Cheap Imports and the Loss of U.S. Manufacturing Jobs
Thomas Kemeny ∗†, David Rigby†† and Abigail Cooke††
†Department of Geography & Environment, London School of Economics
††Department of Geography, UCLA
Please cite published version:
Kemeny, T., Rigby, D., & Cooke, A. (2014). Cheap Imports and the Loss of US
Manufacturing Jobs. The World Economy. DOI: 10.1111/twec.12238.
This paper examines the role of international trade, and speciﬁcally imports from low-
wage countries, in determining patterns of job loss in U.S. manufacturing industries between
1992 and 2007. Motivated by intuitions from factor-proportions-inspired work on oﬀshoring
and heterogeneous ﬁrms in trade, we build industry-level measures of import competition.
Combining worker data from the Longitudinal Employer-Household Dynamics dataset, de-
tailed establishment information from the Census of Manufactures, and transaction-level
trade data, we ﬁnd that rising import competition from China and other developing econ-
omies increases the likelihood of job loss among manufacturing workers with less than a high
school degree; it is not signiﬁcantly related to job losses for workers with at least a college
JEL Classiﬁcation: F14; F15; F16; F6; J31
Keywords: international trade, import competition, job loss, inequality, manufacturing
Disclaimer: Any opinions and conclusions expressed herein are those of the author(s)
and do not necessarily represent the views of the U.S. Census Bureau. All results have been
reviewed to ensure that no conﬁdential information is disclosed
Acknowledgements: Kemeny and Rigby gratefully acknowledge support from the Na-
tional Science Foundation Grant BCS-0961735; the authors also thank Olmo Silva and the
LSE Spatial Economics Research Center. This research uses data from the Census Bureau’s
Longitudinal Employer Household Dynamics Program, which was partially supported by
the following National Science Foundation Grants SES-9978093, SES-0339191 and ITR-
0427889; National Institute on Aging Grant AG018854; and grants from the Alfred P. Sloan
∗Corresponding author. Contact information: Department of Geography & Environment, London School of
Economics, Houghton Street, London, WC2A 2AE, UK; e-mail: firstname.lastname@example.org.
During the 1960s, nearly one in three jobs in the U.S. were in manufacturing. As of January 2013,
less than 9 percent of all American workers held manufacturing jobs. Manufacturing’s relative
importance has undergone a sharp decline in most other high-wage developed economies as well,
including Britain, France, Germany, Italy and Japan.1In the U.S., as in many of these other
economies, the mid-20th century preponderance of well-paid manufacturing jobs underpinned
a society deﬁned by its large middle class; the disappearance of these jobs has accompanied
large and persistent increases in wage inequality, particularly in terms of the gap between those
in the middle of the income distribution and those at the top (Autor et al., 2008). By the
early 1990s, the continued decline of manufacturing was linked to growing global integration; in
some quarters, ‘globalization’ became a synonym for a “giant sucking sound” of jobs allegedly
ﬂeeing the U.S. in favor of Mexico and other low-wage labor markets. However, researchers
at that time concluded that there was simply not enough trade with low-wage economies to
produce the observed changes in the domestic labor market (Lawrence et al., 1993; Haskel and
Slaughter, 2001).2Though it was acknowledged that some production was relocating from
the U.S. to low-wage countries, the consensus view was that the chief cause of both rising
inequality and declining manufacturing employment was the widespread adoption of computers
in the workplace. Computers and other new technologies, researchers argued, were skill biased:
they complemented individuals whose jobs required high levels of skill and abstract thinking,
while they substituted for those performing routine, less-skill-intensive work, including many
workers on manufacturing production lines (Haskel and Slaughter, 2002; Levy and Murnane,
2004; Goos and Manning, 2007). This technology-based explanation remains highly inﬂuential
today (Edwards and Lawrence, 2013).
There is at least one major reason to re-consider the labor market impacts of international
trade. In a word: China. Accounting for just under 5% of the total value of U.S. imports from
low-wage economies in 1992, by 2007, imports from China had risen to 75%.3China’s growing
role occurred as imports from low-wage economies rose from a modest 9% of the total value of
U.S. imports in 1992 to 23% of total imports in 2007, and, in absolute terms, as overall U.S.
imports nearly tripled. In short, U.S. imports from low-wage countries have grown dramatically
in relative and absolute terms, and China is the primary driver of this change. Observing these
trends, a number of prominent researchers and public intellectuals have called for revisiting
the impact of trade on wages, job loss and overall welfare (Blinder, 2006; Krugman, 2008;
Feenstra, 2010). This call has been taken up in a new wave of empirical work examining how
low-wage import competition may be aﬀecting wages, employment, long-term welfare, plant
survival, occupational structure (Bernard et al., 2006; Liu and Treﬂer, 2008; Ebenstein et al.,
2009; Autor et al., 2012b,a), but it has not investigated the links between trade and job loss.
This paper contributes to this empirical literature by examining the relationship between
low-wage import competition and manufacturing job loss. If imports from developing countries
1On the basis of industrial employment ﬁgures from the OECD STAN database (ISIC 15-37).
2Nor did the wage changes occur where standard trade theory predicted they would: across industries.
3Figures in this paragraph are authors’ calculations made using UN COMTRADE data, with low-wage coun-
tries deﬁned according to World Bank classiﬁcations in 1992. See Table 1 for a list of countries. Values have
been deﬂated to base-year 2007.
are low-skill intensive, then growth in these imports should increase relative job loss in work of
a similar character in the U.S. In other words, rising imports from China and other low-wage
countries should result in job losses for American workers with low skill levels. Meanwhile,
the growth of low-wage imports may have neutral or even positive eﬀects on more highly-
skilled workers, as the economy re-orients toward production that more closely resembles U.S.
comparative advantage. This diﬀerential job displacement could be the result of a number of
diﬀerent processes. It might be driven by the closure of manufacturing establishments that
produce low-skill produce varieties Bernard et al. (2006). Alternatively, rather than shutting
down existing plants may simply shed much of their low-skill labor force as they concentrate
on high-sophistication varieties. It could also be the result of strategic decisions to shift the
boundaries of the ﬁrm’s domestic activities in response to shifting trade costs for particular tasks
in a vertical production chain (Grossman and Rossi-Hansberg, 2006, 2008; Baldwin, 2006).
Firms keep tasks subject to high trade costs onshore, while parts of the production process
capturing tasks for which trade costs are suﬃciently low will be imported, either through arms-
length contracting or multinational subsidiary arrangements. To the extent that skilled work
is systematically associated with higher trade costs, ﬁrms could oﬀshore low-skill tasks while
strengthening their specialization in functions that require higher skill levels. These diﬀerent
‘task trade’ and ‘product variety’ explanations, each reﬂecting diﬀerent dimensions of shifting
comparative advantage, could be simultaneously operating.
To explore the link between low-wage import competition and job loss, transaction-level
Census estimates of U.S. imports and exports are ﬁrst used to estimate the extent to which in-
dividual U.S. manufacturing industries are exposed to import competition from low-wage coun-
tries. We then relate changes in import competition between 1992 and 2007 to the probability
of manufacturing job loss, conditional upon individual demographic and establishment-level
characteristics. We consider a worker to have lost her job between two periods when she both
switches establishments and suﬀers a decline in salary. We focus on two categories of work-
ers: low- and high-skill workers. Low skill workers are represented by those with less than a
high school degree, while the category of high-skill workers is represented using workers with at
least a Bachelor’s degree. Establishment-speciﬁc characteristics permit modeling of potential
heterogeneity in organizational responses to import competition.
The data to capture these individual-, establishment-, ﬁrm-, and industry-speciﬁc dynamics
comes from three restricted-use Census Bureau sources: the Foreign Trade Imports and Exports
data; the Longitudinal Employer–Household Dynamics (LEHD) database; and the Census of
Manufactures. The LEHD program integrates administrative records from state-speciﬁc un-
employment insurance (UI) programs with Census Bureau economic and demographic data,
providing a nearly universal picture of workers, employers, and their interaction in 30 U.S.
states (McKinney and Vilhuber, 2011). The Census of Manufactures, meanwhile, considerably
enriches the range of establishment characteristics available in LEHD. These data cover the uni-
verse of manufacturing establishments, with surveys conducted quinquennially in years ending
in 2 and 7. Among other important establishment characteristics, the Census of Manufactures
measures spending on computer equipment, such that we might distinguish between the ef-
fects of trade and technological change. Linked together, the trade, worker and establishment
datasets yield a panel covering 1992, 1997, 2002 and 2007.
In our preferred speciﬁcations, we account for potential endogeneity bias arising from the
possibility that imports are driven by labor demand shocks by instrumenting for import com-
petition using both an industry- and year-speciﬁc measure of low-wage imports into the EU-15,
and an industry-year measure of trade costs. Doing so, we ﬁnd that workers with less than
a high-school degree are vulnerable to the rise of imports from low-wage economies, while job
security for those with at least a Bachelor’s degree is largely unrelated to both levels of and
changes in import competition. This diﬀerentiated result conforms to the theoretical explana-
tions guiding this research. In keeping with recent heterogeneous-ﬁrm extensions of the theory
of factor-proportions, as well as the literature on oﬀshoring and task trade, growing imports
from China and other low-wage economies results in manufacturing job displacement among
those workers with low levels of education. It does not appear to threaten jobs held by more
highly-educated manufacturing workers.
The remainder of this paper is organized as follows. Section 2 describes the theory motivating
this inquiry, and reviews the ﬁndings of the existing empirical literature. Section 3 discusses
low-wage import competition and how it has changed over the study period. Section 4 describes
the empirical approach and the data. Section 5 presents results and Section 6 concludes.
A principal insight of the Heckscher-Ohlin (H-O) trade model is that, given a suﬃciently open
world economy, countries should specialize in activities that reﬂect factors of production that
they hold in relative abundance. Assuming the U.S. is abundant in skilled labor, and low-wage
economies like China are abundant in unskilled labor, then a decline in trade costs should re-
orient production in the U.S. toward industries that intensively demand workers with higher
skill levels, while importing goods from China and other low-wage economies that intensively
require less-skilled workers.4Standard H-O models consider that such shifts in specialization
and trade occur at the scale of industries. But recent trade models emphasize within-industry
heterogeneity, product diﬀerentiation, and job search frictions which can nonetheless lead to
outcomes that are broadly in line with the theory of factor proportions (Melitz, 2003; Yeaple,
2005; Verhoogen, 2008; Helpman et al., 2010). Bernard et al. (2010), for instance, yield such
predictions by assuming that the intensity with which plants employ factors of production
signals the factor-intensity of the resulting goods. Concretely, this means that plants whose
input intensities reﬂect a country’s comparative advantage in skilled labor will be more resilient
in the face of low-wage import competition, while those that abundantly use low-skill labor
are likely to be most challenged. In a major empirical contribution, Bernard et al. (2006)
ﬁnd that plant employment and survival are negatively related to the level of industry-speciﬁc
import competition, and within those industries, more capital-intensive plants are more likely
grow and survive. Moreover, they show that some plants respond to import competition by
switching industries, shifting to those that more intensively require capital and skill. All of this
suggests a ‘horizontal’ adjustment process to low-wage import competition that is consistent
4Assuming trade with such developing economies forms a large and/or growing proportion of total U.S. trade.
with factor proportions theory. Both within and between industries, employment shifts toward
products that better reﬂect U.S. comparative advantage. We should expect this adjustment
process to reduce manufacturing jobs for less-skilled workers whose wages will increasingly be,
in the words of Richard Freeman (1995), set in Beijing.
While the H-O model and its descendants emphasize that local industries or ﬁrms grow or
decline to the extent that they reﬂect comparative advantage, the recent ‘oﬀshoring’ theoretical
framework takes a ‘vertical’ view in which the production process of a given good can be split
into a sequence or continuum of tasks, with each task subject to speciﬁc trade costs (Feenstra and
Hanson, 2001; Antr`as et al., 2006; Blinder, 2006; Baldwin, 2006; Grossman and Rossi-Hansberg,
2006, 2008). The theory of oﬀshoring suggests that, on the basis of these activity-speciﬁc trade
costs, organizations will determine which stages of production to produce domestically and
which to oﬀshore, with oﬀshoring itself accomplished either within the organization through a
multinational aﬃliate or through arms-length transactions. While the results of these various
models are strongly dependent on assumptions regarding country size, preferences, the nature
of technological change, and the number of production factors, models like that of Grossman
and Rossi-Hansberg (2008) suggest that, following factor-proportions intuitions, organizations
should oﬀshore those activities in which relevant factors are not held in relative abundance. In
the case of high-wage economies like the U.S., this means oﬀshoring production tasks that are
intensive in their use of low-skill work. In these models, therefore, oﬀshoring works in a manner
akin to technological innovation that replaces the work of low-skilled laborers (Feenstra, 2008).
Instead of mechanizing these workers’ labor, oﬀshoring sends it abroad. Under this hypothesis,
demand for less-skilled workers will fall, and their employment and wages will decline relative
to employment and wages for highly skilled workers.
Substantial evidence supports this idea. Studies have related changes in industry-speciﬁc
measures of oﬀshoring, sometimes restricted to the imports of intermediates from some deﬁ-
nition of low-wage countries, to either the wage- or employment-share of skilled workers. In
survey of empirical work, Crin`o (2009) ﬁnds that, across more than 25 articles studying 11
ﬁrms and industries in high-wage countries, including the U.S., Canada, UK, France, and Ger-
many, oﬀshoring to low-income economies is positively and signiﬁcantly related to the share of
employment or wages (or both) of highly-skilled workers in high-wage countries. This result
is strongest for studies examining arms-length imports; it exists but is weaker in studies that
model the eﬀects of intra-MNC production transfer, as distinct from broader trade ﬂows (ibid).
Nonetheless, Harrison and McMillan (2011) show that the employment outcomes associated
with such transfers have employment impacts, and these impacts diﬀer depending on whether
the production is being transferred to high- or low-income country aﬃliates. Speciﬁcally, the
authors ﬁnd that transfers to low-income aﬃliates is associated with job losses in the U.S. While
this is not an uncontroversial ﬁnding, few studies in this area have had access to detailed worker
characteristics that would permit a sense of the kinds of jobs created or workers displaced.5
Among the few exceptions, Ebenstein et al. (2009) ﬁnd that, while manufacturing wages are
not strongly aﬀected by oﬀshoring to low-wage countries, across educational categories, wage
5The aggregate employment eﬀects of MNC oﬀshoring is a controversial subject, with eﬀects ranging from
positive (Brainard and Riker, 1997; Mankiw and Swagel, 2006; Navaretti et al., 2010), negative (Harrison and
McMillan, 2006, 2011) and none (Braconier and Ekholm, 2000).
eﬀects for workers who leave their manufacturing jobs are high.
Direct links between imports or oﬀshoring and job loss have been less well studied. Looking
at U.S. manufacturing industries over the period 1977–1987, Revenga (1992) ﬁnds that trade
shocks, in the form of import price changes, chieﬂy aﬀect employment not wages, suggesting
that worker displacement is an important by-product of international trade. Kletzer (2000,
2001) uses Displaced Worker Surveys to show that job losses due to plant closures, mass layoﬀs
or other non-performance reasons are positively linked to import competition, though not to
trade in intermediates. Several other studies conﬁrm this positive relationship between import
competition and displacement (Haveman, 1994; Addison et al., 2000). These are all industry-
level results, however – they do not account for individual diﬀerences, especially skill.
A handful of recent studies have followed up on this work, mostly examining the likelihood
of speciﬁc, short-run labor-market transitions between sectors, and from employment to un-
employment. The variable of interest in these studies has chieﬂy been trade in intermediates.
Among the strengths of this work is the use of individual-level data that account for bias due
to compositional changes, as well as various sources of unobserved heterogeneity. For instance,
Egger et al. (2007) follows 38,000 Austrian males, and ﬁnds that various measures of oﬀshoring
and import competition are negatively related over the short run to the likelihood of an individ-
ual staying in his job in an industry in which Austria does not hold a comparative advantage.
Munch (2010) considers year-to-year transitions directly from manufacturing employment to
unemployment, and ﬁnds a modest positive relationship between intermediate-goods imports
and job loss for low-skill workers. Bachmann and Braun (2011) ﬁnd similar results for Germany.
By contrast, Geishecker (2008) exploits monthly worker data, ﬁnding that job tenure, not skill,
is the chief determinant of German workers’ job security.
Several other studies look beyond intermediates trade to focus on job loss due to other forms
of internationalization. For instance, Becker and Muendler (2008) ﬁnd that multinationals’
foreign expansions do not reduce home-country job security, as compared with similar non-
multinational ﬁrms. Meanwhile, (Hummels et al., 2011) ﬁnd that ﬁrm-speciﬁc importing shocks
reduce wages and employment for Danish workers with only a high school education. Menezes-
Filho and Muendler (2011) examine the eﬀects of tariﬀ reductions in Brazil, and conclude that
they are positively related to manufacturing job losses, often resulting in worker transitions
to services and longer-term unemployment. And Autor et al. (2012b) examine the long-run
impacts of Chinese import competition on U.S. manufacturing workers, ﬁnding that workers in
sectors with high levels of Chinese export exposure spend less time at their initial employer and
in their initial industry.
This paper adds to the scholarship linking trade and job loss in a few speciﬁc ways. First,
it focuses explicitly on trade relationships with low-wage countries, which is where we would
expect the clearest factor-proportions-type eﬀects. Second, it takes a broader focus beyond
intermediates, to examine low-wage import competition from all sources. Third, it accounts
not just for relevant individual characteristics, but also for ﬁrm-speciﬁc factors that can bear
upon job loss. Of particular importance is our inclusion of measures of technology investment,
which could play a considerable and perhaps overlapping role with trade. Fourth, rather than
focusing on aggregate ﬁrm or industry job losses, this paper diﬀerentiates results by skill level,
focusing on workers with at least Bachelor’s degree, and those with less than a high school
diploma. Last, like Autor et al. (2012b) but unlike most of recent research in this area, it
focuses on the links between trade and job in the United States.
3 Trends in Low-Wage Import Competition
To describe low-wage import competition, we rely on a measure proposed by Bernard et al.
(2006), deﬁned as follows:6
LW I COM Pit =ML
Mit +Pit −Eit
it and Mit are the value of imports in industry iand time tfor low-wage countries
Land all countries, respectively; Prepresents total domestic production, or shipments; and
Emeasures U.S. exports. As Bernard et al. (ibid) observe, low-wage import competition is
therefore a function of the share of low-wage imports in the total value of industry imports, and
overall import penetration. Equation (1) can also be modiﬁed to describe sector-speciﬁc levels
of import competition from countries that are expected to pay higher wages, substituting the
existing numerator for (Mit −ML
Nonetheless, our primary interest in this paper is exploring impacts of trade with China
and other low-wage economies. Countries are classiﬁed as low-wage on the basis of the World
Bank country classiﬁcation scheme, which annually slots national economies into four groups:
low income; lower middle income; upper middle income; and high income. We use 1992, the
ﬁrst year available in our data, as the base classiﬁcation year, and deﬁne the group of low-wage
countries in terms of membership in the low income group. Eﬀectively, this means that the
group of low-wage countries had Gross National Incomes (GNI) per capita in 1992 of less than
or equal to US$545.7Table 1 lists the 51 economies that fall into this category, that include
China and India.
[Table 1 about here.]
[Figure 1 about here.]
We calculate LW I COM P by combining data from two U.S. Census Bureau data sources.
Trade information comes from the Foreign Trade Imports and Exports data, which capture
the universe of ﬁrms operating in the U.S. that import goods from abroad, compiled using
information gathered by U.S. Customs and Border Protection. Products in these data are
identiﬁed using ten-digit Harmonized System (HS) codes, with records going back to 1992. Using
a crosswalk produced by Pierce and Schott (2012), we aggregate from the ten-digit HS product
level to as many as 472 six-digit North American Industry Classiﬁcation System (NAICS)
manufacturing industries.8Data on domestic production come from the Census of Manufactures
6, Autor et al. (2012b) modiﬁes this measure somewhat to account for dynamics.
7GNI per capital calculated using the World Bank’s Atlas methodology. For details, see
8Manufacturing is deﬁned as those NAICS codes between 31–33.
(CM). These survey data are gathered in each year ending in 2 or 7, and cover nearly all the
private manufacturing activity in the United States. To conform with our trade data, we settle
on four waves of the CM: 1992; 1997; 2002; and 2007. This results in a dataset that begins as
low-wage imports start their rapid ascent, and ends just before the start of the Great Recession.9
Figure 1 presents the growth of import competition from low-wage and non-low-wage coun-
tries, with industry-level measures weighted by import shares. Between 1992 and 2007, import
competition from low-wage economies has grown by more than 400%, while from countries that
are not classiﬁed as low-wage, it has expanded by one third. In absolute terms, import com-
petition from higher-wage countries remains considerably higher than for low-wage economies,
but as evidenced by diﬀerences in growth rates, this gap is closing.
[Table 2 about here.]
Table 2 presents more detailed descriptive evidence of the rising importance of U.S. imports
from these low-wage economies. In the leftmost columns, for each of the 21 two-digit NAICS
codes that comprise manufacturing, the table describes ﬁve-year changes in the share of low-
wage imports in total imports. In the middle columns, at those same ﬁve-year intervals, it
presents the results of equation (1) constructed at the two-digit NAICS level. The rightmost
columns present absolute changes in employment and the percentage change in employment.
Means and standard deviations for manufacturing on the whole are provided in the bottom
rows. From Table 2, several broad points are evident. First, manufacturing industries vary in
their exposure to trade from low-wage economies. In 2007, for instance, Beverage and Tobacco
imports from low-wage economies accounted for only one percent of all U.S. imports, while
they accounted for 74 percent of all imports in the Leather sector. At the same time, the table
shows that low-wage imports in nearly all manufacturing industries have risen over the 15-year
study period. The import competition measure LW IC OM P shows both this variation and the
common rising trend. Having a high share of low-wage imports plays a role in determining levels
of import competition, but considerable variation exists, due to diﬀerences in the importance of
trade in domestically-consumed production, the other component of LW IC OM P . Additionally,
while most manufacturing sectors have shed considerable fractions of their labor force, with a
24 percent decline overall, some of the largest declines in employment are found in industries
facing substantial low-wage import competition. Apparel, for instance, has current labor force
that is only a quarter of its level in 1992. Some computer component manufacturing has also
experienced both high levels of low-wage import competition and considerable employment
losses. Meanwhile, sectors like Food and Fabricated Metals have not experienced large declines
in employment, with small absolute amounts of import competition (though, in each case there
has been some growth). Clearly, low-wage imports are not the only factor shaping domestic
employment dynamics, but from this table it is plausible that they have played a nontrivial
9Plants in the 1992 CM are classiﬁed using the prior Standard Industrial Classiﬁcation (SIC) system. To
assign NAICS codes to these establishments, we take three steps. First, for continuing plants, we assign 1997
NAICS codes to 1992 iterations. Second, for non continuing plants, we use the standard Census SIC-NAICS
crosswalk. Any unassigned plants after these two procedures are compared to the list of establishments in
LEHD’s Longitudinal Business Dynamics Bridge. Any matching ﬁrms are assigned corresponding NAICS codes.
These three procedures account for over 90% 1992 plants.
role. We now turn to the more systematic examination of this relationship that is the focus of
4 Empirical Approach
The aim of this paper is to measure the extent to which low-wage import competition is related
to job loss for more- and less-skilled workers. Speciﬁcally, we deﬁne a binary outcome capturing
whether an individual loses his or her job between two time periods, and relate this to industry-
speciﬁc low-wage import competition. To pursue this strategy, we ﬁt two kinds of models. In
the ﬁrst, we relate job loss between periods tand t+ 1 to levels of import competition in t,
ijs =β0+β1LW I CO MPit +β2X0
ikt +δi+δt+δg+εjt (2)
where X0is a vector of worker characteristics, including age, sex, nativity, and race, for worker
jwith skill level sworking in industry i;Z0includes features of establishment k, notably com-
puter investment. δiis an industry ﬁxed eﬀect that accounts for sector-speciﬁc characteristics
unrelated to import competition; δtis a year dummy variable that captures business cycle dy-
namics; δgabsorbs shocks that are state-speciﬁc; and εis an error term that satisﬁes classical
regression assumptions. We believe that the lagged structure of our model, with prior levels of
import competition aﬀecting employment decisions over the next ﬁve years, is warranted on the
basis that trade adjustment will not occur instantaneously.
The second model is similar to the ﬁrst, but exploits temporal variation within industries
and establishments. Speciﬁcally, we adopt a ﬁxed eﬀects panel approach, with job loss between
periods t+ 1 and t+ 2 related to changes in import competition, along with changes in plant-
and ﬁrm-speciﬁc dynamics between tand t+ 1, as follows:
ijs =β0+β1LW I CO MPi,t→t+1 +β2Z0
ik,t→t+1 +δi+δt+δs+εjt (3)
In theory, equation (3) could include worker characteristics; in practice, no time-varying indi-
vidual demographics are available in our data. The chief advantage of this model is the control
for worker-level unobserved heterogeneity that is constant over time. This is particularly im-
portant, given that unobserved (but likely fairly stationary) diﬀerences in ability, motivation,
attitude, teamwork could signiﬁcantly aﬀect the likelihood of retaining one’s job.
In both approaches, the variable of interest is LW I COM P . In equation (2), a coeﬃcient
for β1greater than zero indicates that higher levels of low-wage import competition raise the
likelihood of job loss for workers in a particular skill category. For equation (3), positive values
for β1indicate that growth in import competition raises the likelihood of job loss.
Across these speciﬁcations, results obtained using LW I CO MP could be biased to the ex-
tent that this variable is correlated with sector-speciﬁc changes in U.S. demand or productivity.
Though our motivation is to explore how rising imports from China and other low-wage econ-
omies are aﬀecting job loss for workers with diﬀerent skills, our results could instead reﬂect
unobserved factors that have more to do with internal U.S. industrial dynamics than with low-
wage imports themselves. To account for this potential endogeneity bias, we instrument for our
measure of import competition using two variables. The ﬁrst captures the industry- and year-
speciﬁc value of exports from the same group of low-wage economies to countries in the EU-15,
calculated using United Nations COMTRADE data. Assuming that other high-wage countries
will face similar exposure to low-wage import competition when imports reﬂect factors inherent
in low-wage economies, and that local demand-side shocks ought to be relatively uncorrelated
across countries, this instrument ought to be helpful in providing estimates of the exogenous
contribution of low-wage import competition to job loss. In order to produce test statistics of
instrument exogeneity that require more instruments than potentially endogenous regressors,
we also use a second instrument: an industry-year measure of trade costs, which is the sum of
ad valorem tariﬀ and freight rates, constructed by Bernard et al. (2006).10
Worker-level data for the analysis come from the Census Bureau’s Longitudinal Employer-
Household Dynamics (LEHD). The LEHD program integrates administrative records from state-
speciﬁc unemployment insurance (UI) programs with Census Bureau economic and demographic
data. LEHD oﬀers a number of crucial advantages over surveys of displaced workers. First,
it tracks workers as they ﬂow from job to job, combining them with characteristics of the
establishments at which they work. Second, with only minor exceptions, the data cover the
universe of workers in participating states. Jobs covered by UI include 90% of total wage and
salary jobs in the civilian sector of the U.S. economy (Stevens, 2002; McKinney and Vilhuber,
2011).11 The 30 states available to researchers in Census Bureau Research Data Centers through
the LEHD program at the time this research project was approved include populous states like
California, Florida, Texas, and Illinois; together they constitute 65% percent of the total U.S.
workforce. These and other states in which LEHD data is available, such as Georgia, North
Carolina and Wisconsin are also among the largest states in terms of manufacturing employment
One limitation of LEHD is its scant information regarding establishments. We address
this issue by combining LEHD with detailed plant data from the Census of Manufactures. This
means we focus exclusively on individuals whose initial appearance in the LEHD data is in a job
in a manufacturing establishment, though workers who transition from such employment to jobs
in any sector remain in the analytical sample. Linking LEHD to the Census of Manufactures is
a nontrivial task that necessitates a few accommodations. First, it requires a shift from LEHD’s
quarterly frequency to a quinquennial panel. Additionally, we aggregate from the plant level to
the scale of unique State Employer Identiﬁcation Numbers, or SEINs, which are state-speciﬁc
unemployment-insurance taxpayer identiﬁers. An SEIN will correspond to an individual plant
for organizations that operate only a single plant in a particular state, while for ﬁrms with
multiple plants in the same state, it represents the aggregate of those establishments. Although
this aggregation may mask some heterogeneity for some multi-plant employers, this level also
oﬀers two beneﬁts. First, individual workers in LEHD cannot be directly assigned to individual
establishments in the case of multi-plant ﬁrms.12 More importantly, match rates between plants
10In addition to Bernard et al. (2006), instruments like these have been used in other recent studies, including
Autor et al. (2012b) and Hummels et al. (2011).
11Unemployment insurance records typically do not include many government workers; agricultural workers;
and members of the armed forces.
12It is possible to assign workers to establishments using Census Bureau-created imputations, such that each
employee in an organization that has multiple plants in the same state is assigned 10 possible plant matches.
Hence, estimates could in theory be produced using imputation commands such as Stata’s mi suite. Given the
size the data, however, this approach creates computational challenges, since an imputation-ready dataset would
eﬀectively contain nearly ten times the original number of records.
in the LEHD and Census of Manufactures improve dramatically when we restrict matching
to state and Employer Identiﬁcation Numbers (EINs). Eﬀectively, our analytical sample is
dramatically larger when matching on this basis.13
The addition of data from the Census of Manufactures means we can include a number of
key plant-level predictors. Most importantly, we estimate establishment-speciﬁc technological
change, by calculating investment in high technology, using the fraction of spending on com-
puter equipment in total equipment expenditures.14 As discussed earlier, it is widely held that
technological change has been the primary driver of labor market reallocation over the past
decades. If computers and other high-technology investments function to substitute for workers
with low skill levels, we should expect it to be positively related to job loss for those work-
ers. Several additional establishment-level controls are included in estimates of equations (2)
ands (3). We include a measure of total shipments to capture establishment size; the ratio of
capital to value-added to account for overall diﬀerences in capital-intensity; a dummy variable
that takes on a value of one when an establishment is part of a multi-unit ﬁrm; a measure of
establishment exports, since research indicates that exporters and nonexporters behave diﬀer-
ently (Bernard et al., 1995; Bernard and Jensen, 1999); and ﬁnally a measure of the age of the
We retain only workers between the ages of 16 and 65 who are employed at the same
establishment for all four quarters and who do not simultaneously work at another job, and who
appear in at least the subsequent wave of data, such that it may be possible to observe a change
in their job status. Among these workers, job loss is deﬁned as a change in employer between
two consecutive periods, such that nominal salary in the second position is less than the salary a
worker held in the ﬁrst. This is a diﬀerent approach toward identifying job loss than is common
in the literature. In survey data, workers report being ‘displaced’ or ‘dislocated’ when their
departure was involuntary, due to layoﬀs or plant closures, as distinct from being a function
of individual performance (Kletzer, 1998). With administrative data like LEHD, it is not
possible to determine deﬁnitively whether a departure from a job was voluntary or involuntary.
However, it is unlikely that a transition from one full-year fully-employed job to another would
be undertaken voluntarily if such a switch entailed losses in annual income. Moreover, a focus
on the sample of workers who are more likely to be fully-employed also limits bias that might
13We match LEHD and the Census of Manufactures using two bridges: the Business Register Bridge (BRB)
and the Longitudinal Business Dynamics Bridge (LBDB). This is required since BRB covers 1992, while LBDB
tracks the period 1997 to 2004. We ﬁrst adjust the BRB, assigning NAICS codes to EINs, using three methods
in sequence: ﬁrst we assign NAICS codes when there are 1:1 SIC-NAICS matches; second, for matches that are
1:many, we collect industry codes from subsequent appearances of the ﬁrm in LBDB; third, we assign NAICS
codes to any remaining ﬁrms based on the most probable NAICS on the basis of employment ratios. This adjusted
BRB is combined with LBDB to form a master bridge between LEHD and the Census of Manufactures. Due to
LBDB’s termination in 2004, we lose ﬁrms that appear only after this date. However, such ﬁrms would not be
included in the analysis, given the temporal structure of equations (2) and (3). In terms of granularity, matching
using state and industry results in more than a 60 percent match rate, while using the most restrictive matching,
we retain only approximately 17 percent of the ﬁrms.
14Note that computer investment data are unavailable in 1997. We ran models excluding 1997 and subsequently
using computer investment data for 1997 that we interpolated from surrounding observations. There were no
qualitative diﬀerences in the results of our models; results presented are from the interpolated 1997 values of
15For equation (2), we include a direct measure of plant age. For ﬁxed eﬀects estimates of equation (3), which
do not allow time-invariant predictors, we create a ‘relative plant age’ indicator, which is the ratio of the ﬁrst
year of plant operations and the current year.
arise if workers vary considerably in terms their labor-force attachment. For workers with less
than a high school degree, designated as low-skill workers, this results in an analytical sample
for equation (2) of 1,000,400, and a sample of 565,000 for equation (3). Samples for equation (3)
are smaller because of the added restrictions it imposes in terms of continuity across multiple
periods. For workers with at least a Bachelor’s degree, constituting the high-skill subset of
workers, the sample for equation (2) consists of 850,000 workers, while 550,000 college-educated
workers are available to estimate equation (3).16 In samples for equation (2), 13.9 percent of
workers with less than a high school degree lost their jobs over the study period; the comparable
ﬁgure for workers with at least a Bachelor’s degree is 14.4 percent.
Equation (2) is a pooled cross-sectional model that can be estimated using both ordinary least
squares as well as binomial logistic regression.17 Estimates produced using the logistic model are
preferable, given the limited dependent variable. All models are estimated with standard errors
clustered at the level of the establishment, on the basis that the likelihood of a worker losing their
job to trade competition will be related to that of their co-workers. Three kinds of estimates of
equation (2) are produced for low- and high-skill workers: ordinary least squares, logistic, and
a two-stage least-squares regression. It is possible to estimate equation (3) using standard ﬁxed
eﬀects and ﬁxed eﬀects logit models; all things equal the latter is preferable. However, panel
logit models systematically failed to converge; despite a wealth of diagnostics, this problem
could not be solved. Hence, a linear probability ﬁxed eﬀects model is presented, along with
estimates produced using a ﬁxed eﬀects instrumental variables approach. Estimation using the
linear probability model should not be biased, though it will be ineﬃcient due to potential
heteroscedasticity; however, the clustered standard errors should correct for this issue.
[Table 3 about here.]
5.1 Low-skill workers
For workers with less than a high school education, Table 3 reports estimates of the relationship
between low-wage import competition and job loss. The ﬁrst three columns report results from
equation (2), in which levels of import competition at time tpredict the likelihood that a worker
loses her job between tand t+ 1. The ﬁrst column presents estimates produced using ordinary
least squares. For these and all other estimates, standard errors have been clustered at the level
of the establishment, on the basis that one should expect an individual’s likelihood of job loss to
be related to that of her co-workers.18 Low-wage import competition, LW I COM P , is positively
related to the outcome, suggesting that greater threat from low-wage imports is associated with
greater risks of job loss for less-skilled workers; however the coeﬃcient is not signiﬁcant in
16These sample sizes have been rounded to the nearest 1,000 to facilitate disclosure through the U.S. Census
Bureau’s Center for Economic Studies.
17The latter implies a somewhat rewritten version of equation (2).
18Results were also produced using simple heteroscedasticity-robust standard errors. The broad contours of the
results did not vary dramatically, however import competition was positive and signiﬁcant across all estimators.
These results are available upon request
this speciﬁcation. Interestingly, computer investment is negatively and signiﬁcantly related
to job loss for this group of workers. The higher the share of computer equipment in total
investment, the lower the likelihood of a low-skill worker losing one’s job. Job loss is also
negatively related to ﬁrm size, as measured by shipments, as well as to the ratio of capital to
value-added. Being a low-skill worker in a multi-unit ﬁrm increases the likelihood of job loss;
exporting is also positively related to job loss, though insigniﬁcantly in this model. Among
demographic characteristics, older workers and female workers are more likely to lose their jobs,
while immigrants and white workers face lower odds of job loss. This same pattern of results
holds for the second column, that estimates a logistic regression that ought to be more eﬃcient,
given the limited dependent variable.
In the third column, we account for potential endogeneity bias by instrumenting for LW IC OM P
using both exports from our sample of low-wage countries to EU-15 economies, and the measure
of trade costs, each time- and six-digit NAICS industry-speciﬁc. We report several diagnostics
on the suitability of these instruments. The Kleibergen-Paap F-statistic reports on instrument
strength under the assumption that errors are not independent and identically distributed. The
value of 202.13 is well above the Stock-Yogo critical values, from which we conclude that the
instrument set is not weak. The Hansen-J p-value means that we fail to reject the hypothesis
that the instruments are not exogenous. In other words, we can assume that our instruments
are not correlated with the error term, permitting us to estimate the exogenous contribution
of low-wage import competition to job loss. The second-stage results shown in column 3 are
closely related to results in the ﬁrst two columns, except that the coeﬃcient on low-wage import
competition, still positive, is now highly signiﬁcant.
[Table 4 about here.]
The fourth and ﬁfth columns report results from equation (3), in which we better exploit
the dynamics available in the data. In these estimates, changes in import competition between
tand t+ 1 are related to the likelihood of losing one’s job between t+ 1 and t+ 1. As described
above, these models are ﬁtted using the ﬁxed eﬀects estimator, whose chief beneﬁt in this
context is to account for any time-invariant individual unobserved heterogeneity. We therefore
expect these estimates to better reﬂect our relationship of interest. Worker demographics are
not included in the model, since these are time invariant (or in the case of worker age or job
tenure, they change in lockstep with years, which we capture with a year dummy variable).
Column 4 presents results that broadly resemble those from the other low-skill models. Low-
wage import competition is positively and signiﬁcantly related to job loss. Exporting enters as
negative and signiﬁcant, suggesting that growth in exports reduces the likelihood of job loss for
low-skill workers. Neither multi-unit status nor plant age are signiﬁcantly related to job loss in
this model. In column 5 we present results produced using a two-stage least-squares ﬁxed eﬀects
estimator. F-statistic and Hansen Jvalues again suggest that the instrument set is not weak
and can be considered exogenous. Results are in line with those produced using the standard
ﬁxed eﬀects estimator, with import competition positively and signiﬁcantly related to job loss,
while the coeﬃcient on computer investment is negative and signiﬁcant.
5.2 High-skill workers
The second worker subgroup of interest is high-skill workers, deﬁned as those who have obtained
at least a Bachelor’s degree. We follow the same sequence of models as for low-skill workers,
beginning with models that predict job loss as a function of levels of import competition, before
shifting to models that estimate how changes in import competition inﬂuence the odds of job
loss. Columns 1 and 2 present results produced using OLS and Logit estimators. In both cases,
import competition is not signiﬁcantly related to job loss for this worker subgroup. Computer
investment remains negative and signiﬁcant: all else equal, highly skilled workers have more job
security in establishments that invest relatively more on computer equipment. This negative
relationship is true for overall capital investment as well. Highly-skilled workers face greater
odds of job loss in smaller and younger establishments, as well as in establishments that are part
of multiunit ﬁrms. Exports appear to be unrelated to job loss in these initial models. The same
patterns remain in the 2SLS model in column 3, except that import competition is positive and
weakly signiﬁcant (at the 10 percent level).
The fourth and ﬁfth columns again present results in which changes in import competition
over a prior period predict potential job loss over a subsequent period. Results are broadly
consistent with those obtained using equation (2). Most importantly, for highly skilled workers,
import competition appears to be unrelated to the likelihood of losing one’s job. Meanwhile,
computer investments appear to reduce the odds of job loss. Exports are now negatively and
signiﬁcantly related to job loss, as is plant age. Instruments in the ﬁxed eﬀects instrumental
variables estimates are not weak, and pass the test of orthogonality.
U.S. imports of manufactured goods from China and other developing countries have risen
dramatically over the last several decades, while domestic manufacturing employment has con-
tracted from a peak of over 30 percent to less than 10 percent today. The goal of this paper has
been to consider the impact of industry-speciﬁc low-wage import competition on the likelihood
of losing one’s manufacturing job. We have been particularly interested in examining how the
risks of job loss may be disproportionately born by workers who have lower skill levels, as one
would expect from contemporary ‘horizontal’ and ‘vertical’ updates of the neoclassical theory
of factor proportions. We consider the links between trade and job loss while also accounting
for potential roles for individual and establishment characteristics, especially computerization.
Our primary ﬁnding is that low-wage imports have a diﬀerent relationship to job loss de-
pending on a worker’s skill level. We ﬁnd low-wage import competition raises the likelihood of
job loss among manufacturing workers who have completed less than a full high school educa-
tion. Controlling for worker characteristics, as well as features of the establishments where they
work, low-skill workers are more likely to lose their jobs when they work in sectors with higher
initial levels of low-wage import competition, as well as industries where this import compe-
tition has grown over the preceding period. By contrast, among workers who have completed
at least a Bachelor’s degree, industry-speciﬁc pressure from low-wage imports appears to be
largely unrelated to the likelihood of job loss. Over the period 1992 to 2007, these high-skill
U.S. manufacturing workers appear to be sheltered from the eﬀects of China’s rise.
Technology is typically considered to be a major, if not the most important driver of labor
reallocation, and it has been considered to operate in a skill-biased manner. We do not ﬁnd
a skill-biased link between technology and job loss. Across both high- and low-skill workers,
establishment-speciﬁc investments in computer equipment are associated with job retention.
This is a challenging ﬁnding, since a great deal of theory suggests that computers are substitutes
for less-skilled workers. As expected, in both worker groups, larger, older and more capital-
and export-intensive establishments are associated with lower odds of job loss, though these
relationships are not statistically signiﬁcant in all models.
Contrary to a great deal of the literature on the labor market impacts of globalization, this
paper’s ﬁndings suggest that imports from low-wage economies have adversely aﬀected labor
market outcomes for low-skill workers. Consistent with factor proportions, import competition
from countries whose comparative advantage lie in low-skilled labor has meant that less-skilled
workers in the U.S. have had less job security. The positive association between imports and
the risk of job loss for less-skilled workers may well represent a lower bound, since our focus
on workers who are full-year and singularly employed ignores part-time workers and others
with lower levels of labor market attachment. It is not implausible that these workers may be
more vulnerable to the eﬀects of import competition, moreover they are also likely to be better
represented among less-skilled workers. Less-skilled manufacturing workers on the whole, in
other words, may be more vulnerable than suggested by this study. It should be said that these
results say nothing about the overall gains from trade, which may well outweigh losses in this
worker group. However, it provides further evidence that trade has played some role in the poor
recent labor market situation of these domestic workers, and likely in the larger expansion of
Addison, J. T., Fox, D. A., and Ruhm, C. J. (2000). Technology, trade sensitivity, and labor
displacement. Southern Economic Journal, pages 682–699.
Antr`as, P., Garicano, L., and Rossi-Hansberg, E. (2006). Oﬀshoring in a knowledge economy.
The Quarterly Journal of Economics, 121(1):31–77.
Autor, D., Dorn, D., and Hanson, G. H. (2012a). The China Syndrome: Local labor market
eﬀects of import competition in the United States. National Bureau of Economic Research
Working Paper 18054.
Autor, D., Dorn, D., Hanson, G. H., and Song, J. (2012b). Trade adjustment: Worker level
Autor, D., Katz, L., and Kearney, M. (2008). Trends in US wage inequality: Revising the
revisionists. The Review of Economics and Statistics, 90(2):300–323.
Bachmann, R. and Braun, S. (2011). The impact of international outsourcing on labour market
dynamics in germany. Scottish Journal of Political Economy, 58(1):1–28.
Baldwin, R. (2006). Globalisation: the great unbundling(s). Finnish Prime Minister’s Oﬃce
for EU Presidency.
Becker, S. O. and Muendler, M.-A. (2008). The eﬀect of fdi on job security. The BE Journal
of Economic Analysis & Policy, 8(1).
Bernard, A. and Jensen, J. (1999). Exceptional exporter performance: cause, eﬀect, or both?
Journal of International Economics, 47(1):1–25.
Bernard, A. B., Jensen, J. B., and Lawrence, R. Z. (1995). Exporters, jobs, and wages in us man-
ufacturing: 1976-1987. Brookings Papers on Economic Activity. Microeconomics, 1995:67–
Bernard, A. B., Jensen, J. B., and Schott, P. K. (2006). Survival of the best ﬁt: Exposure to low-
wage countries and the (uneven) growth of us manufacturing plants. Journal of International
Bernard, A. B., Redding, S. J., and Schott, P. K. (2010). Multiple-product ﬁrms and product
switching. American Economic Review, 100(1):70–97.
Blinder, A. (2006). Oﬀshoring: The Next Industrial Revolution. Foreign Aﬀairs, 85(2):113–128.
Braconier, H. and Ekholm, K. (2000). Swedish multinationals and competition from high-and
low-wage locations. Review of International Economics, 8(3):448–461.
Brainard, S. L. and Riker, D. A. (1997). Are us multinationals exporting us jobs? Technical
report, National Bureau of Economic Research.
Crin`o, R. (2009). Oﬀshoring, multinationals and labour market: a review of the empirical
literature. Journal of Economic Surveys, 23(2):197–249.
Ebenstein, A., Harrison, A., McMillan, M., Phillips, S., and Hall, G. (2009). Estimating the
Impact of Trade and Oﬀshoring on American Workers Using the Current Population Surveys.
NBER Working Paper.
Edwards, L. and Lawrence, R. Z. (2013). Rising Tide: Is Growth in Emerging Economies Good
for the United States? Peterson Institute for International Economics.
Egger, P., Pfaﬀermayr, M., and Weber, A. (2007). Sectoral adjustment of employment to shifts
in outsourcing and trade: evidence from a dynamic ﬁxed eﬀects multinomial logit model.
Journal of applied Econometrics, 22(3):559–580.
Feenstra, R. (2008). Oﬀshoring in the Global Economy (The Ohlin Lectures 2008). Stockholm
School of Economics.
Feenstra, R. and Hanson, G. (2001). Global production sharing and rising inequality: A survey
of trade and wages. NBER working paper.
Feenstra, R. C. (2010). Oﬀshoring in the global economy: microeconomic structure and macroe-
conomic implications. Cambridge Univ Press.
Freeman, R. B. (1995). Are your wages set in beijing? The Journal of Economic Perspectives,
Geishecker, I. (2008). The impact of international outsourcing on individual employment secu-
rity: A micro-level analysis. Labour Economics, 15(3):291–314.
Goos, M. and Manning, A. (2007). Lousy and lovely jobs: The rising polarization of work in
Britain. The Review of Economics and Statistics, 89(1):118–133.
Grossman, G. M. and Rossi-Hansberg, E. (2006). The rise of oﬀshoring: its not wine for cloth
anymore. The New Economic Geography: Eﬀects and Policy Implications, pages 59–102.
Grossman, G. M. and Rossi-Hansberg, E. (2008). Trading tasks: A simple theory of oﬀshoring.
American Economic Review, 98(5):1978.
Harrison, A. and McMillan, M. (2011). Oﬀshoring jobs? multinationals and us manufacturing
employment. Review of Economics and Statistics, 93(3):857–875.
Harrison, A. E. and McMillan, M. S. (2006). Dispelling some myths about oﬀshoring. The
Academy of Management Perspectives, 20(4):6–22.
Haskel, J. and Slaughter, M. (2001). Trade, technology and UK wage inequality. The Economic
Haskel, J. and Slaughter, M. (2002). Does the sector bias of skill-biased technical change explain
changing skill premia? European Economic Review, 46(10):1757–1783.
Haveman, J. (1994). The inﬂuence of changing trade patterns of displacements of labor. Purdue
University, Center for International Business Education and Research (CIBER).
Helpman, E., Itskhoki, O., and Redding, S. (2010). Inequality and unemployment in a global
economy. Econometrica, 78(4):1239–1283.
Hummels, D., Jorgensen, R., Munch, J., and Xiang, C. (2011). The wage and employment
eﬀects of oﬀshoring: Evidence from danish matched worker-ﬁrm data.
Kletzer, L. G. (1998). Job displacement. The Journal of Economic Perspectives, 12(1):115–136.
Kletzer, L. G. (2000). Trade and job loss in us. manufacturing, 1979-1994. In The Impact of
International Trade on Wages, pages 349–396. University of Chicago Press.
Kletzer, L. G. (2001). Job loss from imports: measuring the costs. Peterson Institute.
Krugman, P. R. (2008). Trade and wages, reconsidered. Brookings Papers on Economic Activity,
Lawrence, R., Slaughter, M., Hall, R., Davis, S., and Topel, R. (1993). International trade and
American wages in the 1980s: giant sucking sound or small hiccup? Brookings Papers on
Economic Activity. Microeconomics, 1993(2):161–226.
Levy, F. and Murnane, R. (2004). The new division of labor: How computers are creating the
next job market. Princeton University Press.
Liu, R. and Treﬂer, D. (2008). Much ado about nothing: American jobs and the rise of service
outsourcing to China and India. NBER Working Paper.
Mankiw, G. and Swagel, P. (2006). The politics and economics of oﬀshore outsourcing. Journal
of Monetary Economics, 53(5):1027–1056.
McKinney, K. and Vilhuber, L. (2011). LEHD infrastructure guiles in the census RDC: Overview
of S2004 snapshot. U.S. Census Bureau Center for Economic Studies Working Paper 11-13.
Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and aggregate industry
productivity. Econometrica, 71(6):1695–1725.
Menezes-Filho, N. A. and Muendler, M.-A. (2011). Labor reallocation in response to trade
reform. National Bureau of Economic Research Working Paper 17372.
Munch, J. R. (2010). Whose job goes abroad? international outsourcing and individual job
separations. The Scandinavian Journal of Economics, 112(2):339–360.
Navaretti, G. B., Castellani, D., and Disdier, A.-C. (2010). How does investing in cheap labour
countries aﬀect performance at home? ﬁrm-level evidence from france and italy. Oxford
Economic Papers, 62(2):234–260.
Pierce, J. R. and Schott, P. K. (2012). A concordance between ten-digit us harmonized sys-
tem codes and sic/naics product classes and industries. Journal of Economic and Social
Revenga, A. L. (1992). Exporting jobs? the impact of import competition on employment and
wages in us manufacturing. The Quarterly Journal of Economics, 107(1):255–284.
Stevens, D. (2002). Employment that is not covered by state unemployment insurance laws.
U.S. Census Bureau Longitudinal Employer – Household Dynamics Technical paper No. TP-
Verhoogen, E. A. (2008). Trade, quality upgrading, and wage inequality in the mexican manu-
facturing sector. The Quarterly Journal of Economics, 123(2):489–530.
Yeaple, S. R. (2005). A simple model of ﬁrm heterogeneity, international trade, and wages.
Journal of international Economics, 65(1):1–20.
List of Figures
1 The Rise of Low-Wage Import Competition . . . . . . . . . . . . . . . . . . . . . 20
Figure 1: The Rise of Low-Wage Import Competition
Import Competition Growth (1992=1)
1990 1995 2000 2005 2010
Lowwage Import Competition Other Import Competition
List of Tables
1 Low-WageCountries.................................. 22
2 Low Wage Import Competition Across U.S. Manufacturing Industries . . . . . . 23
3 Low-Wage Import Competition and Job Loss Among Workers with a High School
DegreeOrLess ..................................... 24
4 Low-Wage Import Competition and Job Loss Among Workers with a Bachelor’s
Table 1: Low-Wage Countries
Afghanistan Comoros Haiti Maldives Sao Tome
Bangladesh Congo Honduras Mali Sierra Leone
Bhutan Egypt India Mauritania Solomon Isl.
Benin Equatorial Guinea Indonesia Mozambique Somalia
Burkina Faso Ethiopia Kenya Myanmar Sri Lanka
Burundi Gambia Laos Nepal Sudan
Cambodia Ghana Lesotho Niger Tanzania
Central African Rep. Guinea Liberia Nigeria Togo
Chad Guinea-Bissau Madagascar Pakistan Uganda
China Guyana Malawi Rwanda Vietnam
Note: Classiﬁed according to the World Bank for the year 1992. N=51
Table 2: Low Wage Import Competition Across U.S. Manufacturing Industries
Low-wage Imports in Total (%) LW I COM P Empl Empl
Industry 1992 1997 2002 2007 1992 1997 2002 2007 ∆ ∆ (%)
Food 6 9 8 13 0.4 0.5 0.5 1.1 -23.2 -1.5
Beverage & Tobacco 1 2 2 1 0.0 0.2 0.2 0.2 -118.9 -33.9
Textile Mills 16 19 20 27 3.2 3.5 5.2 8.5 -322.1 -65.5
Textile Prod. Mills 44 45 53 73 8.6 7.7 14.1 32.8 -77.9 -33.1
Apparel 31 36 39 67 16.4 21.0 28.2 57.5 -688.2 -76.2
Leather 40 57 66 74 28.2 42.3 58.4 70.3
Wood 10 8 10 20 1.7 1.6 2.1 3.8 -25.9 -4.8
Paper 1 3 6 13 0.2 0.4 0.9 2.1 -189.0 -29.2
Printing & Related 10 11 22 36 2.2 1.7 5.3 14.2 -186.5 -23.1
Petroleum & Coal 4 9 4 3 0.4 1.3 0.8 0.6 -38.3 -25.1
Chemicals 2 4 4 6 0.4 0.7 0.9 1.8 -174.8 -16.9
Plastics & Rubber 7 15 20 31 0.8 1.6 2.8 6.1 -67.6 -8.2
Nonmetallic Minerals 8 16 24 30 1.4 2.6 4.5 6.3 -27.9 -5.3
Primary Metals 2 3 5 10 0.6 0.8 1.4 3.8 -232.8 -33.8
Fabricated Metals 7 11 19 29 1.1 1.6 3.5 6.9 -47.2 -2.9
Machinery 2 5 9 16 0.6 1.4 2.8 6.4 -222.7 -15.8
Computer & Electric 4 8 17 36 2.1 3.6 9.3 23.1 -630.0 -33.1
Electrical Equipment 11 19 28 34 2.9 5.3 10.1 16.5 -203.7 -32.2
Transport Equipment 0 1 1 3 0.1 0.2 0.5 1.7 -422.6 -19.8
Furniture 9 22 43 62 1.5 4.0 11.1 19.9 -72.3 -12.0
Miscellaneous 24 36 41 48 10.5 14.4 17.7 26.0 -44.0 -6.4
Mean 11 16 21 30 4 6 9 15 -191 -24
Standard Deviation 13 15 18 23 7 10 13 19 193 20
Note: Employment change and growth rates based on authors’ calculations using BLS data between 1992 and
2007. Trade and import competition based on authors’ calculation using Foreign Trade Imports and Exports
data as well as data from the Census of Manufactures. Low wage countries are deﬁned as being those deﬁned as
“Low-income’ in 1992, according to the World Bank’s country classiﬁcations.
Table 3: Low-Wage Import Competition and Job Loss Among Workers with a High School
Degree Or Less
Outcome: Job Loss (0=Kept Job; 1=Lost Job)
(1) (2) (3) (4) (5)
OLS Logit 2SLS FE FE 2SLS
LW I COM P 0.0508 0.4611 0.2661 0.1044 0.4273
(0.0370) (0.3336) (0.0955)*** (0.0622)* (0.1656)***
Computer Investment -0.0284 -0.2644 -0.0289 -0.0727 -0.0744
(0.0056)*** (0.0578)*** (0.0056)*** (0.0130)*** (0.0131)***
Shipments ($mil) -0.0109 -0.0931 -0.0108 -0.0200 -0.0194
(0.0033)*** (0.0345)*** (0.0033)*** (0.0064)*** (0.0060)***
Capital/Value-Added -0.0001 -0.0008 -0.0001 -0.0002 -0.0002
(0.0001)** (0.0002)*** (0.0001)** (0.0001)*** (0.0001)***
Multi-Unit Firm 0.0484 0.4192 0.0483 0.0343 0.0348
(0.0022)*** (0.0181)*** (0.0022)*** (0.0218) (0.0220)
Exports ($mil) -0.0182 -0.2290 -0.0172 -0.0893 -0.0871
(0.0187) (0.2497) (0.0187) (0.0203)*** (0.0193)***
SEIN Age/Rel. Age -0.0012 -0.0101 -0.0012 -2.0235 -2.0691
(0.0002)*** (0.0014)*** (0.0002)*** (1.7917) (1.8050)
Worker Age 0.0009 0.0077 0.0009
(0.0000)*** (0.0004)*** (0.0000)***
Female 0.0092 0.0764 0.0093
(0.0012)*** (0.0096)*** (0.0012)***
Foreign Born -0.0153 -0.1315 -0.0154
(0.0017)*** (0.0153)*** (0.0017)***
White -0.0086 -0.0717 -0.0087
(0.0014)*** (0.0116)*** (0.0014)***
Year Fixed Eﬀects Yes Yes Yes Yes Yes
Industry Fixed Eﬀects Yes Yes Yes Yes Yes
State Fixed Eﬀects Yes Yes Yes Yes Yes
Clustered SE @ SEIN Yes Yes Yes Yes Yes
Observations 1,00,4000 1,004,000 1,004,000 565,000 565,000
R-Squared 0.02 0.03 - 0.13 -
F-statistic (K-P) - - 202.13 - 95.6
Hansen J- - 0.021 - 0.778
Hansen J p-value - - 0.884 - 0.3777
Note: * signiﬁcant at 10%; ** signiﬁcant at 5%; *** signiﬁcant at 1%
Table 4: Low-Wage Import Competition and Job Loss Among Workers with a Bachelor’s Degree
Outcome: Job Loss (0=Kept Job; 1=Lost Job)
(1) (2) (3) (4) (5)
OLS Logit 2SLS FE FE 2SLS
LW I COM P -0.0057 0.0381 0.2207 0.0207 0.2824
(0.0469) (0.4057) (0.1172)* (0.0752) (0.2451)
Computer Investment -0.0213 -0.1770 -0.0220 -0.0563 -0.0577
(0.0079)*** (0.0724)** (0.0079)*** (0.0172)*** (0.0173)***
Shipments ($mil) -0.0104 -0.0883 -0.0102 -0.0174 -0.0171
(0.0032)*** (0.0341)*** (0.0032)*** (0.0078)** (0.0076)**
Capital/Value-Added -0.0002 -0.0010 -0.0002 -0.0001 -0.0001
(0.0001)*** (0.0003)*** (0.0001)*** (0.0001) (0.0001)
Multi-Unit Firm 0.0456 0.3826 0.0455 0.0531 0.0530
(0.0028)*** (0.0222)*** (0.0028)*** (0.0385) (0.0386)
Exports ($mil) -0.0206 -0.2632 -0.0196 -0.0792 -0.0780
(0.0153) (0.2010) (0.0153) (0.0244)*** (0.0237)***
SEIN Age/Rel. Age -0.0014 -0.0119 -0.0014 -3.4635 -3.5066
(0.0002)*** (0.0016)*** (0.0002)*** (1.9609)* (1.9647)*
Worker Age 0.0011 0.0090 0.0011
(0.0001)*** (0.0005)*** (0.0001)***
Female 0.0104 0.0835 0.0104
(0.0012)*** (0.0100)*** (0.0012)***
Foreign Born -0.0169 -0.1423 -0.0170
(0.0019)*** (0.0159)*** (0.0019)***
White -0.0086 -0.0689 -0.0086
(0.0016)*** (0.0127)*** (0.0016)***
Year Fixed Eﬀects Yes Yes Yes Yes Yes
Industry Fixed Eﬀects Yes Yes Yes Yes Yes
State Fixed Eﬀects Yes Yes Yes Yes Yes
Clustered SE @ SEIN Yes Yes Yes Yes Yes
Observations 850,000 850,000 850,000 550,000 550,000
R-Squared 0.02 0.02 - 0.14 -
F-statistic (K-P) - - 106.52 - 26.265
Hansen J- - 0.031 - 0.295
Hansen J p-value - - 0.891 - 0.5867
Note: * signiﬁcant at 10%; ** signiﬁcant at 5%; *** signiﬁcant at 1%. Standard errors for estimates for
Shipments and Exports are unavailable due to researchers’ error in disclosure process, however coeﬃcients and
signiﬁcance levels are correct, and the missing standard errors can be added in revision phase.