Offshoring and the Skill‐premium: Evidence from Individual Workers’ Data

• Published date: 01 October 2016
• Author: Tommaso Tempesti
• Date: 01 October 2016
• DOI: http://doi.org/10.1111/twec.12298

Offshoring and the Skill-premium:

Evidence from Individual Workers’ Data

Tommaso Tempesti

Department of Economics, University of Massachusetts Lowell, Lowell, MA, USA

1. INTRODUCTION

DURING the 1980s, wage inequality in the United States went up markedly. One of the

reasons behind this surge in inequality was the increase in the relative demand for skill.

1

Even though labour economists have not agreed on the relative importance of the possible

explanations for this phenomenon, they tend to rule out that international trade was an impor-

tant factor. In their undergraduate textbook, Ehrenberg and Smith (2005, p. 504) write that

the ‘findings among economists who have analyzed the effects of trade on inequality are not

unanimous, but the predominant conclusion is that the contributions of international trade to

the changes in wage inequality after 1980 were rather small’.

2

Indeed, early studies found the

importance of trade in final goods to be minor.

3

However, in a seminal paper, Feenstra and

Hanson (1999) (FH henceforth) find a sizable effect of trade in intermediate manufactured

inputs on the relative demand for skill in the United States during the 1980s. Given that this

recent result stands in contrast to the previous literature on the subject, the ‘case’ concerning

the importance of trade on wage inequality has been reopened. This work aims at bringing

new empirical evidence to this debate.

This paper studies the effect of offshoring on the skill-premium, using individual workers’

data from the March Current Population Survey (March CPS henceforth).

4

The advantage of

this strategy is twofold. First, the March CPS has been widely used by labour economists to

document the evolution of the skill-premium over time (See e.g. Katz and Autor, 1999). How-

ever, except for Ebenstein et al. (2009), whose contribution I discuss below, the Current Pop-

ulation Survey has not been used to address the importance of the effect of offshoring of

I thank the members of my Ph.D. dissertation committee at Vanderbilt University, Robert Driskill,

Yanqin Fan, Andrea Moro and Joel Rodrigue for their help and support on this project. I also thank

two anonymous referees, Eric Bond, Bob Hammond, Sangsoo Park, seminar participants at Vanderbilt

University, participants to the 3rd ACE International Conference in December 2009 in Hong Kong and

participants to the April 2011 Midwest Trade Meetings at the University of Notre Dame for their useful

comments and suggestions on previous versions of this paper. All errors are mine.

1

Acemoglu (2002a) documents the increase in the US skill-premium, defined as the wage difference

between workers with college education and workers with only a high-school degree. He also argues that

this increase is mainly due to an increase in the relative demand for skill rather than to a decrease in the

relative supply of skill.

2

In his review of the literature on the subject, Acemoglu (2002a, p. 52) argues that ‘increased interna-

tional trade by itself is not the cause of the changes in the US wage structure’ even though he then adds

that trade may have had a more indirect role on wage inequality by affecting the skill bias of

technological change.

3

See the studies discussed in Acemoglu (2002a, p. 53–54) An exception, not discussed in Acemoglu

(2002a), is Revenga (1992) who finds a significant impact of import prices of final goods on

employment and wages.

4

I use the term offshoring to indicate trade in intermediate inputs.

©2015 John Wiley & Sons Ltd

1628

The World Economy (2016)

doi: 10.1111/twec.12298

The World Economy

manufacturing on the skill-premium.

5

This may also have contributed to the fact that, even

after the contribution of FH, many labour economists seem to be still sceptical about the role

of trade on wage inequality in the United States.

6

It is natural to address this gap in the litera-

ture by making use of the March CPS data set.

Second, and more substantially, individual workers’ data allow me to assess a limitation of

FH’s contribution. Their work relies on workers’ data that are aggregated at the industry level,

which does not allow one to control for demographic compositional changes of the labour

force within industries that may be spuriously correlated with offshoring. To illustrate the

potential importance of this, consider the following example. We know as a stylised fact that

wages are increasing in work experience (see, for example, Murphy and Welch, 1990). Sup-

pose that, for reasons unrelated to offshoring, young graduates do not enter a certain industry

where offshoring happens to go up: we observe offshoring going up and the skill-premium

going up, but this is just a spurious correlation. In the above scenario, the estimat es in FH of

the effect of trade on wage inequality would be biased upwards.

To overcome this problem, I merge the March CPS data set, a large representative sample

of the US population, with industry-level variables such as offshoring trade and a proxy for

skill-biased technological change, which is usually indicated as one of the main factors driv-

ing the increase in the skill-premium. As FH, I focus on the 1979–90 period and on offshor-

ing trade in manufacturing, abstracting from the services sector.

7

In order to motivate my

regression approach, I use the simple model in Feenstra (2004, Ch. 4) to show how offshoring

may affect the relative wage of skilled labour.

Related papers are Lovely and Richardson (2000), Kosteas (2008) and Crin

o (2010).

Lovely and Richardson (2000) and Kosteas (2008) use, respectively, the Panel Study of

Income Dynamics and the National Longitudinal Survey of Youth to study the effect of inter-

national trade and offshoring on wages. These longitudinal data sets allow them to control for

individual fixed effects that may affect wages.

8

Using individual fixed effects, however, has

also its limitations, in that the skill-premium is identified by a small subset of workers, that

is, those workers whose education changes in the data set. Moreover, some of the recorded

5

The first paper that uses the March CPS to address the importance of globalisation on the skill-

premium in the United States is, to my knowledge, Goldberg and Tracy (2003). They study the effect of

exchange rate fluctuations, and not of offshoring trade per se, on the US skill premium. Liu and Trefler

(2011) study the effect of offshoring of services, not of manufacturing, on various labour market out-

comes. Liu and Trefler (2011) also focus on a more recent period than I do.

6

Autor et al. (2006), however, suggest that offshoring may be important to explain the pattern of wage

inequality during the 1990s and beyond. See also Autor et al. (2010).

7

Autor et al. (2008, p. 3) document that the pattern of wage inequality in the United States has changed

in the 1990s with respect to the 1980s. In the 1980s, the increase in inequality was monotonic: higher

incomes rose and lower incomes fell. In the 1990s, the gap between the 90th wage percentile and the

50th wage percentile continued to grow but the gap between the 50th wage percentile and the 10th wage

percentile stabilised. This evidence suggests that at the end of the 1980s, there may have been a struc-

tural break in the evolution of the US wage inequality. Autor et al. (2008) also document a similar pat-

tern for the skill-premium, not just for overall wage inequality. For this reason, I choose to use a trade

model that has the potential of explaining the increase in the skill-premium during the 1980s. Research

points to the importance of service offshoring for the explanation of US wage inequality since the 1990s

[see Crin

o (2009 p. 223), Ahmed et al. (2011) and Liu and Trefler (2011)].

8

The idea is that individuals are endowed with some constant productivity that is unobservable to the

econometrician. If more productive individuals also obtain more schooling, then the estimated return to

education will be biased. See Heckman et al. (2003) for a discussion of this point.

©2015 John Wiley & Sons Ltd

OFFSHORING AND THE SKILL-PREMIUM 1629