Protection and skills: Does trade policy favor low‐skill workers?

• DOI: http://doi.org/10.1111/roie.12406
• Author: Scott Bradford
• Published date: 01 August 2019
• Date: 01 August 2019

Rev Int Econ. 2019;27:981–1000. wileyonlinelibrary.com/journal/roie © 2019 John Wiley & Sons Ltd

|

981

1

|

INTRODUCTION

Many people believe that international trade unduly benefits the rich and that trade barriers provide

extra help for less‐skilled workers. Also, the empirical endogenous protection literature strongly im-

plies that protection decreases with skill levels. This paper presents evidence against this conventional

wisdom.

Regressions based on an explicit political economy model with search frictions and endogenous

unemployment imply that protection and average skill levels in U.S. industries have a mound‐shaped

relationship. The analysis indicates that, for the set of sectors whose average skill level is below a

certain cut‐off, protection increases with average skills, because, for these sectors, the hiring response

from a given price hike increases with that sector's skill level. This greater hiring response results from

the fact that jobs with higher skills give more value to those workers, making them more inclined to

accept an offer. The greater hiring response to a price increase, in turn, creates more political support

Received: 1 April 2018

|

Revised: 7 February 2019

|

Accepted: 12 April 2019

DOI: 10.1111/roie.12406

ORIGINAL ARTICLE

Protection and skills: Does trade policy favor low‐

skill workers?

ScottBradford

Economics Department,Brigham Young

University, Provo, Utah

Correspondence

Scott Bradford, Economics Department,

Brigham Young University, Provo, UT

84602.

Email: bradford@byu.edu

Abstract

Many believe that international trade barriers in rich nations

help low‐skill workers. Also, the empirical endogenous pro-

tection literature strongly implies that protection decreases

with skill levels. This paper presents evidence against this

conventional wisdom. Regressions based on a political

economy model with search frictions and unemployment

imply that protection and skill levels actually have a mound‐

shaped relationship: industries with skills in the middle

range get more protection than industries at either end of the

distribution. Apparently, the politics of trade policy trump

social concerns, rendering protection less “progressive”

than is commonly believed.

JEL CLASSIFICATION

F13; D72; D78

982

|

BRADFORD

for policy makers. Thus, they grant more protection to higher‐skilled industries, not less. It appears

that the politics surrounding trade policy can trump social concerns, rendering the structure of pro-

tection less “progressive” than many would like to believe. The recent tariffs imposed by the Trump

Administration against steel, aluminum, and Chinese goods provide anecdotal evidence in favor of this

conclusion, since those tariffs seem to favor workers with skills in the middle range, as opposed to the

lowest‐skilled workers. In constrast, I do find that the conventional wisdom holds for sectors above a

certain skill cut‐off: For these sectors, higher skill levels lead to less protection.

Many empirical studies have analyzed the skill–protection connection. Ray (1981), Marvel and

Ray (1983), Baldwin (1985), Trefler (1993), Gawande (1998), and Gawande and Bandyopadhyay

(2000) all find a negative correlation between skill levels and protection across U.S. industries. I know

of none that finds a significantly positive correlation. These regression analyses are enlightening,

but they are not derived from explicit models and thus have a higher chance of suffering from spec-

ification error than do analyses based on explicit modeling.1

Esfahani (2005) formally incorporates

skills into a protection model. In his model, sectors with lower skill levels face more vulnerability

to negative shocks, owing to incomplete insurance markets, and thus gain more from protection. He

finds, in accordance with the other literature, that having more workers whose skills are below that of

the highest group leads to higher protection. His paper, though, like all others known to me, considers

only a linear protection–skills relation.2

In this paper, I extend the previous literature by examining the possibility of a nonlinear rela-

tion between protection and skills within an estimating framework derived from an explicit protec-

tion model. The model follows Bradford (2006) and includes four groups: the government chooses

protection levels to maximize votes, producers maximize rents through lobbying, and workers and

consumers vote according to protection's effect on their economic welfare. Thus, this model has two

channels of government influence: lobbying and voting.3

Search‐generated unemployment affects vot-

ing. Politicians can win voting support, in addition to lobbying support, by imposing protection and

reducing unemployment. The model extends Bradford (2006) by incorporating skills: in the search

framework that describes the labor market, a price increase's effect on unemployment depends on

workers' average skill level.

Another concern with previous studies is how protection is measured. Older studies have used tar-

iff rates while more recent ones have generally used nontariff barrier (NTB) indices. Tariff rates alone

are inadequate, and NTB indices have troubling flaws. This study uses newer measures of protection

that I believe are more reliable.

2

|

THE MODEL

2.1

|

The economy

As with much of the rest of the literature, I assume that there are

n+1

goods and that all consumers

have identical quasilinear tastes given by

u

=z

0

+

∑n

i=1

u

i

(z

i)

, where

z

denotes the quantity consumed

of each good. Each

ui

is differentiable and strictly concave, and

z0

is a background numeraire good.

Each person has enough income to consume all goods. The quasilinear tastes imply that demand for

each good depends only on its price. I normalize the population to one. Thus, total consumer surplus

for each good is

cs(pi)=ui[di(pi)]−pidi(pi)

, where pi is the price, and

di(pi)

is demand.

For production, I modify the model in Bradford (2006), which, like Davidson, Martin and Matusz

(1999) (DMM), adapts the continuous time search model of Mortensen and Pissarides (1999) (MP)

to an international trade setting. Production results from workers matching with firms, and those

matches break up from time to time, requiring labor and capital to search again. In this paper, I plow