The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Earnings Losses of Job Losers During the
2001 Economic Downturn*
SHIGERU FUJITA AND VILAS RAO
J
ob losses may involve not only lost earnings
during unemployment but also declines in
earnings at subsequent jobs. After a timeconsuming job search, workers may need to
restart their careers from scratch, accepting a lower wage.
Workers may also need time to acquire new skills, and
total earnings lost during such a period of re-adjustment
can be considerable. But experiences may vary widely.
In this article, using a novel data set, Shigeru Fujita
and Vilas Rao provide evidence on earnings losses after
unemployment. Although the usefulness of the evidence
is limited by the short sample period, the data set allows
us to ask some important questions, the answers to which
may help inform us about important macroeconomic
issues such as the cost of business-cycle fluctuations and
the benefits of policies intended to avoid such fluctuations.
During economic downturns,
more workers become unemployed
and finding a new job becomes harder.
Consequently, unemployment rises.
Higher unemployment also means that
there is a more intensive reallocation
of workers from one job to another
during downturns.1
Shigeru Fujita
is a senior
economist in
the Research
Department of
the Philadelphia
Fed.
www.philadelphiafed.org
The main reason policymakers
and economists are concerned about
job losses is that job losses may involve
not only lost earnings during the period of unemployment but also declines
in earnings at subsequent jobs. It is
conceivable that the experiences of job
losers are painful and costly. After a
time-consuming job search, the worker
may need to restart his or her career
from scratch in a new job, accepting
a lower wage. Furthermore, working
in a new environment might involve
*The views expressed here are those of the
authors and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.
acquiring new skills, establishing a new
personal network of business associates, and so on, all of which may take a
significant amount of time to accomplish. This re-adjustment period can be
quite long, and thus total earnings lost
can be considerable.
This painful story would be
relevant for at least some workers. But
experiences may vary widely across
individuals. In contrast to the example above, it is possible to imagine
a situation in which workers make the
same amount of money (or more) after
a short unemployment spell or one
where workers make less at the new job
initially, but the losses are recovered
quickly as a result of subsequent earnings growth. In these cases, earnings
losses associated with the job loss are
minor relative to one’s lifetime earnings, and unemployment may not be
as costly and painful as the previous
example suggests.2
1
See the 2007 Business Review article by Shigeru
Fujita.
2
The process of destroying less productive jobs
and replacing them with more productive jobs
is important for long-run economic growth
and provides the opportunity for workers to
find higher paying jobs. See the 2008 Business
Review article by Shigeru Fujita.
When he
co-wrote this
article, Vilas
Rao was a
research analyst
in the Research
Department of
the Philadelphia
Fed. He is now a
graduate student
at the Kennedy School of Government,
Harvard University.
Business Review Q4 2009 1
This article provides evidence on
earnings losses after unemployment,
using a novel data set that traces the
labor market experiences of a large
number of workers over a three-year
period that encompasses the recession
in 2001. Although the usefulness of
the evidence is limited by the short
sample period, the data set allows us to
ask important questions such as: What
is the average individual loss (or gain)
due to unemployment? Who loses
the most? What are the sources of
earnings losses? While not definitive,
the answers to these questions may, in
turn, help inform us about important
macroeconomic issues such as the cost
of business-cycle fluctuations and the
benefits of policies intended to avoid
such fluctuations.
A PANEL DATA SET ON
EARNINGS LOSSES (OR GAINS)
FOLLOWING UNEMPLOYMENT
To obtain information on earnings
losses due to unemployment, it is
necessary to trace the earnings history
of a large number of workers over
some length of time. Furthermore,
since workers may lose and find new
jobs within a relatively short period of
time (say, within months), this history
needs to be collected frequently, say,
monthly.
Fortunately, the Census Bureau
maintains a data set called the
Survey of Income and Program
Participation (SIPP) that satisfies
these requirements. The SIPP
2001 panel keeps track of labor
market experiences of a nationally
representative sample of 73,205
workers over the roughly three-year
period from October 2000 through
December 2003.
With this data set in hand, we can
look at workers’ experiences during
the U.S. economic downturn of 2001.
We select the events in which a worker
moves from one job to a new job with
2 Q4 2009 Business Review
an unemployment spell in between.
The data set includes 1,380 such cases.
(For details of the sample selection, see
The SIPP and Other Data Sets Used in
Previous Studies.)
MONTHLY EARNINGS
DROP IMMEDIATELY AFTER
UNEMPLOYMENT
Figure 1 presents the distribution
of earnings losses after unemployment
in the early 2000s. It shows that, on
average, a worker’s monthly earnings
immediately after unemployment drop
roughly 7 percent compared with the
monthly earnings immediately before
unemployment.3 The three bars next
to the average correspond respectively
to 25th percentile, median, and 75th
percentile of the sample of employees
in our sample.
We can make a couple of
important observations here. First,
there is a huge variation across
individual workers in terms of changes
in earnings after unemployment.
Related to this are a large number
of workers whose incomes actually
increase after unemployment. The
earnings gains can occur for two
reasons. First, the outcome of a job
search is affected by luck. That is,
some workers are simply lucky to find
an employer that is a “good match.”
Second, some workers become
3
All calculations using the SIPP are based on
the comparison of average monthly earnings
over the three-month periods before and after
the unemployment spell. Earnings include only
salary from the main job and do not include
benefits.
FIGURE 1
Changes in Earnings After an
Unemployment Spell
Percent Change
in Earnings
50
40
30
20
10
0
-10
-20
-30
-40
-50
Average
25th Percentile
Median
75th Percentile
Notes: Based on 2001 SIPP panel. Sample is restricted to workers who have been employed
for three months continuously before and after an unemployment spell and switch firms after
unemployment. A total of 1,380 unemployment experiences are included. This chart gives the
distribution of earnings losses across unemployment experiences in our sample.
www.philadelphiafed.org
The SIPP and Other Data Sets Used in Previous Studies
T
he Survey of Income and Program
Participation (SIPP) is a monthly
survey conducted by the U.S. Census
Bureau that follows the participation
of individuals and households in
income maintenance programs. Using a
nationally representative sample of individuals 15 years
of age and older from the civilian noninstitutionalized
population, the SIPP gathers a variety of information:
demographic characteristics, labor force participation,
amounts and types of earned and unearned income,
government program benefits, assets, and health
insurance.
As a panel survey, the SIPP tracks the same
individuals over a period of time. For this study, we used
the 2001 SIPP panel, which tracked the labor market
experiences of a nationally representative sample of
73,205 workers over the roughly three-year period from
October 2000 through December 2003. Sample members
who move to a new address are interviewed at their new
address. This characteristic of the SIPP makes it a useful
vehicle for exploring unemployment’s impact on earnings,
since we are able to comprehensively track an individual’s
earnings and employment status for an extended period
of time.
We use each individual’s labor force status after the
second week of each month as his or her labor status
for that month. Unemployment is defined as either not
having a job but looking for work or having a job and
on layoff or absent from work. Individuals who do not
have a job and are not looking are considered not in the
labor force. Individuals with a job who are not on layoff
are considered employed. The same definitions are used
in the BLS’s Current Population Survey, which is the
official source of the national unemployment rate, the
employment population ratio, etc.
For this study, we restrict the sample in a few
significant ways. First, only individuals 25 and older are
included in our analysis. We look only at the primary job
of individuals with multiple jobs, and we exclude workers
who returned to the same job after unemployment.
Finally, we require that a worker be employed for at least
three months on either end of his or her unemployment
spell. Our analysis is based on 1,380 events that satisfy all
requirements.
A handful of papers study earnings losses using
different data sets covering different time periods, but
our data set has many unique features that are absent
from other data sets used in other studies. Previous
studies have used the Displaced Worker Survey (DWS),
a supplement to the Current Population Survey that
has been administered every two years since 1984. The
DWS collects relevant information on the experience of
job losers, such as changes in earnings. However, it asks
only about a single job loss in the past three years due to
business decisions such as a plant closing or the abolition
of a job position. While the information gathered is quite
useful, it may not represent the experience of the average
unemployed worker.
Other studies have used the data set called Panel
Study of Income Dynamics (PSID). This data set also
provides useful pieces of information on the experience
of job losers. However, the interview is conducted only
once a year, and thus it possibly misses many job-loss
experiences that occurred between the two interview
dates. One advantage of the PSID over the SIPP is that
the PSID traces workers over a much longer time than
the SIPP. This feature allows researchers to examine the
long-run effects of job loss. See the discussion in the text
on page 7, under the heading Long-Lasting Effects of Job
Loss.
*
In fact, quite a few workers return to the same employer after unemployment. In our 2001 SIPP sample, 46 percent of workers returned to the same
employer.
unemployed because they chose to
quit their previous job in order to look
for a better one. This result implies
that the overly pessimistic view about
“unemployment” may not necessarily
www.philadelphiafed.org
be an accurate description of the
reality.
At the same time, despite the fact
that some workers experience earnings
gains after unemployment, it is true
that unemployment is, on average,
accompanied by a drop in earnings.
Furthermore, Figure 1 shows that the
average change is below the median
change (2 percent drop), implying that
Business Review Q4 2009 3
the distribution of the earnings losses
is skewed to the left. That is, some of
the losses experienced are very large.
For example, 25 percent of the workers
have earnings losses of more than 40
percent.4
The average drop in earnings here
appears smaller than that reported
in previous studies. For instance,
an article by Henry Farber reports
that the average earnings losses that
occurred between 2001 and 2003 were
more than 13 percent. A plausible
reason behind this difference is that
Farber uses the Displaced Worker
Survey (DWS), which focuses on
a certain type of job separation,
namely, displacement. (See The SIPP
and Other Data Sets Used in Previous
Studies for further explanation of the
DWS.) In the DWS, “displacement” is
defined as job separations associated
with business decisions such as a
plant closing or the abolition of a job
position. The sample in our study,
on the other hand, is selected based
on whether workers experience
unemployment regardless of underlying
reasons and thus is broader than the
DWS. The displacement events in
the DWS are likely to correspond
to the ones on the left-hand side of
the distribution, i.e., ones with large
earnings losses.
There are a few caveats to
remember in our calculation. First,
our calculations ignore the forgone
earnings of job losers. That is, the
job loser might have enjoyed growth
in earnings had he not lost his job.
But this part of the losses is likely to
be small in our sample because we
compare earnings between two dates
that are relatively close, and thus
4
Of course, drops in earnings that many individual workers experience may again simply be
due to luck. However, the facts that the average
change in earnings is negative and that the
distribution is skewed to the left imply that luck
cannot be the only reason.
4 Q4 2009 Business Review
potential growth during that short
period of time would be relatively
small.5 Second, the SIPP 2001 data
set keeps track of individual workers
for only about three years, and thus,
it is difficult to assess whether the
initial losses are recovered later and,
if they are, how long it takes. The
past literature suggests that the loss is
persistent. We will come back to this
issue later. Finally, we know that the
size of earnings losses varies across the
business cycle. Farber’s article presents
the average earnings losses for different
time periods and shows that they
increase significantly during recessions
and decrease significantly during
booms and that the deeper recessions
tend to result in larger earnings losses.
The latter fact implies that earnings
losses in the current downturn may be
significantly larger than those for the
mild recession in 2001.6
With these caveats in mind, we
will explore sources of earnings losses
using the SIPP 2001 panel. Looking
at how worker characteristics are
correlated with their earnings losses is
useful for this purpose.
NO CLEAR RELATIONSHIP
WITH EDUCATION OR RACE
Are there any differences in
earnings losses across different
educational or racial groups? While
we know that earnings levels are
strongly correlated with these worker
characteristics, there is a priori no
reason to believe that the size of
earnings losses is related to these
worker characteristics because these
characteristics do not change before
and after the unemployment spell.
Figure 2 confirms this prediction:
While there are some variations in
the size of earnings losses across races
and educational levels, it is not the
case that workers with a lower level
of earnings lost more in percentage
terms.7 In fact, the reality is quite
the opposite. If we simply look at
the relationship between the level of
earnings at the pre-unemployment
job and the size of earnings losses (in
percentage terms), we find a strong
positive correlation between the two.8
DURATION OF
UNEMPLOYMENT WAS
POSITIVELY RELATED TO
EARNINGS LOSSES
One way to identify the sources
of earnings loss is to look at the
differences in worker characteristics
before and after unemployment.
First, let’s see whether the length of
unemployment has any relationship
to earnings losses. If we assume that
staying on the job plays an important
role in the growth of earnings,
say, reflecting the accumulation of
human capital, we can expect that
as unemployment duration becomes
longer, human capital depreciates more
and hence earnings losses become
larger.9
7
In our data, almost 80 percent of workers
found new jobs within six months.
In Figure 2, the losses of high school graduates
and college graduates are roughly the same.
Similarly, the average earnings losses of white
workers are roughly the same as that of black
workers, although white workers, on average,
make considerably more than black workers.
6
8
5
We also find the same pattern in the SIPP.
The average earnings losses in the SIPP 1996
panel, which traces workers from the end of
1995 through late 2000, a period of economic
expansion, are quite small (-1.7 percent),
whereas the SIPP 1990 panel, which covers the
three-year period encompassing the recession in
the early 1990s, shows average earnings losses of
-15.3 percent.
The correlation coefficient is 0.46.
9
Of course, another possibility is that unemployed workers run down their wealth over time
and thus are less selective about their jobs, and
consequently, they accept jobs that pay less. But
whether this story is important or not, it does
not appear to change our overall conclusion
below.
www.philadelphiafed.org
FIGURE 2
Changes in Earnings by Race and Education
Percent Change
in Earnings
0
-2
-4
-6
-8
-10
-12
-14
-16
-18
-20
White
Hispanic
Black
Other
High School
College
Notes: Based on 2001 SIPP panel. Sample is restricted to workers who have been employed for
three months continuously before and after an unemployment spell and switch firms after unemployment. A total of 1,380 unemployment experiences are included. “High School” includes those
with education up to a high school diploma. “College” includes those with some college experience, a college degree, or postgraduate study.
FIGURE 3
Changes in Earnings by Unemployment Duration
Percent Change
in Earnings
10
5
0
-5
(38.5%)
(41.1%)
-10
-15
Figure 3 presents earnings
losses for workers with the following
unemployment durations: one to two
months, three to five months, and
six months or more. The numbers
below each bar represent the fraction
of workers for each duration of
unemployment.10 First, note that
the distribution of workers over the
duration of unemployment implies that
the average worker found a job fairly
quickly during the sample period. This
is consistent with the evidence found
elsewhere.11 For those who found a job
within two months, earnings losses
tended to be smaller than the average
loss of 7 percent reported above.
However, earnings losses increased
with duration of unemployment.
In particular, when workers were
unemployed for six months or more,
the average loss was more than 15
percent. This finding is consistent
with the notion that workers who
are unemployed for a longer time
experience a larger decline in their
stock of human capital. But what kind
of human capital has the worker lost?
Is it human capital that is useful in
any job? Or is it human capital that is
useful only for a certain firm or certain
occupation?
To answer these questions, note
that if human capital is tied entirely to
a particular firm, there is no reason to
expect a positive relationship between
earnings losses and unemployment
duration, given that workers are not
returning to the same firm, as is the
case in our sample. Therefore, the
(20.4%)
-20
-25
10
-30
1-2 months
3-5 months
6+ months
Unemployment Length
Notes: Based on 2001 SIPP panel. Sample is restricted to workers who have been employed for
three months continuously before and after an unemployment spell and switch firms after unemployment. A total of 1,380 unemployment experiences are included. The numbers in parentheses
indicate fractions of workers in each duration category.
www.philadelphiafed.org
Note that our data miss those workers who
became unemployed in the sample period but
could not find a new job. This censoring problem causes downward bias to our results. However, the bias is likely to be small given that,
in our sample, 80 percent of these unemployed
workers found a new job within five months, as
shown in Figure 3.
11
See the 2007 Business Review article by
Shigeru Fujita.
Business Review Q4 2009 5
evidence above does not appear to
support the idea that firm-specific
skills played a dominant role in
earnings losses.
One way to assess the importance
of occupation-specific human capital
is to split the sample used in Figure
3 into those who stayed in the same
occupation and those who switched
occupations after unemployment.12
The result, which is shown in
Figure 4, is quite striking. The
correlation between the duration of
unemployment and earnings losses
above was largely accounted for by
those who switched occupations. For
example, earnings losses for those
who stayed in the same occupation
were actually smaller than the average
earnings losses of all job losers, and
thus overall earnings losses of those
who were unemployed more than six
months were entirely accounted for by
those who switched occupations. We
will now investigate the robustness of
this result further by slicing the data
differently.
HIGH-TENURE WORKERS WHO
SWITCHED OCCUPATIONS
HAD LARGER EARNINGS
LOSSES
If occupation-specific human
capital is the dominant determinant
of earnings, a larger drop in earnings
is expected to follow when a worker
is forced to switch occupations
after a long career in a certain
occupation. Unfortunately, we were
unable to obtain information on
occupation-specific tenure from the
SIPP. However, the SIPP contains
information on how many years
FIGURE 4
Changes in Earnings with Occupation Switch
(By Unemployment Duration)
Percent Change
in Earnings
10
5
0
-5
-10
-15
-20
Switches to Different Occupation
-25
Occupations are divided based on the twodigit census codes that include categories such
as professional specialty, sales, administrative
support, and so forth. We also considered the
case with finer occupational codes (three-digit
census codes) and the results are similar.
6 Q4 2009 Business Review
Stays in Same Occupation
-30
1-2 months
3-5 months
6+ months
Unemployment Length
Notes: Based on 2001 SIPP panel. Sample is restricted to workers who have been employed
for three months continuously before and after an unemployment spell and switch firms after
unemployment. A total of 1,380 unemployment experiences are included. Jobs are divided into 14
occupation groups.
workers have worked for a particular
firm. To the extent that the firmspecific tenure is correlated with
occupation-specific tenure, this
information can be useful to further
infer the importance of occupationspecific human capital.13
First, let’s look at earnings losses
for workers with different firm tenures
13
12
Total
The assumption regarding the correlation
between firm-specific tenure and occupationspecific tenure seems plausible. For example,
using monthly data from the Current Population Survey over the period 1994 to 2006,
Giuseppe Moscarini and Kaj Thomsson show
that of those who stay at the same firm from the
previous month, only 1.3 percent, on average,
experience a change in their occupation (see
Table 9 of their article).
(Figure 5). The figure shows that those
who had longer tenure (five years or
more) lost much more (19 percent)
than those who had shorter tenure
(2.5 percent). This evidence by itself
appears to suggest the importance
of firm-specific human capital in
determining earnings. However, this
correlation between firm tenures and
the size of earnings losses disappears
when we split the sample of hightenure workers into those who stayed
in the same occupation and those who
switched occupations. The results
are displayed in Figure 6. The large
decline in earnings among high-tenure
workers is accounted for by the even
larger decline in earnings (more than
www.philadelphiafed.org
FIGURE 5
Changes in Earnings by Firm Tenure
Percent Change
in Earnings
0
-5
-10
-15
-20
-25
0-4 years
5+ years
Tenure Length
Notes: Based on 2001 SIPP panel. Sample is restricted to workers who have been employed for
three months continuously before and after an unemployment spell and switch firms after unemployment. A total of 1,380 unemployment experiences are included.
35 percent) among those who switched
occupations. On the other hand, those
who stayed in the same occupation
experienced much smaller earnings
losses, suggesting the relevance of
occupation-specific human capital
instead of firm-specific human capital.
The result here conforms to the
conclusions in previous studies. Using
DWS data on displaced workers in
the 1980s, Derek Neal shows that
earnings losses are strongly associated
with industry tenure as opposed to
firm tenure. While Neal emphasizes
the role of industry-specific human
14
Daniel Parent obtained results similar to
Neal’s using the PSID.
www.philadelphiafed.org
capital, the subsequent research has
shifted emphasis to the occupational
specificity of human capital.14 For
example, Gueorgui Kambourov and
Iourii Manovskii estimate regression
models of earnings growth using the
Panel Study of Income Dynamics
(PSID) and find that once occupation
tenure is included in the regression,
neither firm tenure nor industry tenure
remains significant, while occupation
tenure is highly significant.15
15
Note that Kambourov and Manovskii’s
approach is different from looking at earnings
losses of job losers in that they directly estimate
the return to experience in a certain occupation by considering workers who are employed
throughout the sample period.
LONG-LASTING EFFECTS OF
JOB LOSS
As we mentioned before, the SIPP
2001 panel covers only the three-year
period 2001 through 2003, and thus,
it is difficult to assess how persistent
the effect of job loss is. The question
is whether the lower earnings level
immediately after unemployment
recovers quickly and, if not, how long
it takes to regain earnings. Christopher
Ruhm considered this issue by using
the PSID, which allows him to trace
workers from 1969 through 1982.
He found that even four years after
displacement, job losers make 10 to 13
percent less than their nondisplaced
counterparts.16
An important point to note here
is that the persistence can take two
forms. First, it may take a long time to
regain earnings after an unemployment
spell even if the worker keeps his or
her new job for a long time. Second,
the initial unemployment spell may
raise the risk of subsequent job losses.
The latter may happen because new
workers are the ones who tend to get
laid off when a firm runs into difficult
times. A study by Ann Huff Stevens
attempts to sort out the two effects.
She traces workers’ labor market
experience from 1968 through 1988
using the PSID and shows that much
of the persistence of earnings losses
is actually explained by the latter
16
Louis Jacobson, Robert LaLonde, and Daniel
Sullivan, who use a unique comprehensive data
set derived from administrative records of the
state of Pennsylvania, have done influential
research in this area. While their study has
important limitations — for example, their
results are based on high-tenure workers (with
firm tenure of more than six years) in Pennsylvania who were displaced during the early and
mid-1980s — their data set offers important
advantages in the form of very large sample
sizes and detailed information on workers’ predisplacement employers. They also find that job
losers in their sample experienced large initial
earnings losses followed by very slow recovery of
the earnings.
Business Review Q4 2009 7
FIGURE 6
Changes in Earnings With and Without
Occupation Switch for High-Tenure Workers
Percent Change
in Earnings
0
-5
-10
-15
-20
-25
-30
-35
-40
Stays in Same Occupation
Switches to Different Occupation
Notes: Based on 2001 SIPP panel. Sample is restricted to workers who have been employed for
three months continuously before and after an unemployment spell and switch firms after unemployment. A total of 1,380 unemployment experiences are included. “High Tenure” is defined as
five years with a firm or longer.
effect, i.e., an increased likelihood of
multiple job losses. Specifically, her
study shows that six or more years after
job loss, earnings of job losers remain
approximately 9 percent below those
of their nondisplaced counterparts,
but workers who avoid additional
displacements have earnings losses of
only 1 to 4 percent six or more years
after job loss.
Note that the persistence found
in the literature may not apply to all
unemployed workers. In particular,
the PSID is an annual survey and
thus may possibly miss the majority
of unemployment spells that occur
within a year. As we noted above, one
8 Q4 2009 Business Review
of the advantages of the SIPP is that
it provides high-frequency data that
include short-term unemployment. But
the findings in the earlier studies do
suggest that the earnings of at least
some workers are affected even in the
long run.
CONCLUSION
This article has summarized the
experience of unemployed workers
during and after the 2001 recession,
focusing on changes in earnings
following a period of unemployment.
We found that most of the workers
experienced earnings losses after
unemployment. This is consistent with
earlier findings in the literature, even
though our data set focuses on a short
period of time. Further, larger earnings
losses were associated with loss of
occupation-specific human capital,
a finding that is also consistent with
the results of earlier studies. While
the SIPP does not allow us to assess
the long-term effects of job loss, the
literature suggests that job loss can
have a significant long-term impact
on workers’ earnings and that the
long-term impact takes the form of
an increased likelihood of further job
losses.
From an individual worker’s point
of view, the human capital “specificity”
particularly linked to the worker’s
occupation represents the “human
capital risk.” For instance, in a rapidly
changing economic environment, a
seemingly secure job may not be secure
five years from now. At that point,
workers may be forced to find a job in
a different occupation, in which case
they may need to accept a much lower
wage.
From a macroeconomic point
of view, the presence of significant
earnings losses and “specificity” of
human capital implies that increased
intensity of worker reallocation
during economic downturns is
not simply a reshuffling of workers
between employers. For many workers,
reallocation involves a costly and
time-consuming re-building of human
capital.
Despite the evidence presented
in this article and elsewhere, the
costly and time-consuming nature of
worker reallocation is often ignored
in the typical macroeconomic models
often used in monetary or fiscal
policy analysis. One of the few recent
attempts includes the work by Tom
Krebs. His study focuses on quantifying
the cost of economic fluctuations
when workers face the risk of earnings
losses, such as those discussed in
www.philadelphiafed.org
this article, and finds that businesscycle fluctuations can be quite costly
once earnings losses associated with
job losses are incorporated into the
analysis, which implies a potentially
large benefit of macroeconomic
stabilization policies.
Note, however, that there is
an important caveat regarding the
potential benefit of stabilization
policies: Although stabilization
policies may improve the welfare of the
economy in the short run by avoiding
the costly job reallocation process, they
could potentially interfere with longrun economic growth. For example,
another branch of the literature finds
that labor market policies that impose
costs on firing workers may potentially
have a large negative impact on long-
run growth because such restrictions
allow firms to retain less profitable
jobs.17 It is important that this
consideration also be an integral part
of the discussion. BR
17
See, for example, the article by Hugo Hopenhayn and Richard Rogerson.
REFERENCES
Farber, Henry. “What Do We Know About
Job Loss in the United States? Evidence
from the Displaced Workers Survey, 19842004,” Federal Reserve Bank of Chicago
Economic Perspective (Second Quarter
2005), pp. 13-28.
Fujita, Shigeru. “What Do Worker Flows
Tell Us About Cyclical Fluctuations in
Employment?” Federal Reserve Bank of
Philadelphia Business Review (Second
Quarter 2007).
Fujita, Shigeru. “Creative Destruction and
Aggregate Productivity,” Federal Reserve
Bank of Philadelphia Business Review
(Third Quarter 2008).
Hopenhayn, Hugo, and Richard Rogerson.
“Job Turnover and Policy Evaluation: A
General Equilibrium Analysis,” Journal of
Political Economy, 101:5 (October 1993),
pp. 915-38.
www.philadelphiafed.org
Jacobson, Louis, Robert LaLonde, and
Daniel Sullivan. “Earnings Losses of
Displaced Workers,” American Economic
Review, 83:4 (September 1993), pp.
685–709.
Kambourov, Gueorgui, and Iourii
Manovskii. “Occupational Specificity of
Human Capital,” International Economic
Review, 50:1 (February 2009), pp. 63-115.
Krebs, Tom. “Job Displacement Risk and
the Cost of Business Cycles,” American
Economic Review (June 2007), pp. 664-86.
Moscarini, Giuseppe, and Kaj Thomsson.
“Occupational and Job Mobility in the
US,” Scandinavian Journal of Economics,
109:4, pp 807-36.
Parent, Daniel. “Industry-Specific Capital
and the Wage Profile: Evidence from the
National Longitudinal Survey of Youth
and the Panel Study of Income Dynamics,”
Journal of Labor Economics, 18:2 (2000),
pp. 306-21.
Ruhm, Christopher. “Are Workers
Permanently Scarred by Job
Displacements?” American Economic
Review, 81:1 (March 1991), pp. 319-24.
Stevens, Ann Huff. “Persistent Effects
of Job Displacement: The Importance
of Multiple Job Losses,” Journal of Labor
Economics, 15:1 (1997), pp. 165–88.
Neal, Derek. “Industry-Specific Human
Capital: Evidence from Displaced
Workers,” Journal of Labor Economics, 13:4
(1995), pp. 653-77.
Business Review Q4 2009 9
China’s Emergence as a Manufacturing
Juggernaut: Is It Overstated?*
BEHZAD KIANIAN AND KEI-MU YI
C
hina’s emergence as a manufacturing
juggernaut selling so many goods to so
many countries has attracted enormous
attention from academics, policymakers,
and the media. In this article, Behzad Kianian and
Kei-Mu Yi put China’s manufacturing performance into
a broader context. They emphasize two key themes:
The wages of China’s manufacturing workers are rising
rapidly; and China’s production of export goods relies
heavily on imported inputs and the final exported goods
face large mark-ups in their destination markets. The
first theme implies that China will lose global market
share in some categories of goods. The second implies
that China’s trading relationship with many countries
is complementary, not competitive, and that the
omnipresence of China’s goods exaggerates the extent of
its manufacturing performance. The authors conclude
that China’s emergence as a global manufacturing power
should not be overstated, and concerns that China will
“take over” all manufacturing markets are unfounded.
These days it’s difficult to think of
manufactured goods that are not made
in China. If a product is smaller than
Kei-Mu Yi is a
vice president
and economist
in the Research
Department of
the Philadelphia
Fed.
10 Q4 2009 Business Review
an automobile, it seems, it must have
been made there. China has indeed
become an important, if not dominant,
supplier in global markets for literally
thousands of goods, ranging from
dolls to athletic shoes, from bicycles
to furniture, from steel to air
*The views expressed here are those of the
authors and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.
conditioners, and from telephones to
personal computers. China’s emergence
as a manufacturing juggernaut
selling so many goods to so many
countries has, of course, attracted
enormous attention from academics,
policymakers, and the media. Much
of the media coverage conveys a tone
of concern and consternation at this
rapid emergence.
The purpose of this article
is to put China’s manufacturing
performance into a broader context.
The key themes we emphasize are that
the wages of China’s manufacturing
workers are rising rapidly and that
China’s production of export goods
relies heavily on imported inputs;
these goods also face large markups in their destination markets.
The first theme implies that China
will – and, in fact, has already begun
to – lose global market share in some
categories of goods. The second theme
implies two important points. First,
China’s trading relationship with
many countries is a complementary
one, as opposed to a competitive one.
Second, the omnipresence of China’s
goods exaggerates the extent of its
manufacturing performance. Hence,
we conclude that China’s emergence
as a global manufacturing power
should not be overstated, and concerns
Behzad
Kianian is a research assistant
in the Research
Department of
the Philadelphia
Fed.
www.philadelphiafed.org
that China will “take over” all
manufacturing markets are unfounded.
OVERVIEW OF CHINA’S
ECONOMIC PERFORMANCE
Before delving into our two
primary themes, we believe it is
useful to review China’s economic
performance overall and in
manufacturing, including production
and exports, during the past three
decades. We will not discuss the
theories and hypotheses for China’s
performance; an explanation for
China’s success is a very important but
as yet unanswered question. However,
see Three Important Policy Reforms, for
a brief description of three key reforms
that facilitated the rapid development
of China’s manufacturing sector.
GDP and Manufacturing
Production and Exports. The
broadest measure of a nation’s
economic performance is its gross
domestic product (GDP). GDP can
be measured in three ways. We find
it useful to mention the “product”
approach to GDP, which defines GDP
as the sum of each firm’s “value-added”
– the market value of production
minus the cost of materials and inputs
– in a country in a particular time
period, such as a quarter or year.1 A
1
GDP also includes the economic “output” of
local, state, and federal government. For the
alternative accounting of GDP, it is useful to
think of the government as a large firm that
produces services, such as education, police and
fire protection, and so forth.
Three Important Policy Reforms
C
hina has implemented numerous
economic policy reforms since 1978. We
give a brief overview of three important
trade and foreign direct investment
reforms that have been the most relevant
for China’s manufacturing production
and trade. Much of the description below is from the
study by Nicholas Lardy and another by Lee Branstetter
and Lardy.
Probably the single most important trade policy
reform was the establishment of an export processing
regime. In an export processing regime, raw materials,
parts and components, and other intermediate goods
can be imported duty-free as long as they are used to
produce export goods. According to Lardy, this regime
was developed between the late 1970s and the late 1980s.
This regime greatly facilitated the ability of China’s
domestic and foreign-owned firms to compete in world
markets.
Second, China has reduced its tariff barriers, and
it has become integrated into the official world trading
system. During the 1980s, official tariff rates were as
high as 56 percent, but because of the export processing
regimes, actual tariff collections fell sharply. By 1992,
actual tariffs collected represented less than 5 percent of
total imports. China began sharply reducing its official
tariff barriers during the 1990s. They fell to 15 percent by
2001.
In addition, the U.S. granted China “most favored
nation” status in 1980. This was important because most
favored nation status meant that China had the same
access to U.S. markets as Canada, Mexico, Europe, Japan,
and other countries that were signatories to the General
Agreement on Tariffs and Trade (GATT).a China
officially joined the World Trade Organization (WTO) in
2001. Hence, China lowered its own tariffs, and its most
favored nation status and entry into the WTO meant
that its goods faced lower tariffs.
Third, China implemented policies to encourage
foreign direct investment beginning in 1979. That year,
a legal framework for joint ventures was established,
along with four special economic zones in which “foreign
firms were offered preferential tax and administrative
treatment.” b In 1984, the number of special economic
zones was expanded by 14. In 1986, foreign direct
investment that was export-oriented and technologically
advanced became eligible for additional special benefits.
A key feature of these reforms is that machinery and
equipment could be imported duty-free, as well. These
policies facilitated a surge in inflows of both financial
capital and physical capital so that in recent years China
received more foreign direct investment than the United
States. Importantly, the influx of technology associated
with this foreign direct investment allowed China to
produce more sophisticated products more rapidly than
otherwise.
a
The General Agreement on Tariffs and Trade was the precursor to the World Trade Organization.
b
See the article by Lee Branstetter and Nicholas Lardy, p. 11.
www.philadelphiafed.org
Business Review Q4 2009 11
second way to measure GDP is the
more familiar – to anyone who has
taken a course in macroeconomics
– expenditure approach, which
measures GDP as the sum of four
major categories of spending on final
goods: consumption (C), investment
(I), government purchases (G), and
net exports, or exports – imports
(X-M). These two ways are related
in that the market value of goods and
services produced in a given period
must equal the amount that is spent
on those goods. The measurement of
China’s GDP has sometimes generated
controversy. (See Measurement of
China’s Real GDP, for a discussion of
some of the issues.)
With this caveat in mind, we
will proceed. The growth rate of a
country’s GDP is a simple way to
measure how rapidly a country is
developing. Also, the growth rate of a
country’s GDP per capita is a simple
way to measure how rapidly a country’s
living standards are rising.2
Since 1978, when major economic
reforms were first introduced, China
has experienced very high growth rates
of its GDP. In at least 14 of these years,
annual GDP growth exceeded 10
percent. Since 1980, China’s economy
has increased more than 10-fold and
more than 400 percent since 1987
alone. By comparison, from 1987 to
2006, the economies of the United
States and Japan grew only 76 percent
and 46 percent, respectively.
How large a share of the world
economic pie does China produce?
When converted to dollars at current
exchange rates, China’s GDP as a
2
Both growth rates are important and useful indicators. However, they are imperfect indicators
of a country’s overall development, which would
also include indicators of health, poverty, and
education, for example. For more on China’s
contribution to global economic inequality, see
the article by Keith Sill.
12 Q4 2009 Business Review
Measurement of China’s Real GDP
I
n emerging market economies, the prices of many goods
and services tend to be lower than in the United States, for
example, when the prices are converted at current exchange
rates into dollars. To appropriately compare standards of living
across countries, we must compute an alternative measure
of GDP — a purchasing power parity (PPP) measure. The PPP measure of
GDP adjusts for the price differences across countries. Because prices tend to
rise with income, so that high-income countries also have high prices, PPP
measures have the effect of raising the GDP estimate for countries not as rich
as the United States. For example, using the yuan/dollar exchange rate in
2005, China’s GDP that year was about 5 percent of global GDP. Measured in
PPP terms, it was 14 percent of global GDP.
Recently, the World Bank issued revised PPP estimates of national GDPs
for more than 100 countries for 2005. The revised estimates were based on
more complete data on prices of goods and services. A key result of these
revisions was that China (and India) had its PPP estimates of GDP revised
considerably downward; China’s share of global GDP is now estimated to be 10
percent.
To the extent that prices and changes in prices are understated, this
will overstate the level and growth rate of China’s real GDP. However,
mismeasurement of prices is unlikely to be occurring in the manufacturing
sector because a large fraction of manufactured goods are goods sold on
world markets. The possibility that prices are not measured correctly would
presumably be less than for goods or services that are sold only in China.
share of world GDP has more than
tripled since 1987, reaching 5.5 percent
in 2006 (Figure 1). This made it the
fourth largest economy in the world,
after the United States, Japan, and
Germany. Note, however, that China’s
economy is still considerably smaller
than that of the United States. Also,
the increase in China’s share of world
GDP is not unprecedented, as Japan
experienced a similar jump between
1960 and 1979.
Fueling much of China’s
growth performance has been its
manufacturing sector. As with
countries like Japan, Taiwan, South
Korea, Singapore, and Ireland
before it, China’s rapid economic
development has gone hand-in-
hand with extraordinary growth in
manufacturing. According to the
World Bank’s World Development
Indicator (WDI) database, China’s
share of the world’s manufacturing
value-added increased from 2.8 percent
to 9.9 percent between 1991 and 2005.3
The United States remains the world’s
largest producer of manufactured
goods, but the data suggest that
China will surpass the United States
by 2009 or 2010. Moreover, China is
exporting an increasingly large share
3
To put this large growth in perspective,
consider another large emerging country, India.
During the same period, India’s share of world
manufacturing value-added rose slightly, from
0.9 percent to 1.6 percent.
www.philadelphiafed.org
FIGURE 1
China’s Share of World GDP
World GDP, 1987
World GDP, 2006
United States
27.2%
United States
28.2%
Rest of the
World
58.4%
Rest of the
World
55.7%
Japan
14.5%
Japan
9.0%
China
1.6%
China
5.5%
Current Exchange Rates
Source: World Development Indicators
FIGURE 2
China’s Share of World Manufacturing Exports
2006
1980
Rest of the
World
29.6%
China
0.8%
United
States
13.0%
Rest of the
World
26.3%
Japan
11.2%
United
States
10.0%
Japan
7.1%
NICs* 5.4%
Germany
14.8%
Other
Euro Area
25.1%
China
10.8%
Germany
11.6%
Other
Euro Area
21.9%
NICs*
12.1%
Note: NICs = China, Hong Kong, Singapore, S. Korea, Taiwan
Source: World Trade Organization
of its manufacturing production.
Manufacturing exports as a share of
GDP have more than doubled: from
14.5 percent in 1991 to 33.8 percent in
2006.
On the world market for
manufactured goods, Germany, the
United States, and Japan have long
been the dominant exporters of the
www.philadelphiafed.org
world’s manufactured goods. Figure
2 shows how much the make-up of
the world’s manufactured exports has
changed in the past quarter century.
In 1980, China accounted for less than
1 percent of the world’s manufactured
exports, ranking 21st in the world,
according to the World Trade
Organization. In 2006, China exported
almost $900 billion of manufactured
goods, more than Japan and the
United States, and 10.8 percent of the
world total. Only Germany exported
more.
Composition of China’s Exports.
Moving beyond the aggregate
picture, it is useful to present the
evolution of the composition of
China’s manufacturing exports over
time. China’s abundance of labor
has made it an appealing country for
the production and export of laborintensive goods such as footwear, toys,
and apparel. In 1994, exports of these
and similar goods accounted for over
one-third of China’s total exports and
were almost three times the amount
of China’s exports of computers,
telecommunications equipment, and
other electric machinery (Figure 3).
Over the next 12 years, exports of
the traditional labor-intensive goods
grew rapidly – China continues to be
the world’s leading exporter of many
of these goods – but exports of the
more high-tech goods grew even more
rapidly — so much so, that by 2006,
exports of the high-tech goods were
almost twice as large as exports of
traditional goods.
Tables 1, 2, and 3 present China’s
top 10 exports for three years: 1992,
1999, and 2006.4 They clearly show the
evolution of China’s exports from toys,
footwear, and clothing to electronics
and other telecommunication devices,
including cell phones and computers.
Thus, China’s manufacturing
performance of the past 15 years has
been characterized not only by an
enormous increase in exports and in
4
These data come from the United Nations
Comtrade database. They are categorized under
the Standard International Trade Classification (SITC) system, Revision 3. We use the
four-digit level of categorization, which contains
approximately 600 categories.
Business Review Q4 2009 13
FIGURE 3
China’s Exports Shifting from
Traditional Goods to Machinery
2006
1994
Other
Goods
52.8%
Clothing,
Footwear,
Toys, and
Similar Goods
34.5%
Other
Goods
44.3%
Clothing,
Footwear,
Toys, and
Similar Goods
19.5%
Computers,
Telecom, and
Other Machinery
36.2%
Computers,
Telecom, and
Other Machinery
12.7%
Source: UN Comtrade, SITC, Rev. 3
TABLE 1
China’s Top Exports, 1992
Top Chinese Exports to World, 1992
% of Exports
Petroleum oils, crude oil
3.3%
Toys
2.8%
Jerseys & similar articles
1.9%
Footwear, leather uppers
1.6%
Pants, men’s
1.5%
Other maize, unmilled
1.4%
Other radio-broadcast receivers
1.4%
Bed, table, toilet, and kitchen linen
1.3%
T-shirts & vests
1.3%
Dress shirts, men’s
1.2%
Other footwear
1.1%
Source: UN, SITC, Rev. 3
world market share but also by a steady
shift toward exporting more high-tech
goods in the realm of electronics and
telecommunications equipment.
There have been more formal
and in-depth analyses of the
14 Q4 2009 Business Review
transformation of China’s exports over
time. In his article, Peter K. Schott
uses highly disaggregated U.S. import
data to examine the “sophistication” of
China’s export bundle and how it has
changed over time. Schott compares
the Chinese export bundle to the
U.S. with that of the Organization
for Economic Cooperation and
Development (OECD), a group
of developed economies. Schott’s
measure of sophistication is an export
similarity index, which is equal to 1
if two countries in a given year have
the same set of export goods and each
good’s share of total exports is the
same across the two countries. At the
other extreme, the index is 0 if the
two countries have no export goods
in common. China’s index number
quadrupled between 1972 and 2005,
rising from 0.05 to 0.21. Moreover,
China’s rank in sophistication among
developing countries rose from 19th in
1972 to fourth in 2005. Schott finds
that while China’s sophistication is
consistent with a country of its size,
it is unexpected given China’s level of
development.
Dani Rodrik finds equally
compelling results about the relative
sophistication of China’s exports.
Rodrik uses an indicator that
“measures the productivity level
associated with a country’s export
basket.” Rodrik finds that compared
to other countries, China is a major
outlier. In 1992, for example, the
productivity level associated with
China’s export basket corresponded
to countries with six times the per
capita income of China. Though the
number has shrunk over time, to three
times the per capita income of China
in 2003, Rodrik finds that the initial
high level has been fundamentally
important to China’s enormous
growth.
The common thread in both
Schott’s and Rodrik’s work is that
the story of China’s emergence as a
manufacturing juggernaut is more
than just an enormous increase in
exports. Just as important, if not more
so, is the increasing sophistication
of China’s exports. We now turn to
www.philadelphiafed.org
TABLE 2
China’s Top Exports, 1999
Top Chinese Exports to World, 1999
% of Exports
Children's toys
2.6%
Input or output units for computers
2.5%
Jerseys, pullovers, cardigans, waistcoats
2.0%
Parts, data processing machines
2.0%
Parts, telecommunications equip.
1.9%
Footwear, leather uppers
1.4%
Footwear, rubber/plastic soles/uppers
1.3%
Trunks, suitcases, etc.
1.2%
Plastic articles
1.1%
Storage units, data processing
1.1%
Pants, men’s
1.1%
Source: UN, SITC, Rev. 3
TABLE 3
China’s Top Exports, 2006
Top Chinese Exports to World, 2006
% of Exports
Computers, etc.
4.5%
TV, radio transmitters, etc.
3.7%
Parts, data processing machines
3.4%
Parts, telecommunications equip.
3.2%
Input or output units for computers
2.7%
Electronic microcircuits
2.2%
Sound, video recording, etc.
2.2%
Liquid crystal devices; lasers
1.4%
Jerseys, pullovers, cardigans, waistcoats
1.3%
Television receivers, color
1.3%
Storage units, data processing
1.2%
population means there is an almost
limitless supply of labor available to
work in factories. According to this
story, the large labor force holds down
wages and allows China to extend
its manufacturing tentacles into ever
more categories of goods – from the
most labor-intensive to the most high
tech – and ever more markets abroad.
Figure 4 illustrates the flaw in this
argument because, measured in dollars,
China’s manufacturing wage has risen
at an extremely rapid rate since 1983.
For example, it has increased by 232
percent between 1996 and 2006. By
contrast the manufacturing wage in
the United States rose by only 36
percent in the same period; wages
in two countries competing more
directly with China, Mexico and South
Korea, rose by 60 and 81 percent,
respectively.5
While this trend may be
surprising to some, it is, in fact,
a natural outcome of a rapidly
growing economy supported by
strong manufacturing. A hallmark
of such economies is increased labor
productivity, that is, output per
worker. Increased labor productivity
at the national level can arise from
two broad channels. First, existing
goods can be manufactured in a more
efficient manner, or existing goods can
be manufactured with more capital
per worker. Second, there can be an
increase in the production of “new”
goods – goods that have not been
manufactured by the country before –
whose production makes very effective
use of labor, so that labor productivity
is high. These two channels result
Source: UN, SITC, Rev. 3
5
our first theme in putting China’s
manufacturing performance in
context.
www.philadelphiafed.org
MANUFACTURING WAGES
One of the concerns expressed
in the media is that China’s huge
The source for the U.S., Mexico, and South
Korea wage data is the Bureau of Labor Statistics’ hourly compensation costs in U.S. dollars
in manufacturing: www.bls.gov/news.release/
ichcc.t02.htm. To facilitate a comparison with
Chinese wages, these wages are not adjusted for
inflation.
Business Review Q4 2009 15
FIGURE 4
China’s Manufacturing Wages Rising Rapidly
U.S. $ (annual)
2200
1700
Manufacturing Wages
1200
700
200
1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005
Source: CEIC, IMF
FIGURE 5
Share of Nike Footwear Production
Percent
50
45
China
40
35
30
25
20
15
Vietnam
10
5
0
1995
1997
1999
2001
2003
2005
2007
Source: Nike Annual Reports
in higher wages for workers in most
market-oriented economies. China’s
manufacturing sector tends to be more
market-oriented than other sectors of
its economy because a large fraction of
its production is sold in world markets.
What does the trend of rising
manufacturing wages imply for China’s
16 Q4 2009 Business Review
manufacturing performance? The
primary consequence is that China
is becoming less competitive at
producing goods that other countries,
such as Vietnam or Bangladesh, are
also producing. This would apply, in
particular, to clothing, footwear, and
toys, and similar types of goods.
To illustrate this phenomenon,
consider one of the most prominent
athletic shoe companies, Nike. During
the 1980s a large share of Nike’s
production took place in countries
such as South Korea. However, as
South Korean wages rose, Nike sought
other countries in which to produce
its products. In the 1990s, Nike
increasingly located its production
in China (Figure 5). However, after
2000 – likely owing to China’s rising
wages – this share has declined.
Meanwhile, Nike has found Vietnam
increasingly attractive. From 1995 to
2007, the share of Nike’s production in
Vietnam rose from less than 1 percent
to 31 percent; its current share is now
second to China’s.
Looking at footwear more broadly,
we see a similar pattern. While
China’s share of total U.S. imports of
footwear continues to grow, it is at a
slower rate than before. On the other
hand, U.S. imports of Vietnamesemade footwear are growing rapidly.
Between 2002 and 2006, for example,
China’s share of U.S. imports rose
from 67 percent to 73 percent, while
Vietnam’s share more than tripled
from 1.5 percent to 5.1 percent. The
evolution of both China’s share of
Nike’s production and of China’s
overall footwear share is illustrative of
a larger phenomenon in which rapidly
growing economies like China are also
experiencing rapidly growing wages.
To summarize, the picture we
want to paint in this section has
two brush strokes. The first is that
a key effect of China’s increasing
manufacturing prowess is that
manufacturing wages are rising
rapidly. The second is that rising
manufacturing wages are leading
China to lose market share for some
types of manufactured goods (and also
likely leading China to develop the
ability to produce and export more
sophisticated goods). To be sure, the
www.philadelphiafed.org
types of goods for which China is
losing market share are not the areas
in which the U.S. competes with
China. Our main point is that the
gains in manufacturing prowess overall
lead almost inevitably to declines in
some types of manufactured goods.
CHINA’S IMPORTED INPUTS
AND EXPORT MARK-UPS
We begin by first providing
two examples of China’s market
penetration that some commentators
find worrisome. The total number of
goods that the United States imported
from China doubled between 1989
and 2001, and the share of the total
number of goods increased from 40
percent to 62 percent.6 In other words,
China had a presence in more than
three-fifths of all U.S. markets for
goods by 2001. Only three countries
had a larger presence in U.S. markets.
To be sure, a presence in a large
number of markets does not necessarily
mean that China is exporting a
large dollar amount in each market.
However, in dollar terms, only three
countries exported more to the United
States in 2001, suggesting that China’s
presence in many or most markets is
comparable to that of the other major
countries exporting to the United
States.
Tables 4, 5, and 6 show the top
10 exports by China and South Korea
for 1992, 1999, and 2006.7 They have
become more similar over time. In
1992, only one industry was a top 10
export industry in both countries. In
1999, there were three industries, and
in 2006, there were six industries that
were in the top 10 in both countries.
China’s Imported Inputs. The
above examples suggest that China is
increasingly competitive with countries
such as the U.S. and South Korea, as
well as other countries such as Japan
and Germany. However, a key feature
of Chinese production of its export
goods is that the production relies
heavily on imported intermediate
goods, such as parts and components.
We present two pieces of data on this
issue. Data from the CEIC database
indicates that in recent years, about
40 percent of China’s imports are
intermediate goods that are used
directly to produce China’s exports.8 In
addition, a study by Robert Koopman,
7
8
6
A “good” is defined as a Harmonized Tariff
Schedule (HTS) 10-digit good. See www.usitc.
gov. We thank Christian Broda for providing
these data to us. See the article by Broda and
David Weinstein.
As with the earlier top 10 export data on
China, these data are from the United Nations
Comtrade database and follow the SITC, Rev. 3
categorization.
This is a comprehensive database of national
accounts, trade, industry, financial, employment, and other data for many countries. See
www.ceicdata.com.
TABLE 4
China and South Korea in 1992
Top S.K. Exports to World, 1992
% of Exports
Top Chinese Exports to World, 1992
% of Exports
Electronic microcircuits
8.1%
Petroleum oils, crude oil
3.3%
Ships, boats, etc.
5.4%
Toys
2.8%
Fabric, synthetic yarn
4.1%
Jerseys & similar articles
1.9%
Motor vehicles
3.3%
Footwear, leather uppers
1.6%
Footwear, leather uppers
2.4%
Pants, men’s
1.5%
Input or output units for computers
2.1%
Other maize, unmilled
1.4%
Television receivers, color
1.9%
Other radio-broadcast receivers
1.4%
Sound, video recording, etc.
1.8%
Bed, table, toilet, and kitchen linen
1.3%
Leather apparel, accessories
1.7%
T-shirts & vests
1.3%
Flat-rolled products of iron and steel
1.7%
Dress shirts, men’s
1.2%
Containers
1.5%
Other footwear
1.1%
Source: UN, SITC, Rev. 3
www.philadelphiafed.org
Business Review Q4 2009 17
TABLE 5
China and South Korea in 1999
Top S.K. Exports to World, 1999
% of Exports
Electronic microcircuits
12.4%
Top Chinese Exports to World, 1999
% of Exports
Children's toys
2.6%
Motor vehicles
6.9%
Input or output units for computers
2.5%
Ships, boats, etc.
4.6%
Jerseys, pullovers, cardigans, waistcoats
2.0%
TV, radio transmitters, etc.
2.6%
Parts, data proc. machines
2.0%
Input or output units for computers
2.6%
Parts, telecommun. equip.
1.9%
Fabric, synthetic yarn
2.3%
Footwear, leather uppers
1.4%
Parts, data proc. machines
2.2%
Footwear, rubber/plastic soles/uppers
1.3%
Gas oils
2.1%
Trunks, suitcases, etc.
1.2%
Gold, nonmonetary excl. ores
2.1%
Plastic articles
1.1%
Parts, telecommun. equip.
1.8%
Storage units, data processing
1.1%
Liquid crystal devices; lasers
1.7%
Pants, men’s
1.1%
Source: UN, SITC, Rev. 3
TABLE 6
China and South Korea in 2006
Top S.K. Exports to World, 2006
% of Exports
Top Chinese Exports to World, 2006
% of Exports
Motor vehicles
9.4%
Computers, etc.
4.5%
Electronic microcircuits
7.8%
TV, radio transmitters, etc.
3.7%
Ships, boats, etc.
6.1%
Parts, data proc. machines
3.4%
TV, radio transmitters, etc.
5.3%
Parts, telecommun. equip.
3.2%
Parts, telecommun. equipt.
4.5%
Input or output units for computers
2.7%
Liquid crystal devices; lasers
4.5%
Electronic microcircuits
2.2%
Parts of motor vehicles
2.9%
Sound, video recording, etc.
2.2%
Parts, data proc. machines
2.7%
Liquid crystal devices; lasers
1.4%
Input or output units for computers
1.8%
Jerseys, pullovers, cardigans, waistcoats
1.3%
Cyclic hydrocarbons
1.5%
Television receivers, color
1.3%
Electrical machines and apparatus
1.4%
Storage units, data proc.
1.2%
Source: UN, SITC, Rev. 3
18 Q4 2009 Business Review
www.philadelphiafed.org
Zhi Wang, and Shang-Jin Wei develops
a methodology to accurately compute
the value of imported intermediate
goods directly and indirectly embodied
in exports. Indirect embodiment
occurs if, for example, an imported
intermediate is used to produce
another intermediate good, which is
then used as an input to produce the
exported good. Koopman, Wang, and
Wei report two interesting findings.
First, up to 50 percent of the value of
China’s exports consists of imported
intermediates. Second, the imported
intermediate content is higher for
more sophisticated products, such as
computers and telecommunication
equipment.
These data suggest the following
interpretation. Because China’s
production of its export goods relies
so heavily on imported intermediate
goods, the economic relationship
between China and many of its
trading partners may be more
complementary than competitive.
Indeed, in recent years, the export
data suggest the development of an
East Asian trading network in which
intermediate goods are produced
and exported from “emerging Asia”
and Japan to China, where they
are used to make final goods; the
final goods are then exported to the
United States.9 Between 1994 and
2006, emerging Asia’s and Japan’s
exports to China rose 389 percent
and 191 percent, respectively (Figure
6). In addition, China’s exports to
the United States rose 345 percent.
By contrast, exports shipped directly
from emerging Asia and Japan to the
United States increased by far smaller
percentages. These increases are so
large that emerging Asia now exports
9
Emerging Asia includes South Korea, Taiwan,
the Philippines, Singapore, Indonesia, Malaysia,
and Thailand.
www.philadelphiafed.org
FIGURE 6
China Integrating into Asian Trading Network
Geography of Asian Trade
percent change, 1994 - 2006
Emerging Asia
78%
389%
Greater China
United
States
345%
191%
Japan
24%
Source: IMF Direction of Trade Statistics, National Statistics (Taiwan)
FIGURE 7
China Integrating into Asian Trading Network
Geography of Asian Trade
$U.S. billions, 2006
Emerging Asia
$171
$281
Greater China
$252
United
States
$129
Japan
$147
Note: Emerging Asia = S. Korea, Indonesia, Malaysia, the Philippines, Singapore, Taiwan,
and Thailand
Source: IMF Direction of Trade Statistics, National Statistics (Taiwan)
almost twice as much to China as it
does to the United States (Figure 7).
For example, China has replaced the
United States as South Korea’s largest
trading partner.
This network interpretation
suggests that even though Table 6
suggests that South Korea and China
are now heavily in competition with
each other, the two countries are, in
Business Review Q4 2009 19
fact, exporting different goods; that
is, South Korea exports computer
chips to China, which then uses them
to produce computers. Furthering
the complementary nature of its
trade with other countries, China
has also required capital goods such
as machinery and equipment to fuel
its growth in manufacturing. A large
fraction of these goods are imported
from its richer trading partners.
Mark-Ups on China’s Export
Goods. In addition to their high
content of imported inputs, Chinese
exports often have large mark-ups
once they arrive in their destination
country. Mark-ups include wholesale
distribution costs, retailing costs, and
profit margins. Each of these mark-ups
is an essential part in the process of
coordinating the delivery and ensuring
the quality of a manufactured good
to a consumer. The profit margins
can be thought of as the return to
investment in the good’s intangible
asset capital. The investment could be
the costs associated with developing
a new type of shoe, for example, and
the intangible asset capital would be
the shoe’s brand name. Many Chinesemade goods carry U.S. brands.
Footwear is an excellent example
of this. In 2007, U.S. consumers spent
$59.2 billion on shoes. Close to 100
percent of U.S. expenditure was on
imports.10 As discussed above, about
three-fourths of the imports are from
China. But U.S. imports of shoes in
2007 were about $20.4 billion. The
difference between the U.S. consumer
expenditure value and the value of
imports is the retail and wholesale
costs, transportation costs, and profit
margins. The numbers indicate that
these costs and margins are about
twice as large as the value of the
imported shoes!
Putting the imported input
content and large mark-up forces
together suggests that Chinese “valueadded” – the value of production less
the cost of inputs, that is, wages to
workers plus the rents paid to capital –
in “made in China” goods is not large.
For footwear, Chinese value-added on
the roughly $45 billion of expenditure
on Chinese-made shoes was on the
order of $7.5 billion or less, or about 17
percent of the expenditure.
It is likely that the retail,
wholesale, and transportation costs
and profit margins are not as high
for other U.S. imports from China
as they are for footwear. Consider a
hypothetical case in which these costs
and margins are the same as – rather
than twice as large as – the value
of the imported goods. In 2007, the
U.S. imported $322 billion worth of
goods from China. Hence, in this
hypothetical case, U.S. consumers
spent $644 billion on “made in China”
goods, equivalent to 36 percent of
all U.S. consumer expenditure on
merchandise other than food, fuel, and
automobiles. However, only about $160
billion of this expenditure represents
Chinese value-added (Figure 8).
CONCLUSION
Our main theme is that while
China’s manufacturing growth
has been spectacular – China will
undoubtedly become the largest
manufacturing nation in the world
within a few years – some of the
existing data on its performance
overstate the extent of China’s current
importance in the world economy.
We demonstrated this by showing
that China’s manufacturing wages are
rising rapidly, both in absolute terms,
and relative to other nations, which
means it is losing its status as the
preferred location of production for
some categories of goods, such as Nike
shoes. Moreover, as wages continue to
rise, China will need to continually
produce more sophisticated goods that
require the use of highly productive
labor. We also showed that Chinese
FIGURE 8
U.S. Consumer Expenditure on
Made in China Goods, 2007
Total: $644 billion
+$322
Wholesale Mark-Up
Retail Mark-Up
Domestic Shipping
Profit Margin
+$161
Imported Inputs
+$161
Chinese Value-Added
10
In 2003, 99 percent of all footwear purchased
in the U.S. was imported. In 2002, it was 98
percent. In 2004, the U.S. Census Bureau
discontinued its surveys of U.S. footwear
production.
Source: BEA, Authors’ Calculations
20 Q4 2009 Business Review
www.philadelphiafed.org
export goods rely heavily on imported
inputs and that these goods may have
large mark-ups added on, so that the
final sale price to the consumer of a
“made in China” good exaggerates
by several times the Chinese valueadded to that price. Our conclusion is
that the extent of the manufacturing
juggernaut is overstated.
Our conclusion does not in any
way diminish the current and growing
importance of China in global markets.
To give one example, between 2002
and 2007, the demand for oil in China
– fueled by China’s high GDP growth
rates – rose by two-and-one-half
times more than it did in the United
States. This contributed to the global
increase in the price of oil during this
period. Moreover, China’s wages are
not rising by accident. They are rising
because the country is becoming more
productive, as well as more capable of
producing a wider range of goods, as
well as higher quality goods.
If China follows the pattern of
many other countries, eventually
the manufacturing juggernaut may
actually diminish in size. Many
countries go through a “structural
transformation” as they develop, in
which large numbers of workers leave
the agricultural sector and the share of
agriculture in GDP falls. In addition,
the services sector grows as a share of
total employment and GDP. Finally,
the manufacturing sector typically
increases in importance during the
high growth years, but then falls
in importance when the economy
matures. For example, in the United
States, manufacturing’s share in GDP
has fallen steadily during the postWorld War II period. In Japan, the
manufacturing share of GDP peaked
in the early 1970s and has fallen
steadily since then. It is likely, then,
that as China’s per capita income
and wages rise to developed-country
status, manufacturing will decrease in
importance. BR
Branstetter, Lee, and Nicholas Lardy.
“China’s Embrace of Globalization,” NBER
Working Paper 12373 (July 2006).
Lardy, Nicholas R. Integrating China into
the Global Economy. Washington, D.C.:
Brookings Institution Press, 2002.
Sill, Keith. “The Evolution of the World
Income Distribution,” Federal Reserve
Bank of Philadelphia Business Review
(Second Quarter 2008).
Broda, Christian, and David Weinstein.
“Are We Underestimating the Gains from
Globalization for the United States?”
Federal Reserve Bank of New York Current
Issues (April 2005).
Rodrik, Dani. “What’s So Special About
China’s Exports?” NBER Working Paper
11947 (January 2006).
REFERENCES
Koopman, Robert, Zhi Wang, and ShangJin Wei. “How Much of Chinese Exports Is
Really Made in China? Assessing Domestic
Value-Added When Processing Trade Is
Pervasive,” NBER Working Paper 14109
(June 2008).
www.philadelphiafed.org
Schott, Peter K. “The Relative
Sophistication of Chinese Exports,”
Economic Policy (January 2008), pp. 5-49.
Business Review Q4 2009 21
News About the Future and
Economic Fluctuations*
BY KEITH SILL
I
n the late 1990s, as tech-stock prices were
surging, we often heard discussion about
a “new economy” in which advanced
communications technologies would lead to
higher future productivity growth and greater economic
efficiency. But the boom times largely came to a halt after
August 2000, and in March 2001, the economy entered
a recession that lasted eight months. Economist A.C.
Pigou argued that news about the future or changes in
expectations are important drivers of the business cycle.
His theory seems to offer a plausible explanation of what
happens in boom-bust cycles. But is his theory consistent
with how modern macroeconomic models account for
business cycles? In this article, Keith Sill investigates some
of the empirical evidence for the economic importance
of news shocks, discusses the failings of the standard
macroeconomic model in accounting for the role of news
in business cycles, and touches on what the news view of
business cycles means for the conduct of monetary policy.
Our expectations about how the
future will unfold can have important
implications for the choices we make
today. An expectation of future
Keith Sill is an
assistant vice
president and
director of the
Real-Time Data
Research Center
in the Research
Department of the
Philadelphia Fed.
22 Q4 2009 Business Review
unemployment might result in reduced
consumption and higher savings
today. Or an expectation of a future
promotion and higher salary may
induce higher consumption and lower
saving today, even before the higher
salary is realized. This rather obvious
feature of individual behavior may
have important implications for the
*The views expressed here are those of the
author and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.
economy as a whole. Macroeconomic
aggregate variables such as
consumption and investment could
rise in response to a collective belief
that the economy will experience
higher productivity in the future.
A recent example of how
collective beliefs can influence
economic variables is the dot-com
boom and bust of the late 1990s. In
the late 1990s, as tech-stock prices
were surging, we often heard discussion
about an impending “new economy”
in which advanced Internet and
communications technologies would
lead to higher future productivity
growth and greater economic
efficiency. We could argue that those
collective beliefs about the future
became embedded in stock prices
and led to dramatic gains in the
equity prices of technology-related
companies. In turn, higher stock
prices made households feel wealthier,
which induced increased consumption.
Businesses began investing in the
emerging technologies in the hopes of
generating higher future profits.
These boom times, seemingly
driven at least in part by overly
optimistic expectations about
the future, largely came to a halt
after August 2000. After the fact,
expectations proved to be optimistic.
Over the next three years, the stock
market declined on the order of 40
plus percent. In March 2001, the
economy entered a recession that
lasted eight months. The level of real
private nonresidential fixed investment
(business fixed investment) declined
16 percent from the fourth quarter of
2000 to the first quarter of 2003. The
boom was followed by a bust.
www.philadelphiafed.org
In 1927, A.C. Pigou, an economics
professor at Cambridge University who
studied business cycles, wrote a book
called Industrial Fluctuations. In that
book, Pigou argued that news about
the future or changes in consumers’
and businesses’ expectations are
important drivers of the business
cycle and economic fluctuations. In
particular, when firms and suppliers
of capital are optimistic about the
future, they decide to invest more
today in order to accumulate capital to
meet higher expected future demand.
If it turns out that expectations are
overly optimistic, firms pull back on
investment and consumers retrench,
leading to an economic downturn or
recession.
This seems to be a plausible
explanation of what happens in boombust cycles like the dot-com episode.
But do the data really support this
story, and are Pigou cycles pervasive
features of modern economies?
Also, is Pigou’s theory consistent
with how modern macroeconomic
models account for business cycles?
An emerging body of empirical
evidence supports the view that news
about the future is an important
factor in explaining fluctuations in
output and employment. However,
the standard workhorse model used
by macroeconomists predicts that
good news about the future leads
to what looks a lot like a recession
today! If good news about the future
results in booms today, the standard
macroeconomic model needs some
modification if it is to explain such
behavior.
We will investigate some of the
empirical evidence for the economic
importance of news shocks and how
they affect the economy. We will also
discuss the failings of the standard
macroeconomic model when it comes
to accounting for the role of news
in business cycles. A recent line
www.philadelphiafed.org
of research explores this issue and
examines the features necessary to get
models to predict booms in response to
good news about the future economy.
Finally, we will touch on what the
news view of business cycles means for
the conduct of monetary policy.
stock market is a key component of
the analysis because it is generally
perceived to be forward-looking in the
sense that news that people receive
about future prospects for the economy
should be reflected right away in stock
prices, since participants trade on that
information.
Is there any hard evidence that changes
in expectations about the future
in economic activity today?
EMPIRICAL EVIDENCE
ON EXPECTATIONS AND
FLUCTUATIONS
Is there any hard evidence that
changes in expectations about the
future lead to significantly large
changes in economic activity today?
The key problem that must be
addressed when deciding whether
news about the future affects the
economy is separating the scenario
“economic booms lead to changes
in expectations” from the scenario
“changes in expectations lead to
economic booms.” That is, we have
to account for the fact that changes
in current economic activity also give
rise to changes in expectations about
the future economy. Once we control
for that possibility, we can investigate
the extent to which changes in
expectations can drive economic
fluctuations.
A recent paper by Paul Beaudry
and Franck Portier provides some
evidence of the importance of news
about the future to fluctuations
in economic variables such as
consumption and hours worked.
Beaudry and Portier undertake
a statistical analysis of data on
productivity and stock market prices
to investigate this question. The
Beaudry and Portier are able to
tease shocks out of the data on stock
prices and productivity that give
insight into how expectations about
the future affect today’s economy.1
They find that their shocks contain
information about future productivity
growth that is also reflected in
current stock prices. In addition,
they find that long-run changes in
productivity are reflected in stock
prices before these changes show
up in near-term productivity. These
findings are consistent with the view
that financial market participants can
anticipate productivity improvements,
perhaps because there is a long delay
between receiving news about a new
productivity-enhancing technology
and the realization of higher
productivity once the technology is
implemented. Beaudry and Portier call
this the “news view.”
We can interpret the shocks
that Beaudry and Portier identify
as news shocks because they
represent unpredicted or unexpected
information that shows up in
productivity and stock prices. This
1
For our purposes, a shock can be thought of as
the difference between a predicted outcome and
the actual outcome.
Business Review Q4 2009 23
is what news is: information that
wasn’t previously available that tells us
something about final outcomes.
With news shocks in hand, we
can now investigate whether changes
in news about the future have an
impact on current variables such as
consumption, investment, and hours
worked. Figure 1 shows how stock
prices, per capita consumption, and
hours worked per capita respond to
a positive news shock. The figure
shows the response of these variables
over time to two different measures
of the news shock, though we see
that it makes little difference which
shock we focus on, since they both
imply the same paths for stock prices,
consumption, and hours worked. If
there were no response to the news
shocks, the lines in the figure for stock
prices, consumption, and hours would
stay near zero. What we see instead
is that stock prices, consumption,
and hours worked all jump up right
away in response to positive news.
Consumption and hours worked
continue to rise for about five quarters
and then give up some of their gains
in apparent recession-like behavior.
Eventually, consumption resumes its
general upward trend. Hours worked
flatten out because hours per capita
tend not to rise over time. (People
do not work more and more hours as
productivity increases — leisure is
valuable, too!)
Beaudry and Portier also
investigate how much of the variation
in consumption and hours worked
can be explained by their identified
news shocks.2 This is a measure of how
economically important such shocks
might be. They find that news shocks
FIGURE 1
Response of Stock Prices, Consumption, and
Hours Worked to Two Measures of Positive News
About the Future
Percent Deviation
10
Stock Prices
8
6
4
2
0
-2
-4
0
5
10
15
20
25
30
35
40
25
30
35
40
25
30
35
40
Quarters
Percent Deviation
1.4
Consumption
1.2
1
0.8
0.6
0.4
0.2
0
0
5
10
15
20
Quarters
Percent Deviation
1.4
Hours
1.2
1
0.8
0.6
0.4
0.2
2
More precisely, they compute how much of
the variance of forecast errors for consumption,
hours, and investment can be explained
by news shocks. These are called variance
decompositions.
24 Q4 2009 Business Review
0
0
5
10
15
20
Quarters
From Beaudry & Portier (2006). Used with permission. Dotted gray lines indicate 95 percent
confidence interval
www.philadelphiafed.org
account for 40 to 80 percent of the
variation in consumption, investment,
and hours worked over the postwar
period. This is a huge number and
suggests that news about the future
may be an important determinant of
the economy’s fluctuations.
A second piece of evidence on
the importance of news shocks for
economic fluctuations can be found
in recent research I conducted with
Sylvain Leduc. We use data from
the Philadelphia Fed’s Livingston
Survey and the Survey of Professional
Forecasters (SPF) to identify news
shocks and to assess their effect on
variables such as the unemployment
rate, stock prices, and inflation. Both
the Livingston and the SPF are surveys
of professional forecasters who are
asked to make forecasts of a range
of macroeconomic variables. The
Philadelphia Fed then tabulates and
publishes the forecasts.3
Survey data give us a unique
insight into expectations of the future,
since they are a direct measure of such
expectations. Since we know the time
at which the surveys are conducted,
we can use that information to help
us identify news shocks. That is,
broadly speaking, we know which
realizations of economic variables the
forecasters already knew or had in
hand when they made their forecasts.
So forecasters for a June survey would
know May unemployment rates but not
June unemployment rates, since those
data would not have been released yet.
We can use that type of information to
identify news shocks and assess their
impact on economic variables. More
specifically, we analyze a statistical
model that contains forecasts of
future unemployment rates, current
3
A description of the surveys and survey data
are available on the Philadelphia Fed’s website
at http://www.philadelphiafed.org/research-anddata/real-time-center/.
www.philadelphiafed.org
unemployment rates, interest rates,
and inflation. We align the data in
such a way as to help us identify shocks
to forecasts of future unemployment
that are not driven by the other
variables in the system. We interpret
these shocks as news about the future
that changes people’s expectations
because the shocks are the difference
unemployment. The second row shows
how actual, or current, unemployment
responds to the shock. We see that
the unemployment rate falls, so that
in response to expectations of future
bad times, current times turn better.
The third row shows the response of
inflation to the news shock. Consistent
with the current upturn story, near-
Survey data give us a unique insight into
expectations of the future, since they are
a direct measure of such expectations.
between what we expect the predicted
unemployment rate to be and what the
prediction actually turns out to be. In
this context, a positive shock is bad
news because it implies that forecasts
of future unemployment rates were
higher than we would have predicted
them to be.
Once we have these news shocks,
we can examine how they affect the
other variables in the model, such as
the unemployment rate, CPI inflation,
and interest rates.4 Figure 2 presents
a set of responses from such a shock.
The two columns from the figure
show which survey measure was used
in the model: The left side shows the
Livingston Survey and the right side,
the Survey of Professional Forecasters.
Each row represents the response of a
different variable to a news shock that
decreases the expected unemployment
rate (what economists call impulse
responses). The top row of the figure
shows how forecasts of six-monthahead unemployment rates evolve
when there is a news shock that leads
forecasters to expect lower future
4
We chose these variables because they are not
generally subject to data revisions over time.
term inflation rises. The next row
shows the response of stock returns,
measured using the S&P 500, to the
news shock. Here we see that when
news about expected good times
arrives, the current stock market rises.
Finally, the last row shows the response
of short-term interest rates to the news
shock. Here, we see that monetary
policy tightens as the economy booms
in the near term in response to the
good news shock.
The impulse responses in Figure
2 suggest that the current economy
surges when the future economy is
expected to be better. But how strong
are the results? Plotted around the
impulse responses are confidence
bands that indicate how sure we
are that a response is different from
zero. We see from the figure that
the responses of all the variables
are significantly different from zero.
To further assess the strength of
the results, we performed variance
decompositions similar to those in the
Beaudry and Portier paper described
above. We find that about 50 percent
of the variability of forecast errors for
our variables can be explained by news
shocks, a result that is in line with the
findings in Beaudry and Portier. So, in
Business Review Q4 2009 25
FIGURE 2
Economic Response to Good News About the Future
Livingston Survey
Survey of Professional Forecasters
Expected Unemployment Rate
1
1
0.5
0.5
0
0
-0.5
-0.5
-1
-1
-1.5
1
2
3
4
5
-1.5
6
1
2
3
4
5
6
1
2
3
4
5
6
1
2
3
4
5
6
1
2
3
4
5
6
1
2
3
4
5
6
Actual Unemployment Rate
1
1
0.5
0.5
0
0
-0.5
-0.5
-1
-1
-1.5
1
2
3
4
5
-1.5
6
CPI Inflation
3
3
1.5
1.5
0
0
-1.5
-1.5
1
2
3
4
5
6
S&P 500 Index
0.2
0
0
-0.2
0.2
1
2
3
4
5
6
-0.2
3 - Month Treasury Bill Rate
3
3
2
2
1
1
0
0
-1
1
2
3
4
5
6
-1
The responses were generated from a VAR with expected unemployment percent, actual unemployment, inflation, equity prices, the 10-year T-bill
rate, the 3-month T-bill rate, and dummy variables for oil and fiscal shocks. All of the responses are expressed in percentage terms. The x-axis
denotes years. In each chart, the darker area represents the 68 percent confidence interval, while the sum of the darker and lighter areas denotes the
90 percent confidence interval.
26 Q4 2009 Business Review
www.philadelphiafed.org
sum, we find that the economy surges
in response to expectations of better
times ahead and that the response
of the unemployment rate, inflation,
and the short-term interest rate are
different enough from zero and explain
enough of the variance of the series
that we can be pretty confident that it
is not a statistical fluke.5
THE PIGOU CYCLE
We have seen some of the
empirical evidence that suggests
that changes in expectations about
the future can alter aggregate
economic outcomes today. That is,
news about the future seems to be a
significant driver of current economic
fluctuations. What does economic
theory have to say about how we might
interpret the statistical evidence?
In the early 1900s, A.C. Pigou
wrote: “The varying expectations
of businessmen ... constitute the
immediate cause and direct causes or
antecedents of industrial fluctuations.”
In other words, Pigou believed that
changes in expectations about the
future were a principal cause of
business cycles in the economy.
If people were optimistic about
the future, current consumption,
investment, and output would rise. If
they were overly optimistic, once they
realized that their expectations were
too rosy, the economy would go into
recession as businesses and households
pulled back on their spending. A
Pigou cycle then can arise when
5
See also the paper by Robert Barsky and
Eric Sims. They examine how output and
consumption respond to innovations using
responses from the Michigan Survey. They
also find that changes in people’s expectations
about the future lead to significant changes in
current output, consumption, and productivity.
Stephanie Schmitt-Grohe and Martin Uribe
estimate an equilibrium model with news
shocks and find that news about the future can
account for a substantial fraction of economic
fluctuations.
www.philadelphiafed.org
output, consumption, investment,
and hours worked jointly increase
in response to an anticipated rise in
productivity. When the anticipated
increase fails to materialize, a recession
ensues.
This view of booms and busts
seems consistent with the way
events unfolded during the dotcom bubble. Expectations about
higher future productivity driven
News about the
future seems to be
of current economic
does economic
theory have to
say about how we
might interpret the
statistical evidence?
by Internet-related technologies led
to an investment boom in products
such as fiber-optic cable. The stock
market value of technology stocks
rose to unprecedented highs. The
rationalization for such high valuations
was that the economy was entering
a new era of high productivity that
should be reflected in future stock
earnings and dividends. After the
fact, these expectations turned out to
be overly optimistic, and the dot-com
bust dovetailed into the recession that
began in 2001.
Anecdotal evidence suggests
that people receive and process news
about the future and that such news
can affect behavior. For example,
stock prices and consumer confidence
measures are thought to lead the
business cycle. We can tell stories that
seem consistent with the Pigou cycle
theory of booms and busts. However,
it turns out that this view of the world
does not work particularly well in the
standard workhorse model of modern
macroeconomics.6 In fact, in the
standard model, good news about the
future in the form of higher expected
productivity can lead to a drop in
hours worked, output, and investment
today. In the standard model, expected
booms lead to what look a lot like
recessions today!
To develop some insights into
why the standard model gives this
result, consider first a case in which
households observe an increase in
current productivity that they expect
will persist into the future. With
persistently higher productivity,
households are wealthier, since their
current and expected future real
incomes are higher (for example, real
wages rise with productivity in the
standard model). In this case, output,
employment, consumption, and
investment all rise today.
Two forces are at work behind
this result. The first is a wealth effect.
Higher productivity means higher real
income in the standard model. Thus,
household wealth increases, and being
richer induces more consumption
today, since households like to smooth
out their consumption over time.
But higher wealth also means that
households want to consume more
leisure; so the wealth effect predicts
that hours worked will fall. Offsetting
the impact of the wealth effect on
work effort is a substitution effect. The
substitution effect says that households
6
By a standard macroeconomic model I am referring to the neoclassical growth model. That
model is one of a representative household that
maximizes its consumption and leisure, subject
to the constraint that consumption and investment are no greater than what can be produced
with capital on hand and labor effort. For a very
accessible discussion of the neoclassical growth
model, see the Business Review article by Satyajit
Chatterjee.
Business Review Q4 2009 27
work harder when productivity is
higher and then invest the proceeds
to attain higher consumption in the
future when productivity is lower.
Thus, the substitution effect indicates
that in response to higher productivity
today, households work harder,
consume less, and save more.
Which effect is dominant: wealth
or substitution? It depends first on how
persistent the increase in productivity
is expected to be. The more persistent
the rise in productivity, the stronger
the wealth effect. Also important
is how responsive labor supply is to
changes in the real wage. If labor
supply increases a lot in response to
an increase in wages, the substitution
effect becomes stronger. Figure 3 shows
how consumption, investment, output,
and hours worked respond in the
standard model to a productivity shock
calibrated in the standard way — a
fairly persistent shock. We see that for
labor supply, the substitution effect
dominates the wealth effect and hours
worked increase. In addition, output,
consumption, and investment all rise
in response to a positive productivity
shock.
Consider now what the model
predicts if the productivity shock is
expected to affect the economy in
the future but not directly today. In
anticipation of higher real wages in the
future, households feel wealthier today
and so spend more on consumption
and leisure. Because the productivity
shock hits in the future, there is not
a strong substitution effect today.
(Households are not more productive
today; they only expect to be in the
future.) Consequently, households have
little incentive to work harder today,
since they are no more productive
than before. Thus, in response to
higher expected future productivity,
current consumption rises and hours
worked fall. With lower hours worked,
output falls. Since output falls and
consumption rises, investment must
fall, since output equals consumption
plus investment (we are ignoring
net exports). Thus, an expected
FIGURE 3
Standard Model: Response to a Positive Productivity Shock
Consumption Response
Investment Response
0.009
0.008
0.007
0.006
0.005
0.004
0.003
0.002
0.001
0.004
0.0035
0.003
0.0025
0.002
0.0015
0.001
0.0005
0
-0.001
0
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
Quarters
Quarters
Output Response
Hours Response
0.012
0.002
0.01
0.0015
0.008
0.001
0.006
0.0005
0.004
0
0.002
-0.0005
0
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
Quarters
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
Quarters
Panels show the response of consumption, investment, output, and hours to a percent increase in the productivity shock at time 1.
28 Q4 2009 Business Review
www.philadelphiafed.org
boom in productivity leads to lower
output and hours worked today (but
higher consumption). This intuition
is revealed in the impulse responses
shown in Figure 4. Here, we show the
response of consumption, investment,
hours worked, and output to a shock
that signals that productivity will rise
one year from now.
So it seems that the standard
model does not deliver a result about
the effect of news on the economy that
agrees with the empirical evidence
we presented earlier. Remember, that
evidence suggested that in response
to good news about the future
economy, there is a boom today, with
consumption, output, and investment
all increasing. Is there a model whose
predictions agree with that evidence?
It turns out that a modified
version of the standard model
can predict a boom in response to
expectations of good times in the
future. The standard model has to be
modified so that the wealth effect on
labor supply is not strong. In addition,
various other frictions must be added
to the model so that both consumption
and investment respond positively to
good news about the future. These
modifications are detailed in a recent
paper by Nir Jaimovich and Sergio
Rebelo. They allow firms to vary
the intensity with which they use
capital, which is important because
it increases the extent to which
output can respond to news about
the future. They also assume that
it is costly for firms to adjust their
stock of capital, which gives firms an
incentive to respond immediately to
future productivity changes in order to
smooth out costs over time.
A somewhat different approach
is taken in a recent paper by Wouter
den Haan and Georg Kaltenbrunner.
They postulate that in order to
benefit from future productivity
gains, firms and households have to
invest resources today. In den Haan
FIGURE 4
Standard Model: Response to News Today That Productivity Will
Increase in Four Quarters
Consumption Response
Investment Response
0.0035
0.01
0.008
0.004
0.003
0.006
0.0025
0.004
0.002
0.002
0.0015
0
0.001
-0.002
0.0005
-0.004
0
-0.006
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
Quarters
Quarters
Output Response
Hours Response
0.012
0.002
0.01
0.0015
0.008
0.001
0.006
0.0005
0.004
0.002
0
0
-0.0005
-0.002
-0.001
-0.004
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
Quarters
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39
Quarters
Panels show the response of consumption, investment, output, and hours to a 0.01 percent increase in the productivity shock at time 1 that is
realized at time t=4.
www.philadelphiafed.org
Business Review Q4 2009 29
and Kaltenbrunner’s model, firms
and workers that are not already
engaged in production when news
about higher future productivity is
revealed need to get together today
and form productive relationships.
Since building productive relationships
requires both time and resources,
firms start investing in new projects
right away and immediately begin
looking for new workers with whom to
build productive relationships. Thus,
employment, investment, and output
rise in response to expectations of
higher future productivity growth in
their model.
So we see that there are several
reasonable approaches we might take
in order to get a coherent theoretical
model of the economy that has the
feature that expectations of good
times in the future lead to booms
today. Discriminating among these
alternative modeling strategies is only
at the earliest stages in the economics
profession. Time will tell which
modeling strategy most closely aligns
with the regularities found in the data.
EXPECTATIONS, BUSINESS
CYCLES, AND MONETARY
POLICY
If economic variables such as
stock prices, output, employment,
consumption, and investment
do respond in a meaningful and
important way to expectations about
the future, what are the implications
for policymakers? Recently, the
economy has experienced an unusual
amount of asset-price volatility
whose source can perhaps ultimately
be traced to overly optimistic
expectations about continued increases
in house prices. When house prices
began falling instead of rising, financial
markets began to perform badly, and
a downturn in real economic activity
ensued. This episode is not unique.
Over the past 20 years or so, several
30 Q4 2009 Business Review
boom-bust cycles have unfolded
around the world, including Japan in
the late 1980s and East Asia in the late
1990s. These episodes have generated
debate about the importance of the
role played by monetary policy in
booms and busts: Often the episodes
were accompanied by heightened
criticism of central banks for fueling
the booms by keeping monetary policy
too easy for too long.
Asset-price run-ups and asset-price
close to some level — say, 2 percent
— over a suitably defined length of
time. There is some reason to expect
that such a monetary policy will act as
a natural stabilizing force with respect
to boom-bust cycles. The inflationtargeting approach to monetary
policy dictates that monetary policy
should be adjusted to offset emerging
inflationary or deflationary pressures.7
Bernanke and Gertler argue that by
focusing on inflation, central banks
Recently, the economy has experienced
an unusual amount of asset-price volatility
whose source can perhaps ultimately be
traced to overly optimistic expectations
about continued increases in house prices.
volatility seem to be key features of
expectations-driven booms in practice.
This raises a question about the extent
to which monetary policymakers
should take asset prices into account
when setting policy. Unfortunately,
it is difficult to determine the extent
to which asset prices are aligned with
“true fundamentals” or are being
driven by nonfundamental factors.
For monetary policymakers who are
concerned with stabilizing inflation
and employment growth, determining
the “right” level of asset prices seems
a tall order. However, it may be the
case that by focusing on stabilizing
inflation and employment growth,
policymakers can stabilize asset prices
as a byproduct. This is the message
of a study by Ben Bernanke and Mark
Gertler.
Consider the case of a central
bank that operates monetary policy
in such a way as to try to achieve an
inflation target. That is, the central
bank’s mandate is to keep inflation
in effect respond to the bad effects of
booms and busts without having to
take an explicit stand on whether asset
prices are valued fairly (according to
economic fundamentals) in booms. For
purposes of the ensuing discussion, we
will say that, in a boom, asset prices
are rising, an assumption that agrees
with most definitions of booms in the
data.
How can inflation targeting end
up “getting it right” with respect to
policy and expectations-driven cycles?
7
Bernanke and Gertler actually argue for a regime of flexible inflation targeting, which, they
contend, has three characteristics. The first is
that monetary policy is committed to attaining
a target level of inflation in the long run and
price stability is the overriding goal of monetary
policy. Second, within the constraints imposed
by achieving a long-run inflation target, policymakers have some flexibility in the short run to
achieve other objectives, such as stable output
and employment. Third, there is a commitment
to transparency and openness on the part of
monetary policymakers so that private-sector
expectations about policy and the economy are
well grounded.
www.philadelphiafed.org
Bernanke and Gertler argue that
inflation targeting leads policymakers
to automatically adjust interest rates in
such a way as to stabilize the economy
in the face of booms. The idea is that
booms are associated with increases
in demand; that is, consumption,
investment, and ultimately output
rise. In Bernanke and Gertler’s view,
increases in demand are in turn
associated with rising inflation. But an
inflation-targeting central bank will
raise the interest rate in response to
rising inflation. In effect, the central
bank leans against the wind. This
reins in the increase in demand and
stabilizes financial markets as well.
Financial markets are likely to
stabilize for several reasons. The first
is that the stability of the broader
economy is, in itself, stabilizing for
financial markets. Second, suppose the
economy starts to go into recession and
asset prices start to decline — which
will tend to erode the balance sheets
of banks (and other firms, as well).
The falloff in demand and declining
inflation call for policymakers to lower
the interest rate, which can reduce
the economy’s vulnerability to further
bad shocks. Finally, if financial market
participants expect policymakers to
act in this way, it may mean that the
overreaction of asset prices might
be moderated. Overreaction could
occur if asset prices are in part driven
by a market psychology or some
other factor, such as poor regulatory
practices, not directly fundamental to
determining asset prices.
Bernanke and Gertler’s paper
is really about monetary policy and
asset-price volatility. They note that
financial stability is becoming an
increasing concern for monetary
policymakers because, over the past
25 years, a number of countries have
experienced major boom-bust cycles
in the prices of assets such as equities
and real estate. Associated with the
www.philadelphiafed.org
bust part of the cycles, as asset prices
are falling, real economic activity
is declining significantly. We have
presented evidence that changes in
expectations that can influence real
activity also show up in asset prices,
such as stock prices. So expectationsdriven cycles fit naturally into the
asset-price boom-bust cycles with
which Bernanke and Gertler are
concerned.
If asset prices
fall, the amount of
collateral falls, which
raises the ratio of
borrowing relative
to assets, worsens
balance-sheet
positions, and makes
it harder to borrow.
In the Pigou cycles story, the bust
part of the cycle comes about when
overly optimistic expectations are not
realized and firms and households
cut back on their consumption and
spending. Bernanke and Gertler
point out another negative force at
work in the bust part of the cycle:
negative balance-sheet effects on
firms and households from declines
in asset prices. This channel can be
important because credit markets
are characterized by problems such
as differential information between
parties to a contract, problems of
contract enforcement, and misaligned
incentives between lenders and
borrowers, or managers and investors.
Because these problems exist, credit is
most widely available and on the best
terms to institutions and households
that have strong balance sheets (i.e.,
are in good financial shape with
respect to their assets and liabilities).
So balance-sheet conditions become
important determinants of borrowing
and lending. But falling asset prices
can have an adverse impact on balance
sheets because firms and households
may use the assets they own as
collateral for borrowing. If asset prices
fall, the amount of collateral falls,
which raises the ratio of borrowing
relative to assets, worsens balancesheet positions, and makes it harder to
borrow. In turn, the reduced borrowing
lowers demand in the economy and
may also adversely affect supply by
reducing working capital for firms and
inhibiting investment. These factors
work to further slow down economic
activity and worsen economic
downturns.
Thus, it can be quite important for
monetary policymakers to recognize
the downside of an expectationsdriven boom-bust cycle. If there
is a significant decline in asset
prices, households and firms face
greater difficulty in financing their
consumption and investment decisions,
which lowers aggregate demand and
can make economic downturns more
severe. The good news is that in the
Bernanke and Gertler story, central
banks can help alleviate these policies
not by focusing policy on movements
in asset prices but rather by focusing
on inflation. Asset prices stabilize as a
consequence.
However, there may be
circumstances in which inflation
targeting does not lead to a good
outcome in the face of asset booms.
Lawrence Christiano, Roberto Motto,
and Massimo Rostagno make this
point in their paper. They look at
asset price swings since the 1870s and
argue that there were three significant
boom-bust episodes: one that began
in 1920 and ended with the Great
Depression, one that began in the
mid 1950s and ended in the 1970s,
Business Review Q4 2009 31
8
Sticky prices are prices that are slow to
respond to changes in supply or demand. Similarly, sticky wages are wage levels that are slow
to respond to changes in the labor market.
9
See the paper by Lawrence Christiano, Martin
Eichenbaum, and Charles Evans for a more
detailed discussion of model features that lead
to a better match with the data.
32 Q4 2009 Business Review
FIGURE 5
5
4
Nominal S&P 500
3
Real S&P 500
2
1
PCE Inflation
0
-1
47
19 -Q2
49
19 -Q1
50
19 -Q4
52
19 -Q3
54
19 -Q4
56
19 -Q1
57
19 -Q4
59
19 -Q3
61
19 -Q2
63
19 -Q1
64
19 -Q4
66
19 -Q3
68
19 -Q2
70
19 -Q1
71
19 -Q4
73
19 -Q3
75
19 -Q2
77
19 -Q1
78
19 -Q4
80
19 -Q3
82
19 -Q2
84
19 -Q1
85
19 -Q4
87
19 -Q3
89
19 -Q2
91
19 -Q1
92
19 -Q4
94
19 -Q3
96
19 -Q2
98
19 -Q1
99
20 -Q4
01
20 -Q3
03
20 -Q2
05
20 -Q1
06
20 -Q4
08
-Q
3
-2
19
and one that began in the mid 1990s
and ended in the early 2000s. Their
model includes inflation-targeting
monetary policymakers in an economy
with sticky wages and prices as well
as adjustment costs to investment.8
In that environment, boom-bust
cycles can easily arise. A feature that
distinguishes their paper is sticky
wages, which means that nominal
wages are not fully flexible in response
to the shocks hitting the economy, but
rather take time to adjust to the new
equilibrium level. Some researchers
have argued that this feature of the
model is important for matching
certain features of the data on the
economy.9
Suppose then that nominal wages
are sticky. How does this cause a
problem for an inflation-targeting
central bank? When the boom phase
starts, it is typical in macroeconomic
models for real wages (defined as the
nominal wage divided by a general
price index) to rise to induce people to
work harder. But with sticky nominal
wages, the only way that happens is
if prices start to fall. An inflationtargeting policymaker sees the drop in
inflation and so eases monetary policy
by reducing interest rates in order to
stimulate demand and push inflation
back up to the target level. But this
stimulative action ends up feeding the
already-present optimism about the
economy and generates even faster
growth of consumption, investment,
and output. Monetary policy ends up
making the boom even bigger, and the
eventual bust, worse.
Each variable is normalized to 1 in 1947Q2. The chart shows 100 times the log of the resulting
series.
If inflation targeting is counterproductive in this environment, what
should a monetary policymaker do?
Christiano, Motto, and Rostagno
argue that policymakers also need to
monitor credit market conditions as
well as inflation because credit growth
is correlated with booms. Consequently, if policymakers observe strong credit
growth and declining inflation, they
should still “lean against the wind”
and raise interest rates to slow the
economy and temper the boom.
Bernanke and Gertler and
Christiano, Motto, and Rostagno have
different takes on whether inflation
targeting helps stabilize an economy
that experiences a boom. The key
difference between the conflicting
accounts is how inflation behaves
during the boom phase of the cycle.
If inflation rises in the boom phase,
Bernanke and Gertler’s stabilization
argument holds and inflation targeting
will be stabilizing for the economy. If
inflation falls during the boom phase,
Christiano, Motto, and Rostagno’s
argument holds and inflationtargeting policy is destabilizing for
the economy. Unfortunately, the
data do not give a clear-cut answer
about the relationship between stock
market booms and inflation. The big
problem is defining what constitutes
a boom in asset prices: There is no
completely objective measure. Figure
5 plots the log of the S&P 500 index
in both nominal and real terms and
the rate of inflation measured by the
personal consumption expenditures
(PCE) index.10 Clearly, the correlation
between inflation and asset-price
10
Since the index is plotted in logs, a change in
the level of the index gives the percent change
in the index.
www.philadelphiafed.org
booms depends in part on how booms
are identified. For example, we might
try to define a boom as above-trend
growth in the stock market index. But
then we would have to decide how
to measure trend growth in the stock
price index. Deviations from a linear
trend look different than deviations
from a trend that varies smoothly over
time or a linear trend that has breaks
in it.
For a more general look at the
data, we can go back to Figure 2. Here
we have not defined booms or busts
but instead relied only on the postwar
data (although we have also made
some identification assumptions as
detailed above). The figure shows that
in response to higher expectations of
future unemployment, stock prices
decline and inflation declines. Flipping
that around, we can say that when
expectations for the future economy
are unusually good, stock prices rise
as does inflation. At least over the
postwar period, the response of asset
prices and inflation seems to line up
better with the view in Bernanke and
Gertler. Indeed, Figure 2 also shows
that the Federal Reserve tended to
tighten policy in booms and ease
policy in bad times. That is not to say,
though, that the Christiano, Motto,
and Rostagno story is without merit.
It is hard to argue against the view
that monetary policymakers would
be well served by monitoring credit
market conditions as well as inflation
in setting policy. Indeed, the Federal
Reserve looks at a broad array of
indicators when making decisions
about the appropriate stance of
monetary policy, even if low and stable
inflation is a principal goal of policy.
Christiano, Lawrence, Martin
Eichenbaum, and Charles Evans. “Nominal
Rigidities and the Dynamic Effects of
a Shock to Monetary Policy,” Journal of
Political Economy, 113:1 (2005), pp. 1-45.
Leduc, Sylvain, and Keith Sill. “Do
Changes in Expectations Affect Economic
Activity?” manuscript (2008).
SUMMARY
Expectations play an important
role in decision-making at the
individual level, and there is increasing
evidence that expectations about the
future are important in accounting for
fluctuations in economic aggregates.
New economic models are attempting
to explicitly model the expectations
channel for business cycles. With the
recent housing-related boom and bust
in the U.S. and its manifestations
across the globe, it seems even more
important that macroeconomists
develop models that can help us
understand this episode and guide
monetary policymakers in their
decision-making. BR
REFERENCES
Barsky, Robert, and Eric Sims.
“Information Shocks, Animal Spirits, and
the Meaning of Innovations in Consumer
Confidence,” manuscript (2006).
Beaudry, Paul, and Franck Portier. “Stock
Prices, News, and Economic Fluctuations,”
American Economic Review, 96:4 (2006),
pp. 1293-1307.
Bernanke, Ben, and Mark Gertler.
“Monetary Policy and Asset Price
Variability,” NBER Working Paper 7559
(2000).
Chatterjee, Satyajit. “Productivity Growth
and the American Business Cycle,” Federal
Reserve Bank of Philadelphia Business
Review (September/October 1995).
www.philadelphiafed.org
Christiano, Lawrence, Roberto Motto,
and Massimo Rostagno. “Monetary Policy
and Stock Market Boom-Bust Cycles,”
manuscript (2006).
den Haan, Wouter, and Georg
Kaltenbrunner. “Anticipated Growth and
Business Cycles in Matching Models,”
manuscript (2007).
Pigou, Arthur C. “Industrial Fluctuations,”
Volume 6 of A.C. Pigou: Collected
Economic Writings. Basingstoke: Macmillan
(1927).
Schmitt-Grohe, Stephanie, and Martin
Uribe. “What’s News in Business Cycles,”
NBER Working Paper 14215 (2008).
Jaimovich, Nir, and Sergio Rebelo. “Can
News About the Future Drive the Business
Cycle?” American Economic Review, 99:4
(September 2009) pp. 1097-1118.
Business Review Q4 2009 33
RESEARCH RAP
Abstracts of
research papers
produced by the
economists at
the Philadelphia
Fed
You can find more Research Rap abstracts on our website at: www.philadelphiafed.org/research-and-data/
publications/research-rap/. Or view our working papers at: www.philadelphiafed.org/research-and-data/
publications/.
WHAT DETERMINES LOCAL
PATENT RATES?
The authors geocode a data set of
patents and their citation counts, including
citations from abroad. This allows them to
examine both the quantity and quality of
local inventions. They also refine their data
on local academic R&D to explore effects
from different fields of science and sources
of R&D funding. Finally, they incorporate
data on congressional earmarks of funds for
academic R&D.
With one important exception,
results using citation-weighted patents
are similar to those using unweighted
patents. For example, estimates of the
returns to density (jobs per square mile)
are only slightly changed when using
citation-weighted patents as the dependent
variable. But estimates of returns to city size
(urbanization effects) are quite sensitive to
the choice of dependent variable.
Local human capital is the most
important determinant of per capita rates of
patenting. A 1 percent increase in the adult
population with a college degree increases
the local patenting rate by about 1 percent.
With few exceptions, there is little
variation across fields of science in the
contribution of academic R&D to patenting
rates. The exceptions are computer
and life sciences, where the effects are
smaller. There is greater variation in the
contribution of R&D funded by different
34 Q4 2009 Business Review
sources — academic R&D funded by
the federal government generates smaller
increases in patenting rates than R&D
funded by the university itself. This effect
is somewhat stronger for federally funded
applied R&D than for basic R&D. The
authors also find small negative effects for
cities with greater exposure to academic
R&D allocated by congressional earmarks.
The authors discuss the implications of these
results for policy and future research.
Working Paper 09-12, “What Explains the
Quantity and Quality of Local Inventive
Activity?” Gerald Carlino, Federal Reserve
Bank of Philadelphia, and Robert Hunt, Federal
Reserve Bank of Philadelphia
A NEW CLASS OF CONFIDENCE SETS
FOR DSGE MODEL PARAMETERS
The authors show that in weakly
identified models (1) the posterior mode
will not be a consistent estimator of the
true parameter vector, (2) the posterior
distribution will not be Gaussian even
asymptotically, and (3) Bayesian credible
sets and frequentist confidence sets will not
coincide asymptotically. This means that
Bayesian DSGE estimation should not be
interpreted merely as a convenient device
for obtaining asymptotically valid point
estimates and confidence sets from the
posterior distribution. As an alternative, the
authors develop a new class of frequentist
confidence sets for structural DSGE model
www.philadelphiafed.org
parameters that remain asymptotically valid regardless
of the strength of the identification. The proposed set
correctly reflects the uncertainty about the structural
parameters even when the likelihood is flat, it protects
the researcher from spurious inference, and it is
asymptotically invariant to the prior in the case of weak
identification.
Working Paper 09-13, “Frequentist Inference in
Weakly Identified DSGE Models,” Pablo GuerronQuintana, Federal Reserve Bank of Philadelphia; Atsushi
Inoue, North Carolina State University; and Lutz Kilian,
University of Michigan and CEPR
VACANCIES, HIRES, AND VACANCY YIELDS
IN THE JOB OPENINGS AND LABOR
TURNOVER SURVEY (JOLTS)
The authors study vacancies, hires, and vacancy
yields (success rate in generating hires) using the Job
Openings and Labor Turnover Survey, which provides
job opening and labor turnover data collected from
a large representative sample of U.S. employers. The
authors also develop a simple framework that identifies
the monthly flow of new vacancies and the jobfilling rate for vacant positions, which is the employer
counterpart to the job-finding rate for unemployed
workers. The job-filling rate moves counter to
employment at the aggregate level but rises steeply with
employer growth rates in the cross section. It falls with
employer size, rises with the worker turnover rate, and
varies by a factor of four across major industry groups.
The authors’ analysis also indicates that more than
one in six hires occurs without benefit of a vacancy, as
defined by JOLTS. These findings provide useful inputs
for assessing, developing, and calibrating theoretical
models of search, matching, and hiring in the labor
market.
Working Paper 09-14, “The Establishment-Level
Behavior of Vacancies and Hiring,” Steven J. Davis,
University of Chicago and NBER; R. Jason Faberman,
Federal Reserve Bank of Philadelphia; and John C.
Haltiwanger, University of Maryland and NBER
DETERMINING A FIRM’S EXPORT STATUS
Exporters are few — less than one-fifth among
U.S. manufacturing firms — and they are larger than
nonexporting firms — about four to five times more
total sales per firm. These facts are often cited as
support for models with economies of scale and firm
www.philadelphiafed.org
heterogeneity as in Melitz (2003). The authors find that
the basic Melitz model cannot simultaneously match
the size and share of exporters given the observed
distribution of total sales. Instead, exporters are
expected to be between 90 and 100 times larger than
nonexporters. It is easy to reconcile the model with the
data. However, a lot of variation independent of firm
size is needed to do so. This suggests that economies
of scale play only a minor role in determining a firm’s
export status. The authors show that the augmented
model also has markedly different implications in the
event of trade liberalization. Most of the adjustment is
through the intensive margin, and productivity gains
due to reallocation are halved.
Working Paper 09-15, “Economies of Scale and the
Size of Exporters,” Roc Armenter, Federal Reserve Bank
of Philadelphia, and Miklós Koren, Central European
University, IEHAS, and CEPR
CONCENTRATION OF R&D ACTIVITY IN
THE U.S.
This study details the location patterns of R&D
labs in the U.S., but it differs from past studies in
a number of ways. First, rather than looking at the
geographic concentration of manufacturing firms (e.g.,
Ellison and Glaeser, 1997; Rosenthal and Strange,
2001; and Duranton and Overman, 2005), the
authors consider the spatial concentration of private
R&D activity. Second, rather than focusing on the
concentration of employment in a given industry, the
authors look at the clustering of individual R&D labs
by industry. Third, following Duranton and Overman,
the authors look for geographic clusters of labs that
represent statistically significant departures from spatial
randomness using simulation techniques. The authors
find that R&D activity for most industries tends to
be concentrated in the Northeast corridor, around
the Great Lakes, in California’s Bay Area, and in
southern California. They argue that the high spatial
concentration of R&D activity facilitates the exchange
of ideas among firms and aids in the creation of new
goods and new ways of producing existing goods. They
run a regression of an Ellison and Glaeser style index
measuring the spatial concentration of R&D labs
on geographic proxies for knowledge spillovers and
other characteristics and find evidence that localized
knowledge spillovers are important for innovative
activity.
Business Review Q4 2009 35
Working Paper 09-16, “The Geography of Research
and Development Activity in the U.S.,” Kristy Buzard,
University of California-San Diego, and Gerald Carlino,
Federal Reserve Bank of Philadelphia
AGGLOMERATION ECONOMIES’
ROLE IN APPLYING NEW KNOWLEDGE
TO PRODUCTION
Where does adaptation to innovation take
place? The author presents evidence on the role of
agglomeration economies in the application of new
knowledge to production. All else equal, workers are
more likely to be observed in new work in locations
that are initially dense in both college graduates
and industry variety. This pattern is consistent
with economies of density from the geographic
concentration of factors and markets related to
technological adaptation. A main contribution is to
use a new measure, based on revisions to occupation
classifications, to closely characterize cross-sectional
differences across U.S. cities in adaptation to
technological change. Worker-level results also provide
new evidence on the skill bias of recent innovations.
Working Paper 09-17, “Technological Adaptation,
Cities, and New Work,” Jeffrey Lin, Federal Reserve Bank
of Philadelphia
TRANSMISSION OF CREDIBLE
INFORMATION BY A BENEVOLENT
CENTRAL BANK
The authors study credible information
transmission by a benevolent central bank. They
consider two possibilities: direct revelation through
an announcement versus indirect information
transmission through monetary policy. These two
ways of transmitting information have very different
consequences. Since the objectives of the central
bank and those of individual investors are not always
aligned, private investors might rationally ignore
announcements by the central bank. In contrast,
information transmission through changes in the
interest rate creates a distortion, thus lending an
amount of credibility. This induces the private investors
to rationally take into account information revealed
through monetary policy.
Working Paper 09-18, “Money Talks,” Marie Hoerova,
European Central Bank; Cyril Monnet, Federal Reserve
Bank of Philadelphia; and Ted Temzelides, Rice University
36 Q4 2009 Business Review
TRADE REFORM POLICIES, TARIFF
REDUCTIONS, AND OUTPUT PER WORKER
IN KOREA’S MANUFACTURING SECTOR
South Korea’s growth miracle has been well documented. A large set of institutional and policy reforms
in the early 1960s is thought to have contributed to the
country’s extraordinary performance. In this paper, the
authors assess the importance of one key set of policies — the trade policy reforms in Korea — as well as
the concurrent GATT tariff reductions. They develop a
model of neoclassical growth and trade that highlights
two forces by which lower trade barriers can lead to
increased per worker GDP: comparative advantage and
specialization, and capital accumulation. The authors
calibrate the model and simulate the effects of three
sets of tariff reductions that occurred between early
1962 and 1995. Their main finding is that the model
can explain up to 32 percent of South Korea’s catch-up
to the G7 countries in output per worker in the manufacturing sector. The authors find that the effects of the
tariff reductions taken together are about twice as large
as the sum of each reduction applied individually.
Working Paper 09-19, “How Much of South Korea’s
Growth Miracle Can Be Explained by Trade Policy?,”
Michelle Connolly, Duke University, and Kei-Mu Yi,
Federal Reserve Bank of Philadelphia
TECHNOLOGY, UNCERTAINTY, AND
FLUCTUATIONS IN REAL EXCHANGE RATES
This paper investigates the extent to which
technology and uncertainty contribute to fluctuations
in real exchange rates. Using a structural VAR and
bilateral exchange rates, the author finds that neutral
technology shocks are important contributors to the
dynamics of real exchange rates. Investment-specific
and uncertainty shocks have a more restricted effect on
international prices. All three disturbances cause shortrun deviations from uncovered interest rate parity.
Working Paper 09-20, “Do Uncertainty and Technology Drive Exchange Rates?” Pablo A. Guerron-Quintana,
Federal Reserve Bank of Philadelphia
SECURITIZATION AND THE POOR
PERFORMANCE OF MORTGAGES IN THE
FINANCIAL CRISIS
The academic literature, the popular press, and
policymakers have all debated securitization’s contribution to the poor performance of mortgages originated in
www.philadelphiafed.org
the run-up to the current crisis. Theoretical arguments
have been advanced on both sides, but the lack of suitable data has made it difficult to assess them empirically. We examine this issue by using a loan-level data
set from LPS Analytics, covering approximately threequarters of the mortgage market from 2003-2007 and
including both securitized and nonsecuritized loans. We
find evidence that privately securitized loans do indeed
perform worse than similar, nonsecuritized loans. Moreover, this effect is concentrated in prime mortgage markets; for example, a typical prime ARM loan originated
in 2006 becomes delinquent at a 20 percent higher rate
if it is privately securitized, ceteris paribus. By contrast,
subprime loan performance does not seem to be worse
for most classes of securitized loans.
Working Paper 09-21, “Securitization and Mortgage
Default: Reputation vs. Adverse Selection,” Ronel Elul,
Federal Reserve Bank of Philadelphia
HOUSING SHOCKS, HOUSE PRICES, AND
DEFAULT: A QUANTITATIVE MODEL
FOR EXPLORING THE IMPACT OF THE
FORECLOSURE PREVENTION POLICY
The authors construct a quantitative model of the
housing market in which an unanticipated increase in
the supply of housing triggers default mortgages via its
effect on house prices. The decline in house prices creates an incentive to increase the consumption of housing space, but leverage makes it costly for homeowners
to sell their homes and buy bigger ones (they must
absorb large capital losses). Instead, leveraged households find it advantageous to default and rent housing
space. Since renters demand less housing space than
homeowners, foreclosures are a negative force affecting
house prices. The authors explore the possible effects
of the government’s foreclosure prevention policy in
their model. They find that the policy can temporarily
reduce foreclosures and shore up house prices.
Working Paper 09-22, “Foreclosures and House Price
Dynamics: A Quantitative Analysis of the Mortgage Crisis
and the Foreclosure Prevention Policy,” Satyajit Chatterjee,
Federal Reserve Bank of Philadelphia, and Burcu
Eyigungor, Koç University
BANK CAPITAL REQUIREMENTS AND
BUSINESS-CYCLE EFFECTS
This paper attempts to quantify the business-cycle
www.philadelphiafed.org
effects of bank capital requirements. The authors use a
general equilibrium model in which financing of capital
goods production is subject to an agency problem. At
the center of this problem is the interaction between
entrepreneurs’ moral hazard and liquidity provision by
banks as analyzed by Holmstrom and Tirole (1998).
They impose capital requirements on banks and
calibrate the regulation using the Basel II risk-weight
formula. Comparing business-cycle properties of the
model under this procyclical regulation with those
under hypothetical countercyclical regulation, the
authors find that output volatility is about 25 percent
larger under procyclical regulation and that this
volatility difference implies a 1.7 percent reduction of
the household’s welfare. Even with more conservative
parameter choices, the volatility and welfare differences
under the two regimes remain nonnegligible.
Working Paper 09-23, “Time-Varying Capital
Requirements in a General Equilibrium Model of Liquidity
Dependence,” Francisco Covas, Federal Reserve System
Board of Governors, and Shigeru Fujita, Federal Reserve
Bank of Philadelphia
MORTGAGE SALES, MORTGAGE
INVENTORIES, AND TRADE
Consider the sale of mortgages by a loan originator
to a buyer. As widely noted, such a transaction is
subject to a severe adverse selection problem: The
originator has a natural information advantage and
will attempt to sell only the worst mortgages. However,
a second important feature of this transaction has
received much less attention: Both the seller and the
buyer may have existing inventories of mortgages
similar to those being sold. The authors analyze how
the presence of such inventories affects trade. They
use their model to discuss implications for regulatory
intervention in illiquid markets.
Working Paper 09-24, “Why Do Markets Freeze?”
Philip Bond, University of Pennsylvania, and Yaron
Leitner, Federal Reserve Bank of Philadelphia
IMPLICATIONS OF RELAXED BORROWING
CONSTRAINTS IN THE PRESENCE OF
HYPERBOLIC DISCOUNTING
Is the observed rapid increase in consumer debt
over the last three decades good news for consumers?
This paper quantitatively studies macroeconomic
and welfare implications of relaxing borrowing
Business Review Q4 2009 37
constraints when consumers exhibit a hyperbolic
discounting preference. In particular, the author
constructs a calibrated general equilibrium life-cycle
model with uninsured idiosyncratic earnings shocks
and a quasi-hyperbolic discounting preference and
examines the effect of relaxation of the borrowing
constraint, which generates increased indebtedness.
The model can capture the two contrasting views
associated with increased indebtedness: the positive
view, which links increased indebtedness to financialsector development and better insurance, and the
negative view, which associates increased indebtedness
with consumers’ over-borrowing. He finds that while
there is a welfare gain as large as 0.4 percent of flow
consumption from a relaxed borrowing constraint,
which is consistent with the observed increase in
aggregate debt between 1980 and 2000 in the model
with standard exponential discounting consumers,
there is a welfare loss of 0.2 percent in the model with
hyperbolic discounting consumers. This result holds in
spite of the observational similarity of the two models;
the macroeconomic implications of a relaxed borrowing
constraint are similar between the two models.
Cross-sectionally, although consumers of high
and low productivity gain and medium productivity
consumers suffer due to a relaxed borrowing constraint
in both models, the welfare gain of low-productivity
consumers is substantially reduced (and becomes
negative in the case of strong hyperbolic discounting)
in the hyperbolic discounting model due to the welfare
loss from over-borrowing. Finally, the author finds that
the optimal (social welfare maximizing) borrowing limit
is 15 percent of average income, which is substantially
lower than both the optimal level implied by the
exponential discounting model (37 percent) and the
level of the U.S. economy in 2000 implied by the model
(29 percent).
Working Paper 09-25, “Rising Indebtedness and
Hyperbolic Discounting: A Welfare Analysis,” Makoto
Nakajima, Federal Reserve Bank of Philadelphia
A STUDY OF BANKING USING MECHANISM
DESIGN
The authors study banking using the tools of
mechanism design, without a priori assumptions
about what banks are, who they are, or what they do.
Given preferences, technologies, and certain frictions
— including limited commitment and imperfect
38 Q4 2009 Business Review
monitoring — they describe the set of incentive feasible
allocations and interpret the outcomes in terms of
institutions that resemble banks. The bankers in the
authors’ model endogenously accept deposits, and their
liabilities help others in making payments. This activity
is essential: If it were ruled out, the set of feasible
allocations would be inferior. The authors discuss
how many and which agents play the role of bankers.
For example, they show that agents who are more
connected to the market are better suited for this role,
since they have more to lose by reneging on obligations.
The authors discuss some banking history and compare
it with the predictions of their theory.
Working Paper 09-26, “Banking: A Mechanism
Design Approach,” Fabrizio Mattesini, University of Rome
Tor-Vergata; Cyril Monnet, Federal Reserve Bank of
Philadelphia; and Randall Wright, University of Wisconsin,
and Visiting Scholar, Federal Reserve Bank of Philadelphia
IMPLEMENTING MONETARY POLICY:
STANDING FACILITIES AND OPEN MARKET
OPERATIONS
The authors compare two stylized frameworks
for the implementation of monetary policy. The first
framework relies only on standing facilities, while
the second framework relies only on open market
operations. They show that the Friedman rule cannot
be implemented when the central bank uses standing
facilities, while it can be implemented with open
market operations. For a given rate of inflation, the
authors show that standing facilities unambiguously
achieve higher welfare than just conducting open
market operations. They conclude that elements of
both frameworks should be combined. Also, their
results suggest that any monetary policy implementation
framework should remunerate both required and excess
reserves.
Working Paper 09-27, “Monetary Policy
Implementation Frameworks: A Comparative Analysis,”
Antoine Martin, Federal Reserve Bank of New York, and
Cyril Monnet, Federal Reserve Bank of Philadelphia
ASSESSING THE PRECISION OF ECONOMIC
PREDICTIONS: EARLY RELEASE DATA AND
DEFINITIONAL CHANGES
In this paper, the authors empirically assess
the extent to which early release inefficiency and
definitional change affect prediction precision. In
www.philadelphiafed.org
particular, they carry out a series of ex-ante prediction
experiments in order to examine the marginal
predictive content of the revision process, the tradeoffs associated with predicting different releases of
a variable, the importance of particular forms of
definitional change, which the authors call “definitional
breaks,” and the rationality of early releases of economic
variables. An important feature of their rationality
tests is that they are based solely on the examination of
ex-ante predictions, rather than on in-sample regression
analysis, as are many tests in the extant literature.
Their findings point to the importance of making realtime datasets available to forecasters, as the revision
process has marginal predictive content, and because
predictive accuracy increases when multiple releases
of data are used when specifying and estimating
prediction models.
The authors also present new evidence that
early releases of money are rational, whereas prices
and output are irrational. Moreover, they find that
regardless of which release of their price variable
one specifies as the “target” variable to be predicted,
using only “first release” data in model estimation and
prediction construction yields mean square forecast
error (MSFE) “best” predictions. On the other hand,
models estimated and implemented using “latest
available release” data are MSFE-best for predicting
all releases of money. The authors argue that these
contradictory findings are due to the relevance of
definitional breaks in the data-generating processes of
the variables they examine. In an empirical analysis,
they examine the real-time predictive content of money
for income, and they find that vector autoregressions
with money do not perform significantly worse than
www.philadelphiafed.org
autoregressions when predicting output during the past
20 years.
Working Paper 09-28, “Real-Time Datasets Really Do
Make a Difference: Definitional Change, Data Release,
and Forecasting,” Andres Fernandez, Rutgers University
and Universidad de Los Andes, and Norman R. Swanson,
Rutgers University, and Visiting Scholar, Federal Reserve
Bank of Philadelphia
TESTING THE ACCURACY OF PREDICTIVE
DENSITIES DERIVED FROM DIFFUSION
MODELS
This paper develops tests for comparing the
accuracy of predictive densities derived from (possibly
misspecified) diffusion models. In particular, the
authors first outline a simple simulation-based
framework for constructing predictive densities for
one-factor and stochastic volatility models. Then,
they construct accuracy assessment tests that are in
the spirit of Diebold and Mariano (1995) and White
(2000). In order to establish the asymptotic properties
of their tests, the authors also develop a recursive
variant of the nonparametric simulated maximum
likelihood estimator of Fermanian and Salanié (2004).
In an empirical illustration, the predictive densities
from several models of the one-month federal funds
rates are compared.
Working Paper 09-29, “Predictive Density
Construction and Accuracy Testing with Multiple Possibly
Misspecified Diffusion Models,” Valentina Corradi,
University of Warwick, and Norman R. Swanson, Rutgers
University, and Visiting Scholar, Federal Reserve Bank of
Philadelphia
Business Review Q4 2009 39
@FedFRASER