The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
/conomic
-A f
^ ^ ^ ^
161992
September/October 1992
Volume 77, Number 5
Federal Reserve
Bank of Atlanta
In This Issue:
Human Capital Investment and
Economic Growth: New Routes
In Theory Address Old Questions
Regional Employment by Industry:
Do Returns to Capital Matter?
Tracking Manufacturing: The Survey of
Southeastern Manufacturing Conditions
Review Essay
^
>_"» . V. V
Jconomie
September/October 1992, Volume 77, Number 5
panarne
Federal Reserve
Bank of Atlanta
President
Robert P. Forrestal
Senior Vice President and
Director of Research
Sheila L. Tschinkel
Vice President and
Associate Director of Research
B. Frank King
Research Department
William Curt Hunter, Vice President, Basic Research
Mary Susan Rosenbaum, Vice President, Macropolicy
Thomas J. Cunningham, Research Officer, Regional
William Roberds, Research Officer, Macropolicy
Larry D. Wall. Research Officer, Financial
Public Affairs
Bobbie H. McCrackin, Vice President
Joycelyn T. Woolfolk, Editor
Lynn H. Foley, Managing Editor
Carole L. Starkey, Graphics
Ellen Arth, Circulation
The Economic Review of the Federal Reserve Bank of Atlanta presents analysis of economic
and financial topics relevant to Federal Reserve policy. In a format accessible to the nonspecialist, the publication reflects the work of the Research Department. It is edited, designed, produced, and distributed through the Public Affairs Department.
Views expressed in the Economic Review arc not necessarily those of this Bank or of the Federal Reserve System.
Material may be reprinted or abstracted if the Review and author are credited. Please provide the
Bank's Public Affairs Department with a copy of any publication containing reprinted material.
Free subscriptions and limited additional copies are available from the Public Affairs Department, Federal Reserve Bank of Atlanta, 104 Marietta Street, N.W., Atlanta, Georgia 30303-2713
(404/521-8020). Change-of-address notices and subscription cancellations should be sent directly to the Public Affairs Department. Please include the current mailing label as well as any newinformation. ISSN 0732-1813
Contents
September/October 1992, Volume 77, Number 5
//uman Capital Investment
And Economic Growth:
New Routes in Theory
Address Old Questions
Since the 1700s economists have questioned how economies
grow. From the 1950s the dominant theoretical framework for explaining growth focused on changes in the quantity of capital and
raw labor inputs, with a large residual, unexplained in the theoretical framework, attributed to technological progress. Recent research suggests that technological advances might themselves result
from individuals and society investing in education to enhance the
quality of physical and especially human capital. The authors conclude that the diffusion of education may be essential to modem
economic development, calling for greater educational investment
and subsidies.
/Regional Employment
By Industry: Do Returns
To Capital Matter?
Through a variety of programs, U.S. states and communities
have long sought to enhance local and regional economic development. Effective programs—those that improve a region's long-term
relative wealth and well-being—grow out of an understanding of
the economic forces that shape an area's evolving industry composition. This article discusses the influence of differing returns to
capital on movements of jobs and capital among regions. Empirical
tests confirm the influence of relative profitability on changes in
employment in various industries across U.S. states from 1969
through 1986.
FYI—Tracking Manufacturing:
To enhance its monitoring of the economy in its region, the
Atlanta Fed last year initiated a formal survey of southeastern manufacturers. This article outlines the development and construction
of the survey, explains how the data are compiled and used to calculate diffusion indexes, and describes some applications of the survey data.
Ellis W. Tallman and Ping Wang
[^
Stacy E. Kottman
26
The Survey of
Manufacturing
R. Mark Rogers
34
-Review Essay
. .
Aruna „Srinivasan
Southeastern
Conditions
The Future of Banking
by James L. Pierce
Jïï/uman Capital Investment
And Economic Growth:
New Routes in Theory
Address Old Questions
r
Tollman is an economist in the
macropolicy section of the
Atlanta Feds research department. Wang is an assistant
professor in the Department of
Economics of The Pennsylvania
State University and visiting
senior economist at the Federal
Reserve Bank of Dallas.
Federal Reserve Bank of Atlanta
Ellis W. Tallman and Ping Wang
he question of how economies grow has been a central topic of research for economists since the time of Adam Smith. The formal
analysis of growth focuses on long-run economic progress, and
since the mid-1950s the dominant theoretical framework for inquiry into economic development, the model proposed by Robert M.
Solow (1956) and T.W. Swan (1956), has provided a paradigm for analyzing
growth that formalizes how inputs of physical capital and raw labor combine to create real (inflation-adjusted) output. In related empirical work Edward F. Denison (1962) employed statistical methods in a procedure
referred to as growth accounting to investigate the sources of income
growth. Denison focused on the growth rates of physical capital and raw labor to determine how much income growth can be explained by the growth
in inputs.
Unfortunately, in applications of the Solow-Swan model as well as growth
accounting the empirical results show that the input growth rates fail to explain most of the variation in output. In general, the unexplained portion of
output growth has been attributed to the area of technological progress,
where measures of inputs do not capture improvements in output creation
such as new methods of production and innovations in transportation.
Theodore W. Schultz (1961) attributed the fact that the growth in output
exceeds that of the measured inputs—the components of production, physically reproducible capital and raw labor worker-hours—to investment in human capital. His analysis focused on the concept that individuals invest in
learning skills, gaining knowledge, and otherwise enhancing their physical
or mental abilities. According to Schultz, improvements embodied in inputs,
such as technological changes, are likely the products of human capital in
Eco n o m ic Review
1
action, especially in the United States, where much
leisure time is spent enhancing skills and gaining
knowledge. His intuitive argument implies that technological change, often viewed as an exogenous or external factor affecting the economy, may be explainable
within a model embodying the human capital investment decision. While Schultz sought to address existing research on economic growth and development, he
also related his comments to studies examining the returns to education, which now make up a voluminous
body of literature but were new discoveries at the time
of his writing.
The literature on human capital and growth highlights the concept of human capital as a mechanism to
advance technology, improve productivity, and generate growth. This article reviews that research and elaborates Schultz's assertion appending human capital to
The crises in U.S. cities among the unskilled unemployed and in education systems illustrate the need for
investments in support of human capital formation, it
is appropriate that the topic of human capital has received attention not only from U.S. policymakers and
the economics profession but also from the business
world (see, for example, Bruce Nussbaum et al. 1988
and Leonard Silk 1992).
7Tie Solow-Swan Growth
Model and Growth Accounting
The Solow-Swan model in many ways revolutionized the theory as well as the measurement of economic
growth. Its impact has been widespread across various
economic disciplines. The characteristics described
here will paint at least a superficial picture of this influential model.
The Solow-Swan growth model begins with a basic
assumption that output (Y) can be produced using
combinations of physical capital (K) (including reproducible machines, equipment, and buildings) and labor
(L) in variable proportions. The model assumes constant returns to scale {CRTS), implying that doubling
the amount of each input in combination will double
the output. However, if either input is held fixed and
the other is doubled, output will increase by less than
double. In addition to using labor and capital as the
only inputs, the economy's production function is subject to a technological factor (A) in the standard
Solow-Swan model representation. The production
function for the economy, often referred to as the aggregate production function, is given below:
1
Technological change, often viewed as an external factor affecting the economy, may be
explainable within a model embodying the
human capital investment decision.
2
3
the Solow-Swan growth model. A discussion of the relationship between education and lifetime earnings
presents relevant empirical evidence. The article then
synthesizes the recently developed endogenous growth
literature—that is, the studies examining forces within
the economy that may generate growth. The analysis
considers the role of human capital in enhancing output growth, emphasizing how the rational decisions of
individuals choosing to invest in education may make
a difference. Both theoretical and empirical deficiencies in the existing studies are summarized. In conclusion, the article points out policy implications of the
findings concerning human capital and its relationship
to output growth.
Observing that international trade and competition
have drastically reduced earnings potential for lesserskilled workers, policy-oriented research has proposed
that future job growth will be in technical fields, requiring workers to have advanced education levels.
2
Economic Review
Y = F(K, AL),
(l)
where the production function, F, determines how
much output, Y, can be made by combining inputs of
capital and labor, K and L, in the production process.
The factor A represents the labor-augmenting technological advancement, which is exogenous. The laboraugmenting character of the technological factor
means that it is like simply adding more labor to the
production function. This specification of the factor is
referred to as Harrod-neutral technological change.
The model's steady-state relationships—in which
all variables grow at constant rates—highlight key
implications that have been a source of criticism.
One important equation can be derived directly from
the definition of the growth rate of capital and the
September/October 1992
equation for aggregate investment/savings. The steadystate equation is below:
s(Y/L) = (n + X)K/L,
(2)
where s is the savings rate, n is the rate of population
growth, and A is the rate of technological advancement.
With constant returns to scale, per capita output (Y/L)
is a function of the capital-labor ratio (K/L) alone. Thus
the equation can be used to determine the steady-state
accumulation of physical capital in per capita units (also see Chart 1). In the equation, two exogenous processes (/?, A) determine the economy's growth rate. The
savings rate determines the equilibrium output/labor
and capital/labor ratios, but increased savings cannot
alter the rate of economic growth in the model. In addition, the population growth rate cannot alter the per
capita growth in output. The only way to achieve higher per capita output growth is to have a greater rate of
technological advancement, something presumed to be
outside the control of the agents in the economy.
To explore the role of technological change for explaining growth in the United States, Solow (1957)
employed a simple empirical analogue to the growth
model. In it, he estimated a series he referred to as
"technological change" (often called the "Solow residual") that reflects the concept of an external force generating economic growth. The notable finding was that
almost 90 percent of the growth in the U.S. economy
from 1909 to 1949 could be attributed to the technological factor.
Chart 1
Steady-State Capital Accumulation
In the Solow-Swan Model
Denison (1962, 1974) provided alternative estimation of economic growth, referred to as growth accounting. The empirical method begins with the basic
assumption that output depends on numerous determinants (inputs) and that changes in these inputs cause
output changes. How much growth a particular source
contributes depends upon its importance to output
growth as shown in the amount that output growth
changed. The statistical methods attempt to measure
the sources of economic growth using data on relevant
inputs. The typical growth accounting equation is
4
AY/Y=a(&L/L
+ AA/A) + (l - a ) A K / K ,
(3)
where A A/A represents the growth rate of laboraugmenting technological change, a is the labor
income share (that is, the return to labor from the production of output), and (1 — a) is the capital income
share. Denison measured the growth in labor hours
and capital stock and compared these variables with
the growth in output. In his 1962 study Denison found
that in the United States over the 1909-57 period real
output, employed worker hours, and capital stock
grew an annual average rate of 2.9 percent, 1.4 percent, and 2.4 percent per capita, respectively. That
part of output growth not accounted for by input
growth, the residual measure that he interpreted as
"advances in the state of knowledge," is comparable to
the rate of technical progress in Solow (1957). Many
researchers have found the significant proportion of
output attributed to an external (exogenous) process of
knowledge or technological advancement an obvious
flaw in the theory and its application. The model appeared to leave an inordinate amount of variation in
economic growth unexplained.
5
6
/education and Human
Capital Investment
Schultz (1961) emphasized a linkage between earnings and education, suggesting that investment in human capital accounts for most of the observed rise in
real earnings. In addition, he proposed that the value of
human capital may be as big or bigger than the value
of the tangible physical capital stock, with a rate of
growth in excess of capital growth. Schultz also acknowledged that improvements in factor input quality
generally affect productivity. For example, suspension
bridge construction methods illustrate how input quality has improved over time, increasing productivity. The
7
Federal Reserve Bank of Atlanta
Eco no m ic Review
3
construction of the Brooklyn Bridge took fourteen
years in the 1880s; during the 1960s the VerrazanoNarrows Bridge, more than twice the span of the
Brooklyn Bridge, was built in five years. Such improvements in productivity may result from advances
in physical technology (new capital goods), the stock
of knowledge, or human capital embodied in individual
workers. The matter is a key issue in research.
Although the idea of investment in human capital
may seem intangible and difficult to quantify for scientific analysis, measuring the individual's education
levels has been a successful method. Gary Becker
(1975), for instance, investigated the returns to education in the United States. The theory analyzes the educational choice of consumers, using a model in which
individuals choose the level of education that they desire on the basis of expected returns to their investment of time, effort, and expense. Becker explicitly
linked the education level of an individual to his or her
productivity as a worker, implying that those workers
with higher education are more productive and therefore receive higher wages. The human capital investment function is as follows:
HCINV = G(R, T, B, H\
(4)
where IICINV is the rate of investment in human capital, G is the human capital investment function, R is
the input of other resources (capital and labor), T is the
input of time toward education, B is the physical and
mental powers of the individual, and H is the input of
human capital. Additional amounts of any of these inputs result in increased production of human capital—
that is, human capital investment is a positive function
of all the inputs.
in 1976 Sherwin Rosen published a breakthrough
article that introduced dynamics to the theory of life
earnings relying on human capital investment. The
lifetime earnings function is below:
W{t) = \"E(H,H,s)e- - ds,
r{s t)
current work effort and wages. Clearly, there is a
trade-off between current earnings and increased future productivity by further investment in human capital today, and the model takes account of these factors
and determines an optimal level of human capital investment.
Chart 2 reflects the earnings capacity (EC) and actual earnings (E) over an individual's lifetime. Earnings
capacity exceeds actual earnings, and the difference
between them reflects human capital investment. The
chart illustrates the tendency of the lifetime pattern of
earnings to peak during middle age and to show the
greatest difference between earnings and earnings capacity in the early years. Human capital investment
during the early stages of a lifetime has a long time to
generate returns. An individual who invested less in
human capital would have flatter curves on this chart.
Rosen's work is important not only for adding the
temporal dimension to the analysis of returns from education but also because it elaborates on the various
ways that education enhances human capital. Certain
processes that take place on the job—for example,
learning by doing and specific job training—add to human capital by increasing worker productivity. Rosen's
theory emphasizes that education both improves an individual's knowledge once and for all and strengthens
the individual's capacity to learn on the job, thereby increasing the worker's speed of human capital accumulation. This insight implies that the labor productivity
enhancement from investment in human capital may
exceed that recognized by Becker (1975).
Chart 2
Earnings Capacity versus Lifetime Earnings
EC
(5)
where W(t) represents lifetime earnings, N is the predetermined length of the work life, E is the current earnings
function, 5 measures time, and r is the discount factor.
The model takes into account future as well as current earnings, discounting future earnings and presenting a lifetime earnings function. In this framework,
current earnings are positively related to the accumulated human capital stock (//) but are negatively related to increased current investment in human capital
(//) because it takes time and resources away from
4
Economic Review
September/October 1992
.Empirical Evidence on
Returns to Education
9
The evidence on returns to investment in education
is voluminous, and the selective sample offered here
attempts only to raise some issues that relate, directly
and indirectly, to inquiries on economic growth. The
work of T. Paul Schultz (1988), for example, raises additional questions about our understanding of the
sources of growth, suggesting that although it may be
comfortably assumed that the changing quality of capital and labor are likely sources of growth, accounting
for such sources explicitly remains unresolved. Schultz
presented a useful survey of the literature on the returns to education as well as the link between education and growth and concluded that "the record of
sustained modern economic growth in real per capita
income cannot be accounted for by the accumulation
of conventional units of physical capital or by the increased application of hours of labor per capita" (1988,
544).
B e c k e r ' s (1975) empirical work supports his
premise that more education is correlated with higher
earnings. At the same time, his interpretations of several results hint at more general issues that have become central in the recent endogenous growth literature.
For example, he suggested that gains from college education are not fully quantified by earnings analysis
because college graduates are only partially compensated for their effect on the development and spread of
knowledge. He observed that the accumulation of education (by individuals) may be measured and conceived of as separate from the growth in knowledge
(or technology), although increases in both education
and knowledge improve productivity. Becker's interpretations perceive part of education's effects as an external effect—that is, beyond the primary positive
effect on an individual's productivity—and not taken
into account by an individual when choosing a desired
education level. In reference to the "Solow residual"
Becker questioned the size and composition of the
technological advance and what portions are attributable to growth in human capital, the stock of society's
knowledge, or education of individuals.
In a cross-sectional sample of developed, developing, and less-developed countries George Psacharopoulos (1984, 1985) found stable returns to education. His
findings suggest that the returns to investments in primary education are greater than those to secondary and
higher education. Notably, the overall returns profile
declines as the education level increases. The results,
8
Federal Reserve Bank of Atlanta
however, cannot capture the potential external effects
of higher education implied by Becker.
Richard A. Easterlin (1981) examined a cross-section
of countries and found a link between widespread
public education and economic growth. In agreement
with Becker's ideas on the external effects of education, Easterlin hypothesized that modern economic
growth relies on the diffusion and the advancement of
knowledge. He viewed the spread of mass education,
separate from the growth of science and technology, as
a key to economic development, offering upward mobility to a wider segment of the world population. His
work suggests that a populace characterized by at least
basic education is a precondition for economic growth
and that widespread expansion of schooling reflects a
voluntary move on the part of informed governments
toward economic growth through education.
One of the more fundamental external effects from human capital investment lies in
the advancement of knowledge and the development of new applications of knowledge.
Dale W. Jorgenson and Barbara M. Fraumeni (1991)
presented measurement of investment in human capital from a perspective different from previous studies: they measured the investment in terms of the income produced by human capital rather than using
the more typical method of examining the outlays for
education. Their study found that the overwhelming
portion of economic growth in the United States is
based on investment in both human and physical capital. Estimates show that in the United States investment in education dwarfs other kinds of investment.
Jorgenson and Fraumeni also noted that their estimate of the stock of human wealth, derived from the
measures of educational output, is ten times greater
than previous estimates (see John W. Kendrick
1973)."
As Schultz (1988) pointed out, the existing evidence on education and returns suffers from some
10
Eco n om ic Review
5
measurement problems. First, the studies overlook
immeasurable components like effort and innate ability. Second, although there is a correlation between education and income, the nature of the underlying link
between them is uncertain. For example, the rich are
generally well paid and well educated. Does higher income result from better education or from nonhuman
(financial) wealth?
An additional concern is whether, in certain environments, there may be overinvestment in education.
Psacharopoulos (1985), for instance, argued that certain underdeveloped nations have placed too much
emphasis on higher education without enough attention having been placed first on primary education; the
profile of estimated returns suggests that the composition of education expenditures should be skewed
toward widespread elementary education first. However, the assumption that there is overinvestment in
education may overlook the potential external effects
of human capital (which will be further examined below) so that the returns resulting from additional human capital are not linked to investments made in
human capital accumulation.
12
/ / u m a n Capital Accumulation
And Output Growth
A relatively new direction of study has arisen to
explain economic growth and development without
appealing to an exogenous source of technological advance as the main source of economic growth. The
models offer a way to identify the role of human capital in enhancing output growth, to emphasize individuals' decisions to invest in acquiring skills, and to
rationalize how these actions allow the economy to
grow endogenously—that is, as a result of the actions
of individuals represented in the model. In this framework human capital accumulation provides the engine
of growth by achieving the technological advance that
previous models assumed to be exogenous. Thus, the
endogenous growth theories attempt not only to identify the main sources of technological change (such as
endogenous human capital accumulation) but also to
design models in which economic incentives (such as
greater returns to higher levels of education) explain
what drives economic advancement.
The aggregate production function that incorporates
the endogenous growth feature is
Y = AF(K, HL),
6
Econom ic Review
(6)
where the disembodied technological factor, A, is outside the function and human capital, H, is laboraugmenting technical change. The function is assumed
to display constant returns to scale in two reproducible
capitals—physical, K, and human, H—in contrast to
diminishing returns in the Solow-Swan model.
One of the main issues of contrast between the
Solow-Swan and the endogenous growth models
concerns the predicted growth rate of output per effective unit of input in the steady state, or the long
run. The Solow-Swan growth model predicts a zero
growth rate of output per effective unit input because
output growth is entirely determined by exogenous
factors like the population growth rate (affecting
labor input) and the labor-augmenting technology
shock. The long-run capital and effective labor
growth rates are the same as the exogenous rate of
growth—the sum of the growth rates of population
and the technology shock. If a Solow-Swan growth
model is examined without the exogenous rate of
technological advancement, the rate of per capita
growth in the model economy is zero. On the other
hand, for endogenous growth models the growth rate
of output per capita is a positive constant because the
advancement in technology results from the choice
of individuals to invest in human capital. As human
capital accumulates, technology improves. Rather
than exogenous factors determining growth, the technological advances that enhance productivity are attained endogenously.
The rate-of-returns equation from an endogenous
growth model is useful for distinguishing between the
extended, optimizing Solow-Swan and endogenous
growth models (also see Chart 3):
13
/• = MPK = p + n + (g/a),
(7)
where MPK is the marginal product of capital (which
equals the real rate of return, /•), p is the rate of time
preference (which measures a consumer's preference
for present rather than future consumption), n is the
rate of population growth, g is the rate of (endogenous) economic growth, and a is the intertemporal
elasticity of consumption substitution (which measures the willingness of consumers to substitute between current and future consumption).
The characterization of the marginal product of
capital is a key difference between the two theories.
In both models the rate of population growth, the
time preference rate, and the intertemporal elasticity
of substitution are generally assumed to be exogenous. To determine the equality, either the rate of
September/October 1992
Chart 3
Exogenous versus Endogenous Growth
r
line at r = T indicates how the endogenous growth
models fix the rate of return based on constant returns to reproducible capitals but can determine the
(endogenous) growth rate.
The amount of research in this field has grown
rapidly. While a comprehensive survey of the literature is beyond the scope of this article, it should be
useful to present several of the main ideas that characterize the role of human capital in an aggregate
production function and to provide some empirical evidence that relates to the predictions of theory.
Robert E. Lucas (1988) provided a clear-cut linkage between the aggregate production models of
growth (Solow-Swan) and the idea that human capital
levels directly affect output. The model suggests that
human capital accumulation is the main driving force
of economic growth. Moreover, Lucas's work emphasizes an external effect of human capital—that the average level of human capital can magnify the impact
of individual human capital and lead to greater output.
For example, the concentrated, collaborative effort of
great scientists on the Manhattan Project produced the
atomic bomb perhaps more quickly than if they had
not benefited from the simultaneous group effort. The
individual deciding to invest in human capital fails to
account for this external effect of human capital,
though, basing his or her choices solely on the perceived private returns. Thus, one of the main implications of Lucas's findings is that human capital
investment is likely to be below the socially optimal
level unless there is market intervention in the form of
a subsidy for accumulating additional human capital.
Unlike Lucas (1988), Paul Romer (1990) envisioned a model in which human capital is an essential
source of economic growth but human capital levels
have no external effect. Human capital has two different definitions—cognitive skills tied to individuals
and aggregate knowledge not tied to individuals.
Cognitive skills are often measured by variables like
individuals' years of education, which cannot grow
perpetually and are perceived as unlikely to explain
the sustained growth of per capita output. However,
the stock of human knowledge (scientific understanding) may grow without bound, and the key to growth
and development is its exploitation. In Romer's model, effective use of human knowledge leads to developing new capital goods that are more productive than
previous versions. Old capital goods may still be useful, but the new goods are more efficient. Consider, for
example, the personal computer. Five years ago, the
80286-based PC was state-of-the-art; it remains useful
now, but it is far less productive than either the 80386- or
14
economic growth or the rate of return to capital must
be determined a priori. The key difference between
the models involves the choice of which rate will be
considered endogenous (determined by the equation).
The Solow-Swan model assumes diminishing returns to capital, implying a marginal product of capital
that falls as additional capital accumulates, holding
other inputs fixed. As a result, the rate of return to
capital is endogenous, depending on the level of the
capital-labor ratio. However, as described above, the
rate of growth in the model economy is constant and
externally (exogenously) determined. In contrast, endogenous growth models imply that the marginal
product of capital is constant—that is, the rate of return on capital, r, is fixed and determined solely by the
production technology. This restriction is implied by
the assumption of constant returns to scale in the reproducible factors so that additional amounts of capital do not reduce the rate of return. These assumptions
allow the model to determine the growth rate (g) endogenously. Thus, the Solow-Swan model allows the
marginal product of capital to vary but fixes the rate of
economic growth while endogenous growth models
fix the marginal product of capital but allow the rate of
economic growth to be endogenous.
In Chart 3 the upward-sloping line is the KeynesRamsey formula (equation [7J) for the real interest
rate. The vertical line at g = g~ reflects how the
Solow-Swan model fixes the growth rate but can determine the rate of return to capital because of diminishing returns to capital. In contrast, the horizontal
Federal Reserve Bank of Atlanta
Eco 11 o m ic Revieu>
7
80486-based machines. Thus, improved machines
essentially embody the human capital of advanced
knowledge.
Nancy Stokey (1991) brought together two key
points from the Lucas and Romer models. The Stokey
model includes the external effect of education suggested by Lucas but also separates the stock of human
knowledge from individuals as in Romer (1990). The
external effect is seen as coming about from individuals' investment in human capital. Human capital accumulation is a conscious decision of an individual to
invest and enter school, but as individuals accumulate
human capital the aggregate stock of knowledge increases as well. Additional human capital improves labor quality, increases the stock of knowledge, and
directly affects the product pattern, leading to the development of new and higher-quality goods.
The Lucas, Romer, and Stokey models present numerous insights into the characterization of human
capital as the engine of growth in an endogenous
growth model. The key aspects of each study revolve
around the treatment of human capital in an aggregate production function and the question of how
human capital enhances output. The empirical implications of the models remain somewhat limited, but
the empirical work is enough to give a general impression of the evidence and the direction research is
likely to take.
15
16
Empirical Evidence on Human
Capital and Growth
however, that enrollment rates for 1950 and 1970 are
not significant for explaining average output growth,
suggesting that the positive results based on 1960 enrollments may be not be robust.
Costas Azariadis and Allen Drazen (1990) related
human capital and income growth in a framework that
presents a threshold level of human capital beyond
which a country may experience accelerating growth.
Essentially, beyond the threshold level the social returns to scale for human capital investment increase.
For empirical application, Azariadis and Drazen used
a literacy rate of 40 percent as an initial threshold and
found that this proxy is positively related to output
growth in their sample of countries. Thus, the private
yield on education is higher in more developed countries, and additional education precedes but is not sufficient to cause accelerating output growth.
The empirical studies of Romer (1990), Barro
(1991), and Azariadis and Drazen (1990) found, in
varying degrees, some support for the idea that human
capital has significant explanatory power for output
growth. Their perspective is that human capital provides for endogenous growth. In contrast, N. Gregory
Mankiw, David Romer, and David N. Weil (1992) designed human capital accumulation as an exogenous
process rather than a function of individual decisions.
Their evidence supports a role for human capital, but
it is contrary to the idea that human capital has external effects and to the framework of perpetual growth.
Using a transformation of secondary school enrollment as the human capital proxy, Mankiw, Romer, and
Weil show that the role of human capital in the aggregate production function is consistent with diminishing returns to scale in all reproducible factors.
Unlike in the endogenous growth models, their results
imply that perpetual output growth cannot emerge as a
result of physical and human capital accumulation.
They argue that the Solow-Swan growth model is
preferable for analyzing economic growth.
While suggestive, the existing empirical results do
not offer conclusive evidence in support of human
capital having a significant role in economic growth.
Ross Levine and David Renelt (1992) examined the
robustness of correlations between long-run growth
rates and policy variables (several of which are human
capital proxies) found in cross-country empirical studies. Their findings indicate that regression results that
capture a positive relationship between human capital
and growth are not robust to the inclusion of other relevant variables. As a consequence, they suggest a reasonable degree of skepticism about inferences from
empirical studies linking human capital and growth.
17
18
Romer (1990) investigated whether the literacy rate
in 1960, a proxy measure of the initial level of human
capital, affects the growth experience of a crosssection of countries in the subsequent twenty-five
years. The results suggest that the literacy rate fails to
affect output significantly once the rate of investment
is included in the analysis. On the other hand, Romer
found that the initial level of human capital (and the
change in literacy) does have a significant effect on
the rate of investment, providing an indirect link between human capital and output growth through physical capital investment.
Robert J. Barro (1991) used enrollment rates in primary and secondary schools as human capital proxies
in a cross-country empirical study. He found significant positive effects for primary and secondary enrollment rates for 1960 on output growth averaged over
the period from 1960 to 1985. It should be noted,
8
liconomic Review
19
September/October 1992
In another recent study, Jess Benhabib and Mark
M. Speigel (1991) estimated an aggregate production
function specification of growth by constructing physical and human capital series for a selection of countries. Consistent with Romer (1990), their results
suggest that human capital does not enter significantly
into the explanation of aggregate output growth but remains important in explaining capital accumulation.
These empirical studies embody a number of issues to be addressed by future research. Measurement
problems in particular are commonly pointed to as a
criticism of these initial studies. One concern is that
the studies examine cross-sections of data, and the data—how they were collected, how reliable the numbers are, and so forth—may not be consistent across
countries. In addition, the human capital proxies are
necessarily crude. For instance, although the literacy
rate may be a fairly consistently measured variable, it
may only tangentially measure the human capital concept of interest (that is, a measure of knowledge or
achievement, a more advanced human capital variable). Further, enrollment rates are flow variables that
measure the proportion of the population attending educational institutions but do not necessarily pick up
relevant movements in the stock of human capital or
knowledge. Another limitation is that simply comparing average output growth rates with initial levels of
some human capital proxy fails to examine some of
the interesting dynamics that take place over time between human capital and growth.
In an alternative empirical strategy, Ellis W. Tallman and Ping Wang (1992) have focused on the growth
experience of an individual country, Taiwan, over time
to examine the effect of human capital on output. This
approach has several advantages. First, there is a more
appropriate measure of human capital that focuses on
achievement levels—that is, an aggregate human capital measure created by weighting levels of education
completed. The human capital measure is used in
combination with the raw labor measure to form an effective labor input. Estimating an aggregate production function yields the finding that effective labor
added to measured physical capital directly affects
output growth. In addition, it is shown that the income
shares from most estimates are consistent with constant returns to scale and are robust to adding variables
that are typically correlated to output in the output regressions. The research complements cross-country
studies that attempt to find more general human capital/growth relationships.
Even among endogenous growth theories there are
differences on a number of issues. Distinctions between
20
21
Federal Reserve Bank of Atlanta
human and physical capital are not made entirely clear,
for instance. In Romer, human capital accumulation occurs, but it is new capital goods that provide the key
input toward output growth. If human capital in the
form of knowledge represents potential capital but only becomes a driving force of growth when it is embodied in physical capital, which of these factors is
the true driving force? Another question centers on
whether human capital depreciates. Capital embodied
in individuals likely depreciates, but it is unlikely that
the stock of knowledge declines in value (although it
might be subject to shocks, such as the complete destruction of the Mayan libraries).
A further shortcoming is that the predictions of
both the Solow-Swan and endogenous growth models
appear somewhat unrealistic regarding observed behavior of economies. Instead of either zero or constant
positive growth rates, there are variable growth rates
of output per effective unit input depending on a country's stage of development. Less-developed countries
appear to experience a low growth rate until a crucial
(perhaps human capital level) threshold is passed. Newly industrialized countries may have high, sustained
growth rates whereas developed countries may experience slower output growth.
It is also true that the two model frameworks appear difficult to disentangle empirically despite some
obvious contrasts. The Solow-Swan model predicts
that output levels across countries possessing the same
parameters of technology and preferences will converge over time. According to this assumption, poor
countries should grow faster than rich countries. In
general, however, endogenous growth models offer no
such prediction. Empirical tests of this "convergence
hypothesis" have not reached a consensus. The results
of Mankiw, Romer, and Weil (1992) and Barro and
Xavier Sala-i-Martin (1992) show evidence consistent
with convergence; Barro (1991) finds evidence of income nonconvergence. Unfortunately, the evidence
supporting convergence suffers from a problem in the
estimation methods that renders the empirical results
equivocal. To make things more complicated, some
endogenous growth models imply convergence, and
some extended Solow-Swan models predict nonconvergence.
It appears that in the literature on endogenous
growth models theory may be ahead of the measurement. Applications of the endogenous growth literature are only beginning to confirm a significant
relationship between human capital and growth (but
not all implications from the models are directly
testable). Improvements in formulating human capital
22
23
24
Eco n o m ic Review
9
measures may establish a stronger link between human capital and growth. In addition, more intensive
examinations of specific case studies may be a promising path toward understanding the role of human capital in economic development.
Numerous studies in the economics literature propose that through a variety of roles human capital provides increased earnings for individuals and generates
economic growth. Although the empirical evidence in
support of a link between education measures and output growth is equivocal, all studies suggest a beneficial
effect of education on output or individual earnings.
Few, if any, would argue that an economy can have
too much human capital or too much education. Increased demands for technically skilled workers make
it clear that improved human capital is essential for
economic progress. The economics literature offers an
overview of the particular characteristics of human
capital that may be important for growth, some of
which may require activist policies to achieve the
most desirable outcomes.
First, the diffusion of education may be essential to
the spread of modern economic growth. Widespread
public education at at least a basic level sets the stage
for development, as in the idea of a threshold, discussed above (Azariadis and Drazen 1990). Such a
policy is likely to be more appropriate for an economy
in the early stages of development, when primary education should be a first priority. Educational improvements can be expected to allow a higher level of future
growth because workers would be better able to operate and exploit modern, sophisticated physical capital.
In developed nations with widespread education
systems, like the United States, the emphasis should
be on the quality of education at the primary as well as
higher levels. Greater investment in human and physical capital appears necessary for the United States to
regain sustained economic growth. Recognizing problems in urban public schools may help galvanize a focus on the long-term economic benefits to be realized
by improving the quality of our school systems nationwide. Unfortunately, because human capital involves a
long gestation period and is costly considering the
time value of the investment, no immediate solutions
are at hand for the United States. With substantial investment and considerable patience we may at least
anticipate future returns.
An important point to keep in mind is that individuals do not take into account the positive external effects of human capital. The overriding implication is
that individuals are likely to underinvest in education
and, in terms of enhancing growth, there may need to
be subsidizing policies that encourage the accumulation of human capital. Such policies may help explain
the rapid development of newly industrialized countries like Taiwan. More specifically, one of the more
fundamental external returns to human capital investment lies in fundamental research in basic sciences.
The advancement of knowledge and the development
of new applications of knowledge—technological advancements—provide for future economic growth,
and in the long run the United States would benefit
from directing individuals into the fields of scientific
research and engineering.
It is generally agreed that there are potential gains
from greater emphasis on higher education, which
improves learning efficiency on the job and yields
significant positive external effects. This improvement in on-the-job learning is also important for
promoting perpetual economic growth, adding significantly to individual human capital stock as well as
to the stock of society's knowledge that may improve
the quality of life.
1. Notably, in the analysis of real business cycles Prescotl
(1986) employs Solow's (1957) method for accounting for
technological change. The Nobel Prize given to Solow indicates the significance of his work.
2. Economists refer to this characteristic of a production function as displaying the "diminishing marginal product" of
the single factor.
3. The factor grows over time but may be subject to random
shocks. This factor, proposed as an exogenous (that is, determined outside of the model) rate of technical change, has
been interpreted as the rate of knowledge advancement
(Denison 1962) and also as the "measure of our ignorance"
(Abramovitz 1956). The large proportion of growth ascribed
to this factor led to inquiries into its composition. See note 8.
Conclusion
25
26
10
Economic Review
September/October 1992
4. A shortcoming of this process is that Dcnison attributes a
proportion of economic growth to numerous sources that
make sense theoretically but that may not represent direct
inputs into the aggregate production function.
5. The term AX/X represents percentage changes, or growth
rates in X.
6. See Denison (1962, 142, table 18). Denison (1974) refines
the input measures, weighting the labor input, for example,
to account for the distribution of education level and, in
turn, for the increase in labor productivity. Despite the refinement, he still finds a large residual component—that is,
that advances in exogenous technology contribute significantly to output growth.
7. Weisbrod (1961) estimates the capitalized value of human
capital.
8. Note that Denison (1974) attributes 25 percent of U.S.
growth since 1930 to the economic effects of education and
still has more than 40 percent of U.S. output growth explained by the residual, which lie interprets as the growth in
knowledge.
9. Lucas (1988) and Stokey (1991) also suggest that external
effects from human capital and knowledge accumulation
may be a major source of economic growth.
10. Jorgenson and Fraumeni estimate the value of increased lifetime earnings related to further educational attainment by
comparing the incomes of individuals that are the same age
(and sex) but with different levels of education. This differential becomes the basis for a measure of the output of education. The procedure is similar to Denison (1962, 67-69).
11. Jorgenson and Fraumeni (1991) conclude that their labor
input measure accounts for more than 61 percent of the estimated growth in U.S. output; the capital input accounts for
more than 22 percent, and the residual (technology) input
accounts for only about 17 percent. Theirs are time-series
estimates versus Denison's long-run averages.
12. Easterlin (1981) appears to take this perspective toward developing nations but also hints at external effects of education by discussing the growth in knowledge separate from
educational investment.
13. The analysis and graph are taken directly from Sala-i-Martin
(1990a, 1990b). The rate of return equation is often referred
to as the Keynes-Ramsey equation.
14. Useful surveys of the endogenous growth literature include,
for theory, Sala-i-Martin (1990b) and van de Klundert and
Smuldcrs (1991). See Renelt (1991) for a review of the empirical work as well.
15. Romer's empirical work, which will be examined later, investigates his contention that the initial level of human capital may explain subsequent growth (similar to the ideas of
Easterlin 1981).
Federal Reserve Bank of Atlanta
16. This discussion, of course, fails to convey the technical difficulty of creating these models and deriving the results.
For the purposes of this study, however, the main results
and implications are the relevant portions of the papers.
17. Romer (1990) suggests that this finding may be the result of
measurement error. Azariadis and Drazen ( 1990) do not include the rate of investment in their regressions to examine
whether their human capital proxy would retain its significance in the output regression.
18. They use the average (from 1960 to 1985) percentage of
secondary school-aged individuals in school as the human
capital proxy. The output variable is the natural logarithm
of output per working-aged individual.
19. They employ a version of Learner's variance bounds tests
that examines how much a coefficient estimate changes
when the set of explanatory variables is altered (see Learner
1978). The variables that they examine include the 1960 literacy rate used in Romer (1990) and Azariadis and Drazen
(1990) as well as the 1960 levels of primary and secondary
school enrollment rates used in Barro (1991).
20. Benhabib and Speigel (1991) provide a recent exception.
21. The study examines several different weighting schemes, in
each assigning greater weight to better-educated workers
(assuming more productive workers). Additionally, a consistent capital stock time series is created using the estimated capital stock from 1975 and the aggregate investment
series.
22. Lucas (1988) suggests that the growth rates of countries
may converge but there is no sense of convergence in the
level of output per capita.
23. Levinc and Renelt (1992) find a convergence result that appears robust in a specification that includes a human capital
measure in the regression.
24. Quah (1990) emphasizes the problem known as Gallon's
fallacy of regression toward the mean. The tests use the output level at the beginning of a period as an explanatory
variable for the subsequent output growth rate. Quah shows
that the estimated coefficient can be negative, positive, or
zero for the same cross-sectional distribution. As a result,
the sign of the estimated coefficient of initial levels provides no information about whether the cross-section of
country outputs converges or diverges. The negative coefficient on initial output level found in the studies cited above
is viewed as evidence in support of convergence.
25. Learning on the job and specific job training, additional important sources of human capital investment, appear more
difficult to examine empirically.
26. Although some research (cited above) argues for a better allocation of educational effort toward primary education,
there is no call for a net reduction in total education level.
Eco n o m ic Review
11
References
Abramovitz, Moses. "Resources and Output Trends in the
United States since 1870." American Economic Review Papers and Proceedings 46 (May 1956): 5-23.
Azariadis, Costas, and Alien Drazen. "Threshold Externalities
in Economic Development." Quarterly Journal of Economics 105 (May 1990): 501-26.
Barro, Robert J. "Economic Growth in a Cross-Section of
Countries." Quarterly Journal of Economics 106 (May
1991): SI03-25.
and Xavier Sala-i-Martin. "Convergence." Journal of
Political Economy 100 (April 1992): 223-51.
Becker, Gary. Human Capital. Chicago: University of Chicago
Press, 1975.
Benhabib, Jess, and Mark M. Speigel. "Growth Accounting
with Physical and Human Capital Accumulation." New
York University, photocopy, November 1991.
Denison, Edward F. The Sources of Economic Growth in the
United States and the Alternatives before Us. New York:
Committee for Economic Development, 1962.
. Accounting for United States Economic Growth. 19291969. Washington, D.C.: The Brookings Institution, 1974.
Easterlin, Richard A. "Why Isn't the Whole World Developed?" Journal of Economic History 61 (March 1981): 1-19.
Jorgenson, Dale W„ and Barbara M. Fraumeni. "Investment in
Education and United States Economic Growth." Harvard
Institute of Economic Research Discussion Paper 1573 October 1991.
Kcndrick, John W. Postwar Productivity Trends in the United
States, 1948-1969. New York: National Bureau of Economic Research, 1973.
Learner, Edward E. Specification Searches: Ad Hoc Inference
with Nonexperimental Data. New York: John Wiley and
Sons, 1978.
Levinc, Ross, and David Renclt. "A Sensitivity Analysis of
Cross-Country Growth Regressions." American Economic
Review 82 (September 1992): 942-63.
Lucas, Robert E. "On the Mechanics of Economic Development." Journal of Monetary Economics 22 (July 1988): 3-42.
Mankiw, N. Gregory, David Romer, and David N. Weil. "A
Contribution to the Empirics of Economic Growth." Quarterly Journal of Economics 107 (May 1992): 407-37.
Nussbaum, Bruce, Aaron Bernstein, Elizabeth Ehrlich, Susan
B. Garland, and Karen Pennar. "Needed: Human Capital."
Business Week, September 19, 1988, 100-41.
Prescott, Edward C. "Theory Ahead of Business Cycle Measurement." Federal Reserve Bank of Minneapolis Quarterly
Review (Fall 1986): 9-22.
Psacharopoulos, George. "The Contribution of Education to
Economic Growth: International Comparisons." In International Comparisons of Productivity and Causes of the Slow-
12
Economic Review
down, edited by John W. Kendrick, 335-55. Cambridge,
Mass.: Ballinger Publishing Company, 1984.
. "Returns to Education: A Further International Update
and Implications." Journal of Human Resources 20 (Fall
1985): 583-604.
Quah, Danny. "Gallon's Fallacy and Tests of the Convergence
Hypothesis." Massachusetts Institute of Technology photocopy, 1990.
Renelt, David. "Economic Growth: A Review of the Literature." World Bank Working Paper WPS 678, May 1991.
Romer, Paul. "Human Capital and Growth: Theory and Evidence." Carnegie-Rochester
Series on Public Policy ^52
(1990): 251-86.
Rosen, Sherwin. "A Theory of Life Earnings." Journal of Political Economy 84 (August 1976): S45-67.
Sala-i-Martin, Xavier. "Lecture Notes on Economic Growth
(I): Introduction to the Literature and Neoclassical Models."
National Bureau of Economic Research Working Paper
3563, December 1990a.
. "Lecture Notes on Economic Growth (II): Five Prototype Models of Endogenous Growth." National Bureau of
Economic Research Working Paper 3564, December
1990b.
Schultz, T. Paul. "Education Investments and Returns." In
Handbook of Development Economics, 544-630. Amsterdam: North-Hoi land, 1988.
Schultz, Theodore W. "Investment in Human Capital." American Economic Review 61 (March 1961): 1-17.
Silk, Leonard. "Riots Put Focus on Economic Ills." New York
Times, May 6, 1992.
Solow, Robert M. "A Contribution to the Theory of Economic
Growth." Quarterly Journal of Economics 70 (February
1956): 65-94.
_ . "Technical Change and the Aggregate Production Function." Review of Economics and Statistics (August 1957 V
312-20.
Stokey, Nancy. " H u m a n Capital, Product Quality, and
Growth." Quarterly Journal of Economics 106 (May 1991V
587-616.
Swan, T.W. "Economic Growth and Capital Accumulation."
The Economic Record 32 (November 1956): 334-43.
Tallman, Ellis W„ and Ping Wang. "Human Capital and Endogenous Growth: Evidence from Taiwan." Federal Reserve Bank of Atlanta unpublished paper, 1992.
Van de Klundert, Theo, and Sjak Smulders. "Reconstructing
Growth Theory: A Survey." CentER, Tilburg Universit>*
Working Paper 9146, July 1991.
Weisbrod, Burton. "The Valuation of Human Capital." Journal
of Political Economy 69 (October 1961): 425-36.
September/October 1992
R e g i o n a l Employment
By Industry:
Do Returns to
Capital Matter?
Stacy E. Kottman
f
^ ^
•
The author is an economist
in the regional section of the
Atlanta Fed's research
department. He is grateful to
Ellis Tallman and Tom
Cunningham for helpful
comments and Hing Kwong
Law for indefatigable
research assistance.
Federal Reserve Bank of Atlanta
T tates and communities in the United States have long pursued economic development. A variety of programs, including tax incentives, direct subsidies, and advertising, have met with mixed
W success over the years. Whatever the strategy—recruiting firms
from other regions, promoting new foreign investment, or attempting to stimulate local entrepreneurship—the goal of these communities is to
influence the regional distribution of industries across the country. Effective
development efforts, those that improve a region's long-term relative wealth
and well-being, should be grounded on an understanding of the economic
forces that shape an area's evolving industrial composition.
Stories about regional growth and interregional shifts in industrial composition do not make the news as frequently as in the heyday of competition
between the Rust Belt and Sun Belt during the 1970s. Nonetheless, regional
shares of national employment continue to change, with the extent of the
changes varying widely from industry to industry regardless of what is happening to total industry employment at the national level. These movements
of employment and capital among regions are critical in explaining differences in regional growth rates. In addition, regional shifts in industrial composition play a key role in determining changing patterns of per capita
income and labor productivity.
Several fundamental cause and effect questions surround the issue of interregional industrial migration: In an economy in which capital and labor
have long been able to migrate across regions without restriction, why do
interregional employment and capital shifts continue? Is this migration a
form of arbitrage, exploiting fundamental economic imbalances among regions? Or are U.S. regions generally in economic equilibrium only to be
Eco n o m ic Review
13
buffeted by economic "shocks" with region-specific
effects, such as dramatic changes in oil prices, defense
expenditures, or state development initiatives?
From the viewpoint of conventional economic
models used to study regional growth, it is puzzling
that interregional capital flows have persisted despite
the fact that workers appear content to receive different wages as compensation for regional differences in
cost of living or amenities (that is, in technical terms,
labor markets are in equilibrium across regions). The
continued interregional employment shifts under these
conditions suggest that relative profitability—differentials in returns to capital—may be the source of the
capital flows.
If so, there are implications not only for theoretical
models of regional growth and development policy but
also for regional econometric models used to forecast
economic activity or evaluate the impact of development initiatives. In conventional regional econometric
models, changes in industry employment are primarily
dependent on changes in national or local product demand. These top-down, demand-oriented models typically assume away the potential impact of regional
differentials in returns to capital (see the box on page
17) and, consequently, disregard interregional capital
flows, which may be more important to long-run regional growth and development than existing industrial composition.
The purpose of this study is to address the fundamental question of whether regional differentials in
returns to capital drive changes in regional employment. Empirical tests assess the influence of relative
profitability on changes in industry employment across
U.S. states from 1969 through 1986.
The results confirm that for industries producing
goods and services for export to other regions there is a
positive, significant relationship between relative returns to capital and a state's share of national employment. Regional differentials in profitability were not
found to be significant in explaining employment share
for industries that produce goods or services for local
consumption. For these industries local market size is
the key determinant of employment share over time.
1
Interregional Mobility by Industry
In this study interregional mobility is broadly
defined as the rate at which regional employment
shares—industry ratios of regional employment to national employment—change over time. (In any year
2
14
Economic Review
the sum of regional shares for an industry will equal
1.0.) Considering regional employment mobility in
these terms largely neutralizes the influence of business cycles and the longer-term effects of changes in
technology or international competition. For instance,
in a declining industry in which employment is falling
faster nationally than in a region, the region will gain
national employment share.
An example contrasting employment mobility in
four industry categories in two U.S. regions will illustrate the concept. Using Bureau of Economic Analysis
(BEA) regions, national employment shares for the
Mideast and Southeast are calculated for total employment, services, retail trade, and electronic equipment
over the 1969-90 period. These regions were selected
because they had similar total employment levels in
1969 and exhibited divergent employment trends. Employment share for each category is plotted in Chart 1.
Over the twenty-one-year period, total employment
share drifted steadily lower in the Mideast and higher
in the Southeast. The relative shift in services was
smaller than for total employment in both regions. In
fact, the Southeast's share of national service employment has remained relatively steady. In retail trade, the
regional shifts proceeded in the same direction as total
employment but at a modestly faster pace.
At the same time, a dramatic shift occurred in national employment share for the electronic equipment
industry. In 1969 the Mideast employed 24 percent of
the nation's electronic equipment workers, more than
twice as many as the Southeast. By 1990, however, the
Southeast's electronic equipment work force was onethird larger than the Mideast's. Mideast employment
levels fell nearly in half while the Southeast's electronic equipment work force rose by 44 percent.
To assess whether such variations in industry mobility are observable across all eight regions, Table 1
offers a simple aggregate measure of interregional mobility for thirty-four private-sector industries from
1969 to 1990. By industry, national employment
shares for the eight BEA regions were calculated for
1969 and 1990. For each region, 1969 employment
share was subtracted from 1990 employment share.
The absolute value of these regional employment
share differences were then summed by industry.
The resulting measure is a crude but revealing index of interregional mobility across industries. By design, the index does not provide information on
industry growth, contraction, or cyclical performance.
It does offer a comparable measure of net regional employment mobility for industries in national decline as
well as expansion.
3
4
September/October 1992
Two regularities are immediately apparent from
Table 1. First, manufacturing industries are the most
mobile, with their higher numbers dominating the bottom half of the ranking, while services, trade, and utilities are the least mobile, appearing in the top half of the
list. Second, the manufacturing industries that exhibit
relative immobility are typically restrained by what
economists call fixed effects. For example, the production of chemicals, lumber, and food products all depend heavily upon geographically fixed raw resources.
To the extent that these employment shifts serve as
a proxy for regional investment, this evidence suggests that interregional capital flows vary systematically by type of industry. Empirically identifying the
fundamental economic incentives that determine this
variable migration requires a conceptual framework,
or economic model.
/interregional Capital Migration
Do Returns to Capital Matter? Conceptually, it is
neither innovative nor imaginative to assume that capital will seek its highest return. It is relative profitability
and changes in the regional profit topology that interest
owners of capital (James A. Chalmers and Terranee L.
Beckhelm 1976). Unfortunately, because of limited data on regional capital stock and returns to capital, the
determinants and pattern of capital migration have
Chart 1
Employment Share in the Mideast and Southeast, 1969-90
(Total and by selected
Total
Percent
25
Services
Percent
25
Southeast
20
industries)
^.Mideast
20
Mideast
N
Southeast
15
10
1969
1973
1977
1981
1985
1989
Retail Trade
10 - 1 1 1 1 1 1 1 1 i 1 ! 1 1 : 1 1 !
•H—1—
1969
1973
1977
1981
1985
1989
Percent
25
2 0
"-- ^
Mideast
--
Southeast
Mideast
x
V
15
15
10
1969
Electronic Equipment
\
10
1973
1977
1981
1985
V
1969
1973
1977
1981
1985
1989
S o u r c e : C a l c u l a t e d by the Federal Reserve Bank of Atlanta using B u r e a u of E c o n o m i c Analysis data.
Federal Reserve Bank of Atlanta
Eco nom ic Review
15
Table 1
Regional Employment Mobility by Industry
Mobility
lndexa
0.115
0.133
0.134
0.137
0.147
0.156
0.162
0.168
0.177
0.185
0.187
0.193
0.196
0.196
0.197
0.201
0.201
0.205
0.225
0.252
0.258
0.258
0.261
0.267
0.276
0.278
0.288
0.308
0.337
0.353
0.375
0.376
0.432
0.510
a
/\ low mobility
BEA
Industry
Code
Industry
Chemicals and Allied Products
Lumber and W o o d Products
Health Services
Personal Services
Wholesale Trade
Amusement Services
Construction
Insurance Agents and Brokers
Retail Trade
Electric, Gas, and Sanitary Services
Communication
Hotels and Lodging
Food and Kindred Products
Trans. Equip., except Motor Vehicles
Trucking and Warehousing
Legal Services
Paper and Allied Products
Furniture and Fixtures
Textile Mill Products
Business Services
Fabricated Metal Products
Transportation by Air
Real Estate Services
Stone, Clay, and Glass Products
Printing and Publishing
Motor Vehicles and Equipment
Primary Metals Industries
Transportation Services
Miscellaneous Manufacturing Industries
Machinery, except Electrical
Rubber and Miscellaneous Products
Electric and Electronic Equipment
Apparel and Other Textile Products
Instruments and Related Products
index suggests relatively
less interregional
change
ment share across eight U.S. regions during the 7 969-90 period.
reported for those industries that are evaluated
empirically
260
320
621
580
510
611
180
554
520
500
490
570
210
380
460
622
240
330
220
601
350
482
555
410
250
390
340
484
430
360
290
370
230
420
in employThe index is
in the article.
Source: C a l c u l a t e d by the Federal Reserve Bank of Atlanta using Bureau of
E c o n o m i c Analysis data.
received relatively little attention compared with the
voluminous research on wages and interregional labor
migration. Additionally, data constraints have limited
research on interregional capital flows to the manufacturing sector. However, existing empirical research
provides an important foundation from which this
study proceeds.
3
6
16
Economic Review
Robert F. Engle (1974) regressed an approximation of relative returns to capital on an estimate of total investment in Massachusetts for four highly aggregated manufacturing sectors over a sixteen-year
period. Although his results confirm a positive and
significant role for differentials in returns to capital,
he did not test his proposition across a cross-section
of states.
Lynne E. Browne, Peter Mieszkowski, and Richard
F. Syron (1980) concluded that the South's rapid capital expansion was not attributable to the existing
industrial composition but rather to interregional capital migration. In addition, their study indicates that
low nominal wages played the most important role in
attracting net investment into the South. Using surveys, other researchers have confirmed that low
wages and a low level of unionization are important
factors in investment location decisions (Robert A.
Nakosteen and Michael A. Zimmer 1987; Leonard F.
Wheat 1986; Robert J. Newman 1983; W.C. Carlton
1979). These surveys also reveal that proximity to
markets is a key factor in location decisions.
Edward M. Miller (1981) tested whether nominal
wage differentials compensate for differences in labor
productivity. If so, unit labor costs could be in equilibrium across regions despite nominal wage differentials. He found that variations in labor productivity
only partially offset wage differentials. Not surprisingly, after adjusting for the differences in labor productivity, Miller found a positive correlation between
lower wages and profitability. However, he makes no
attempt to correlate capital migration to estimates of
relative profitability.
A synthesis of these results suggests that capital migration does respond to profit differentials that may
originate from regional variation in production costs.
Assuming that product price and the cost of capital are
nationally determined, imbalance in regional wages
appears to play a primary role in determining relative
regional profitability and, by extension, interregional
capital flows in the manufacturing sector.
However, two issues remain unresolved. These results do not immediately explain why the pace of interregional capital migration would vary widely
among manufacturing and nonmanufacturing industries. More intriguing is the question of how these results square with the growing body of evidence that
the regional labor markets are in equilibrium.
How Firms and Labor View Regional Wages.
There is an emerging consensus in the literature that
regional nominal wage and income differentials have
all but vanished once adjusted for cost-of-living and
September/October 1992
Labor Demand in Regional Econometric Models
While excellent labor market data are available, the
absence of consistent capital stock, net investment, and
output data has rendered direct production function estimation for regional models nearly impossible. In 1972
Donald Ratajczak proposed a methodology to specify
state-specific labor demand without capital stock data.
This methodology has since become widely used. Assume that (1) the production relationship can be approximated by constant returns to scale; (2) firms exhibit
profit-maximizing behavior in perfect competition; (3)
the labor market is in equilibrium; and (4) regional returns to capital are in equilibrium.
Under these assumptions, it can be shown that labor
is paid its marginal product and the derived demand for
labor can be estimated without capital data. The legitimacy of this theoretical specification rests heavily on
its assumptions. Ratajczak (1972) notes that differential
capital returns, or profitability, could stimulate interregional capital flows. However, the consideration of disequilibrium is dismissed because of data constraints.
In practice, the conventional labor demand is presented as follows:
1
lL = a + b(l Q) + c(lW) + d(T),
n
t
where l is log, L is demand for labor services in time /,
Q is output (national or local demand or both), W is the
real wage, and T is the level of technology (with l L
often added as a partial adjustment term) (Max E. Jerrell
and James M. Morgan 1988; Roger Bolton 1985; William
J. Milne, Norman J. Glickman, and F. Gerard Adams
1980; Ratajczak 1972).
The limitation of this equilibrium specification in determining relative regional employment growth among
n
1
t
f
t
n
[{
labor force characteristics (see John A. Bishop, John P.
Formby, and Paul D. Thistle 1992; Gerald A. Carlino
1992; Mark Dickie and Shelby Gerking 1987; Gary
M. Fournier and David W. Rasmussen 1986; Leonard
G. Sahling and Sharon P. Smith 1983; Don Bellante
1979). The labor market, from the viewpoint of a
worker, appears to be in equilibrium.
However, industries continue to migrate across regions in response to profit differentials generated by
the apparent lack of wage equilibrium. The conundrum lies in the fact that real wages from labor's viewpoint are not equivalent to real wages f r o m the
viewpoint of the firm (see Daniel H. Garnick and
Howard L. Friendenberg 1982; Stacy E. Kottman
1990; Carlino 1992). From the viewpoint of labor, for
Federal Reserve Bank of Atlanta
states can be seen by examining the paths through which
relative labor demand could be expressed. Under equilibrium conditions, changes in aggregate industry demand would result in proportional, not relative, changes
in labor demand.
Technology or innovation would not affect relative
labor demand by industry unless it was region-specific.
The assumption of limited diffusion is not credible in today's economy unless it is dependent on a region-specific
resource, such as raw material. Generally, technology is
assumed to displace labor. Therefore, the expected sign
of d is negative.
The real wage, which in most regional models is simultaneously dependent on labor demand and supply
conditions, inversely influences labor demand in the
standard specification. T h e r e f o r e , the c o e f f i c i e n t c
should be negative. In fact, empirical estimation of this
specification often results in a positive wage coefficient
(Gordon L. Clark, Meric S. Gertler, and John E. Whiteman 1986).
Adjusted for aggregate demand, change in the derived demand for labor under this specification is limited
to wage-driven capital-labor substitution and technological change. The equilibrium specification does not
consider employment shifts driven by differentials in interregional returns to capital.
Note
1. Technically, a constant elasticity of substitution (CES)
production function, homogeneous of degree one, with
disembodied technological change is specified for estimation purposes.
example, a 10 percent wage differential may compensate for regionally determined differences in cost of
living or amenities. However, to a firm producing a
product whose price is determined nationally, a 10
percent wage differential may represent a 10 percent
differential in labor costs (assuming similar labor productivity across regions).
For the firm, labor cost differentials may create an
opportunity for more profitable production in the lowwage region and may drive capital across regions.
Whether this migration results in a narrowing of regional wage differences depends on the resultant
change in relative industrial composition. In fact,
Browne, Mieszkowski, and Syron (1980) concluded that
the Northeast had become relatively more concentrated
Eco n o m ic Review
17
in high-wage, or capital-intensive, industries as lowerwage, or labor-intensive, production moved South. Incorporating this insight helps to explain the remaining
unresolved issue—the varying rate of interregional mobility across industries.
Expectations of Mobility Vary by Industry.
Conceptualize the following industrial spectrum. Toward one end is an industry—the neighborhood dry
cleaner, for example—whose total regional capital investment and employment is directly proportional to
local market size. The product or service is produced
primarily for local customers, and both input prices,
such as labor costs, and output prices are determined
by local markets. If it is assumed that capital requirements for firm entry are not prohibitive, competition
quickly squeezes out opportunities to set high prices
and raise profits. In such an industry, differentials in
returns to capital, or profitability, would not vary appreciably across regions. Therefore, once adjusted for
local market size, locally oriented industries do not
engage in interregional migration in search of higher
returns to capital.
Toward the opposite end of the spectrum is an industry like automobile assembly that, after using local
inputs, exports all production. Product demand and
price are largely determined in national markets. Although the cost of capital is determined nationally,
other important input prices—primarily wages—are
determined locally and could vary across regions, allowing regional differentials in total returns to capital.
This relative profitability most likely would initiate
interregional capital flows resulting in shifting regional employment shares. Local market size should by
comparison be unimportant in determining a region's
employment share for export-oriented industries.
All thirty-four industries in Table 1 lie somewhere
along this conceptual spectrum. A casual evaluation of
industry mobility as ranked by the table suggests that
export-oriented manufacturing industries do exhibit
more mobility than locally oriented industries such as
services, retail, and utilities. To test this proposition on
an industry-by-industry basis, an empirical model was
estimated to examine the importance of relative returns to capital as a determinant of employment share
across states from 1969 through 1986.
7
ES = a + b(RRC _ ) + c(MS^)
jt
jt
{
+ d(FXD.) + <?.,
where j represents the state, t is the year, ES is national
employment share, RRC is relative returns to capital,
MS is share of national market, FXD specifies state
fixed effects, and e is random error.
For any industry this model specifies that state employment share (ES ) is a lagged function of relative
returns to capital (RRC^, state market size (MS), and
state fixed effects (FXD.) that are not lagged'. This
specification evaluates the role of relative profits for
industries in absolute national decline as well as expansion. The model is not uniquely concerned about
firms physically moving from state A to state B. By
design, it considers relative regional shifts in net investment and employment, regardless of the source.
The dependent variable, ES.., was defined above,
but a brief review of the explanatory variables, their
empirical representations, and their expected sign or
significance by industry should be helpful in interpreting the results.
Relative Returns to Capital. Conceptually, RRC
is defined as the ratio of state returns to capital to national returns to capital. The model assumes a lagged
effect on employment share by considering the nature
of capital investment. Once the decision for additional
investment is made on the basis of observed changes
in relative profitability, it takes time to locate a site,
construct facilities, and hire a work force. Lags should
vary with the complexity or fixity of the capital investment under consideration.
Unfortunately, no consistent annual corporate-profits
or capital-stock series exists at the regional level. Previous empirical studies used value-added minus labor
earnings as an admittedly imperfect proxy for total returns to capital in manufacturing industries. However,
the BEA's recently released gross state product (GSP )
series by industry offers an opportunity to calculate
a more serviceable measure of relative profitability
across manufacturing and nonmanufacturing industries.
Gross state product by industry is defined as the
sum of employee compensation, proprietors' income,
indirect business taxes, and capital-related charges
(BEA 1985, 1988, 1991; see the appendix for a detailed discussion). The BEA's estimate of capital-related
charges (CRC ), which include adjusted corporate
profits, rental income, net interest, and depreciation, is
significantly more independent and consequently less
distortive than the gross residual between value-added
and labor earnings. Therefore, the ratio of capitalrelated charges to gross state product provides an
r
r
r
;j
.Model Specification
The influence of relative returns to capital is modeled for each industry as follows:
18
Economic Review
September/October 1992
improved measure of relative returns to capital in each
state. Because capital-related charges can be negative,
relative returns to capital (RRC .) by industry was calculated as the difference between the state and national ratios in each year:
(/
RRC = (CKC-./GS/-..) - ( C R C J G S P J .
tj
If the model holds, the estimated coefficient for
R R C b , should be positive and significant for
expoit-oriented industries. In fact, for export-oriented
industries, differentials in returns to capital should be
more important than local market share {MSp. Relative returns to capital should not be significant for locally oriented industries, where local market share is
expected to dominate.
Market Share. Surveys confirm that proximity to
market is a key factor in location and investment decisions. Therefore, to reveal the true role of RRC^ the
influence of local market share, MS.,j must first be accounted for. Local market share is calculated as the ratio of state to national total personal income in each
year.
MS should be more significant for those industries
that retain a local orientation in production, pricing,
and consumption. Competitive pressures inherent in
such industries would serve to limit persistent differentials in returns to capital among regions. The coefficient, c, for MS may be positive and significant for
some export-oriented industries, but it should clearly
dominate the results for locally oriented industries,
where differentials in returns to capital are not expected to persist.
Fixed Effects. Fixed, but unobservable, effects—
such as access to natural resources, climate, amenities,
culture, political environment, and historical accident—are assumed to be significant in determining
employment share by state. If such effects are assumed
to exist and are correlated with other explanatory variables, ignoring their presence would introduce specification bias (Cheng Hsiao 1986). Because it is reasonable
to assume that a state's fixed effects are correlated with
relative returns to capital, they were accounted for by
including dummy variables for each state.
A major limitation of using dummy variables to capture fixed effects is that they do not identify which factors are responsible for the fixed effect. However, this
characteristic has some value because it makes it possible to compare the significance of fixed-effect coefficients across industries and states, providing additional
insight on mobility. State fixed effects are expected to
be more important in some industries than others.
i j f
j
j
8
Federal Reserve Bank of Atlanta
Estimation Issues
The Economic Importance of Relative Returns
to Capital. Estimating the model with the relative ratios as defined may verify whether RRC is statistically significant by industry but not whether it is economically important in comparison with local market
share. Because of the wide cross-sectional variation in
state sizes, employment share and market share will be
highly correlated. Consequently, the estimated coefficient on MS should be statistically significant and
large in value when the model is estimated with the ratios directly.
In contrast, the cross-sectional range of values for
RRC is relatively narrow. Although its statistical significance may be confirmed, the estimated coefficient
on RRC would likely be small relative to the coefficient on M S It would thus be inappropriate to compare standardized coefficients for RRCand MS in
evaluating relative economic significance. However,
estimating the model in first-difference form—that is,
how the ratios change from year to year—and comparing the standardized coefficients provides an evaluation of the relative economic importance of each
variable.
State Sample Selection. GSP data are available for
all fifty states and the District of Columbia. However,
it is difficult to defend the proposition that these fiftyone areas represent independent samples drawn from a
single population. In addition, several small states
report a very small employment or GSP share. The
smaller a state's share, the more vulnerable its data are
to errors in measurement or dominant firms that raise
disclosure problems (BEA 1985).
Because of these considerations, the model was estimated across a subset of the states. The selection
rule was simple: All states with less than a 1 percent
share of national total manufacturing employment
were deleted from the sample, leaving thirty states
whose total manufacturing employment share summed
to 93 percent of the national total. These larger states
are more likely to share broad characteristics that
make the underlying assumption of a single population
more defensible, placing a reduced burden on the
fixed-effects variables.
Industry Sample Selection. For estimation, industry selection was limited to private nonagricultural
sectors. Two types of industries were excluded from
testing on an a priori basis. First, industries with significant immobility were not considered: tobacco,
petroleum and coal, mining, rail transportation, water
r
/
if
fj
r
/
9
10
Economic Ret neu>
19
Table 2
The Significance of Relative Returns to Capital
In Explaining Change in State Employment by Industry'
BEA Code
RRC
Export-Oriented
Primary Metals Industries''
Fabricated Metal Products
Electric and Electronic Equipment
Motor Vehicles and Equipment15
Chemicals and Allied Products1'
Miscellaneous Manufacturing Industries
Instruments and Related Products
Machinery, except Electrical
Trans. Equip., except Motor Vehicles'3
Textile Mill Products1'
Lumber and Wood Products
Furniture and Fixtures
Apparel and Other Textile Products
Rubber and Miscellaneous Products
340
350
370
390
260
430
420
360
380
220
320
330
230
290
+HS
+HS
+HS
+HS
+HS
+HS
+S
+S
+S
Both Export- and Locally Oriented
Stone, Clay, and Glass Products''
Transportation by Air
Printing and Publishing
Paper and Allied Products
Food and Kindred Products
410
482
250
240
210
+HS
Locally Oriented
Legal Services
Construction
Real Estate Services
Health Services
Insurance Agents and Brokers
Hotels and Lodging
Electric, Gas, and Sanitary Services
Trucking and Warehousing
Wholesale Trade
Retail Trade
Personal Services
Communications Services
Amusement Services
Transportation Services
Business Services
622
180
555
621
554
570
500
460
510
520
580
490
611
484
601
Industry
This table summarizes
a
lagged
significant),
than
b
the first-difference
RRC^ coefficient
(HS = highly
estimation
significant
the sign and level of significance
i
+1
+1
-I
+s
+1
+HS
+1
+1
+1
+1
-I
-I
+!
+l
+S
+s
-I
-HS
+HS
+ HS
+S
+HS
+HS
+HS
+HS
+HS
+HS
+HS
+HS
+HS
+HS
+HS
+HS
+S
+l
+l
+l
+l
+l
-HS
-HS
+l
+l
-I
at a I percent
for the lagged MS
u
-I
+s
results for thirty states. It shows
[significant
RRC > MS
MS:
coefficient,
level];
the sign and level
S = significant
and whether
of statistical
Isignificant
the standardized
significance
at a 5 percent
coefficient
for the
levell;
for RRC,
I = inis larger
MSj.
For these industries,
changes
RRC,; in level form on changes
in RRC:/ were not significant
in employment
in explaining
changes
in employment
share. Reported
results reflect
regressing
share.
S o u r c e : C a l c u l a t e d by the Federal Reserve B a n k of Atlanta using Bureau of E c o n o m i c Analysis data.
20
Economic Review
September/October 1992
transportation, and pipelines. In these industries fixed
effects in the form of natural resources or capital fixity
limit potential migration. Several other industries were
excluded from consideration on the basis of data considerations."
Aside from these deletions, data were prepared
across thirty-four private, nonagricultural industries.
Because many industries exhibited some degree of
missing or disclosure data problems, a sample of states
with complete data was identified by industry. The
specified variables were then calculated and arranged
into a cross-section time-series format over the 1969-86
period for estimation.
RRCj was significant for two industries (legal services
and construction), the standardized coefficient for
market share was larger in every industry, substantiating the relative importance of market share for
determining employment share in locally oriented industries. In general, these results confirm that regional
differentials in returns to capital are not important in
determining interregional employment shifts for locally oriented industries.
12
Re suits
The results confirm that employment share responds in a dynamic fashion to differentials in lagged
returns to capital for those industries that are exportoriented. Table 2 presents summarized results across
all thirty-four industries, categorized as export-oriented,
locally oriented, or industries that should exhibit characteristics of both.
Fourteen manufacturing industries are labeled
export-oriented. After adjustment for local market
share and fixed effects, RRC was found to be statistically significant in determining employment share in
ten industries. In eight of these ten, the standardized
coefficient on relative returns to capital was larger
than the standardized coefficient on market size.
An examination of the state dummy variable coefficients for the four export-oriented industries for which
RRC was not significant confirms a dominant role for
state fixed effects. Natural resource fixed effects no
doubt play a direct major role in lumber and wood and
an indirect role in furniture and fixtures and rubber
and plastics. Apparel reflected fixed effects in the
form of high preexisting employment concentrations
over few states.
Five industries were thought to exhibit both export
and import characteristics. RRC was only significant
in stone, clay, and glass while MS.. was significant in
air transportation services, printing and publishing,
and paper and allied products. Neither RRC nor MS
were able to explain employment share in food and
kindred products, which exhibited significant fixed effects.
Sixteen individual industries were considered locally oriented. In contrast to export-oriented industries,
MS. was significant in all but two cases. Even though
13
rj
:j
}j
:j
Federal Reserve Bank of Atlanta
Conclusion
This empirical study examines a simple but fundamental question regarding the interregional migration
of capital and employment. Do differentials in returns
to capital matter? For each of thirty-four industries, a
fixed-effects model was estimated with pooled data
over eighteen years and thirty states to evaluate the
role of differentials in profitability on relative employment growth. Appropriate proxies for returns to capital were constructed from gross state product data by
industry, and an appropriate estimation procedure was
applied.
To the extent that capital is not influenced by noneconomic fixed effects, the results suggest that differentials in returns to capital exert a significant and
positive dynamic influence on interregional employment growth for export-oriented industries. For those
industries, differentials in returns to capital are generally more important than local market size in explaining changes in state employment share over time.
As expected, entry and competition preclude a
lagged dynamic relationship between returns to capital
and relative employment growth for locally oriented
industries. In these cases, employment share is primarily a positive function of local market size.
The results are consistent with the proposed theoretical framework that firms and workers hold differing viewpoints on nominal regional wage differentials.
From the viewpoint of labor, the regional labor markets appear to be in general equilibrium. Nominal
wage differentials simply compensate for differences
in regional cost-of-living or amenities. However, from
the viewpoint of the export-oriented firm, lower nominal wages may offer a more profitable investment opportunity that motivates interregional migration of
capital and employment.
These findings yield several implications for regional development policy and regional growth models. First, low-wage states should not count on long-term
Econ om ic Review
21
convergence to national per capita norms of income
and output. Although conventional regional growth
models maintain that convergence should occur as a
result of interregional migration, the conceptual framework and empirical evidence presented in this article
suggest an alternative scenario. Employment in lowwage regions could grow faster than in the nation
without exhibiting wage convergence by absorbing increasing shares of the nation's relatively labor-intensive,
low-wage industries. These results confirm the importance of looking beyond aggregate per capita measures
of incomes and output when discussing regional dynamics.
Second, interregional capital and employment mobility is a rational response to persistent differentials in
returns to capital across regions for export-oriented
firms. As evidenced by the Mideast/Southeast comparison, these interregional shifts in industrial composition have been relatively steady throughout the business
cycles and oil shocks of the past twenty-one years.
Broadly interpreted, this persistent interregional migration represents a search for regional equilibrium
that may be more pervasive than adjustments to periodic exogenous price shocks.
One sobering extension of this interpretation for development policy is that marginal tax incentives or investment subsidies may have little effect on a region's
employment composition over time. While development officials are quick to claim victory when recruiting firms, the long-term costs and benefits of direct
subsidies are not so clear. In fact, direct investment incentives could conceivably deter state progress toward
national norms by subsidizing inefficient firms or by
displacing public investment in policies that could truly lift a region's relative per capita earnings.
For regional economists engaged in econometric
modeling, these results suggest that specification bias
may lurk in conventional specifications of labor demand by industry, which assume away differentials in
returns to capital as a source of labor demand. Although this potential specification bias may present
few problems in short-term forecasting, it could be
troublesome when using demand-driven models for
longer-term policy evaluation.
14
Appendix
Gross State Product and Other Data Series
In 1988 the BEA reported gross state product (GSP)
for each of sixty-one industries at the two-digit Standard
Industrial Classification (SIC) level from 1963 through
1986. GSP is defined as the sum of four independently
generated and reported component series (BEA 1985, 1988,
1991):
1. C o m p e n s a t i o n of e m p l o y e e s , which includes
wages, salaries, employer contributions for social
insurance, and other labor income.
2. Proprietors' income with inventory valuation
and capital consumption adjustment, which is the
income, including income-in-kind, of sole proprietorships, partnerships, and tax-exempt cooperatives.
3. Indirect business taxes (IBT.^ and nontax liabilities, which consist of tax liabilities that arc chargeable to business expense in the calculation of
profit-type income. Normally, this category would
include sales, excise, and property taxes and regulatory or inspection fees. It does not includc corporate income taxes.
4. Capital-related charges ( C R C w h i c h consist of
(a) corporate profits with inventory valuation adjustment, which is the income of corporations
measured before profit taxes, before deduction
of depletion charges, after exclusion of capital
22
Economic Review
gains and losses, and net of dividends received
from domestic corporations;
(b) the rental income of persons from the rental of
real property, imputed net rental income of
owner-occupants of nonfarm dwellings, royalties received from patents, copyrights, and
rights to natural resources;
(c) net interest, which is the interest paid by business less interest received by it;
(d) business transfer payments and subsidies less
current surplus of government enterprises; and
(e) capital consumption allowances, which are depreciation charges and the value of accidental
damage to fixed business capital.
The BEA data differ from previous periodic Census
Bureau estimates in four ways. First, the BEA adjusts the
census product to include value added by central administrative offices. Second, the BEA subtracts the costs of
purchased services from census estimates. Third, the
BEA adjusts for differences in industrial classification
between census and BEA sources for payroll data. Each
of these adjustments relies on general distributional assumptions. However, any distortion caused by these adjustments should be smaller than the distortions that
might exist in their absence. Fourth, the BEA includes
nonmanufacturing industries.
September/October 1992
In general, data for goods-producing and regulated industries offer a higher degree of independence and a lower reliance on distributional assumptions. More extensive
assumptions are required for calculating service and
trade industry components. Nonetheless, in the absence
of complete information these series offer a defensible
approximation of gross product originating by industry.
The potential data distortions should not nullify the inference.
The BEA's new capital-related charges CRC series is
significantly more independent and less distortive than
the gross residual between value-added and labor earnings, which functioned as a C f t C in previous empirical
studies. The CRC^ scries excludes IBT^, which by definition was included in earlier residual estimates. To the extent that //?7V varies across states, its exclusion further
reduces previous distortions.
In this article, capital-related charges ( C R C ) are used
as a proxy for total returns to capital. Does the inclusion
tj
of factors other than corporate profits inappropriately
bias C/?C. as a proxy for capital returns? Used in a comparative fashion across industries, the CRC . proxy could
present potential distortions because the distributional assumptions that are used vary by industry. However, the
hypotheses are to be tested on an industry-by-industry
basis. To the extent that capital-labor ratios are similar
within an industry across regions, CRC is an unbiased
proxy.
Market share (MS ) originates with the BEA's SA-5
personal income series (PL ) and is calculated as follows:
j
tj
jr
f
MS = (PL/PIJ.
jt
The BEA SA-25 series provided annual industry employment data. For estimation purposes, state employment share is calculated as follows:
tj
ES =
ijl
(EM /EM J.
jj
i
Notes
1. Although economists have recently offered more sophisticated applications of neoclassical growth theory (see Barro
and Sala-i-Martin 1992), these extensions remain variations
on a simple neoclassical production model first applied to
U.S. regions by Borts and Stein in 1964. Under standard
neoclassical assumptions, firms in low-wage regions should
exhibit a lower capital-to-labor ratio for any given production process, yielding a higher marginal value product for
capital than is found in high-wage states. As capital migrates from the high-wage region to the low-wage region,
capital-labor ratios in the low-wage region begin to rise, reducing returns to capital and increasing the marginal value
product of labor. Consequently, wages and returns to capital will converge across regions. Once in equilibrium, interstate differences in employment growth rates for a given
manufacturing industry arise solely from the wage impacts
of relative shifts in regional labor supply.
Boris and Stein concluded that low wages persist in some
regions because of the intraregional migration of agricultural workers into manufacturing and the low-wage region's
higher fertility rates. Although this explanation was consistent with demographic and industrial developments during
the period under study (1919-57), it has become increasingly irrelevant since 1960 as the intraregional shift from agriculture to manufacturing has abated.
2. Changes in shares, the focus of this research, should not be
confused with changes in actual employment levels.
3. The Mideast is defined as Delaware, the District of Columbia,
Maryland, New Jersey, New York, and Pennsylvania. The
Southeast encompasses Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina,
Federal Reserve Bank of Atlanta
South Carolina, Tennessee, Virginia, and West Virginia.
Employment data were taken from the BEA s SA-25 annual employment by industry series, which begins in 1969.
4. The BEA regions arc Great Lakes, Mideast, New England,
Plains, Southeast, Southwest, Rocky Mountain, and Far
West.
5. State estimates of value-added, gross product, investment,
or capital stock by industry typically have been restricted to
the manufacturing sector.
6. The studies discussed are not intended to form an exclusive
list. They were selected on the basis of their contribution to
the questions posed by this article.
7. The author is currently reestimating the model with revised
GSP data released in late 1991. The newly released data begin in 1977 and extend through 1989. Attempts are under
way to link the new series with the original data, which extend back through 1963.
8. A random-effects model could introduce specification bias
by ignoring the presence of fixed effects, so a covariance
model is estimated that allows for differential intercepts
over states (Hsiao 1986; Judge et al. 1985; Jakubson 1988).
The FXZX dummy variable holds the value 1 for j and 0 for
all other states to account for the effects of omitted variables that are specific to individual cross-sectional units but
stay constant over time (Hsiao 1986). A joint F-test on the
restricted versus unrestricted model was performed for each
industry and confirmed that dummy variables provided important additional information. Because all variables arc expressed as ratios of state-to-national values, time-specific
effects, which would include national business fluctuations,
interest rates, and inflation, have been discounted by variable
Eco no m ic Review
23
construction and need not be considered (Quan and Beck
1987). A joint F-test confirmed this assumption.
9. Because of data limitations, most empirical studies on this
topic have been cross-sectional or time-series in nature and
have not captured the dynamics of cross-sectional behavior
over time. Although this model exploits new data and offers information on dynamics over time, it presents several
intriguing econometric problems. In pooled estimation,
disturbance terms could exhibit time-series correlation,
heteroskedasticity, or cross-sectional correlation. Diagnostic tests from ordinary least squares estimation suggested
first-order serial correlation and heteroskedasticity across
all industries. Although serial correlation and heteroskedasticity will not bias estimators, it will bias standard errors
and could lead to incorrect inferences from the application
of significance tests. Therefore, a generalized least squares
estimation was specified to correct for these error term
problems, as in Kmenta (1971) and Johnston (1984). Usual
panel data problems, such as sample selectivity, population
inference, and limited degrees of freedom, are not issues
with these data series.
10. Although this paper discusses results from the thirty-state
sample, the model was also estimated across all forty-nine
contiguous areas with consistent but slightly less robust results.
11. Industries were excluded on the basis of size and lack of geographic distribution (for example, leather manufacturing
and motion pictures), assumed level of imputation on returns to capital (such as private household services and miscellaneous professional services), missing data, and where
returns to capital can vary dramatically on the basis of
changes in interest rates or financial market conditions (for
example, holding and investment companies, banks and
other credit agencies, and insurance carriers).
12. The BEA does not disclose state data when a firm's dominance in that state's industry makes its identification possible.
13. These categorizations are based on a priori expectations,
not empirical evidence. Generally, manufacturing industries
are considered export-oriented, and nonmanufacturing, locally oriented.
14. See Krikelas (1992) for a critique of economic base models.
References
Barro, Robert J., and Xavier Sala-i-Martin. "Convergence."
Journal of Political Economy 100 (April 1992): 223-51.
Bellante, Don. "The North-South Differential and the Migration
of Heterogeneous Labor." American Economic Review 69
(March 1979): 166-75.
Bishop, John A., John P. Formby, and Paul D. Thistle. "Convergence of the South and Non-South Income Distributions,
1969-79." American Economic Review 82 (March 1992):
262-72.
Bolton, Roger. "Regional Econometric Models." Journal of Regional Science 25 (November 1985): 495-520.
Borts, George H„ and Jerome L. Stein. Economic Growth in a
Free Market. New York and London: Columbia University
Press, 1964.
Browne, Lynne E., Peter Mieszkowski, and Richard F. Syron.
"Regional Investment Patterns." Federal Reserve Bank of
Boston, New England Economic Review (July/August
1980): 5-23.
Bureau of Economic Analysis. "Experimental Estimates of
Gross State Product by Industry." Staff Paper 42, 1985.
. "Gross State Product by Industry, 1963-86." Survey of
Current Business 68, no. 5 (1988): 30-46.
• "Gross State Product by Industry, 1977-89." Survey of
Current Business 71, no. 12 (1991): 43-59.
Carlino, Gerald A. "Are Regional Per Capita Earnings Diverging?" Federal Reserve Bank of Philadelphia Economic Review (March/April 1992): 3-12.
Carlton, W.C. "Why Do New Firms Locate Where They Do?
An Econometric Model." In Interregional Movements and
Regional Growth, edited by William C. Wheaton, 13-50.
Washington, D.C.: The Urban Institute, 1979.
24
Economic
Review
Chalmers, James A., and Terrance L. Beckhelm. "Shift and
Share and the Theory of Industrial Location." Regional
Studies 10, no. 1 (1976): 15-23.
Clark, Gordon L., Meric S. Gertler, and John E. Whiteman. Regional Dynamics—Studies in Adjustment Theory. Boston:
Allen and Unwin, 1986.
Dickie, Mark, and Shelby Gerking. "Interregional Wage Differentials: An Equilibrium Perspective." Journal of Regional
Science 27 (November 1987): 571-85.
Engle, Robert F. "A Disequilibrium Model of Regional Investment." Journal of Regional Science 14 (December 1974):
367-76.
Fournier, Gary M., and David W. Rasmussen. "Real Economic
Development in the South: The Implications of Regional
Cost of Living Differences." Review of Regional Studies 16
(Winter 1986): 6-13.
Garnick, Daniel H., and Howard L. Friendenberg. "Accounting
for Regional Differences in Per Capita Personal Income
Growth, 1929-79." Survey of Current Business 62 (September 1982): 24-34.
Hsiao, Cheng. "Analysis of Panel Data." Econometric Society
Monograph, no. 11. Cambridge University Press, 1986.
Jakubson, George. "The Sensitivity of Labor-Supply Parameter
Estimates to Unobserved Individual Effects: Fixed- and
Random-Effects Estimates in a Nonlinear Model Using Panel
Data "Journal of Labor Economics 6 (July 1988): 302-29.
Jerrell, Max E., and James M. Morgan. "Modeling Labor Demand in a State Econometric Model." Review of Regional
Studies 18 (Fall 1988): 31-40.
Johnston, J. Econometric Methods. New York: McGraw-Hill,
1984.
September/October 1992
Judge, George G„ William E. Griffith, R. Carter Hill, Helmut
Lutkepohl, and Tsoung-Chao Lee. The Theory and Practice
of Econometrics. New York: John Wiley and Sons, 1985.
Krnenta, Jan. Elements of Econometrics. New York: Macmillan
Publishing Company, Inc., 1971.
Kottman, Stacy E. "Relative Regional Employment Growth:
The Role of Differentials in Returns to Capital." Ph.D. dissertation, Georgia State University, 1990.
Krikelas, Andrew C. "Why Regions Grow: A Review of Research on the Economic Base Model." Federal Reserve Bank
of Atlanta Economic Review 77 (July/August 1992): 16-29.
Miller, Edward M. "Variation of Productivity, Wages, and
Profitability with Location." Review of Regional Studies 11
(Spring 1981): 27-41.
Milne, William J., Norman J. Glickman, and F. Gerard Adams.
"A Framework for Analyzing Regional Growth and Decline: A Multiregion Econometric Model of the United
States." Journal of Regional Science 20 (May 1980): 17389.
Federal Reserve Bank of Atlanta
Nakosteen, Robert A., and Michael A. Zimmer. "Determinants
of Regional Migration by Manufacturing Firms." Economic
Inquiry 25 (April 1987): 351-62.
Newman, Robert J. "Industry Migration and Growth in the
South." Review of Economics and Statistics 65 (February
1983): 76-86.
Quan, Nguyen T., and John H. Beck. "Public Education Expenditures and State Economic Growth: Northeast and Sunbelt
Regions." Southern Economic Journal 54 (October 1987):
361-76.
Ratajczak, Donald. "An Econometric Specification of a Quarterly State Model." Unpublished manuscript, University of
California-Los Angeles, 1972.
Sahling, Leonard G., and Sharon P. Smith. "Regional Wage
Differentials: Has the South Risen Again?" Review of Economics and Statistics 65 (February 1983): 131-35.
Wheat, Leonard F. "The Determinants of 1963-77 Regional
Manufacturing Growth: Why the South and West Grow."
Journal of Regional Science 26 (November 1986): 635-59.
Eco n o m ic Review
25
IFWE
Tracking Manufacturing:
The Survey of Southeastern
Manufacturing Conditions
R. Mark Rogers
The author is forecast coordinator on the macropolicy team
of the Atlanta Fed's research
department. He would like to
thank Diana Cunningham,
Cynthia Bansak, Teresa Beckham, James Mancuso, and
Marguerite Lewis for their
valuable contributions in the
development of the survey.
26
/iconomic Review
he Federal Reserve Bank of Atlanta, like the eleven other Reserve
Banks across the nation, monitors economic conditions in its region. Probably its most important reason for doing so is to contribute to the Federal Reserve System's task of setting appropriate
monetary policy. In the Southeast one of the most important influences on the economy's performance is manufacturing activity, which is
more variable than most other sectors and accounts for a larger employment
share in the region than it does in the nation. Consequently, to augment its
analysis of current conditions in the region's economy, the Atlanta Fed's research department in the fall of 1991 launched the first comprehensive survey to focus solely on changes in indicators of manufacturing activity in the
Southeast.
Because the Atlanta Fed's Survey of Southeastern Manufacturing Conditions has a rapid turnaround—less than three weeks for gathering, compiling, and reporting the data—it provides very recent information on the
southeastern economy not available from other sources. During the past
year the survey has proved increasingly valuable as it has been refined, the
number of participants has steadily increased, and the patterns of responses
have been better understood. In November the Atlanta Fed plans to begin
releasing the survey data as a regular economic report.
This article traces the development and construction of the survey and
explains the methodology used in compiling and calculating indexes from
the data. The article also describes some uses of the survey data.
September/October 1992
Developing the Survey
Before the southeastern manufacturing survey was
developed, the Atlanta Fed monitored regional economic conditions chiefly by collecting anecdotal information and analyzing statistical data collected from
various public sources. Approximately every six weeks
the research department's regional team conducted an
informal survey of business contacts to prepare a summary of current information about economic conditions
in the Sixth Federal Reserve District. (This summary is
still compiled with those from the other Reserve Banks
in a publicly available document, Current Economic
Conditions by Federal Reserve District, known as the
Beigebook. The Atlanta Fed's summary is also used in
briefing documents prepared for the Atlanta Bank president's use at Federal Open Market Committee [FOMC]
meetings.) Such contacts continue to provide valuable,
up-to-the-minute information about regional conditions
in all major sectors of the economy—not just manufacturing. Adding information from a formal manufacturing
survey was seen as a way to corroborate and augment
anecdotal information and provide a basis for comparing
over time the producers' reports about various flows
within the manufacturing process, such as orders, production, shipments, and inventories.
The survey was limited to the manufacturing sector
for several reasons. Manufacturing activity is more
readily quantifiable than that of many services industries. Moreover, because the manufacturing sector is
one of the more cyclical components of the economy
and is a key factor in pulling the overall economy into
recovery or recession, timely information about conditions in the sector is critical for regional economic analysis. Finally, manufacturing is also important as an
export base and as a high-wage sector, especially in
the Southeast.
The methodology of the Survey of Southeastern Manufacturing Conditions (discussed in detail in the following sections) has a long tradition. Both the Philadelphia
and Richmond Reserve Banks have produced manufacturing surveys for a number of years (see John Bell and
Theodore Crone 1986; Christine Chmura 1987/88), and
each is similar to the longer-running survey conducted
by the National Association of Purchasing Management
(1990). While the Atlanta Fed's survey generally adopted the methodology developed by the Philadelphia and
Richmond Feds, it differs in several aspects. After refinements based on responses of a small test pool of plant
representatives surveyed beginning in August 1991, the
fonnat was finalized in December 1991.
1
Federal Reserve Bank of Atlanta
Sampling Criteria. All surveys are developed using some method of drawing respondents from the
complete population. The Atlanta Fed's manufacturing
survey used the manufacturing facility's location as
the first criterion, choosing panelists at production facilities from across the six states that, in whole or in
part, make up the Sixth Federal Reserve District. Confining the survey to production facilities located in the
Southeast eliminated companies headquartered in the
region whose plants are elsewhere and. conversely, included plants operating in the region but owned by
businesses headquartered in other parts of the country.
This restriction makes the survey an indicator of activity in the Southeast exclusively.
Respondents also were chosen to reflect major industries in the broad range of manufacturing facilities
within the region. They represent eighteen of the
twenty industries classified as manufacturing industries according to the two-digit code of the Standard
Industrial Classification (SIC) system (see Table I)/
Dividing the Sample. In addition to classifying industries according to the two-digit SIC code, the Atlanta Fed survey's distribution of production units
attempts to mimic the industry value-added distribution
recorded in the 1987 Census of Manufactures. ("Valueadded" refers to the value added during the manufacturing
process to the raw materials and intermediate products
used as inputs.)" Currently, because its sample is small
in relation to the number of classification categories, the
survey distribution does not match exactly the Census
of Manufactures' stratification. The Atlanta Fed survey
weights each plant equally, and the distribution of respondents by industry varies somewhat from month to
month. Table 1 shows the percentage distribution of
respondents by industry for May-July 1992. As of
September 1992, the survey was sent to approximately
250 manufacturing plants; the average response rate is
slightly below 50 percent for the initial report.
2
4
1
Survey Information
The Atlanta Fed survey collects data on various
types of manufacturing activity as indicators of trends,
current activity, expectations, and plans. Manufacturers
are first asked one question about the general performance of their firms' industry; the remaining questions
relate to current and expected activity at their specific
plants. Respondents are instructed to take seasonal
variation into account and are assured of the confidentiality of their responses. They are questioned about
Eco n o m ic Review
33
Table 1
State's Value-Added Manufactures as a Percentage of State's Total Value Added, 1987
SIC Code
Number
Description
20
21
22
23
24
Food and kindred products
Tobacco products
Textile mill products
Apparel and other textile products
Lumber and wood products
25
26
27
28
29
Furniture and fixtures
Paper and allied products
Printing and publishing
Chemicals and allied products
Petroleum and coal products
30
31
32
33
34
Rubber and misc. plastic products
Leather and leather products
Stone, clay, and glass products
Primary metal industries
Fabricated metal products
35
36
37
38
39
Electric and electronic equipment
Transportation equipment
Instruments and related products
Miscellaneous manufacturing
AL
FL
GA
LA
15.3
D
D
3.1
3.0
10.6
D
D
5.6
3.3
8.4
D
1.2
3.0
2.0
14.0
3.6
9.1
D
1.8
4.1
10.3
8.0
0.4
1.6
8.5
4.9
7.3
D
6.4
2.8
8.6
5.4
2.5
D
4.3
1.2
5.6
7.0
5.6
6.0
1.3
0.8
100.0
6.1
Db
7.8
7.1
5.1
—
Machinery, except electrical
Total
a
Average
for May-July
b
D indicates
by Census because
of disclosure
regulation.
S o u r c e : U . S . Department of C o m m e r c e , Census of Manufactures, 1987.
MS
TN
Region
U.S.
10.7
D
7.2
8.6
12.2
1.2
3.2
5.6
2.2
11.0
0.2
1.7
4.9
3.7
10.4
1.2
2.2
2.8
2.5
4.9
4.7
2.7
0.2
9.2
2.9
41.2
10.3
6.8
7.7
D
6.6
D
3.0
5.8
5.7
13.7
0.3
2.2
7.8
5.4
13.1
1.4
1.7
4.3
7.7
10.4
1.6
3.3
12.1
3.8
9.6
1.1
2.4
D
3.3
2.9
3.2
0.9
5.8
1.0
3.5
3.1
6.1
3.7
0.2
3.2
3.1
4.7
3.8
0.4
4.9
1.6
0.9
2.7
4.5
D
2.8
2.7
5.2
8.6
12.6
7.0
10.2
1.2
4.1
7.1
14.1
1.5
1.0
2.2
4.9
9.2
0.2
0.3
6.4
8.9
9.3
0.6
D
8.0
6.3
8.3
2.0
2.0
6.2
7.7
9.4
3.1
1.1
100.0
100.0
100.0
100.0
100.0
100.0
—
—
—
1992.
that data are withheld
Survey
Distribution
by Units3
Omission
indicates
that industry
is not represented
in that state.
5.9
—
—
2.9
4.0
6.4
2.5
2.7
8.8
10.1
8.2
10.4
11.2
9.9
1.6
11.8
6.1
1.5
1 00.0
—
100.0
production, shipments, new orders, order backlogs,
materials inventories, finished goods inventories, number of employees, the average workweek, finished
goods prices, raw materials prices, capital expenditures, new export orders, and supplier delivery time.
The survey questions refer to three time frames: current
month versus previous month, current month versus
twelve months ago, and expected activity six months
ahead compared with current levels. (See Table 2 for
items and time periods.)
To minimize the reporting burden on panelists and
to keep processing time to a minimum, the survey asks
for the direction of changes for indicators rather than
for specific levels of activity. For instance, instead of
asking for the dollar volume of production for the current month, the survey asks whether production has
decreased, increased, or remained unchanged.
Compiling the Data
The survey questionnaire requesting data pertaining
to a particular month, along with the previous month's
summary data, is mailed to manufacturers on approximately the twenty-fifth of the survey month. This procedure allows responses to be based on about three
weeks of information concerning operations and on
general knowledge about plans and expectations for the
remainder of the month. Replies are requested to be returned by about the fifth of the following month. Each
month, once the data are accumulated, the percentages
of respondents showing improvement, deterioration, or
no change for each activity indicator and time period
are calculated. Final tallies for the initial survey release
are completed by the tenth. Any late responses are incorporated into revised tallies. Summary statistics are
revised only once and only to incorporate late survey
responses. Release data are based solely on respondent
information; there are no imputations or estimates for
firms not replying.
For example, data on monthly directions of change in
new and unfilled orders, along with information on inventories, may indicate whether changes in production
are likely. Data on directions of change in finished goods
prices, raw materials prices, and supplier delivery time
may serve as a rough inflation barometer. Information
on directions of change in number of employees and the
average workweek can provide insight into income
trends as well as what plant managers believe about the
degree to which production trends are sustainable.
Each of the survey's three time perspectives provides insight. Monthly comparisons, for instance,
show progress over the current phase of the business
cycle. Year-ago comparisons delineate longer-term
trends and help indicate possible remaining seasonality in responses. Responses concerning expectations
six months ahead (compared with current levels) can
help ascertain whether recent changes in activity are
expected to be temporary or not.
The regional economists and analysts of the Atlanta
Fed's research department use the survey data to complement other information on the manufacturing sector. As the survey data suggest production trends, for
example, other indicators, including regional employment and income as well as anecdotal evidence from
contacts, can be monitored for confirmation.
Diffusion Indexes
To simplify determining how widespread changes
are among respondents and comparing responses overtime, the Atlanta Fed also calculates a statistical measure known as a diffusion index for each response category. A diffusion index is a tool used to gauge the
similarity of individual reporter's changes in a particular category and time period.
The manufacturing survey's diffusion indexes—
one for each time period of each indicator surveyed,
such as production or new orders—are calculated as
the percentage of total respondents reporting increases
minus the percentage reporting declines. (The percentage of "no-change" responses does not directly
enter into the calculation although the number of "nochange" answers affects the other two percentages.) If
all plants report increased activity for a particular
category then the diffusion index for that category equals
100, and if all plants report declines then the category's
value is —100. Thus, each diffusion index can range in
value from 100 to - 1 0 0 . The logic of this method for
calculating the diffusion index is that positive values
6
t/sing the Data
The aggregate pattern of survey responses showing
decrease, no change, and increase (see Table 2) provides
information about changes in basic demand and supply
flows as well as other data on items such as inflation and
investments. This information should provide insight into manufacturers' current and expected performance.
Federal Reserve Bank of Atlanta
Eco n o m ic Review
29
Table 2
Summary of Southeastern Manufacturing Conditions, July 1992
(Percentage of total respondents)a
Current versus Previous Month
Decrease
No
Change
25.0
51.5
28.5
32.1
27.9
25.4
Current Month versus
Six Months Ahead
Current Month versus Year-Ago
Decrease
No
Change
Increase
Decrease
No
Change
Increase
23.5
25.0
23.5
51.5
15.7
27.6
56.7
45.3
39.4
43.4
53.0
26.3
28.5
28.7
21.6
21.5
20.1
25.6
23.1
22.2
26.1
27.8
41.8
56.3
53.7
46.6
35.1
13.2
13.4
14.3
18.5
39.0
35.8
31.6
45.9
47.8
50.7
54.1
35.6
34.3
35.9
18.8
16.1
16.1
8.3
49.3
45.0
60.9
67.9
74.5
73.5
16.4
19.1
20.3
16.1
9.5
18.2
9.5
6.9
75.9
87.0
14.7
6.1
36.4
39.5
34.1
17.9
30.1
13.4
28.5
10.8
10.9
34.1
28.7
29.6
50.0
41.4
35.8
43.1
64.2
71.9
29.5
31.8
36.3
32.1
28.6
50.7
28.5
25.0
17.2
32.3
32.3
18.2
13.2
12.9
4.4
14.5
2.5
3.9
51.1
46.9
53.3
69.1
59.8
54.8
51.1
70.5
79.8
16.5
20.8
28.5
17.6
27.3
40.7
34.4
27.0
16.3
Increase
Industry Business Conditions
What is your evaluation of business
activity in your industry?
Business Indicators
Production
Volume of shipments
Volume of new orders
Backlog of orders
Inventories
Materials
Plants' finished goods
Number of employees
Average employee workweek
Prices received for finished products
Prices paid for raw materials
Capital expenditures6
New orders for exports
Supplier delivery timec
a
Normal
b
Because
c
Decrease
seasonal
fluctuations
firms plan capital
= slower;
are taken into account.
expenditures
increase
= faster.
Figures may not sum exactly
on a long-term basis, this question
to 100 becausc
is not applicable
of
rounding.
to month-ago
comparisons.
indicate increased activity at a majority of the plants
over a certain time period while negative values indicate decreased activity at a majority of plants.
Interpreting the Indexes. Although the Atlanta
Fed survey does not collect data on specific levels of
activity and cannot be used to deduce such levels, the
7
diffusion indexes provide an indicator of the direction of changes in various types of manufacturing
activity. Moreover, statistical studies indicate that
over an extended period the level of diffusion indexes correlates well with growth rates in the economy
(Ethan Harris 1991). See Table 3 for diffusion indexes
Table 3
Diffusion Index Summary of
Southeastern Manufacturing Conditions, July 1992
1992
June
July
May
April
1991
March
Feb.
Jan.
Dec.
Current versus Previous Month
Industry Business Conditions
-1.5
Plant Indicators
Production
-2.2
Volume of shipments
-3.6
Volume of new orders
0.7
Backlog of orders
-3.7
Inventories
Materials
-17.9
Plants' finished goods
-16.8
Number of employees
1.4
Average employee workweek
0.0
Prices received for finished products - 6 . 6
Prices paid for raw materials
9.8
New orders for exports
5.2
Supplier delivery time1'
0.8
37.4
18.2
35.4
34.8
14.6
3.8
-22.6
33.6
33.0
24.6
10.4
16.8
13.4
23.9
1.8
29.0
27.3
26.5
11.3
40.4
39.4
42.6
19.1
24.1
19.5
13.3
4.9
21.8
11.5
10.3
-7.9
-9.3
-13.0
-20.4
-32.1
-7.2
1.8
15.4
18.8
-0.9
19.1
14.3
8.0
-3.6
0.9
1.8
2.7
3.6
16.4
9.8
0.0
-4.2
-3.1
13.3
16.0
0.0
15.3
21.2
2.1
-3.3
12.5
5.4
8.6
-1.1
15.1
17.1
-4.3
3.8
14.3
3.7
10.0
-2.6
7.6
8.3
10.3
-4.0
6.8
-11.5
1.3
-18.7
10.5
11.7
22.2
-11.5
-13.7
-11.3
-5.8
-5.9
9.6
4.3
3.8
Current Month versus Six Months Ahead
Industry Business Conditions
41.0
Plant Indicators
Production
34.6
Volume of shipments
37.3
Volume of new orders
39.8
Backlog of orders
17.0
Inventories
Materials
-15.8
Plants' finished goods
-11.5
Number of employees
10.2
Average employee workweek
4.4
Prices received for finished products
14.4
Prices paid for raw materials
36.3
Capital expenditures
19.8
New orders for exports
24.6
Supplier delivery time3
-12.4
a
Percent
slower
minus percent
49.1
54.2
59.6
60.0
45.5
47.2
30.9
31.5
29.7
18.9
40.2
42.3
42.3
23.9
53.0
57.1
54.1
25.0
52.2
57.1
58.2
30.0
53.1
56.3
58.5
33.8
52.6
56.0
58.7
32.9
60.4
53.7
60.4
31.5
-11.3
-18.1
8.0
-8.0
24.5
41.1
26.2
20.8
-5.6
-10.9
-24.5
10.7
6.3
26.6
33.9
23.4
18.9
-10.3
-5.3
-19.8
24.0
11.1
28.9
27.6
24.7
27.3
-4.2
-10.5
-11.6
18.7
9.9
27.5
29.7
12.4
33.7
-16.3
-1.3
-15.6
22.0
6.3
21.5
36.3
23.0
28.0
7.4
-16.4
-8.3
25.0
11.8
21.9
31.5
15.7
31.3
11.0
-3.8
-6.0
20.8
9.6
49.0
41.2
27.5
30.0
7.4
faster.
Federal Reserve Bank of Atlanta
41.5
Econ o m ic Review
31
from the Atlanta survey for month-ago and six-monthsahead comparisons during the December 1991-July
1992 period.
By showing to what degree the number of plants
with gains offsets those indicating worsening conditions, diffusion indexes indicate the direction of general trends for particular facets of manufacturing. The
higher the index number in absolute terms (the closer
the index is to 100 or to - 1 0 0 ) , the greater is the similarity of change among the responding firms. As indicated above, the closer index values are to 100, the
more prevalent gains are among respondents; the
closer values are to - 1 0 0 , the more prevalent declines are. An index level of zero indicates that the
number of plants expanding and those contracting is
evenly balanced.
Conclusion
The Survey of Southeastern Manufacturing Conditions has proved useful to the Atlanta Fed in monitoring
regional economic developments. Over time the accumulated survey data can be evaluated statistically along
with other indicators to provide greater depth of information about both the regional and national economies.
In the longer run, because its data base includes background information on plants for a large number of
factors not available in other series, the survey may provide a basis for research on manufacturing's cyclical
behavior. The data aggregated by various background
factors could provide valuable insight into the longrun behavior of manufacturing plants in the Southeast.
8
Notes
1. The Beigebook is sent to Congress and the Federal Reserve
Board of Governors and is made available to the public
through the news media. It is released about a week before
meetings of the FOMC, which meets eight times a year.
2. The Sixth Federal Reserve District encompasses Alabama,
Florida, Georgia, the southern halves of Louisiana and Mississippi, and the eastern two-thirds of Tennessee.
3. Manufacturers of tobacco products and leather are not represented because of their very small share in the region's manufacturing output.
Industrial classification refers to the grouping of reporting
establishments on the basis of their major product or activity
as determined by the establishments' percentage of total sales
or receipts. The Atlanta Fed's manufacturing survey data are
currently classified in accordance with the Standard Industrial Classification Manual (Office of Management and Budget
1987). The SIC codes indicate the level of aggregation of data. A two-digit code signifies a broader level of aggregation
than, say, a five- or.six-digit classification.
4. The Census of Manufactures takes place only every five
years; 1987 data are the most recent. When data for the 1992
Census of Manufactures become available, they will be used
as a basis for stratifying the Atlanta Fed survey.
5. This method of measuring manufacturing production by industry is similar to that used by the Federal Reserve Board's
national index of industrial production (see Board of Governors 1986). This procedure contrasts with counting physical
units coming off an assembly line. Output for a given industry includes the value of the inputs. Counting output for all
industries would involve double counting and overstate the
contribution of manufacturing to overall output.
6. The Atlanta Fed manufacturing survey's diffusion index is
analogous to the fairly well-known diffusion index of em-
32
Econom ic Review
ployment change published by the U.S. Bureau of Labor
Statistics, which measures the percentage of industries that
posted increases in employment over specified time spans
such as monthly and six-month periods. (For example, see
U.S. Department of Labor 1992.)
7. The Bureau of Labor Statistics (BLS) and the National Association of Purchasing Management (NAPM), for example,
calculate their indexes by determining the percentage of positive responses and adding to this one-half of the no-change
responses. This method produces diffusion index ranges from
0 to 100, with 50 being the break-even point. Values above
50 indicate that positive responses outnumber negative responses; values below 50 indicate the opposite.
In both the Atlanta Fed and BLS/NAPM methodologies,
respondents have only three choices for answers (decrease,
no change, increase), so a diffusion index based on a combination of any two answers provides the same information as
any other combination of the answers. (Any combination of
two out of three responses will yield the same information as
long as the combinations are linear in construction. Essentially, three variables are set equal to a constant [100 percent!
and two of the variables arc known.) Using either methodology, one could construct an index based on negative and
no-change responses that would have the same relative
movement.
8. This background information includes industrial classification size of firm by number of employees, geographic location (state), union or nonunion status, domestic or foreign
ownership, and vintage of plant according to the date of plant
construction or last major capital improvement.
September/Octobér 1992
References
Bell, John, and Theodore Crone. "Charting the Course of the
Economy: What Can Local Manufacturers Tell Us?" Federal Reserve Bank of Philadelphia Business Review (July/August 1986): 3-16.
Board of Governors of the Federal Reserve System. Industrial
Production, with a Description of the Methodology. Washington, D.C., 1986.
Chmura, Christine. "New Survey Monitors District Manufacturing Activity." Federal Reserve Bank of Richmond Cross
Sections (Winter 1987/88): 9-11.
Federal Reserve System. Current Economic Conditions by Federal Reserve District. Published eight times annually.
Federal Reserve Bank of Atlanta
Harris, Ethan. "Tracking the Economy with the Purchasing
Managers' Index." Federal Reserve Bank of New York
Quarterly Review 16 (Autumn 1991): 62.
National Association of Purchasing Management. The Report
on Business: Information Kit. Tempe, Ariz.: NAPM Information Center, 1990.
Office of Management and Budget. Standard Industrial Classification Manual. Springfield, Va.: National Technical Information Service, 1987.
U.S. Department of Labor. Bureau of Labor Statistics. "Diffusion
Indexes of Employment Change, Seasonally Adjusted" (Table
B-6). Employment Situation. Washington. D.C., August 1992.
Eco n o m ic Reiñeu>
33
eview Essay
The Future of Banking
by James L. Pierce.
New Haven, Conn.: Yale University Press, 1991.
163 pages. $25.00.
Aruna Srinivasan
/ I
/ T
banks forever be regulated? Has regulation added
/ I
/ M
banks' problems? Should banks be regulated in the same
/ I / K
manner as in the past, perhaps with some minor modifica- Surely the financial system is much different now
/ ^J m
-A- ^
M .
than when the basic public policies toward banks were created in the 1930s. The debate about refonn and redesign of these policies often fails to recognize such changes. Instead, discussions assume that the
economy's credit and liquidity services will continue to be performed by
banks, as they have been historically. An exception to this trend is The Future of Banking, which proposes fundamental changes in banking powers,
deposit insurance, and bank regulation. The Twentieth Century Fund, a research foundation that analyzes economic, political, and social issues, asked
James L. Pierce to write this book to help clarify the issues in the bank reform debate and thus aid policymakers in making better-informed judgments about how to cure the industry's problems. Pierce, who teaches
economics at the University of California at Berkeley, was also asked to offer his own solutions to banking's difficulties.
Pierce's basic analysis hinges on a few key points: Because the banking
industry is critical to the stability of the nation's financial system, regulation
is essential to protect it. However, what has worked in the past is no longer
effective. New regulations should be consistent with economic forces and
should not isolate and protect banks from competitive realities. Although
banks at one time provided essential financial services that were not available elsewhere, now most of these services are not unique. Conversely,
banks have been barred from certain lines of business, such as securities,'
that today seem a logical and beneficial fit with other bank services.
u s t
t o
t i o n s
The reviewer is an economist
in the financial section
of the Atlanta Fed's
research department.
She thanks Larry Wall
for helpful discussions.
34
Uconomic Review
September/October 1992
Pierce advocates regulations based on the type of
financial services provided regardless of the type of
firm that offers these services. Under this scheme any
firm offering a checking account—whether a bank, securities firm, or some other entity—would be regulated in the same way for that particular service. Pierce
would extend the safety net afforded by federal deposit insurance and Federal Reserve discount window
lending only to firms offering monetary services, leaving the nonmonetary services of financial service companies, with uninsured deposits, at the mercy of the
free market in gathering and lending funds.
The changes Pierce outlines are far more sweeping
than the provisions contained in the Federal Deposit
Insurance Corporation Improvement Act of 1991
(FDICIA), but his plan is one of a number of alternative ideas that seek to limit bank regulation and deposit insurance to firms that offer demand deposits and
invest these deposits in "high-quality" assets, such as
Treasury securities.
In contrast to Pierce's proposal, FDICIA attempts
deposit insurance reform within the framework of the
existing regulatory structure. FDICIA's approach assumes that problems in the banking system stem from
regulatory mismanagement and lack of market discipline in the deposit insurance system. The law relies
on prompt, capital-based corrective action, risk-based
deposit insurance, and least-cost resolution to minimize losses to the deposit insurance fund. FDICIA also curtails too big to fail by limiting the FDIC's ability
to cover uninsured depositors and restricting the use of
the Federal Reserve discount window to prop up weak
banks.
This essay suggests that while the new banking
structure proposed by Pierce falls short of the ideal,
FDICIA may pose greater risks to the system because
it significantly increases the regulatory burden on
banks without affording them the protection from competition they have enjoyed in the past. While FDICIA
appropriately seeks to limit moral hazard and introduce market discipline, it also risks creating future
crises by not recognizing the contraction in banks' traditional role.
1
2
What
Services Do Banks Perform?
To understand current banking issues,
starting point is to consider the "special"
services banks provide. Pierce believes that
ing industry developed in response to two
Federal Reserve Bank of Atlanta
a logical
nature of
the bankeconomic
needs. The first he defines as a monetary one, whereby
bankers protect "money," account for its ownership,
and facilitate its use in settling economic transactions.
In this original role, banks did not provide a return to
the holders of funds because they were really providing liquidity services for which they were entitled to
receive compensation.
In their second role—making loans or enhancing
credit—banks evolved into the most efficient institutions for taking funds from surplus units (depositors)
and channeling these funds to deficit units (borrowers). Recent theoretical work on the economic functions of banks suggests that banks provide two major
credit services. First, banks provide information services that aid individual investors. For example, by
purchasing bank liabilities, investors can avoid both
the costly duplication of effort each would make in researching and analyzing credit risks and opportunities
while avoiding the possibility that other investors could
reap the benefits of their analysis without incurring the
costs (free riding). The second credit service that banks
provide is monitoring firms' managements at a lower
cost than individual debtholders can.
Pierce appears to suggest that banks' credit and liquidity services were combined by historical accident,
and he does not explain why bank debt should be of
rather short maturity (for example, demand deposits).
This omission is significant because the rationale for
traditional bank regulation rests on the idea that banks
historically have been firms that combine nonmarketable assets with demandable debt liabilities.
Recent papers by Charles W. Calomiris and Charles
M. Kahn (1991) and Mark J. Flannery (1992) offer
several reasons that banks have historically engaged in
maturity mismatching. Their results suggest that this
mismatching provided important economic benefits.
Calomiris and Kahn show that demandable debt is an
important part of an incentive scheme for disciplining
bankers. Demandable deposits permit depositors to
vote with their feet; they withdraw funds when they
lose confidence in banks. Without the ability to make
early withdrawals, depositors would have little incentive to monitor banks. Flannery demonstrates that maturity mismatching may be optimal in an unregulated
environment. He evaluates optimal means of financing
a portfolio of bank-type loans and shows that uninsured banking firms face asset substitution and investment problems, which are best addressed by shortening
debt maturity.
Financing nonmarketable assets with demandable
liabilities is problematic, however, in that it exposes
banks to depositor runs. To address this problem
3
Eco n om ic Review
35
governments have chosen to intervene by insuring deposits and regulating banks.
lesterday's Solutions
In response to a particularly devastating failure of
the banking system during the Depression, the New
Deal banking reforms of the 1930s were designed to
resolve the problem of depositor runs. The most important of these reforms related to deposit insurance
and the safety net. The Federal Reserve was to act as
lender of last resort, providing liquidity to banks when
depositors wanted their money and thus limiting systemic risk. The FD1C was to insure deposits to protect
individuals' wealth and assure them of the banking
system's safety.
Other notable financial reforms were enacted during this period. The Glass-Steagall Act, which separated commercial banking from investment banking and
prohibited commercial ownership, restricted banks'
ability to compete with each other and with other financial institutions. Pierce argues that restrictions on
granting of new bank charters, combined with limitations on branching, granted banks a monopoly of sorts
in their limited territories. Congress also imposed ceilings on deposit rates in an effort to aid bank profitability. According to Pierce, all of these reforms were
designed to prevent banks from failing and needing
government assistance in the long run.
7Tie Roots of Banking's
Current Problems
The banking system put in place during the New
Deal worked unusually well, Pierce concedes, until the
1960s, when the interaction of an inflexible regulatory
system and changing banking structure led to problems in the financial services industry. At that time,
changes in the structure of money markets and major
improvements in technology allowed competition to
flood into banking. Competitors, including the commercial paper market and money market mutual funds,
offered less expensive, more flexible products that effectively challenged the profitability of the banking industry. The effects of these financial innovations on
banking were so profound, Pierce contends, that understanding them is essential to any meaningful discussion of public policy toward banks.
36
Economic Review
Pierce charges both regulators and bankers with
mismanaging the financial system during the 1960s
and 1970s. Regulators, he argues, refused to adjust
their actions in response to changes in the financial
system and thus hindered banks from developing
products that would compete with the new financial
instruments from nonbank sources until the competitors were well entrenched in the market. Banks also
made substantial errors in judgment, according to
Pierce. Improved technology and access to money
markets implied that large banks could grow rapidly
by managing their liabilities, but this growth was accompanied by increasing leverage and declining asset
quality, exposing banks to greater risks from loan losses and other sources. The substantial increase in leverage was allowed chiefly because bank deposits were
protected by government insurance.
j F D I C I A ' s Objectives
Problems that arose in the banking industry during
the 1960s and 1970s have become more pervasive.
Since the early 1970s depository institutions have
been plagued by persistent financial difficulties. The
problems of the thrift industry and loan losses at large
banks have received widespread publicity. Even as
large banks in New England and Texas have required
massive amounts of federal assistance, the bank insurance fund has sustained losses, raising the possibility
of taxpayer bailouts.
The hiatus in legislative activity relating to federal
deposit insurance finally ended, however, with the
passage of FDICIA. The act represents, among other
things, the first attempt to modify the deposit insurance system created during the 1930s. Under FDICIA
capital becomes the centerpiece of bank regulation.
The act requires banking regulators to divide all
banks into five categories according to their capital
ratios and specifies actions, increasing in severity as a
bank's capital ratio moves down the scale, that regulators must take for banks falling in each category.
These provisions are intended to ensure prompt regulatory action when a bank first experiences difficulty
and "early closure" when those problems (as measured by capital ratios) become severe. Regulators
are also required to revise risk-based capital standards to take into account additional measures of risk
such as interest rate risk, establish a system of riskbased deposit insurance premiums that would presumably rely heavily on capital ratios, and limit the
4
September/October 1992
ability of banks in lower capital categories to acquire
brokered deposits.
FDICIA requires the FD1C to establish a risk-based
deposit insurance scheme by January 1994 and to
study the feasibility of establishing a private reinsurance system. FDICIA also addresses the long-standing
problem created by discount window lending to troubled institutions. It prohibits the Federal Reserve from
making loans to an undercapitalized institution for
more than 60 out of 120 days.
In addition to capital-focused mandates, FDICIA
requires regulators to prescribe for all insured institutions operational and managerial standards covering
such items as executive compensation restrictions,
internal control standards, underwriting standards,
interest rate exposure, and asset growth. The act also
significantly increases banks' reporting requirements
and includes new consumer protection provisions relating to Truth in Savings and CRA disclosures.
While FDICIA makes key changes in deposit insurance, it does so within the framework of the existing regulatory system. FDICIA does not lower the
$100,000 limit on deposit insurance coverage, it does
not categorically prohibit regulators from protecting
uninsured depositors at large institutions, nor does it
privatize deposit insurance. FDICIA is a "narrow"
piece of legislation in that it does not allow interstate
branching, repeal Glass-Steagall provisions, or permit
ownership of banks by commercial firms.
5
Pierce's Solutions
While Pierce acknowledges that there is scope for
further reform within the existing regulatory structure,
he believes that reforms based on this approach are
likely to be effective for only a short time because
competitive forces will continue to erode the traditional role of banks. Federal deposit insurance reform in
particular, he argues, will not work without radical
structural changes. Pierce discusses several reforms
that could pave the way for more fundamental changes
he has in mind.
Pierce wants to restore market discipline to banking
by reducing or eliminating insurance on most categories of deposits. He recommends that banks be required to issue subordinated debt that would force
added monitoring of banks by debtholders. Under
Pierce's plan coinsurance would cover a given percentage of deposit balances in excess of the $ 100,000
statutory limit and would serve to remind large deposiFederal Reserve Bank of Atlanta
tors that they need to apply discipline in selecting
banks. The FDIC and uninsured depositors would
share in the cost of bank failures.
Like other reformers, Pierce advocates restricting
the maximum amount of insurance protection available to depositors. He suggests limiting the number of
banks at which a depositor may carry fully insured deposits. Deposits spread among several banks would be
coinsured rather than fully insured. Pierce contends
that these proposals represent a slow retreat from total
FDIC coverage and would reintroduce market discipline in stages.
Pierce's most distinctive reform would create monetary service companies as separately capitalized
companies within banks or financial services firms.
Monetary service companies would isolate, insure,
and protect monetary functions. Only monetary companies would offer federally insured deposits and provide payments services. These companies would hold
only money market instruments such as Treasury bills,
commercial paper, and other short-term, liquid, highly
rated instruments typical of money market mutual
fund assets. Such a company could not lend to its
owners under any circumstances and would be completely insulated from its parent company's liabilities.
All other activities currently thought of as banking
would become uninsured and unprotected by access to
the safety net.
A separate financial services company, also under
the umbrella of a larger banking or financial services
firm, would provide nonmonetary services such as
time and savings deposits and all lending functions.
Deposits in a financial services company would be
uninsured. Thus, if the company failed and investors
lost their money, it would be solely a private-sector
predicament. Depositors and other creditors of such
companies would be forced to look more closely at
where they placed their funds, Pierce argues, and this
scrutiny would impose a healthy discipline on the
banking industry.
Under Pierce's plan government regulation and insurance would be confined to only about 40 percent of
what is currently defined as banks' liabilities. Commercial banking as it has been traditionally viewed,
Pierce points out, has become a smaller component of
the overall financial system. He estimates that currently no more than 37 percent of U.S. banks' sources of
funds are made up of deposits payable on demand.
Most liabilities are unrelated to the payments system,
he contends. Only 18 percent of bank assets are devoted to commercial and industrial lending—the type of
lending that historically has made banks special. Small
Eco n o m ic Review
37
business lending, the one area in which banks have
special expertise, is probably less than 10 percent of
what the banking industry does. Pierce questions the
necessity of regulating the other 90 percent of bank
activities to ensure provision of this credit.
Although several of Pierce's bank reform proposals—including his narrow bank plan or the idea of
breaking up the existing financial system—are not
new or unique, they also are no longer discussed only
in academic circles. FDICIA's reforms offer a framework against which Pierce's or any alternative proposal can be judged.
Problems Posed by FDICIA
Capital-based regulation, a key element of FDICIA,
draws heavy criticism from Pierce. Such regulations
cannot be implemented effectively, he believes, because they invite circumvention. For example, banks
circumvented capital regulation during the 1980s by
holding riskier assets and by increasing their offbalance-sheet activities. Regulators responded by developing risk-based capital standards, which weight
various asset categories by their degree of risk. Riskbased capital standards are complex and irrational,
Pierce asserts, and they distort banks' portfolio choices because they treat risks as additive. He fears that
FDICIA's capital-based regulations will prompt a similar cycle of banks' finding new ways of circumventing the regulations and regulators' responding with
restrictions on the new techniques for circumvention.
Pierce rightly raises concerns about FDICIA's early
closure provisions. He points out that timely recognition of declines in asset values is crucial to implementing these provisions effectively. While FDICIA calls
for a review of accounting rules for banks and thrifts,
it does not mandate the use of market value accounting. Allen N. Berger, Kathleen Kuester King, and
James M. O'Brien (1991) discuss some of the conceptual, measurement, and verification issues associated
with implementing market value accounting that arise
because of banks' roles in solving information and
monitoring problems. Their study suggests applying
market value principles to cases in which they are
most feasible (for example, for traded securities and
securitized loans) and using a statistical procedure for
situations in which market value accounting is problematic (to correct for changes in credit quality, for instance). Even if market values could somehow be
determined, Pierce questions the benefits of using
38
Economic Review
market value accounting to close institutions that are
viable in the longer run.
Pierce believes that major practical problems hinder
implementation of risk-based deposit insurance. In his
view, pegging deposit insurance assessments to risk
"ascribes to the regulators an ability to identify and
measure risk that they do not possess." Market discipline imposed by large depositors and capital requirements would be more effective, he argues.
Many in the banking industry have raised concerns not only about FDICIA's deposit insurance
provisions but also about other elements of the legislation they consider highly onerous, especially those
that require regulators to scrutinize more carefully
the actual day-to-day business of financial institutions. These concerns stem partly from the fact that
FDICIA maintains or strengthens existing consumer
and social responsibility requirements on banks without liberalizing existing banking laws such as the
Glass-Steagall Act and restrictions on nationwide
branching. These additional requirements amount to
a new tax on banks, while banks remain constrained
in their ability to find more efficient ways to serve
their customers. In Pierce's view, tightening the intensity and scope of bank regulation as FDICIA has
done will only push more banking activities outside
of conventional banks, putting continued pressure on
banks to contract, especially in the traditional loanmaking business.
Problems with Pierce's Solutions
Proposals like the fundamental restructuring of the
financial system Pierce advocates require careful elaboration. Pierce anticipates a number, but certainly not
all, of the "what-ifs."
The biggest problem with Pierce's solution is its extremity. It would abolish traditional banks, even those
that remain viable, and would break up a number of
existing financial institutions.
Pierce examines the impact his narrow bank ideas
would have on lending behavior and credit availability
in the financial system generally. Mortgage and consumer lenders would tend to accelerate their trend toward securitization. Commercial loans that are not
easily securitized would have to be held on financial
services companies' balance sheets, funded by uninsured deposits. As a result, interest rates on such loans
would almost certainly increase significantly. Pierce
does not concede this point, even though he argues that
September/October 1992
current insurance on time deposits acts as a subsidy to
commercial borrowers.
Another concern raised by Pierce's system is that it
might facilitate the diversion of credit from smaller to
larger businesses. Larger financial organizations could
choose to engage only in deposit taking in various regions of the country without offering lending services
through their financial services holding companies. If
these diversified firms were to attract deposits away
from locally based nondiversified banks, smaller businesses that now rely on local institutions could find
less credit available.
While narrow bank proposals like Pierce's would
separate illiquidity risk from the payments system, financial system stability would remain an important
policy concern. Studies by Calomiris and Kahn (1991)
and Flannery (1992) suggest that maturity mismatching
in the existing system provides important market benefits and that the economy may be no more stable with
maturity-matched banks because new firms that move
in to fill the vacuum left by the demise of traditional
banks may inherit the problem of depositor runs. Runs
on financial services companies would have the same
impact on the system that runs on banks had more than
fifty years ago. Flannery argues that, as a result of political pressures, liquidation of financial services companies may take the form of government bailouts and
thereby limit the benefits of narrow bank reform.
Pierce believes that many small banks would survive his restructuring proposal even though they might
have trouble attracting uninsured deposits. Small banks
that have relied on deposit insurance to stay alive
would probably disappear; the more profitable small
banks might be able to survive by securitizing and servicing the loans they originate.
Like any system of reforms, Pierce's proposal has
several weaknesses. Ultimately, however, the most important issue is whether his restructuring scheme raises problems more severe than would occur under
alternative frameworks such as FDICIA.
Policy Alternatives
What are the alternatives to the bank reform approaches discussed in this essay? One approach would
be to weigh carefully all existing regulations and discard any that exceed the minimum regulation needed
to prevent systemic risk and to protect the insurance
fund. This strategy, which would place far greater reliance on market forces, probably has very little chance
of being adopted.
Another solution would be to make some version of
the narrow bank an option for any firm wishing to provide monetary services. This approach would permit
traditional banks to continue operating in situations in
which the costs of regulations are exceeded by the
benefits and would allow for services that cannot be
efficiently performed by banks (under current regulations) to be shifted to more efficient providers. Thus,
monetary service providers would continue to operate
under the safety net.
Overall, The Future of Banking is well worth reading. Pierce provides valuable insights into banking's
problems and focuses the reader's attention on some
fundamental issues relevant to the debate about the future of not only banking but of the entire financial services industry.
Notes
1. Picrce's monetary service companies are a variant of the
core banks proposed by Bryan (1991) and the narrow banks
proposed by Litan (1987).
2. See Carnell (1992) for a general discussion of FDICIA and
Wall (1992) for a discussion of too big to fail and related
provisions.
3. Gorton and Pennacchi (1990) summarize the major studies
in this area.
Federal Reserve Bank of Atlanta
4. See Carnell (1992) for a description of FDICIA's capital categories and rules governing institutions in those categories.
5. Congress required the FDIC to study the feasibility of authorizing insured depository institutions to offer both insured
and uninsured deposit accounts, perhaps leaving the door
ajar for more fundamental reform.
Eco n o m ic Review
39
References
Bcrger, Allen N., Kathleen Kuester King, and James M.
O'Brien. "The Limitations of Market Value Accounting and
a More Realistic Alternative." Journal of Banking and Finance 15 (September 1991): 753-84.
Bryan, Lowell L. Bankrupt: Restoring the Health and Profitability of Our Banking System. New York: Harper Collins,
1991.
Calomiris, Charles W., and Charles M. Kahn. "The Role of Demandable Debt in Structuring Optimal Banking Arrangements." American Economic Review 81 (June 1991): 497513.
Carnell, Richard Scott. "Implementing the FDIC Improvement
Act of 1991." Paper presented at the Conference on Re-
40
Economic Review
building Public Confidence through Financial Reform, The
Ohio State University, Columbus, Ohio, June 1992.
Flannery, Mark J. "Debt Maturity and the Deadweight Cost of
Leverage: Optimally Financed Banking Firms." University
of Florida, unpublished paper, June 1992.
Gorton, Gary, and George Pennacchi. "Financial Innovation
and the Provision of Liquidity Services." The Wharton
School, unpublished paper, May 1990.
Litan, Robert E. What Should Banks Do? Washington D.C.:
The Brookings Institution, 1987.
Wall, Larry D. '"Too Big To Fail' after FDICIA." Federal Reserve Bank of Atlanta unpublished paper, 1992.
September/October 1992
:;fefeäfe^fefe:¿f
ífefeífefesfe
ífefe
-•
«
. f e - . -
^
- -
>~E- ...
fe
.Y
-
-
-fe
fe
- f e "fe - ^ f e ^ f e
7
"
< "
< •!fe
. -.
^
fe:fe;:fefe^
fe
.
.
\
feí: fe^fefe-v-r
-
fefefe;
fe^fefe
-
-fe - ¥T-fefe
>
'
-
.,'jr-
— • , - f e f e -- - -
;
£
fe-fe
,-Ai.:- ^ i
fe
^
•
fe.-;
-
^"Hfe^
-
- ^
;
' -
fe^tfe'^fe/
-
^
-, ;
fe:
\
-fe'.
:
^
: .
. .... ^
•X
wt; -
í,. • "... I , - . •
-fe-fe
./.
• ! .
• < " _ ,- ,fe; - ' " i r -vc.'/í,.- í - f e
:=?
fe,-
-
fe-fe fe^fet^fe
^
1
.' •• 36-'
- •
T
-
•< i
fe--,
-y
.
--
'
'
^
- *
Bulk Rate
U.S. Postage
PAID
Atlanta, GA
Permit 292
BANK OF
Public Affairs Department
104 Marietta Street, N.W.
Atlanta, Georgia 3 0 3 0 3 - 2 7 1 3
(404)
5 2 1 - 8 0 2 0
®
printed on recycled paper-
@FedFRASER