Offshoring and skill overlap: An empirical investigation

• Date: 01 September 2019
• Published date: 01 September 2019
• DOI: http://doi.org/10.1111/roie.12423

Rev Int Econ. 2019;27:1199–1233. wileyonlinelibrary.com/journal/roie

|

1199

© 2019 John Wiley & Sons Ltd

1

|

INTRODUCTION

There are two contrasting theories about the relationship between the amount of offshoring and the

relative factor endowments between two countries. In the traditional Heckscher–Ohlin–Vanek (HOV)

model and the Feenstra and Hanson (1996) model, international trade and offshoring will be greatest

when countries’ relative factor endowments are most different. In contrast, Helpman (1987), who ex-

amines intra‐industry trade, Markusen (1995), who examines foreign direct investment, and Markusen

and Venables (2000), who examine multinational activities, note that the share of intra‐industry trade,

Received: 15 August 2018

|

Revised: 9 May 2019

|

Accepted: 13 May 2019

DOI: 10.1111/roie.12423

ORIGINAL ARTICLE

Offshoring and skill overlap: An empirical

investigation

JaerimChoi

Department of Economics,University of

Hawaii at Manoa, Honolulu, HI

Correspondence

Jaerim Choi, Department of Economics,

University of Hawaii at Manoa, Saunders

Hall 542, 2424 Maile Way, Honolulu, HI

96822.

Email:choijm@hawaii.edu

Abstract

Conventional international trade theory predicts that bilat-

eral offshoring flows will be highest when two countries

have very different relative factor endowments. In con-

trast, the new trade theory contends that offshoring is more

likely to exist when countries’ relative factor endowments

are similar. This paper empirically tests the relationship be-

tween offshoring and relative factor endowments, measured

by the skill overlap index between two countries and finds

evidence that there is an inverted U‐shape relationship. Our

empirical results predict that the rise in educational attain-

ment in China will motivate U.S. multinationals to send

their tasks to China in the short run; over the long run, how-

ever, U.S. multinationals will have fewer incentives to send

their tasks to China. This finding sheds new light on the

current trade tensions between the United States and China

and has implications for trade policy.

JEL CLASSIFICATION

F14; F23; J24

1200

|

CHOI

foreign direct investment, and multinational activities, respectively, are more likely to exist when

two countries have similar relative factor endowments.1

The first theoretical prediction implies that

international trade and offshoring flows will be most significant between developed countries like the

United States and developing countries like Malawi; the second theoretical prediction suggests that

flows will be highest between developed countries.

Kremer and Maskin (1996, 2006) propose a matching model of production by workers of different

skill levels that rebuts the first theoretical prediction. If the skill levels in two countries are sufficiently

different, then it is inefficient for workers in the skill‐abundant country to match with workers in the

skill‐scarce country. Hence, the model predicts that there will be little offshoring between developed

countries and developing countries. With respect to the second theoretical prediction, the matching

model of Kremer and Maskin (1996, 2006) neither supports nor refutes the idea that workers in skill‐

abundant countries are more likely to match with workers in other skill‐abundant countries.

However, when we turn our attention to observed offshoring flows in the U.S. data, we see that

the United States shifts most of its production activities to countries such as Brazil, China, Mexico,

and India, whose relative factor endowments are not too similar or not too disparate. The existing

theories fail to explain this observed pattern of multinational activity. Some may challenge this view

by maintaining that, after controlling for other confounding factors such as quality of institutions, the

relationship between the amount of offshoring and the relative factor endowments between two coun-

tries would still be linear, either increasing or decreasing, which supports the traditional view. Others

may disagree with this view because U.S. offshoring cannot be generalized to global offshoring flows.

In this paper, we set up a formal empirical specification to uncover the relationship between the

amount of offshoring and relative factor endowments, measured by the skill overlap between two

countries. The major challenge in estimating this relationship is the limited data. Unlike for interna-

tional trade flows, there are few available data sources that capture bilateral offshoring flows.2

We

overcome the data limitation by using two different data sources, both of which are relatively new

and unexploited, to capture bilateral offshoring flows: Ramondo, Rodríguez‐Clare, and Tintelnot’s

(2015) multinational production dataset, and the U.S. Bureau of Economic Analysis Direct Investment

& Multinational Enterprises (MNEs) dataset. Both measures are based on multinational activities

and are more appropriate objects to capture offshoring than the foreign direct investment (FDI) flow

measures that are commonly used in the literature, such as the OECD’s Foreign Direct Investment

Statistics.3

The advantage of the Ramondo et al.’s (2015) multinational production dataset is its com-

prehensive coverage of 59 countries with 3,422 (58 × 59) possible bilateral observations. The U.S.

BEA dataset is larger, covering 199 countries, but is limited to observations that are bilateral with the

United States.4

We develop a novel measure, the skill overlap index, using the Barro and Lee (2013) Educational

Attainment Dataset, to proxy for the relative factor endowment variable. The Barro and Lee (2013)

dataset provides data for 146 countries on the distribution of educational attainment of the population

over age 15 at seven levels of schooling: no formal education, incomplete primary, complete primary,

lower secondary, upper secondary, incomplete tertiary, and complete tertiary. For all country pairs, we

calculate the overlapping area of two skill distributions, using a range from 0 (for no skill overlap) to 1

(for complete skill overlap). The newly developed skill overlap index provides a more precise measure

of the relative factor endowment difference between two countries than the average‐years‐of‐school-

ing difference that is widely used in the literature. Suppose, for example, that the skill distributions of

two countries have the same average but different variances. If researchers measure the skill difference

between the two countries based on the average years of schooling, the two countries will be shown

to have the same skill level. However, the skill‐overlap index yields different skill levels for two coun-

tries, which captures beyond the first moment of the distribution.

|

1201

CHOI

In this study, we adopt a structural gravity model of FDI developed by Head and Ries (2008).5

The theoretically based gravity model of FDI yields an equation for bilateral FDI as a function of

source‐country fixed effects, host‐country fixed effects, and a vector of pair‐specific variables such

as distance, contiguity, language, and colonial link.6

We base our estimation framework on Head and

Ries (2008) and incorporate the newly developed skill overlap index and its squared term into the

equation to test the relation between bilateral offshoring flows and relative factor endowments. We

then employ Silva and Tenreyro’s (2006) method of Poisson pseudo‐maximum‐likelihood (PPML) to

estimate the offshoring gravity equation.7

We first specify a linear fitting between offshoring and skill overlap using Ramondo etal.’s (2015)

multinational production dataset and find that the coefficients of interest are not statistically sig-

nificant. This result implies that neither the conventional trade theory nor the new trade theory can

explain the relation. Then we set up a quadratic relation and find compelling evidence that there is an

inverted U‐shape relation between offshoring and skill overlap between two countries. Next we ana-

lyze the relationship using the U.S. BEA data and still find an inverted U‐shape relation. This result

stands up to several robustness checks.

Based on the estimated hump‐shaped pattern, we ask whether U.S. multinationals are more or

less likely to send their tasks to China in the future, given that China is used to being blamed for

taking away U.S. jobs. Our empirical results predict that more educational attainment in China will

motivate U.S. multinationals to send more tasks to China in the short run. However, in the long run,

more human capital accumulation in China will discourage U.S. multinationals to send their produc-

tion tasks to China. Admittedly, the prediction is based on the assumption that everything remains

unchanged except the relative factor endowment difference, which narrows between the United States

and China.

Our empirical offshoring gravity model is related to important early work that uses the grav-

ity equation to investigate the determinants of FDI. Carr, Markusen, and Maskus (2001) present a

“knowledge‐capital model” of multinationals and estimate multinational activity between countries as

a function of country characteristics. Bergstrand and Egger (2007) add physical capital to the “knowl-

edge‐capital model” and test their propositions using gravity equations for FDI flows. Although the

main focus of their studies is not the relation between offshoring and relative factor endowments, their

empirical gravity equation estimation results reveal that the relation between FDI flows and skilled‐

labor abundance between two countries is positive. We extend these earlier studies to incorporate the

squared term of the skill overlap index into the gravity equation and find that the relationship is an in-

verted U‐shape. Also, we expand our dataset to cover a larger number of country pairs than the previ-

ous studies,8

and use a more precise measure of the relative factor endowment between two countries.

2

|

BACKGROUND

Throughout the 2016 U.S. presidential election campaign, candidate Donald Trump appealed to U.S.

voters by saying he would “bring back U.S. jobs” that had been outsourced to foreign countries such

as China. After his election, in early 2018, the United States placed tariffs on products produced out-

side the United States, such as solar panels, steel, and aluminum, with the aim of making imported

goods more expensive and bringing back manufacturing to the United States that had been outsourced

to China and other countries. Trade tension between the United States and China is an ongoing issue,

and whether a protectionism policy is the best way to restore U.S. jobs is also the topic of heated de-

bate. From an academic perspective, multinationals’ shifting of tasks from the United States to other

countries, that is, offshoring, is also an interesting research topic. To investigate this topic further, it