Asian, Hispanic, black, and recent immigrant; and, for each year, attendance rates,
language assessment battery percentiles, free- or reduced-price-lunch eligibility, resource room participation, grade retention, and advancement to a grade
higher than typical; and a set of missing-value indicators. Regressions with school variables include the number of consecutive years a student has been in
the same school. “Recent immigrant” and “advancement to a grade higher than typical” are dropped from the 1998-99 regressions. As of 1998-99, no recent
immigrant student in 1995-96 retained that status. None of the fourth- and fifth-grade-cohort students who advanced to a higher grade than typical in
1998-99 had valid reading or math test scores for that year.
a

Year-specific school controls are total enrollment; number of teachers per 100 students; teachers’ average number of days absent; the percentage of students
who are black, Hispanic, Asian, free-lunch-eligible, limited-English-proficient, recent immigrants, special education, and resource room participants; and
the percentage of teachers fully licensed and permanently assigned, with a master’s degree, with more than five years of experience, and working more than
two years in the same school.
***Statistically significant at the 10 percent level.
***Statistically significant at the 5 percent level.
***Statistically significant at the 1 percent level.

convert large comprehensive high schools with up to 5,000
students into small schools with 500 or fewer students.
Whether the small-schools initiative succeeds depends on its
effectiveness with its own students, the impact on district costs
associated with smaller units and more of them, and the effects
on the larger high schools that remain. Our analysis of the small
schools created in the early phases of the initiative attempts to
address the first issue, using data on school expenditures and

cohort graduation rates in New York City high schools. The use
of cohort graduation rates is key. The New York City
Department of Education tracks students for up to seven years,
beginning in ninth grade—recording whether they graduate,
transfer to another school or system, drop out, or continue past
four years. Thus, we can construct, for each school, the budget
per graduate and examine the way it varies with school size.
The findings are compelling. The small academic high schools,

FRBNY Economic Policy Review / December 2005

169

most like the ones being replicated now, have a better
performance record, deliver a higher cohort graduation rate,
and in the end have similar per-pupil expenditures as the large
schools. Put differently, the small high schools have higher
graduation rates to balance their higher expenditures per pupil.

7. Lessons
New York City, like cities around the United States and the
world, faces particular difficulties providing public education
efficiently and equitably. The student body is heterogeneous
and dynamic. Poverty is common, and limited proficiency in
English challenges many. Further, turnover is high. Each year,
thousands of new students enter the New York City public
schools midway through their school career, many of them
from schools outside the United States. New York City schools
include substantial numbers of students from dozens of
countries, speaking many languages. Together, these factors
pose a formidable challenge to the school system. That said, we
still find much cause for optimism. Our research shows that,
other things equal, immigrant students fare reasonably well.
Their performance on standardized tests is good, their schools
receive resources in the same measure as schools with more
native-born students, and their performance seems to improve
over time as they adjust to their schools and new homes. Thus,

170

Public Education in the Dynamic City

the programs and interventions that the New York City school
system has in place to address the difficulties faced by
immigrant students seem to be working.
Further, the school system itself seems quite dynamic. Each
year brings a wide range of reforms—in curriculum, school
organization, governance, testing, and accountability—and,
while not all of them work, our research on earlier reforms
suggests that it is possible to use evaluation to disentangle those
programs that work from those that do not. Administrative
data are increasingly available, allowing relatively low-cost
evaluations. Even more important, advances in econometric
methods are facilitating efforts to disentangle causality and
distinguish good programs and good schools from bad ones.
At the same time, there is much room for improvement.
While evaluation is possible, it is still far from universal. Too
many reforms are implemented and declared successes or
failures without any investigation, and the largest and most
sweeping reforms are rarely subject to careful evaluation.4
Further, evaluation can be simplified. We make too little use of
randomization and access to data, and the ease of using and
interpreting the data is more limited than it should be. Finally,
there are many inequities and inefficiencies that continue. For
instance, disparities persist between blacks, Hispanics, whites,
and Asians, as well as in the allocation of teachers and resources
across schools, despite significant efforts to close these gaps.
Much more work remains to be done.

Endnotes

1. Notice that the group of students who entered by third grade is a
subset of the continuing students in Table 2 because only a fraction,
roughly 82 percent, of the continuing students from the third-grade
cohort of 1995-96 were in eighth grade (others were in seventh grade,
in special education, or elsewhere).
2. The WWC was established in 2002 by the U.S. Department of
Education’s Institute of Education Sciences to provide educators,
policymakers, researchers, and the public with a central and trusted
source of scientific evidence of “what works” in education. It aims to
promote informed decision making on education through a set of
easily accessible databases and user-friendly reports that provide
consumers with ongoing, high-quality reviews of the effectiveness of
replicable educational interventions (programs, products, practices,
and policies) that aim to improve student outcomes. The WWC is
administered by the Institute of Education Sciences through a contract
to a joint venture of the American Institutes for Research and the

Campbell Collaboration. Both organizations are nationally recognized
leaders in education research and rigorous reviews of scientific
evidence. Subcontractors to the project are Aspen Systems
Corporation, Caliber Associates, Duke University, and the University
of Pennsylvania. (See <http://www.whatworks.ed.gov/whoweare/
overview.html#key>.)
3. Other examples of whole-school reform are Success for All,
Accelerated Schools, Edison Schools, Comer Schools, and New
American Schools. All of these reforms aim to change many parts
of the school at once (some combination of components such as
curriculum, teacher attitudes, time devoted to subjects, use of
technology).
4. The New York City Department of Education has requested
proposals from outside evaluators for reform of its promotion/
retention policy.

FRBNY Economic Policy Review / December 2005

171

References

Berne, R., and L. Stiefel. 1984. The Measurement of Equity in
School Finance: Conceptual and Empirical Dimensions.
Baltimore: Johns Hopkins University Press.
Ellen, I. G., K. O’Regan, A. E. Schwartz, and L. Stiefel. 2002. “Immigrant
Children and Urban Schools: Lessons from New York on
Segregation, Resources, and School Attendance Patterns.”
In W. G. Gale and J. R. Pack, eds., Brookings-Wharton Papers
on Urban Affairs. Washington, D.C.
Schwartz, A. E., and L. Stiefel. 2004. “Immigrants and the Distribution
of Resources within an Urban School District.” Educational
Evaluation and Policy Analysis 26, no. 4 (winter): 303-27.
———. 2005. “Testing the Convergence Hypothesis in Immigrant
Academic Achievement: A Longitudinal Analysis.” Unpublished
paper, New York University.

Stiefel, L., R. Berne, P. Iatarola, and N. Fruchter. 2000. “High-School
Size: Effects on Budgets and Performance in New York City.”
Educational Evaluation and Policy Analysis 22, no. 1
(spring): 27-40.
Stiefel, L., R. Rubenstein, A. E. Schwartz, and J. Zabel. 2005.
Measuring School Performance and Efficiency:
Implications for Practice and Research. 2005 Yearbook
of the American Education Finance Association. New York.
Stiefel, L., A. E. Schwartz, C. Portas, and D. Y. Kim. 2003. “School
Budgeting and School Performance: The Impact of New York
City’s Performance-Driven Budgeting Initiative.” Journal of
Education Finance 28, no. 3 (winter): 403-24.
Suarez-Orozco, M. M. 2001. “Globalization, Immigration, and
Education: The Research Agenda.” Harvard Educational
Review 71, no. 3 (fall): 345-65.

———. Forthcoming. “Is There a Nativity Gap? New Evidence on the
Academic Performance of Immigrant Students.” Education
Finance and Policy.
Schwartz, A. E., L. Stiefel, and D. Y. Kim. 2004. “The Impact of School
Reform on Student Performance: Evidence from the New York
Networks for School Renewal Project.” Journal of Human
Resources 39, no. 2 (spring): 500-22.

The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
172

Public Education in the Dynamic City

Dalton Conley

Commentary

1. Overview

T

he New York City public school system serves as an
amazing laboratory to study issues of immigration and
integration through the educational system. A full one-eighth
of the third-grade cohort that Amy Ellen Schwartz and Leanna
Stiefel study are foreign born; add to that a significant number
who arrive after third grade as well as a large number of
students who were born here to non-native parents and the
influence of immigration on the city’s classrooms is enormous.
Schwartz and Stiefel’s paper provides a rich and textured
portrait of these students as well as a thorough comparison
with their native-born counterparts.
The authors, in fact, present us with a paradox of sorts.
However, it is one that should be familiar to scholars of
immigration: despite more disadvantageous family
backgrounds (in terms of income, at least), immigrant
children—at least those who take standardized tests—
outperform native-born students. (Never mind, for the
moment, that important differences exist within the immigrant
community between, for example, early and late entrants or by
national origin.) Is this immigrant advantage an accurate
reflection of the performance of young immigrant children, or
is it all selection into who takes the tests? If it is not artifactual,
how is it possible that the immigrant success story starts as early
as third grade (or perhaps even earlier; third grade is merely the
first opportunity New York provides to assess its student

Dalton Conley is a professor of sociology and public policy at New York
University and director of New York University’s Center for Advanced
Social Science Research.
<dalton.conley@nyu.edu>

population in a standardized way)? And how do these
immigrant children affect their native-born classmates?
Do they cause positive peer effects because of their superior
performance? Or rather, do they place unique strains on the
system and therefore create negative externalities for
nonimmigrant students?
The New York school system is the largest—and arguably
the most complex—school system in the country. This means
that in addition to managing a high proportion of immigrant
children, the system deals with an incredible amount in
absolute terms. It also means that there is incredible diversity
across schools, largely reflecting New York’s diverse
neighborhoods. For example, the average share of white
children in the system is 16 percent. However, the standard
deviation for this mean is 23 percent. Likewise, in some schools
almost all the teachers are certified and in others only a handful
are. Some schools have thousands of students while others only
a couple of hundred. Financing also varies dramatically across
schools—as does, in fact, almost every measurable
characteristic. (Ironically, immigrants are fairly well
distributed across most of the schools in the system—a fact that
stands in stark contrast to black-white segregation or isolation
by free-lunch status.)
Another result of the size and complexity of the school
system is the fact that one or another reform effort is almost
constantly under way. Most of these reforms are not applied
uniformly or universally because of the unwieldy nature of the

The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve
System.

FRBNY Economic Policy Review / December 2005

173

task. For economists, however, this is a blessing, since it allows
for a quasi-experimental design to track the reforms’ effects
(since most are operationalized in a rather random,
unsystematic manner). Indeed, Schwartz and Stiefel take
advantage of this fact to assess the effect of various reforms on
immigrant students in particular.

2. Which Immigrants Are Thriving?
After demonstrating that there do not appear to be significant
financing differences between schools with greater or smaller
proportions of non-native students (although there perhaps
should be, because of differing levels of academic need),
Schwartz and Stiefel proceed to the main part of their analysis,
which documents, ceteris paribus, which immigrant groups are
performing the best. Most of the differences the study
documents can be attributed to demographic background
factors such as poverty rates or length of time in the school
system. However, even when the authors control for a wide
range of background variables, they find that immigrant
students of Russian or Chinese backgrounds perform especially
well.
The authors then tease out this analysis by looking at changes
over time in student performance using student-fixed effects.
Contrary to the notion that immigrants and native-born
students “converge” in their performance, the authors find that
immigrants as a whole start with a slight advantage in third
grade, which only swells as they move up through the system. Of
course, this does not take into account the students who attrite
(or those who enter late). The more troublesome piece of the
puzzle is the fact that while all racial groups of immigrants seem
to start off with similar slopes of relative improvement over their
native counterparts, as black and Latino immigrants move into
middle school their advantages taper off (and perhaps even
reverse slightly). This suggests that ethnographic evidence on the
effect of race trumping nativity status observed in the
sociological literature on identity (see, for example, Waters
[1999]) may have real impacts on learning curves (and therefore
on downstream outcomes as well). This is a troubling note in an
otherwise optimistic report on the progress of immigrant
children as they make their way through the Byzantine public
school system. It appears—at first blush at least—that the
authors’ worries about these vulnerable students are misplaced.
Immigrant students seem to be thriving despite having parents
who generally enjoy low levels of political clout and perhaps
limited social and cultural capital.
However, the authors’ last line in section 5 still appears
unwarranted: they conclude that “In the end, it seems that

174

Commentary

immigrants may well be good [emphasis theirs] for the New
York City public schools.” Schwartz and Stiefel do not perform
an analysis of the peer effects of immigrants on their nativeborn counterparts, so they really should not make these claims.
I would have loved to have seen just this analysis. For example,
using school-fixed effects, how does the percentage of foreignborn students affect the performance of native students? This is
a big lacuna in the analysis that is sorely needed to determine
whether the authors’ ultimate statement is accurate. It could
be, of course, that immigrants are thriving at the expense of the
rest of the students in the system.
Slowing down even more, one may question even the
conclusion that immigrants themselves are thriving (with the
caveat of the aforementioned within-group differences). The
authors undertake their analysis as if the public schools form a
closed educational system. However, just as a full third of nonnative-born students in eighth grade entered the system post
third grade, we know very little about those native-born
students who are leaving the sample. While there are relatively
few financing distinctions across the public schools, the real
story in New York is the public-private divide. Many elite (that
is, high socioeconomic status) parents tolerate the public
system for a while during elementary school and then move
their children into private schools as they progress through the
ranks. A particularly large exodus may occur in the transition
to middle school. So, in other words, the swelling immigrant
advantage may, in fact, be selection effect on the native-born
population: those least able to escape the system for financial or
ability reasons may be left as the dwindling comparison group.
The authors try to assuage such fears by offering us a means
comparison for “continuing students” (that is, those who stay
in the sample) versus “attriting students.” They show, however,
that the attriters are more likely to be native born and more
likely to have higher test scores at baseline. Therefore, this only
worries me more. What we really need is to see the two groups
broken down by immigrant status to determine the differencein-difference in test scores between attriters and stayers in the
groups.

3. School Reform
The latter part of the paper addresses three school reforms: the
Performance-Driven Budgeting initiative, the New York
Networks for School Renewal project, and the small-schools
initiative. The paper’s results show, on the whole, modest,
positive effects of reform on measurable student outcomes.
The problem, however, is the strong possibility that these
reforms may be endogenous to school quality. Especially in a

system in which schools are opening and closing at such a
frequent rate (in some years, in excess of 5 percent of schools
have closed or opened for business), the notion that these
reforms are distributed randomly is not entirely credible. This
notion is furthered by the authors’ own findings that the
reform that seems to matter least is Performance-Driven
Budgeting—the one that I would argue is most exogenous to
administrator quality. The school renewal project and the
small-schools initiative both appeared to require considerable
entrepreneurship on the part of the school’s leadership team in
order to be enacted. Thus, the results may not entirely be

driven by treatment effects of these reforms, but rather by the
underlying characteristics of the institutions (such as staff,
administration, and PTA, not to mention community and
family characteristics) at the schools that adopt such reforms.
At the very least, I would have liked to have seen Schwartz and
Stiefel provide comparisons on the measurables between the
treatment and control groups; at the very best, I would have
liked to have seen documentation that the reforms were
implemented in a truly random fashion. This is just a final,
cautionary note on what is otherwise a very informative
paper.

FRBNY Economic Policy Review / December 2005

175

References

Waters, M. C. 1999. Black Identities: West Indian Immigrant
Dreams and American Realities. New York: Russell Sage
Foundation.

The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
176

Commentary