Income inequality and industrial composition.
Moore, William S.
INTRODUCTION (1)
When thinking about social class and inequality, the theme for this
symposium publication, it is important to consider not only the
magnitudes and changes in inequality but the determinants of such
phenomena as well. The focus of this paper is specifically on income
inequality in the United States. This issue will first be briefly
examined from a national, state and international perspective. Following
this brief review, a basic model of analysis is presented including a
description of the data used in the analysis. The author utilizes a type
of regression analysis that takes advantage of variations that are
present at the sub-national level (states) in the dependent and
independent variables. The paper then proceeds to a presentation of the
results from estimations applying the regression model to two sets of
independent variables. Finally the paper concludes with some possible
generalizations and their public policy implications.
INCOME INEQUALITY IN THE UNITED STATES
It is clear that there is a growing disparity in the distribution
of incomes over time in the United States. In a report published by the
US Census Bureau (Jones and Weinburg, 2000), the authors note that from
1968 to 1998 income inequality has increased substantially. The measure
of income inequality that they and many others use is the Gini
coefficient (2) which is an index of income concentration. A value of 0
indicates perfect equality and a value of 1 indicates perfect
inequality. Even though there have been some periods in which the Gini
coefficient decreased, the fifty year period from 1947 to 1998 has seen
a net increase in the Gini coefficient or income inequality in the
United States. Figure 1 displays the Gini ratio for three selected years
which correspond to the years utilized for subsequent analysis in this
paper.
As can be seen, the level of income inequality has increased for
every time period noted in Figure 1. The percentage change from 1979 to
1989 was 7.2% and the percentage change from 1989 to 1999 was 4.0%.
Further the percentage change over the twenty year period was 11.6%.
The causes as noted by Jones and Weinburg (2000) for changes in
income inequality are generally associated with labor market changes and
changes in household composition. More specifically Jones and Weinburg
(2000) observe that a shift from manufacturing employment to high paying
technical services jobs and lower paying retail trade jobs appears to be
related to the increase in income inequality. Further Jones and Weinburg
(2000) state that, "But within-industry shifts in labor demand away
from less-educated workers are, perhaps, a more important explanation of
eroding wages than a shift out of manufacturing" (pg. 10). Finally
Jones and Weinburg (2000) note that changes in household composition
such as increased divorce rates, out of wedlock births, and increasing
age of first marriage may also contribute to the increase in income
inequality. While it is clear that income inequality has increased in
the United States as a nation over the last thirty years, the question
remains as to how the variation within states in the United States? This
is the issue that is reviewed next.
INCOME INEQUALITY WITHIN THE UNITED STATES
In a paper by Lynch (2003), the researcher examines income
inequality for the United States and the individual states for the years
1988, 1995, and 1999. Lynch uses a data set he created using statistical
matching techniques. The results on the national trend are consistent
with the previously cited work and will not be reviewed here. In
analyzing the trends in income inequality amongst the states, Lynch
(2003) finds that the overall trends were generally similar to the
nation. The magnitude of the changes however may vary substantially by
region and state, though in some cases the trend is even opposite of the
nation's. Thus the states reveal substantial variation that will be
utilized when estimating the regression model later in this paper.
Lynch's (2003) calculations reveal that for all years
considered (1988, 1995 & 1999) New York State had the highest level
of income inequality. Other states that ranked highly were California,
Connecticut, Florida, Illinois, New Jersey, Nevada, and Texas. States
that consistently ranked low were Iowa, Indiana, North Dakota, Maine,
Vermont, and Wisconsin. With respect to regions within the United
States, Lynch (2003) notes that the Mid-Atlantic and Southern region had
the highest levels of income inequality while the Midwest and Mountain
regions had the lowest levels. Table 1 below displays Gini ratio data
for the years used in this paper for each state. As seen in Table 1 the
US average percentage change period for the twenty years from 1979 to
1999 was 5.6%. All states had a positive average percentage change
indicating increasing income inequality. The states with the highest
rates of change was California (7.9%), Connecticut (10.6%), the District
of Columbia (10.5%), Massachusetts (7.9%), New Jersey (8.2%) and New
York (9.1%). The States with the lowest rate of change were Alaska
(1.1%) and South Dakota (3.0%). Thus far the magnitude and changes in
income inequality on a national scale and at the state level have been
examined. But how does the United States compare with other
industrialized nations?
INCOME INEQUALITY IN THE UNITED STATES AND OTHER INDUSTRIALIZED
NATIONS
Smeeding (2005) provides an international comparison of income
inequality among the industrialized nations or member nations of the
Organization for Economic Cooperation and Development (OECD). Table 2
displays the Gini coefficients (3) for the OECD members.
As the figures from Table 2 clearly show the United States, with
the exception of Russia and Mexico, has the highest level of income
inequality of the OECD member nations. In fact the United States has a
level of income inequality that is about 20% higher than the simple
average of all the industrialized nations. Smeeding (2005) states that,
"Americans have the highest income inequality in the rich world and
over the past 20-30 years Americans have also experienced the greatest
increases in income inequality among rich nations" (pg. 968).
Smeeding's (2005) paper continues with a discussion of why
income inequality in the United States is comparatively higher than the
rest of the industrialized world. Two primary observations are the
relatively low wages at the bottom of the income distribution and a weak
income support system in the United States. In further explaining the
differences between the United States and the rest of the industrialized
world, Smeeding (2005) provides more specifics. First he discusses the
role that government plays in this area pointing to direct effects such
as income redistribution policy and indirect effects such as legal
institutions and regulations associated with labor markets that support
wages particularly for the lower income households. He notes that both
are comparatively weak in the United States. He also observes that there
is a relationship between income inequality and the number of low wage
jobs in a nation. Secondly, Smeeding (2005) continues with other
explanations that he and other researchers have offered. These include
the argument that labor in United States is more productive but as he
notes this would account for only a modest amount of difference.
Nevertheless he does observe that workers in the United State do tend to
work more hours but higher income households also work more hours than
similar households in other countries. Further higher income US workers
are more likely to marry spouses who work comparatively more hours.
Smeeding (2005) also discusses demographic factors that may impact
earnings particularly at the lower end of the income distribution. He
notes that countries with higher levels of immigration and larger
numbers of single parents tend to have more income inequality. However
this factor has only a minor impact on inequality. In fact he continues
by citing a work he co-authored that analyzes any number of demographic
factors and concludes that their impact on inequality is small indeed
and that redistribution policy is much more likely to influence
inequality.
Smeeding (2005) continues his review of the evidence by discussing
the role that labor market institutions play and compares that with
educational attainment, both of which have an impact on inequality. The
labor market institutions he cites are collective bargaining, wage
setting, and minimum wages. He concludes, based on the evidence to date
that while education levels matter with respect to earnings, the
differences in wage setting institutions matter more.
Finally, Smeeding (2005) comments on the issue of globalization as
a force in increasing the earnings gap where earnings are associated
with value-added productivity. Thus as low skill, low paying jobs are
shifted to other nations and high skill, high paying jobs are retained
and/or created, the United States may experience a continuation in the
trend of increasing income inequality unless government intervenes
somehow. Smeeding (2005) makes the following conclusive statement with
respect to income inequality in the United States: "America has
more inequality and less redistribution (lesser amounts and less
effective as an anti-poverty device) than any other rich nation. Large
numbers of low-skilled workers and inadequate safety nets are two
important reasons for these outcomes" (pg. 979). As discussed above
the change in labor markets associated with changes in industrial
composition in the US economy may have an impact on income inequality
here in the United States. It is this issue that is the focus of this
paper.
TRENDS AND CAUSES OF WAGE INEQUALITY IN THE UNITED STATES
Neckerman and Torche (2007), provide an excellent current survey of
the literature on economic inequality including the causes of economic
inequality in the United States. With respect to income inequality in
the United States, Neckerman and Torche (2007), discuss the trend in
wage inequality from World War II to the present. Neckerman and Torche
(2007) note, as others have, that wage inequality in the United States
started to increase in the 1970s, increased more rapidly in the 1980s,
stabilized in the 1990s and has remained stable in the early 2000s
(Neckerman and Torche, 2007, p. 336). Moreover they discuss, as others
have, how changes in household earnings associated with changes in
household composition have influenced income inequality noting how
changes in the percentage of single-adult households and changes in
earnings for households with married couples may have contributed to the
overall disparities (Neckerman and Torche, 2007, p. 336).
In addition, Neckerman and Torche (2007) discuss how the
distribution of earnings has changed over time. They point out that in
the 1980s income inequality increased in both tails of the earnings
distribution. Inequality in the lower end of the distribution stopped in
the late 1980s and then slightly decreased in 1990s. In the meantime,
inequality in the upper end of the earnings distribution continued to
increase (Neckerman and Torche, 2007, p. 336). In fact, the authors
state, "The highest 1% experienced faster income growth than the
next highest 9%, while the highest 0.1% gained more than others in the
top 1%" (Neckerman and Torche, 2007, p. 337 as cited in Piketty and
Saez, 2003). This is not just a matter of the income inequality
associated with reduced income in the lower end of the earnings
distribution as discussed by Smeeding (2005) but also points out an
important change in income inequality in the United States.
(Neckerman and Torche (2007) continue with a thorough review of the
literature regarding the causes of such income inequality and the new
dynamics of such in the United States. The authors point out that there
is some consensus on the causes which include: 1) changes in real
minimum wages, 2) declining male union membership, and 3) rising returns
to higher education (Neckerman and Torche, 2007, p. 337).
Nevertheless there is disagreement on the role of computer
technology with respect to income inequality. Neckerman and Torche
(2007) point out that one hypothesis that the increased use of computers
in the workplace may increase the value of education and computer skills
which would partially explain the increase in income inequality. The
authors point out that this hypothesis may be particularly attractive
among economists but it is not without it critics (Neckerman and Torche,
2007, p. 337). It may be "that computerization enhanced the
productivity of highly educated professionals, undermined the demand for
routine cognitive workers usually located in middle-wage jobs, and had
relatively impact on the lowest skilled blue-collar and service
occupations located in the lower end of the earnings distribution"
(Neckerman and Torche, 2007, p. 338 as cited in Autor et al., 2005,
2006). This would be consistent with the apparent erosion of earnings in
the middle-class and the increase in earnings for well-educated workers.
Finally Neckerman and Torche (2007) discuss the potential impact of
changes in firms and labor markets in the United States on wage
inequality. The authors note the following factors: 1) structural change
in the economy, including a shift from manufacturing to services, 2)
deregulation in many industries, 3) changes in corporate governance, 4)
decline in union representation, and 5) an increase in contingent labor
(Neckerman and Torche, 2007, p. 338). It is the first factor, economic
restructuring, that this research is concerned with. The authors state
however that, "Although the link between changes in economic
organization and growing inequality of wages is plausible, as Morris and
Western (1999) concluded eight years ago, we have little direct evidence
of such a link" (Neckerman and Torche, 2007, p. 338). It is this
issue that this research will now specifically examine.
THE MODEL AND THE DATA
Again, the purpose of this research is to explore the relationship
between industrial composition in the United States and income
inequality. A fixed effects model is used to test the hypothesis that
the change in the industrial structure of the United States economy is
associated with income inequality. The model has the following general
form:
[Y.sub.it] = [alpha] + [beta][T.sub.t] + [D.sub.[chi]] +
[X.sub.[delta]] + [[epsilon].sub.i,t] (0.1)
Y is the Gini ratio for each state and the District of Columbia for
the years 1979, 1989, and 1999 and represents a measure of income
inequality. The [alpha] is a common intercept and T is a common time
counter variable that does not vary between panels. The D is a matrix of
panel specific intercepts, and X is a matrix of variables representing
industrial composition for the years 1978, 1988, and 1998. These years
were used to help control for the possibility of simultaneity conditions
that may be present. Of course the [epsilon] is the usual error term. As
widely noted in the econometrics literature and elsewhere (4) the fixed
effects model allows the researcher to statistically control for
unobserved differences between the panels or states in this case.
Data for the Gini ratios is from the U.S. Census Bureau, Table S4,
Gini Ratios by State, derived from their Current Population Survey. The
data for the industrial composition variables is from the U.S.
Department of Commerce, Bureau of Economic Analysis, State Annual
Tables. Two different sets of independent variables representing
industrial composition were constructed and used to determine if an
association does indeed exist between industrial composition and income
inequality. These two sets of independent variables are the percentage
of total employment by industry and percentage of total earnings by
industry. Tables 3-8 provide some descriptive statistics of interest.
Table 3 seen below shows the distribution of and percentage change
in employment by major industrial sectors for the years 1978, 1988, and
1998. Again the manufacturing and service sectors receive particular
scrutiny given the perception of their impact on income inequality.
Examination of Table 3 reveals that the share of total employment
in manufacturing in the United States has decreased dramatically over
the time frame considered. More precisely the change in manufacturing
employment as a percentage of total employment has decreased by about
twenty percent over twenty years. Manufacturing represented nineteen
percent of total employment in 1978 and by 1998 that figure was down to
just over twelve percent. Further over the same time frame manufacturing
moved from a rank of second, just slightly lower than services, to a
ranking of fourth with respect to the percentage of total employment.
Services, retail trade, and government all are ranked above
manufacturing as of 1998. Conversely, service employment as a percentage
of total employment increased by nearly twenty-two percent over the same
twenty year period. Further services accounted for twenty-one percent of
total employment in 1978 and by 1998 it accounted for slightly over
twenty-one percent of total employment. Clearly there is a significant
increase in the share of service employment to total employment and
conversely a significant decrease in the share of manufacturing
employment to total employment from 1978 to 1998.
In addition Table 4, seen below, shows the overall (across all
states and years) means, and standard deviations for this set of
independent variables and the dependent variable (the Gini ratio). The
average Gini ratio for all states in the three years is about .43. Also
noteworthy is the overall mean for services that is almost .26 followed
by retail trade and government at about .16, and then by manufacturing
at .14. Clearly the data shows the dominance of the service sector in
employment concentration.
Table 5 is the correlation matrix for all the independent variables
in this set. The relationship between manufacturing and services is as
hypothesized in that a one percent decrease in the share of
manufacturing employment is associated with a .49 percent increase in
service employment share. The strongest correlations however are between
the share of government employment and the shares of retail trade
employment (-.59) and wholesale trade employment (-.54), respectively.
Interestingly the relationship between government and manufacturing
indicates that as the manufacturing sector decreases, government
increases as noted by the -.461 correlation statistic.
Descriptive statistics for the other set of independent variables,
the share of earnings by industry, are shown below in Tables 6, 7 and 8.
Table 6 shows the distribution of shares of earnings by major
industry in the United States. Once again particular emphasis is on
manufacturing and services. In 1978 manufacturing accounted for almost
27% of all earnings in the United States. By 1998 it accounted for
slightly over 18% of all earnings. From 1978 to 1998 the share of
manufacturing earnings decreased on average by about 18% which is
similar to the decline in employment shares noted (20%) from in Table 3.
In 1978 services accounted for over 15% of total earnings and by 1998 it
accounted for over 27% of all earnings. The share of earnings from
services increased on average by slightly over 33% from 1978 to 1998.
This is considerably different than the share of employment which had an
average increase of almost 22%.
Table 7 again displays the overall means and standard deviations
for the earnings share set of independent variables. In this set of
variables manufacturing, services, and government have roughly equal
shares of total earnings at .20 which are also the largest shares.
Table 8, below, shows the correlation matrix for this set of
variables. Again the relationship between manufacturing and services is
as hypothesized. For every one percent decrease in the share of
manufacturing employment there is a .463 percent increase in the share
of services earnings. The largest correlation however is between
manufacturing and government. The correlation here is -.582. Both the
relationships between manufacturing and services, and manufacturing and
government appear moderately strong.
ESTIMATIONS OF THE MODEL
Given the above figures the basic regression model described
earlier is applied to try to determine if a relationship exists between
industry composition, first as measured by the share of total employment
for the industrial sectors and secondly by the share of earnings for the
industrial sectors, and with income inequality as measured by the Gini
ratio in both cases.
Shares of Total Employment by Industry
In the first set of estimates using shares of total employment by
industry, the mining sector was dropped to avoid perfect collinearity
among the independent variables. Table 9 below shows the results of the
fixed effects regression.
As can be seen in Table 9 the model explains about 32 percent of
the overall variation in the Gini ratio. Clearly, as the literature
suggests, there are other explanatory variables that have been omitted.
Nonetheless the use of a fixed effects model minimizes omitted variable
bias. The model as expected provides for a high R-square value at .90
within the panels or states. Further the estimates strongly pass the
F-test of joint significance.
In analyzing the parameter estimates it is interesting to note
first that the time trend variable shows increased, though moderately,
income inequality and is statistically significant at above the 99
percent level of confidence. Second, as hypothesized, the result for
manufacturing indicates that a one percent increase in
manufacturing's share of employment leads to a .3448 decrease in
the Gini ratio. This means that indeed increased manufacturing
employment share does reduce income inequality or conversely that a
decrease in manufacturing employment share leads to an increase in
income inequality. Further this result is statistically significant at
about the 99 percent level of confidence.
Unexpectedly, however, the relationship for services employment has
the opposite sign as hypothesized and more importantly the parameter
estimate is statistically significant at just slightly lower than the 95
percent level of confidence. In fact both services and manufacturing
employment appear to have the same relative impact on income inequality
with parameter estimates at approximately -.33. The possible reason for
the unexpected result associated with services employment will be
discussed later in this paper.
Other results that are statistically significant at the 95 percent
level of confidence or above are government and construction employment.
A one percent change in the share of government employment reduces the
Gini ratio by .2891. A one percent increase in the share of construction
employment reduces the Gini ratio by a relatively large .7213. Of the
parameter estimates that are statistically significant at the 95 percent
level of confidence or above, this has by far the largest impact which
is meaningful with respect to policy considerations.
Shares of Total Earnings by Industry
In an attempt to improve the performance of the basic model and
more specifically to insure accurate measurement of the size of the
industrial sectors, a second set of independent variables is utilized.
The second set of independent variables, as noted previously, is the
shares of total earnings for each industry sector.
In this estimation of the basic model, the data on the shares of
earnings by industry at the state level for 1978, 1988, and 1998 was
utilized to test the hypothesis. Again the mining sector was dropped to
avoid perfect collinearity. Table 10 below shows the results of the
estimation.
The results of the R square and F test statistics are similar to
the first model. Again the model only explains about 34 percent of the
overall variation in the Gini ratio indicating the omission of important
independent variables. And the estimates of the model are jointly,
statistically significant above the 99 percent level of confidence.
In examining the parameter estimates once again the overall time
trend variable indicates a statistically significant, above the 99
percent level of confidence, increase in the level of income inequality.
However, in this model only the estimates for the share of earnings in
construction and finance, insurance, and real estate (FIRE) are
statistically significant at about the 98 percent level of confidence.
None the other estimates are statistically significant at even the 90
percent level of confidence.
The estimates for construction and FIRE are -.3017 and .3239,
respectively. So in this model is appears that an increase of one
percent in the share of construction earnings reduces the Gini ratio by
.3017; thus, reducing income inequality. And a one percent increase in
the share of FIRE earnings increases the Gini ratio by .3239; thus,
increasing income inequality.
Finally it appears that the hypothesized notion that the reduction
in the size of the manufacturing sector and the increase in the size of
the service sector lead to increases in income inequality does not hold
for this model.
UNDERSTANDING THE RESULTS IN THE CONTEXT OF THE RESEARCH HYPOTHESIS
As noted in the literature review it was hypothesized that
decreases in the size of the manufacturing sector and increases in the
size of the service sector would lead to increased income inequality.
Nevertheless the results here are at times inconsistent with such
assertions. The size of the manufacturing sector as measured by the
share of employment does follow the research hypothesis but when the
share of earnings is used the result is not consistent and statistically
insignificant. In the case of services, using shares of employment as a
measure of industrial composition, the result is statistically
significant but contrary to the hypothesis; in other words, an increased
share of service employment actually reduces income inequality. Further
when measuring industrial composition by using share of earnings, the
result is statistically insignificant although the sign of estimate
remains the same but again this is contrary to the hypothesis. Finally
nothing is mentioned in the literature about the role of the
construction sector and income inequality and yet there appears to be a
relatively strong and statistically significant relationship between the
two in terms of reducing income inequality. The above results do beg the
question of why one would hypothesize the relationship between income
inequality and the changes in the manufacturing and services sectors in
the United States. In order to understand the possible origins of the
research hypothesis some further analysis is necessary.
Table 11 below shows what the average earnings were for the major
industrial sectors in the United States in 1978, 1988, and 1998. All
figures are in nominal dollars; that is, they were not adjusted for
inflation.
As seen in Table 11 the average change in average earnings for all
industries from 1978 to 1998 was a positive 64%. By contrast the average
earnings for manufacturing increased on average by slightly over 67%
while the service sector's average earnings increased by almost
80%. Thus the increase in average earnings for both manufacturing and
services outperformed the nation as a whole.
The level of average earnings relates a different story, however.
The average earnings in the manufacturing sector in 1978 were $14,519
compared to $9,301 for services. By 1998 the average earnings for
manufacturing was $39,997 and for services it was $29,243. Thus the
difference in the level of average earnings for workers between
manufacturing and services had increased from about $5,000 in 1978 to
approximately $10,000 in 1998. It is this differential that may be one
of the true drivers of income inequality as a result of a change in the
industrial composition of the United States economy.
Additionally, the figures for the construction sector show that the
level of average earnings is somewhat close to the level for
manufacturing. They are also higher than the US average and the service
sector in all years. Nonetheless the average rate of change in average
earnings for construction is less than manufacturing, services, and the
US average.
Given the figures in Table 11, the plausibility of the research
hypothesis is well founded. Nonetheless the results from the estimations
may be telling a bit of different story.
This difference solcits the question of why the inconsistency? One
answer may be in the measurement of the explanatory variable. Using the
major industrial classifications, which are aggregate measures of all
types of manufacturing, may be masking vital information. Indeed, under
the three digit Standard Industrial Classification (SIC) code there are
21 categories of manufacturing
In the case of the services sector there are 16 different
categories of services at the two-digit level of the SIC code. Further
there are considerable differences in the average earnings per worker in
the service sector with respect to the 16 categories. For example, in
1998 the average earnings per worker in legal services were $53,603
while it was $10,656 for private household services. This can be seen in
Table 12 below. As a result it is very likely that the aggregation of
the services variables is again masking valuable information that is
causing measurement error and as such leads to the results that have
been displayed previously.
Cleary aggregation could be masking the important differences as
noted above; however, another factor that can shed some light on the
results particularly with respect to services, may also be at play. If
Table 3, the share of employment by industry, is examined along with
Table 11, average earnings by industry, the estimation results may
appear more plausible than at first glance. It was noted in Table 3,
that employment in services was increasing dramatically over time.
Further it was noted from Table 11 that the average earnings in the
service sector, even though lower than manufacturing and the US average,
were increasing at a higher rate than manufacturing or the US average.
Thus it is conceivable that the effect of the increase in average
earnings along with the increase in share of employment were indeed
decreasing income inequality as shown in the estimates using employment
shares by industry as the independent variables.
CONCLUSION
Summary of Findings
Even with the measurement concerns noted in the previous section
there are some conclusions that can be made at this point. First, there
is evidence that a reduction in the size of the manufacturing sector in
the United States as measured by the share of total employment does seem
to lead to greater income inequality.
Second, not all changes in the manufacturing sector may have the
same impact; however, for example it may be that a reduction in durable
goods manufacturing may be the key to the relationship with respect to
income inequality.
Third, there is evidence that increases in the share of employment
in services, in the aggregate, also reduces income inequality. It seems
fairly obvious that the likely scenario is that some categories of the
service sector may have strong impacts that increase or decrease income
inequality depending on the quality of the jobs in the services
sub-sector in question. As noted above in Table 12, average earnings for
some workers for legal services are much higher than the whole category
of service workers. Indeed they are even substantially higher than the
national average while workers in private household services command
much lower than average earnings compared to the nation or the entire
services category.
Finally positive changes in the construction sector as measured by
employment or earnings shares help to reduce income inequality in not
only a statistically significant but also in a substantial manner with
respect to the other industrial sectors.
Policy Implications
The public policy implications for the United States from this
analysis are reasonably clear. If the policy goal is to reduce income
inequality then policymakers must look beyond the simple or convenient
notion that all that is needed is to protect the manufacturing sector.
Globalization with all its implications for income inequality will be
difficult if not impossible to stop. The United States can promote the
manufacturing sector, in particular manufacturing of durable goods, but
it could also encourage growth in other industrial sectors that show
prospects for generating quality jobs. This would also include certain
services sub-sectors such as legal, business, and other professional
services as these sectors do produce better than average earnings jobs.
Such a strategy would certainly involve more investment in human capital
through education and training.
Moreover, one policy alternative that might be able to promote
equity and economic growth simultaneously would be increased spending on
construction for public infrastructure. The need for such investment is
well documented but beyond the scope of this paper. Increased spending
on public infrastructure if conducted efficiently and effectively could
certainly lead to higher economic growth. Further as shown in this
paper, there is a strong positive relationship between construction
employment or earnings and income equality. Further as noted above
increases in government employment also reduces income inequality.
Increasing government employment associated with public infrastructure
expenditures may provide additional benefits in improved equity and
economic growth.
In the meantime additional research on the determinants of income
inequality including the use of other explanatory variables and
disaggregated measures of industrial composition should be conducted.
Further, the cost and benefits of engaging in some type of industrial
policy aimed at reducing income inequality should be examined carefully
so that policymakers are better informed on what the likely impact of
such policies would be. It could very well be that the best policy for
the United States would be to take actions that minimize the pain of
globalization such as establishing a stronger social safety net and to
promote policies such as investment in human and public capital, through
increased spending on education and public infrastructure, will help the
United States compete successfully in a new world economic order. The
results of this paper clearly support such a proposition.
REFERENCES
Autor, D.H., Katz, L., & Kearney, M. (2005). Trends in US wage
inequality: re-assessing the revisionists. NBER Working Paper 11627,
National Bureau of Economic Research.
Autor, D.H., Katz, L., & Kearney, M. (2006). The polarization
of the US labor market. NBER Working Paper 11986, National Bureau of
Economic Research.
Gujarati, D. N. (2003). Basic econometrics (4th ed.). New York:
McGraw-Hill.
Jones Jr., A. F. & Weinberg, D.H. (2000). The changing shape of
the nation's income distribution (Current Population Reports,
PS60-204). Washington, DC: U.S. Census Bureau
Lynch, R. G. (2003). Estimates of income and income inequality in
the United States and in each of the fifty states: 1988-1999. Journal of
Regional Science. Vol. 43, No. 3, pgs. 571-587.
Morris, M. & Western, B. (1999). Inequality in earnings at the
close of the twentieth century. Annual Review of Sociology. Vol. 25,
pgs. 623-657.
Neckerman, K., & Torche, F. (2007). Inequality: causes and
consequences. Annual Review of Sociology. Vol. 33, pgs. 335-357.
Piketty, T. & Saez, E. (2003). Income inequality in the United
States: 1913-1998. Quarterly Journal of Economics. Vol. 118, pgs. 1-39.
Smeeding, T.M. (2005). Public policy, economic inequality, and
poverty: the United States in comparative perspective. Social Science
Quarterly. Vol. 86, No. 5, pgs. 955-983.
WILLIAM S. MOORE
California State University, Long Beach
(1) Thanks are due to two anonymous reviewers for this paper, to
comments from the participants on the panel on inequality of the General
Economics section at the Western Social Science Association Spring 2008
conference, and to Edward Martin, Symposium Editor, for all his
patience.
(2) Gini ratios or coefficients have a long history of use as a
measure of income inequality. To be sure, their use as a measure of
income inequality can be problematic. Further it is a measure of pre-tax
income concentration which can be altered by tax systems but it can also
be altered by public expenditures including cash income transfer
programs. The limitations of the Gini ratio are beyond the scope of this
paper however. Nonetheless it remains a reasonably plausible measure of
income inequality.
(3) The Gini coefficients are calculated differently which is why
the figure for the United States appears considerable lower than the
figure for 1999 in Figure 1 and Table 1. Most importantly the
calculations are consistently applied to all nations which maintain the
relative magnitudes and rankings.
(4) Most introductory to advanced econometrics texts will contain a
section if not a whole chapter on analysis of panel data including fixed
effects models. For a good introduction see Gujarati (2003). In addition
there are entire books dedicated to analysis of panel data. For example
see: Baltagi, Badi H. (1995) Econometric Analysis of Panel Data, John
Wiley & Sons.
Table 1--Gini Ratios by State, (1979, 1989, 1999)
Source: Table 3, Lynch (2003)
% %
Change Change Average
1979 1989 1999 1979- 1989- %
1989 1999 Change
Alabama 0.427 0.458 0.475 7.3 3.7 5.5
Alaska 0.393 0.397 0.402 1.0 1.3 1.1
Arizona 0.399 0.439 0.450 10.0 2.5 6.2
Arkansas 0.428 0.450 0.458 5.1 1.8 3.4
California 0.408 0.441 0.475 8.1 7.7 7.9
Colorado 0.392 0.426 0.438 8.7 2.8 5.7
Connecticut 0.390 0.434 0.477 11.3 9.9 10.6
Delaware 0.396 0.411 0.429 3.8 4.4 4.1
D. C. 0.450 0.492 0.549 9.3 11.6 10.5
Florida 0.421 0.450 0.470 6.9 4.4 5.7
Georgia 0.421 0.446 0.461 5.9 3.4 4.6
Hawaii 0.390 0.408 0.434 4.6 6.4 5.5
Idaho 0.390 0.421 0.427 7.9 1.4 4.6
Illinois 0.396 0.440 0.456 11.1 3.6 7.3
Indiana 0.379 0.411 0.424 8.4 3.2 5.8
Iowa 0.390 0.412 0.418 5.6 1.5 3.5
Kansas 0.399 0.428 0.435 7.3 1.6 4.4
Kentucky 0.420 0.456 0.468 8.6 2.6 5.6
Louisiana 0.436 0.476 0.483 8.7 1.5 5.0
Maine 0.382 0.414 0.434 8.4 4.8 6.6
Maryland 0.385 0.410 0.434 6.5 5.9 6.2
Massachusetts 0.398 0.428 0.463 7.5 8.2 7.9
Michigan 0.389 0.429 0.440 10.3 2.6 6.4
Minnesota 0.391 0.418 0.426 6.9 1.9 4.4
Mississippi 0.440 0.475 0.478 8.0 0.6 4.2
Missouri 0.408 0.438 0.449 7.4 2.5 4.9
Montana 0.395 0.421 0.436 6.6 3.6 5.1
Nebraska 0.396 0.414 0.424 4.5 2.4 3.5
Nevada 0.387 0.42 0.436 8.5 3.8 6.1
New Hampshire 0.372 0.387 0.414 4.0 7.0 5.5
New Jersey 0.393 0.431 0.460 9.7 6.7 8.2
New Mexico 0.415 0.448 0.460 8.0 2.7 5.3
New York 0.419 0.467 0.499 11.5 6.9 9.1
North Carolina 0.403 0.430 0.452 6.7 5.1 5.9
North Dakota 0.397 0.409 0.441 3.0 4.9 4.0
Ohio 0.385 0.427 0.455 10.9 3.3 7.0
Oklahoma 0.419 0.445 0.438 6.2 2.2 4.2
Oregon 0.394 0.421 0.452 6.9 4.0 5.4
Pennsylvania 0.391 0.435 0.457 11.3 3.9 7.5
Rhode Island 0.397 0.420 0.454 5.8 8.8 7.3
South Carolina 0.406 0.428 0.454 5.4 6.1 5.7
South Dakota 0.409 0.394 0.434 -3.7 10.2 3.0
Tennessee 0.418 0.451 0.465 7.9 3.1 5.5
Texas 0.415 0.457 0.470 10.1 2.8 6.4
Utah 0.371 0.395 0.410 6.5 3.8 5.1
Vermont 0.386 0.385 0.423 -0.3 9.9 4.7
Virginia 0.399 0.425 0.449 6.5 5.6 6.1
Washington 0.388 0.414 0.436 6.7 5.3 6.0
West Virginia 0.406 0.448 0.468 10.3 4.5 7.4
Wisconsin 0.381 0.402 0.413 5.5 2.7 4.1
Wyoming 0.372 0.395 0.428 6.2 8.4 7.3
United States 0.415 0.445 0.463 7.2 4.0 5.6
Table 2--Gini Ratios, OECD Nations
Gini
OECD Nations Coefficients
Denmark (1992) 0.236
Slovak Republic (1996) 0.241
Finland (2000) 0.247
Netherlands (1999) 0.248
Slovenia (1999) 0.249
Norway (2000) 0.251
Sweden (2000) 0.252
Germany (2000) 0.252
Czech Republic (1996) 0.259
Luxembourg (2000) 0.260
Austria (2000) 0.260
Romania (1997 0.277
Belgium (2000) 0.277
France (1994) 0.288
Poland (1999) 0.293
Hungary (1999) 0.295
Taiwan (20000 0.296
Canada (2000) 0.302
Spain (1990) 0.303
Switzerland (1992) 0.307
Australia (1994) 0.311
Japan (1992) 0.315
Ireland (2000) 0.323
Italy (2000) 0.333
United Kingdom (1999) 0.345
Israel (2001) 0.346
Estonia (2000) 0.361
United States (2000) 0.368
Russia (2000) 0.434
Mexico (2002) 0.471
Average 0.300
Source: Smeeding (2005, pg. 958)
Table 3
Distribution of US Employment by Industry
% %
Change Change
1978 1988-
1978 1988 1998 -1988 1998
Agricultural services,
forestry, fishing & other 0.7% 1.0% 1.2% 36.2% 21.6%
Mining 1.0% 0.8% 0.5% -15.8% -34.4%
Construction 5.1% 5.3% 5.5% 4.2% 2.5%
Manufacturing 19.1% 14.8% 12.2% -22.7% -17.4%
Transportation and public
utilities 4.9% 4.6% 4.8% -6.3% 4.5%
Wholesale trade 5.0% 4.8% 4.6% -2.9% -4.0%
Retail trade 15.8% 16.4% 16.4% 4.3% -0.4%
Finance, insurance, and
real estate 7.5% 7.9% 7.8% 5.0% -1.6%
Services 21.0% 26.7% 31.1% 27.0% 16.6%
Government and
government
enterprises 16.4% 15.2% 13.9% -7.8% -8.4%
Table 4
Percentage of Total Employment and Gini Ratio
Variable Mean (Overall) Standard Deviation
(Overall)
giiii .425 .029
agriculture .010 .006
mining .010 .017
construction .055 .012
manufacturing .140 .063
transportation .047 .008
wholesale trade .044 .009
retail trade .162 .017
f.i.re .072 .015
services .255 .056
government .165 .048
Table 5
Correlation Matrix of Employment Variables
A Mn C Mf T W R
A 1
Mn .041 1
C .101 .322 1
Mf -.384 -.335 -.130 1
T .195 .441 .073 -.346 1
W -.266 -.185 -.032 .289 .223 1
R .011 .032 .449 .057 .061 .318 1
F .034 -.250 -.045 -.216 .011 .167 .012
S .149 -.246 -.123 -.491 -.204 -.246 -.091
G .224 .212 -.220 -.461 .154 -.540 -.595
F S G
A
Mn
C
Mf
T
W
R
F 1
S .354 1
G -.124 -.108 1
Key: A = agriculture; Mn = mining; C = construction; Mf = manufacturing
T = transportation; W = wholesale trade; R = retail trade;
F = finance, insurance, real estate; S = services; and G = government
Table 6
Distribution of US Earnings by Industry
% Change
1978 1988 1998 1978-1988
Agricultural services,
forestry, fishing & other 0.4% 0.5% 0.6% 42.1%
Mining 1.5% 1.0% 0.7% -31.0%
Construction 5.8% 5.3% 5.0% -8.1%
Manufacturing 26.9% 21.7% 18.1% -19.3%
Transportation and public
utilities 7.5% 6.6% 6.4% -12.0%
Wholesale trade 6.8% 7.0% 6.9% 2.9%
Retail trade 10.3% 9.9% 9.3% -3.8%
Finance, insurance, and real
estate 5.5% 7.8% 8.8% 43.0%
Services 15.5% 21.7% 27.4% 40.2%
Government and government
enterprises 19.2% 18.0% 16.4% -6.4%
% Change
1988-1998
Agricultural services,
forestry, fishing & other 16.9%
Mining -27.8%
Construction -6.6%
Manufacturing -16.7%
Transportation and public
utilities -4.0%
Wholesale trade -1.0%
Retail trade -5.9%
Finance, insurance, and real
estate 12.1%
Services 26.2%
Government and government
enterprises -8.5%
Table 7
Percentage of Total Earnings and Gini Ratio
Variable Mean (Overall) Standard Deviation
(Overall)
gini .425 .029
agriculture .004 .002
mining .018 .033
construction .057 .017
manufacturing .205 .092
transportation .071 .016
wholesale trade .063 .015
retail trade .102 .014
f.i.re .062 .023
services .205 .064
government .202 .065
Table 8
Correlation Matrix of Earnings Variables
A Mn C Mf T
A 1
Mn -.190 1
C -.158 .388 1
Mf -.333 -.370 -.287 1
T -.253 .505 .268 -.374 1
W -.158 -.256 -.104 .177 .195
R -.042 .007 .404 -.054 .170
F .159 -.363 -.305 -.114 -.311
S .449 -.300 -.164 -.463 -.345
G .250 .251 .049 -.582 .276
W R F S G
A
Mn
C
Mf
T
W 1
R .194 1
F .256 -.263 1
S -.066 -.193 .496 1
G -.482 -.134 -.327 -.115 1
Key: A = agriculture; Mn = mining; C = construction; Mf = manufacturing
T = transportation; W = wholesale trade; R = retail trade;
F = finance, insurance, real estate; S = services; and G = government
Table 9
% of Total Employment by Industry
R-sq:
Within = .9040 F(10,92) = 86.61
Between = .0003 Prob > F = 0.0000
Overall = .3249
Std. P> [absolute
Gini Coef. Err. t value of t]
Time .0235 .0048 4.85 0.000
Agriculture .7262 .5596 1.30 0.198
Construction -.7213 .1913 -3.77 0.000
Manufacturing -.3448 .1120 -3.08 0.003
Transportation -.5018 .2937 -1.71 0.091
Wholesale Trade -.0641 .4235 -.150 0.880
Retail Trade -.0616 .1575 -0.39 0.696
FIRE .1871 .1550 1.21 0.230
Services -.3245 .1655 -1.96 0.053
Government -.2891 .1250 -2.31 0.023
Constant .6138 .1127 5.45 0.000
Table 10
% of Total Earnings by Industry
R-sq:
Within = .8947 F(10,89) = 75.65
Between = .0006 Prob > F = 0.0000
Overall = .3411
P>
Std. [absolute
Gini Coef. Err. t of t]
Time .0171 .0038 4.46 0.000
Agriculture -.1260 .8155 -0.15 0.878
Construction -.3017 .1311 -2.30 0.024
Manufacturing -.0509 .0886 -0.58 0.567
Transportation -.0760 .1831 -0.42 0.679
Wholesale Trade .3016 .2485 1.21 0.228
Retail Trade .2095 .1911 1.10 0.276
FIRE .3239 .1349 2.40 0.018
Services -.0136 .0960 -0.14 0.888
Government -.0199 .0984 -0.20 0.840
Constant .3715 .0814 4.56 0.000
Table 11
Avg. Earnings (US) by Industry
% Change % Change
1978 1988 1998 1978-1988 1988-1998
All 11,827 21,530 31,411 82.0% 45.9%
Agricultural 8,286 14,360 19,912 73.3% 38.7%
Mining 18,688 34,481 51,849 84.5% 50.4%
Const. 14,574 24,529 33,484 68.3% 36.5%
Manufact. 14,519 27,208 39,997 87.4% 47.0%
Transport. 16,993 29,185 39,921 71.7% 36.8%
Wholesale 14,999 27,874 41,883 85.8% 50.3%
Retail trade 7,641 12,306 17,075 61.1% 38.8%
FIRE 12,344 28,146 48,776 128.0% 73.3%
Services 9,301 19,252 29,243 107.0% 51.9%
Gov't. 11,898 21,542 31,016 81.1% 44.0%
Avg. %
Change
All 64.0%
Agricultural 56.0%
Mining 67.4%
Const. 52.4%
Manufact. 67.2%
Transport. 54.3%
Wholesale 68.0%
Retail trade 49.9%
FIRE 100.7%
Services 79.4%
Gov't. 62.5%
Table 12
US Average Earnings by Services Sub-Sectors
1978 1988 1998
All Sectors $ 11,828 $ 21,531 $ 31,411
Services $ 9,301 $ 19,253 $ 29,244
Hotels/lodging $ 7,168 $ 13,375 $ 20,017
Personal services $ 7,621 $ 12,554 $ 18,253
Private Household $ 3,388 $ 5,489 $ 10,656
Business services $ 10,357 $ 17,825 $ 30,020
Automotive repair $ 10,079 $ 17,322 $ 23,706
Misc repair $ 11,625 $ 20,950 $ 29,451
Amusement/rec $ 7,917 $ 14,616 $ 22,381
Motion pictures $ 11,521 $ 20,843 $ 32,051
Health services $ 10,972 $ 22,865 $ 32,697
Legal services $ 14,327 $ 37,049 $ 53,603
Educ. services $ 8,688 $ 16,155 $ 23,774
Social services $ 6,776 $ 11,916 $ 17,349
Museums, $ 8,323 $ 15,050 $ 21,441
Membership orgs $ 8,046 $ 14,612 $ 21,049
Eng./Mngmt. N/A $ 32,332 $ 48,929
Miscellaneous $ 16,091 $ 41,559 $ 56,556
Figure 1--Gini Ratio (All US Households)
Year Gini Ratio
1979 0.415
1989 0.445
1999 0.463
Source: Data from Jones & Weinburg (2000, p. 3)
Note: Table made from bar graph.
