Wage inequality and offshoring: are they related?
Ghosh, Koushik ; Saunders, Peter J. ; Tenerelli, Thomas 等
Abstract
The objective of this paper is to investigate the impact of
offshoring on wage inequality and labor productivity in the U.S.
Short-run and long-run data tests are undertaken to analyze the
relationship among offshoring, wage inequality, and labor productivity
in the U.S. Cointegration tests indicate that these three variables are
related in the long-run. The main contribution of this paper lies in its
focus on the short-run investigation of the relationship among these
three variables. This investigation is conducted using the vector error
correction (VEC) testing framework. VEC tests indicate that offshoring
has had a statistically significant impact on both labor productivity
and wage inequality in the U.S.
Key Words: offshoring, wage inequality, and labor productivity.
JEL Classification: F16
I. INTRODUCTION
Analyzing and explaining the causes of wage inequality have long
been focuses of economic inquiry. Although this issue has been addressed
thoroughly in both theoretical discussions and empirical research, there
is no commonly accepted explanation of the causes of changing wage
inequality. The beginning of this discussion can be traced to the
writings of Adam Smith (1776) who outlined how trade and specialization
can transform societies. Since the first industrial revolution, which
transformed agricultural societies to manufacturing ones, there have
been two more such events. The second revolution was characterized by a
movement from manufacturing to service industries. Currently most
countries are experiencing the third industrial transformation which is
characterized by, among other things, offshoring (commonly referred to
as outsourcing). One of the key concerns of this latest economic
development is its impact on income distribution and wage inequality in
all of the countries operating in the global market place. While the
analyses of the effect of increased offshoring on wage inequality are
potentially ambiguous, this issue remains vigorously alive due to a
paucity of empirical studies and theoretical discussions.
Offshoring typically involves trade in tasks and goods among
participating countries. Trade in tasks occurs when one or more portions
of the production process are offshored. This type of international
trade can have a negative or positive impact on the wages of low-skilled
workers in countries where tasks are outsourced abroad. Grossman and
Rossi-Hansberg (2006 and 2008) defined this issue elegantly. Their
analysis invites further empirical research that can shed additional
light on the impact of offshoring on wage inequality. The authors point
out that trade in tasks rather than trade in goods characterizes the
above mentioned new industrial revolution. They suggest that such trade
is not necessarily detrimental to wages of low-skilled workers. The
authors also point out, however, that their empirical analysis is
relatively crude and should be fine-tuned.
Clearly, controversies surround the issue of the impact of
offshoring on income inequality in countries that participate in
international trade. In order to assess this impact, it is important to
outline initially the key features of the new industrial revolution that
is characterized by trade in tasks. Offshoring has important effects on
U.S. imports and, thereby, it can impact income distribution in the U.S.
It affects not only the volume of U.S. imports, but also their content.
Using OECD data, Grossman and Rossi-Hansberg (2006 and 2008) have
calculated the estimated share of imported inputs in total inputs used
by all goods-producing sectors in the United States. Their estimates
indicate that the share of imported inputs in the gross output of these
sectors has been growing steadily over the last three decades. They find
evidence of an acceleration of this trend after 1995.
Trade in tasks, that is to say offshoring, has played an
increasingly important role in international trade. (1) This type of
trade can have a considerable impact on the wages of skilled and
unskilled workers. The primary objective of this paper is to analyze the
impact of offshoring on income inequality in the U.S. This category of
trade is likely to grow even more in the coming years, as more
"routine" cognitive tasks are increasingly exported overseas
[Autor, D., Levy F., and Murnane, R (2003)]. Offshoring has become an
important feature of today's global economy, and it will continue
to be so. How offshoring affects wages and income inequality in trading
countries is unclear. Only empirical research, such as the present
study, can provide some answers to this critically important and not yet
resolved issue.
In order to understand better the impact of offshoring on income
inequality, it is also essential to analyze the potential effects of
offshoring on the labor supply. Learner (2006) and others have described
how increased opportunities for offshoring can lead to an expansion in
the world supply of low-skilled labor. Citing the properties of the
Heckscher-Ohlin (1933) model with incomplete specialization, Grossman
and Rossi-Hansberg (2006 and 2008) develop a simplified version of their
general equilibrium model that eliminates the relative price effect of
trade. The two authors suggest that the expansion in the world supply of
low-skilled labor may not affect factor prices, since factor growth can
be accommodated without an impact on factor prices. Thus, there need not
be a depressing effect on domestic wages of low-skilled workers, even if
Leamer's (2006) hypothesis is correct.
Grossman and Rossi-Hansberg (2006 and 2008) outline a new paradigm
in trade theory. As mentioned previously, the two authors suggest that
international trade theory should focus on trade in tasks rather than
trade in goods. Given this assertion, they investigate the impact of
offshoring on wages of high-skilled and low-skilled labor. The two
authors suggest that as the cost of offshoring decreases, firms move
L-tasks (low-skill tasks) abroad, thus increasing both productivity (due
to the decreased costs associated with the already offshored L-tasks) as
well as the supply of low-skilled labor in the economy. Productivity
improvement is primarily due to the decreased costs associated with the
already offshored L-tasks. The authors further suggest that the positive
productivity effect on demand for low skilled labor that raises their
wages may indeed dominate the negative labor supply effect on these
wages. This would then imply that the wages of low-skilled workers would
not be affected negatively by the decrease in offshoring costs. This
would also mean that offshoring would not worsen wage inequality. In
their 2006 empirical analysis, the authors conclude that "the data
leave room for a positive effect of offshoring on wages" (p. 30).
These observations notwithstanding, the authors point out that there are
several omitted factors in their analysis, and that their conclusions
"be taken with a grain of s alt until a more thorough empirical
study can be performed." (p. 31).
In this paper, we accept the above stated empirical challenge and
subject the Grossman and Rossi-Hansberg (2006 and 2008) hypothesis to
empirical tests using both long-run and short-run analyses of wage and
trade data. The novelty of our empirical inquiry lies not only in its
focus on analyzing the long-run as well as short- run relationship among
wage inequality and international trade, but also on its analysis of the
impact of offshoring on labor productivity. Therefore, the present paper
provides empirical evidence for both key issues raised by Grossman and
Rossi-Hansberg: the effects of offshoring on labor productivity and on
wage inequality.
There is one additional important objective of our present
research. Our paper is an extension and a further refinement of our
previously published research on the relationship between trade, wage
inequality, and productivity in the U.S. [Ghosh, Saunders, and Biswas
(2000), and Ghosh, Saunders, and Biswas (2002)]. The objective of our
2000 paper was to investigate the relationship between wage
differentials of unskilled and skilled labor (approximated by the
differences between the median incomes of males who completed high
school and the median incomes of four-year college graduates) and trade
(approximated by net exports). We found that trade in the U.S. is
negatively impacted by wage inequality. In our 2002 paper, we expanded
our investigation into the relationship between trade and income
inequality in the U.S. by including the effects of labor productivity
(measured by the output per hour of all persons in the non-farm business
sector) on these variables. We found that while wage differentials and
net exports had a statistically significant impact on labor
productivity, wage inequality was not impacted by the combined impact of
trade and labor productivity. While our present paper is a natural
extension of our previously published research, it adds two new
important dimensions to our previous empirical analyses. First, the
focus is on analyzing the effects of offshoring on wage inequality and
labor productivity/Second, by expanding our empirical analyses
throughout 2011, it is possible to find out whether the basic
relationships between trade, productivity, and wage inequality have
changed in the U.S. in the last decade.
All of the objectives described above are accomplished within a
trivariate time-series testing framework. Our empirical investigation,
below, is divided into four sections. In section II, a thorough
literature review concerning the effects of trade and other factors on
wage inequality is undertaken. In section III, the data and the
methodology used to investigate the relationship among these variables
are outlined. The test results are described and analyzed in section IV.
Section V concludes our paper with final remarks about the relationship
between offshoring, wage inequality, and labor productivity in the U.S.
II. LITERATURE REVIEW
The literature on causes of wage inequality is a rich and varied
one. Several explanations have been posited for the large and growing
inequality in the United States [Katz and Autor, (1999)]. Most
explanations fall under four broad categories: changes in relative
demand for high-skilled versus low-skilled labor [Autor, Katz, and
Kearney, (2008); Juhn, Murphy, and Pierce, (1993)]; changes in the
relative supply of high-skilled versus low-skilled labor; institutional
changes in the labor force; and compositional changes between
high-skilled and low-skilled workers. Common demand based explanations
are international trade and technological change (and related
organizational changes). Common supply based explanations are changes in
college graduation rates [Card and Lemieux, (2001); Goldin and Katz,
(2009); Katz and Murphy, (1992)] and immigration [Altonji and Card,
(1991); Borjas, (1995 and 2003); and Card, (2009)]. Institutional
explanations [DiNardo, Fortin, and Lemieux (1996)] include minimum wage
changes [Lee, (1999)] and changes in unionization [Kahn (2000)].
Finally, compositional explanations [Lemieux (2006)] claim that changes
in inequality are unrelated to prices of high-skilled versus low-skilled
labor, and are, rather, due to shifts in the relative quantities of
high-skilled versus low-skilled labor.
The wage inequality literature has provided two main explanations
for wage differentials. The first explanation is that demand, supply,
and the decline in labor market institutions (unions and the minimum
wage) were important components of the rise in inequality in the
1980's [Autor, Katz, and Kearney (2008)]. The second explanation is
that continued increases in the relative demand for high-skilled labor,
in the face of a stagnant relative supply of high-skilled labor, have
played an important role in increasing wage differentials since the
1980's [Autor, Katz, and Kearney (2008); Goldin and Katz (2009);
and Katz and Autor (1999)]. However, even among those who accept a
demand based argument for continued wage inequality, a vigorous debate
about the ultimate cause of the rise in relative demand for high-skilled
labor remains--with international trade and technology as the primary
candidates. The technology argument posits that skill-biased
technological change has occurred that favors high-skilled workers and
thus raises their wage relative to low-skilled workers. The literature
that emphasizes technology as an explanation for increased wage
inequality has focused on three observations. First, that employment
shifts towards high-skill intensive industries have been small relative
to employment shifts towards high-skilled employment within industries
]Bound and Johnson (1992); Berman, Bound, and Griliches (1994); and
Berman, Bound, and Machin (1998)]. The argument is that this observation
is largely inconsistent with the Heckscher-Ohlin (1933) framework, where
the lowering of trade barriers should lead to an expansion of the
high-skill-intensive sector. Second, there has been significant within
industry substitution towards high-skilled labor despite an increase in
the relative wage of high-skilled labor [Bound and Johnson (1992);
Berman, Bound, and Griliches (1994); Berman, Bound, and Machin (1998);
and Lawrence and Slaughter (1993)]. Third, that measures of
computerization are associated with relative increases in high-skilled
employment [Autor, Katz, and Krueger (1998); and Autor, Levy, and
Murnane (2003)].
Conventional trade theory, which rests on the Heckscher-Ohlin
(1933) model and the Stolper-Samuelson (1941) theorem, also provides a
plausible explanation of wage inequality. According to conventional
trade theory, increased trade with developing countries results in
increased wage inequality in developed countries. However, this
hypothesis has not been uniformly supported in the empirical literature.
Empirical investigation of the impact of trade on wage inequality has
left this issue largely unresolved. Most studies up to date have found
only modest effects of trade on wage inequality [Edwards and Lawrence
(2008); Feenstra and Hanson (1999); Krugman (1995); and Liu and Trefler
(2008)].
It is clear from the above literature review of both theoretical
explanations and empirical research on the causes of wage inequality
that this issue is far from being resolved. We hope that our present
research can provide additional information on this unresolved yet
important issue. The focus of our paper is on analyzing the impact of
offshoring on wage inequality and labor productivity in the U.S. To
accomplish this objective, we deploy reduced form modeling of the
time-series data.
III. DATA AND METHODOLOGY SELECTION
The data selection is determined by the hypotheses under empirical
investigation. As stated above, the objective of our paper is to
investigate the effects of offshoring on wage inequality and labor
productivity in the U.S. One obvious way to measure the wage inequality
is by computing the wage differentials between the L and the H types of
labor wages. In this paper, the wages of L-tasks (those performed by low
skilled labor) are approximated by the total private industry average
weekly earnings of production and nonsupervisory employees (AWEP), while
the wages of H-tasks (those performed by high-skilled labor) are
measured by the average weekly earnings of production and nonsupervisory
employees in the manufacturing sector of the U.S. (AWEM). The
differences between these two wages (WD) are used as the measure of
income inequality.
The selection of a variable that can be used to approximate
offshoring is a challenging task, because there is no uniquely developed
and accepted measure of offshoring. One way to obtain a quantitative
measure of offshoring is to focus on the relationship between exports
and imports through their impact on the terms of trade (TOT). When the
terms of trade for the U.S. improves, it implies that the goods and
services that the U.S. specializes in selling abroad (U.S. exports) by
using its abundant factor (skilled labor) are experiencing a price
advantage over the goods and services that countries with abundant
unskilled labor are exporting to the U.S. (U.S. imports). Consequently,
TOT reflects the relative prices of goods and services that are H-type
goods and services versus L-type goods and services. Therefore, TOT is a
measure of the incentive to offshore, and hence, it can be used to
approximate the impact of offshoring on wage inequality in the U.S. (3)
Using this approach to measure the impact of offshoring on income
inequality makes it possible to test empirically the Grossman and
Rossi-Hansberg (2006 and 2008) hypothesis. In this paper, the TOT
variable is computed as the ratio of the BEA end of use export to import
indexes. The selection of the labor productivity variable (PRODL) is
straight-forward. It is approximated by the output per hour in the
nonfarm business sector of the U.S. (4)
A further novelty of our research lies in its selection of the
empirical methodology and testing specifications. The impact of
offshoring on the wage inequality and labor productivity is investigated
within a time-series testing framework. Quarterly data ranging from the
first quarter of 1990 to the third quarter of 2011 are used in all
subsequent data analyses. (5) The time-series testing framework requires
that several steps be undertaken in econometric data analyses. Initially
it is necessary to determine the stationarity or nonstationarity of each
individual time-series variable. This determination is based upon the
results of unit root tests. These tests determine the order of
integration of each time-series variable under empirical investigation.
If all time-series data are found to be integrated of order one I(1),
then it is possible to find out if a long-run relationship exists among
all of the test variables. This objective is accomplished within a
cointegration testing framework. Cointegration test results determine
the next step in the time-series data analyses. In particular,
cointegration test determine whether vector error correction (VEC) or
vector autoregression (VAR) testing framework is appropriate for further
data analyses.
IV. INTEGRATION, COINTEGRATION, AND VEC TEST RESULTS
All the above outlined steps were followed in the present research.
Initially all the data were subjected to unit root testing. As stated
above, the objective of unit root tests is to determine the degree of
integration of each individual time-series data. (6) Numerous unit root
tests can be used to make this determination. The Phillips-Perron (1988)
(PP) and the Augmented Dickey-Fuller (1976 and 1979) (ADF) tests are the
most commonly used procedures to test the stationarity of time-series
data. In the present paper, the Phillips-Perron test was used to make
this determination. The results of this test are reported in Table 1.
Unit root test results may be sensitive to a particular test
specification, or an arbitrary lag selection. These tests can be
conducted with or without the inclusion of the trend variable. In order
to test the robustness of the PP test, both test specifications were
examined. The PP tests indicate that all test variables, AWEP, AWEM, WD,
PRODL, EXPORTS, IMPORTS, and TOT are nonstationary, and I(1).
Given the fact that all individual time-series are I(1), it is
possible that these variables are related in the long-run. This
information would provide crucial information about the long-run
relationship among offshoring, income inequality, and labor productivity
in the U.S. This determination can be made by deploying cointegration
tests. Although there exist numerous cointegration tests, all of these
tests have one common objective--to find the most stationary linear
combination of the vector time-series. The most commonly used
cointegration tests are the Engle-Granger (1987) test, the Stock and
Watson (1988) test, and Johansen's (1988) procedure.
Johansen's test appears to have several statistical advantages over
the other above mentioned cointegration tests as noted by Gonzalo
(1994). The superior statistical properties of Johansen's test
include its ability to include all prior knowledge about the existence
of unit roots in the time-series data under investigation, as well as
the maximum likelihood estimation method that results in coefficient
estimates that are symmetrically distributed and asymptotically
efficient. Additionally, Johansen's method performs better in cases
of non-normal error distribution and where the dynamics of the model
under investigation are not known. Given the superior statistical
properties of Johansen's cointegration test, it was used to analyze
the long-run relationships between PRODL, WD, and TOT. Test results are
summarized in Table 2 below.
Johansen's (1988) cointegration test results outlined in the
above Table 2 imply that PRODL, WD, and TOT are cointegrated. This
conclusion is reached by analyzing both the trace and the eigenvalue
statistics. In both of these cases, the two test statistics (30.658,
22.664 and 32.226,23.411) are statistically significant at the
conventional five-percent level. Additionally, the likelihood test
indicates the existence of one cointegrating equation at the
five-percent statistical significance level in both cases. The
normalized cointegrating coefficient for PRODL, WD, and TOT is {1.00,
-0.711,237.379} in the 1-2 lag case, and {1.00, -0.810, 217.036} in the
1-4 lag test case. These cointegration tests provide important
information about the relationship between offshoring, wage inequality,
and labor productivity in the U.S. Clearly, these variables are related
in the long-run. Consequently, it is fair to conclude that outsourcing
tasks in goods and services may impact wage inequality and the labor
productivity in the U.S. in the long-run. However, cointegration tests
alone cannot determine the direction of causal flow among the three test
variables. This information may be obtained by analyzing the short-run
dynamics of the relationship between PRODL, WD, and TOT. This analysis
can be undertaken within a VEC estimation framework.
One additional hurdle needs to be overcome in time-series data
analyses. This problem involves lag determination in data testing. Lag
selection has to be made in most time-series data testing procedures. In
some test cases, such as the ADF (1976 and 1979) unit root test, this
selection can be made automatically using numerous criteria, such as the
SIC and the AIC criterion. In other time-series testing procedures, such
as Johansen's (1988) cointegration test, lags can only be selected
arbitrarily. An arbitrary lag selection can influence test results. This
fact is well documented in economic literature [Thornton and Batten
(1985), Hsiao (1979 and 1981), Saunders (1988), and others]. One way to
mitigate this problem is by investigating more than one lag test
structure, and determining whether test results are affected by doing
so. Consequently, Johansen's (1988) cointegration test was repeated
using the second lag structure (1-4 lags). The 1-4 lag specification
test results are reported in Table 3. Clearly, these results were
unaffected by this particular lag specification, as all variables were
still found to be cointegrated. This fact attests to the robustness of
the present model under investigation.
Cointegration test results determine the next step in the empirical
investigation of the relationship between offshoring, wage inequality,
and labor productivity in the U.S. Since the three variables are
cointegrated, the VEC testing framework is the appropriate procedure to
be deployed in the further analyses of this relationship. The main
objective of the VEC modeling is to shed some light on the short-run
dynamics among any number of time-series variables that are
cointegrated. VEC modeling determines whether the system under empirical
investigation is in a state of short-run equilibrium or disequilibrium.
Additionally, in the present case, VEC estimates can provide some
information on the short-run impact of offshoring on wage inequality and
labor productivity in the U.S.
We adopt the Engle-Granger (1987) approach in VEC modeling of our
time-series data. There are two important reasons for adopting this
particular VEC testing procedure. First, the Engle-Granger VEC technique
is ideally suited for use with Johansen's (1988) cointegration
test. If the null hypothesis of cointegration of a group of variables is
rejected by Johansen's test, that is if it is determined that such
variables are cointegrated (as in the present case), then the residuals
from Johansen's cointegration test can be used to estimate the
Engle-Granger VEC model. In the present study, the saved residuals from
the cointegrating equations reported in Table 2 above were used in
subsequent VEC data modeling. (7) Second, using this particular form of
VEC estimation allows us to make meaningful comparisons with our
previously reported conclusions on the relationships between trade,
income inequality, and labor productivity in the U.S. Given these two
considerations, we adopt the Engle-Granger VEC approach in our present
trivariate analyses of the TOT, WD, and PRODL data. Therefore, all of
our test equations are specified as follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
The results of estimations of equations (1)-(3) are reported in
Tables 3 and 4 below.
The estimation results of equations (1)-(3) are outlined in above
Tables 3 and 4. Although the same VEC estimation procedures are followed
in both cases, the lag length is varied from 1-2 to 1-4 lags. The main
reason for adding an additional lag specification is to test the
robustness of our model. As mentioned previously, arbitrary lag
selection can influence time-series tests results. Alleviating this
potential problem, as we have done in the present case, adds to the
strength of our statistical analyses as it is evident that varying the
lag structure does not affect the key test results. An analysis of the
estimates reported in the above two tables provides crucial information
about how offshoring impacts wage inequality and labor productivity in
the U.S. The focus of this analysis must be on the lagged [z.sub.t]
terms in all equations under investigation. The lagged [z.sub.t] terms
are the speed of adjustment residuals from Johansen's cointegrating
tests carried out previously. Engle and Granger (1987) outline the
methodology in this type of VEC data modeling. We follow their approach
in our VEC data modeling. Essentially, VEC estimation determines whether
the system under empirical investigation is in a state of short-run
equilibrium or disequilibrium. This determination is based upon the
statistical analysis of the lagged [z.sub.t] terms. Conventional
"t" tests are used to make this determination. Lagging these
terms implies that disturbances of the last period may impact the
current period. In general, finding a statistically insignificant
coefficient of the [z.sub.t] term implies that the state of the
short-run equilibrium exists and there are no disturbances present. If
this coefficient is statistically significant, then a state of short-run
disequilibrium exists. In such a case, the coefficient estimate gives an
indication of the size and the direction of the impact of explanatory
variables on the dependent variable in the short-run. In the present
case, this type of statistical analysis can provide crucial information
about the effects of offshoring on wage inequality and labor
productivity in the U.S.
The focus of the analysis is on the estimation of equations (1) and
(2). These estimates provide crucial information about the effects of
offshoring on wage inequality and labor productivity in the U.S. The
coefficient of the lagged [z.sub.t] term in equation (1) is
statistically significant and positive. One obvious interpretation of
this result is to conclude that offshoring and productivity growth have
had a statistically significant impact on wage inequality in the U.S.
Since the WD variable measures the gap between skilled and unskilled
workers' wages, it would seem that offshoring has widened this gap.
Equation (2) provides additional empirical evidence on the combined
impact of offshoring and wage differentials on labor productivity. Since
the coefficient of the lagged [z.sub.t] term is both statistically
significant and positive, this result implies that this impact is
positive.
Equations (1) and (2) allow an empirical investigation of the
previously outlined trade theorists' and labor economists'
hypotheses about the causes of wage inequality to be undertaken in one
combined model. Trade theorists maintain that trade can increase wage
inequality. Trade affects wages of workers in exporting and importing
sectors of an economy because it affects the prices of exports and
imports. However, according to labor economists, wage differentials are
primarily due to technological and productivity advances. Our VEC
estimation results seem to support both of these hypotheses. They
indicate that wage inequality has worsened due to the combined effect of
offshoring and labor productivity changes. However, our present
study's results do not support the Grossman Rossi-Hansberg (2006
and 2008) hypothesis about the impact of offshoring on wage inequality.
It appears that although offshoring positively impacts labor
productivity growth, it also widens wage inequality in the U.S. Our
results imply that the positive productivity effect on wage inequality
does not dominate the negative labor supply effect.
It is also worth noting that offshoring has changed the
implications about the impact of trade on wage inequality that we
reached in our previous research of this topic [Ghosh K., and Saunders
P.J. (2002)]. In our 2002 study, we used the same empirical methodology
to investigate the relationship between wage inequality, trade, and
labor productivity. Our finding of a positive combined impact of trade
and wage differentials on labor productivity was consistent with the
results found in the present research. However, in our 2002 paper, we
found no statistically significant combined impact of labor productivity
changes and trade (approximated by net exports) on wage differentials
(measured by the difference between the median incomes of
bachelor's degree holders and those with only high school degrees).
There are two likely explanations for the different conclusions reached
in our present study. First, since our focus is specifically on
investigating the impact of offshoring on wage inequality, the TOT
variable used to approximate this effect is different from the net
exports variable used in our previous study. Second, it is entirely
possible that nine additional years of U.S. trade deficits and ever
growing globalization may have changed the basic relationship between
trade and wage inequality in the U.S. In any case, it appears that
offshoring has lead to increasing income inequality in the U.S.
V. CONCLUDING REMARKS
Our paper investigates the impact of offshoring on wage inequality
and productivity changes in the U.S. There are two key motivations for
our empirical investigation of this topic. First, we provide an
empirical framework for testing a new theory about the impact of
offshoring on wage inequality outlined by Grossman and Rossi-Hansberg
(2006 and 2008). According to their hypothesis, offshoring may not
affect negatively the wages of low-skilled workers, because its positive
productivity effect may dominate the negative labor supply effect and
price effect. Therefore, according to the two authors, offshoring may
not worsen wage inequality in the U.S. Second, our present research is a
natural extension and a refinement of our previously published research
on the relationship between wage inequality, trade, and labor
productivity.
Our present study uses quarterly data ranging from the first
quarter of 1990 to the third quarter of 2011 to analyze the impact of
offshoring on income inequality and productivity changes in the U.S.
Income inequality is measured by computing the wage differences between
the L and the H types of workers. Low skilled labor (L-tasks) wages are
approximated by the total private industry average weekly earnings of
production and nonsupervisory employees (AWEP). High skilled labor
(H-tasks) wages are measured by the average weekly earnings of
production and nonsupervisory employees in the manufacturing sector of
the U.S. (AWEM). The differences between these two wages (WD) are used
as the measure of income inequality in the U.S. One of the key
contributions of our research to the ongoing debate about the effects of
offshoring on income inequality is our development of the measure of
offshoring that can be used to test empirically various outsourcing
hypotheses. We approximate offshoring by the TOT variable that is
computed as the ratio of the BEA end of use export to import indexes.
The TOT variable determines how the gains from trade are distributed
among trading countries. Since TOT reflects the relative prices of
H-type goods and services versus L-type goods and services, this
variable is well suited to measure the distribution of gains from trade
between skilled and unskilled labor in the U.S. Therefore, the TOT
variable is an ideal measure of the impact of offshoring on income
inequality in the U.S. Furthermore, using this measure of income
inequality allows us to test empirically the Grossman and Rossi-Hansberg
(2006 and 2008) hypothesis about the impact of offshoring on low skilled
labor wages in the U.S. The labor productivity variable (PRODL) is
approximated by output per hour in the U.S. nonfarm business sector.
The time-series methodology used in our paper consists of unit
root, cointegration, and vector error correction (VEC) estimation. The
Phillips-Peron (1988) unit root tests indicate that all test variables
are nonstationary, and integrated of order one I(1). Given this outcome,
we deploy Johansen's (1988) cointegration test to analyze the long
term relationships among WD, TOT, and PRODL. The cointegration test
indicates the existence of a stable long-run relationship among these
three variables. These results suggest that offshoring may have a long
term impact on both wage inequality and labor productivity in the U.S.
However, cointegration tests alone cannot make this determination. VEC
estimation can accomplish this objective.
The key contribution of our paper to the ongoing debate of
outsourcing on income inequality and labor productivity lies with its
VEC data analyses. We use the Granger and Engle (1988) VEC test to
analyze the short-run dynamics of the relationship among WD, TOT, and
PRODL. Our test results indicate that outsourcing has had a positive
impact on labor productivity in the short-run. Consequently, our results
support the arguments about the effects of outsourcing on labor
productivity made by both labor and trade economists. They are also
consistent with the conclusions reached in our 2002 research. When
analyzing the impact of offshoring on income inequality, we find that
outsourcing worsens income inequality in the U.S. Therefore, the results
of our present study do not support the new hypothesis about the effect
of offshoring on income distribution proposed by Grossman and
Rossi-Hansberg (2006 and 2008). It appears that outsourcing harms
low-skilled workers in the U.S.
References
Altonji, J. G. and D. Card (1991), "The Effects of Immigration
on the Labor Market Outcomes of Less-skilled Natives," in
Immigration, Trade, and Labor, J.M. Abowd and R.B. Freeman, eds.
Chicago: University of Chicago Press.
Autor, D. H., L.F. Katz, and M.S. Kearney (2008), "Trends in
U.S. Wage Inequality: Revising the Revisionists," Review of
Economics and Statistics, Vol. 90, No. 2, 300-323.
Autor, D. H., L. F. Katz, and A.B. Krueger (1998), "Computing
Inequality: Have Computers Changed the Labor Market?," The
Quarterly Journal of Economics, Vol. 113, No.4, 1169-1213.
Autor, D. H., F. Levy, and R. J. Murnane (2003), 'The Skill
Content of Recent Technological Change: An Empirical Exploration,"
Quarterly Journal of Economics, Vol. 118, No. 4, 1279-1333.
Berman, E., J. Bound, and Z. Griliches (1994), "Changes in the
Demand for Skilled Labor within U.S. Manufacturing: Evidence from the
Annual Survey of Manufactures," The Quarterly Journal of Economics,
Vol. 109, No. 2, 367-397.
Berman, E., J. Bound, and S. Machin (1998), "Implications of
Skill-Biased Technological Change: International Evidence," The
Quarterly Journal of Economics, Vol. 113, No. 4, 1245-1280.
Borjas, G. J. (2003), "The Labor Demand Curve Is Downward
Sloping: Reexamining the Impact of Immigration on the Labor
Market," The Quarterly Journal of Economics, Vol. 118, No. 4,
1335-1374.
Borjas, G. J. (1995), "The Economic Benefits from
Immigration," The Journal of Economic Perspectives, Vol. 9, No. 2,
3-22.
Bound, J. and G. Johnson (1992), "Changes in the Structure of
Wages in the 1980's: An Evaluation of Alternative
Explanations," American Economic Review, Vol. 82, No. 3, 371-392.
Card, D. and T. Lemieux (2001), "Can Falling Supply Explain
the Rising Return to College for Younger Men? A Cohort-Based
Analysis," The Quarterly Journal of Economics, Vol. 116, No. 2,
705-746.
Card, D. (2009), "Immigration and Inequality," American
Economic Review, Vol. 99, No. 2, 1-21.
Dickey, D. A. and W.A. Fuller (1979), "Distribution of the
Estimators for Autoregressive Time Series with a Unit Root,"
Journal of the American Statistical Association, Vol. 74, No. 366,
427-431.
DiNardo, J., N. M. Fortin, and T. Lemieux (1996), "Labor
Market Institutions and the Distribution of Wages, 1973-1992: A
Semiparametric Approach," Econometrica, Vol. 64, No. 5, 1001-44.
Edwards, L. and R. Z. Lawrence (2010), "U.S. Trade and Wages:
The Misleading Implications of Conventional Trade Theory," NBER
Working Paper no. 16106.
E., Walter (1995), Applied Econometric Time Series, John Wiley
& Sons, Inc., New York.
Engle, R. F. and C. W. J. Granger (1987), "Cointegration and
Error-Correction: Representation, Estimation and Testing,"
Econometrica, Vol. 55, No. 2, 251-276.
Feenstra, R. C. (1998), "Integration of Trade and
Disintergration of Production in the Global Economy," Journal of
Economics, Vol. 12, No. 4, 31-50.
Feenstra, R. C. and G. H. Hanson (1999), "The Impact of
Outsourcing and High-Technology Capital Wages: Estimates for the U.S.,
1972-1990," The Quarterly Journal of Economics, Vol. 114, No. 3,
907-940.
Fuller, W. A. (1976), Introduction to Statistical Time Series, John
Wiley & Sons, Inc., New York.
Ghosh, K., P. J. Saunders, and B. Biswas (2000), "Trade and
Wage Inequality: Are they Related?," Atlantic Economic Journal,
Vol. 28, No. 3, 364-374.
Ghosh, K., P. J. Saunders, and B. Biswas (2002), "An Empirical
Investigation of the Relations Between Wage Differentials, Productivity
Growth, and Trade," Contemporary Economic Policy, Vol. 20, No. 1,
83-92.
Goldin, C. and L. F. Katz (2009), "The Race between Education
and Technology: The Evolution of U.S. Educational Wage Differentials,
1890 to 2005," updated chapter 8 of The Race between Education and
Technology.
Gonzalo, J. (1994), "Five Alternative Methods of Estimating
Long-Run Equilibrium Relationships," Journal of Econometrics, Vol.
60, No. 1-2, 203-233.
Grossman, G. M., and E. Rossi-Hansberg (2006), "Trading Tasks:
A Simple Theory of Offshoring," NBER Working Paper no. 12721.
Grossman, G. M., and E. Rossi-Hansberg (2008), "Trading Tasks:
A Simple Theory of Offshoring," American Economic Review, Vol. 98,
No. 5, 1978-1997.
Heckscher, E. (1919), "The Effects of Foreign Trade on the
Distribution of Income," Ekonomisk Tidskrift, Vol. 21, 497-512.
Holden, K. and J. Thompson (1992), "Co-Integration: An
Introductory Survey," British Review of Economic Issues, Vol. 14,
No. 33, 1-55.
Hsiao, C. (1979), "Autoregressive Modeling of Canadian Money
and Income Data," Journal of the American Statistical Association,
Vol. 74, No. 367, 553-560.
Hsiao, C. (1981), "Autoregressive Modeling and Money-Income
Causality Detection," Journal of Monetary Economics, Vol. 7. No. 1,
85-106.
Johansen, S. (1988), "Statistical Analysis of Cointegrating
Vectors," Journal of Economic Dynamics and Control, Vol. 12, No. 2,
231-54.
Juhn, C., K. M. Murphy, and B. Pierce (1993), "Wage Inequality
and the Rise in the Returns to Skill." Journal of Political
Economy, Vol. 101, No. 3, 410-442.
Kahn, L. M. (2000), "Wage Inequality, Collective Bargaining,
and Relative Employment from 1985 to 1994: Evidence from Fifteen OECD
Countries," Review of Economics and Statistics, Vol. 82, No. 4,
564-579.
Katz, L. F. and K. M. Murphy (1992), "Changes in Relative
Wages, 1963-87: Supply and Demand Factors," The Quarterly Journal
of Economics, Vol. 107, No. 1, 35-78.
Katz, L. F. and D. H. Autor (1999), "Changes in the Wage
Structure and Earnings Inequality," in O. Ashenfelter and D. Card,
eds., Handbook of Labor Economics, Vol. 3A, 1463-1555.
Krugman, P. R. (1995), "Growing World Trade: Causes and
Consequences," Brookings Papers on Economic Activity, Vol. 1995,
No. 1, 327-377.
Lawrence, R. Z. and M. J. Slaughter (1993), "International
Trade and American Wages in the 1980s: Giant Sucking Sound or Small
Hiccup," Brookings Papers on Economic Activity: Microeconomics,
Vol. 1993, No. 2, 161-226.
Leamer, E. E. (2007), "A Flat World, A Level Playing Field, a
Small World After All, or None of the Above?," Journal of Economic
Literature, Vol. 45, No. 1, 83-126.
Lee, D. S. (1999), "Wage Inequality in the U.S. during the
1980s: Rising Dispersion or Falling Minimum Wage?," The Quarterly
Journal of Economics, Vol. 114, No. 3, 977-1023.
Lemieux, T. (2006), "Increasing Residual Wage Inequality:
Composition Effects, Noisy Data, or Rising Demand for Skill?,"
American Economic Review, Vol. 96, No. 3, 461-98.
Liu, R. and D. Trefler (2008), "Much Ado About Nothing:
American Jobs and the Rise of Service Offshoring to China and
India," NBER working paper no. 14061.
McCallum, B. T. (1993), "Unit Roots in Macroeconomic Time
Series: Some Critical Issues," Federal Reserve Bank of Richmond
Economic Quarterly, Vol. 79, No. 2, 13-43.
Ohlin, B. (1933), Interregional and International Trade, Harvard
University Press, Cambridge.
Phillips, P. C. B. and P. Perron (1988), "Testing for a Unit
Root in Time Series Regression," Biometrika, Vol. 75, No. 2,
335-346.
Saunders, P. J. (1988), "Causality of the U.S. Agricultural
Prices and the Money Supply," American Journal of Agricultural
Economics, Vol. 70, No. 3, 588-96.
Smith, A. (1776), The Wealth of Nations, reprinted in 1937, Random
House, New York.
Stock, J. H. and M.W. Watson (1988), "Testing for Common
Trends," Journal of the American Statistical Association, Vol. 83,
No. 404, 1097-1107.
Stolper, W. F. and P.A. Samuelson (1941), "Protection and Real
Wages," The Review of Economic Studies, Vol. 9, No. 1, 58-73.
Thornton, D. L. and D.S. Batten (1985), "Lag-Length Selection
and Tests of Granger Causality Between Money and Income," Journal
of Money, Credit, and Banking, Vol. 17, No. 2, 164-178.
* Department of Economics, Central Washington University,
Ellensburg, WA 98926-7486, E-mails: ghoshk@cwu.edu, sauders@cwu.edu,
tenerelt@cwu.edu
Notes
(1.) When analyzing the impact of international trade on services,
it appears that the U.S. imports of Business, Professional, and
Technical services have displayed double-digit growth rates for more
than a decade. Such trade in service tasks, however, lags trade in
manufacturing tasks. For example, in 2005, imports of private services
into the U.S. only accounted for about 13 percent of total U.S. imports.
(2.) One of the novelties of the present paper lies in its
empirical measure of offshoring. The data selection of this variable is
described in detail in section III of this paper.
(3.) The labor content of U.S. exports is mostly skilled labor,
whereas the labor content of U.S. imports is mostly unskilled labor.
Consequently it would be fair to conclude that the unskilled workers in
the U.S. work primarily in the import-competing sector, while the export
sector of our economy primarily employs skilled labor. Since the U.S.
exports goods and tasks that have a high content of skilled labor to the
rest of the world, and imports goods and tasks that are largely produced
by unskilled labor, the TOT variable (defined essentially as the ratio
of exports to imports) can be used as a proxy for the cost of
offshoring.
(4.) All data were obtained from the Bureau of Labor Statistics.
Seasonally adjusted data were used for the AWEP and AWEM variables.
(5.) The selection of the data frequency was determined by the fact
that offshoring only became an issue in the early to mid 1990s.
Consequently, earlier time-series data were not included in this
paper's estimates.
(6.) Statistical inferences about the degree of integration of
individual time-series are based upon the presence or absence of unit
roots in each data series. Unit root tests and their implications to
time-series data analyses are well documented in the econometric
literature. Detailed explanations of these issues are provided by Holden
and Thompson (1992) and McCallum (1993), among others.
(7.) The Engle-Granger (1987) technique requires that several steps
be followed in VEC data modeling. These steps include unit root and
cointegration testing of time-series data. Detailed explanations of
these steps are provided by Enders (1995), pages 373-81. All of these
steps were followed in the present research.
Table 1
Phillips-Perron (PP) Test Results for AWEP, AWEM, WD, PRODL,
EXPORTS, IMPORTS, and TOT
Variable Test Results Test Results
Linear Trend Linear Trend
Not Included Included
AWEP (1) 2.225 -2.097
AWEP (2) -4.982 ** -5.424 **
AWEM (1) 0.569 -3.069
AWEM (2) -7.188 ** -7.182 **
WD (1) -2.149 -2.769
WD (2) -7.117 ** -7.095 **
PRODL (1) 0.721 -2.027
PRODL (2) -8.254 ** -8.322
EXPORTS (1) 1.725 -0.017
EXPORTS (2) -5.014 ** -5.104 **
IMPORTS (1) 0.514 -1.299
IMPORTS (2) -6.139 * -6.441 **
TOT (1) -1.879 -2.955
TOT (2) -8.575 ** -8.693 **
(1) PP test results for the levels of variables
(2) PP test results for the first differences of levels of variables
** Indicates statistical significance at the one-percent level.
Table 2
Johansen Maximum Likelihood Cointegration Test Results for
PRODL, WD, and TOT
Test Statistics Test Results Test Results
Lags 1-2 Lags 1-4
Trace Statistic 30.658 * 32.226 *
Max-Eigen Statistic 22.664 * 23.411 *
* Indicates statistical significance at the five-percent level.
Table 3
VEC Estimates of Equations (1), (2), and (3). Lags 1-2
Equation Dependant Independent Coefficients "t" Statistics
Variable Variables
1 DWD Constant -0.3399 -0.6240
z(-1) 0.1181 2.1092 *
DWD(-1) 0.1411 1.2756
DWD(-2) 0.0064 0.0634
DPRODL(-1) 1.5780 2.5508 *
DPRODL(-2) -0.0714 -0.1118
DTOT(-1) -24.8990 -1.2506
DTOT(-2) 25.0175 1.2519
2 DPRODL Constant 0.5992 5.7029 *
z(-1) 0.0466 4.3103 *
DWD(-1) 0.0140 0.6551
DWD(-2) 0.0082 0.4185
DPRODL(-1) -0.1972 -1.6527
DPRODL(-2) -0.0641 -0.5207
DTOT(-1) -2.1944 -0.5714
DTOT(-2) -2.2629 -0.5870
3 DTOT Constant 0.0002 0.0740
z(-1) -0.0003 -1.0182
DWD(-1) -0.0011 -1.6297
DWD(-2) -0.0009 -1.5531
DPRODL(-1) -0.0008 -0.2137
DPRODL(-2) 0.0007 0.1842
DTOT(-1) 0.2286 1.9577 *
Table 4
VEC Estimates of Equations (1), (2), and (3). Lags 1-4
Equation Dependant Independent Coefficients "t" Statistics
Variable Variables
1 DWD Constant 0.6194 0.7926
z(-1) 0.1805 2.5179 *
DWD(-1) 0.1934 1.6184
DWD(-2) 0.0981 0.8350
DWD(-3) 0.2754 2.3558 *
DWD(-4) -0.1211 -1.1342
DPRODL(-1) 1.1735 1.7791
DPRODL(-2) -0.3566 -0.5481
DPRODL(-3) -0.9189 -1.4129
DPRODL(-4) -0.6019 -0.9319
DTOT(-1) -23.7387 -0.9559
DTOT(-2) 9.6752 0.4100
DTOT(-3) 16.4675 0.7621
DTOT(-4) -17.9732 -0.8461
2 DPRODL Constant 0.8031 5.3657 *
z(-1) 0.0551 4.0114 *
DWD(-1) 0.0167 0.7315
DWD(-2) 0.0312 1.3851
DWD(-3) -0.0155 -0.6993
DWD(-4) -0.0087 -0.4242
DPRODL(-1) -0.2771 -2.1932 *
DPRODL(-2) -0.0924 -0.7412
DPRODL(-3) -0.1732 -1.3902
DPRODL(-4) -0.0931 -0.7528
DTOT(-1) -5.7711 -1.2134
DTOT(-2) -3.8036 -0.8416
DTOT(-3) -0.7817 -0.1889
DTOT(-4) 0.2354 0.0579
3 DTOT Constant 0.0024 0.5222
z(-1) -0.0002 -0.4744
DWD(-1) -0.0010 -1.4775
DWD(-2) -0.0007 -1.0453
DWD(-3) 0.0007 0.9680
DWD(-4) -0.0003 -0.4037
DPRODL(-1) -0.0013 -0.3310
DPRODL(-2) 0.0008 0.2047
DPRODL(-3) 0.0000226 -0.0059
DPRODL(-4) -0.0056 -1.1929
DTOT(-1) 0.2738 1.8553
DTOT(-2) -0.3677 -2.6221 *
DTOT(-3) 0.0677 0.5273
DTOT(-4) -0.0381 -0.3018
* Indicates statistical significance at the five-percent level.
