The Increase in College Premium and the Decline in Low‐Skill Wages: A Signaling Story

• Author: PAU BALART
• DOI: http://doi.org/10.1111/jpet.12146
• Published date: 01 June 2016
• Date: 01 June 2016

THE INCREASE IN COLLEGE PREMIUM AND THE DECLINE

IN LOW-SKILL WAGES:ASIGNALING STORY

PAU BALART

Universidad Carlos III de Madrid

Abstract

The increase in the college premium over the last 30 years in the

United States is to a large extent driven by a reduction in noncollege

wages. We show that the signaling effects triggered by an improvement

in the incentives to attend higher education can explain this fact, as

well as the increase in the number of college graduates. Under im-

perfect credit markets and wealth heterogeneity, higher education is

not only a signal of ability but also of individuals’ (parents’) wealth.

General conditions on the distribution of wealth guarantee that after

an increase in incentives to attend higher education, the absence of a

college degree becomes a more evident signal of low ability. This re-

sults in a reduction in low-skill wages but not necessarily in increased

high skill wages. The increased incentives to enroll in college can ei-

ther arise from a skill-biased technology change or an improvement in

access to higher education. An important difference is that while skill-

biased technology change always results in an increase in the college

premium, that is not necessarily the case when the increased incentives

to enroll arise from improved access to higher education.

1. Introduction

During the last part of the 20th and the beginning of the 21st centuries the supply of

college graduates and the returns to higher education have simultaneously increased in

many developed countries.1

Returns to higher education are measured by the college wage premium, that is, by

differences in earnings between college graduates and individuals with lower levels of

education. A remarkable fact that has received much less attention is that in the United

1The United States case has been widely documented and studied, see Acemoglu and Autor (2011) or

Katz and Autor (1999) among others. We report evidence for other countries in Section 2.

Pau Balart, Universidad Carlos III de Madrid, Madrid, Spain (pbalart@eco.uc3m.es).

I gratefully acknowledge the comments from Antonio Cabrales, Caterina Calsamiglia, Guillermo

Caruana, David Perez-Castrillo, and Pedro Rey-Biel as well as to the participants in the seminars in UAB,

UC3M, URV, SAEE ECORE. All errors are my own responsibility. This research has been supported by

the Spanish Ministry of Economy and Competitivity, through grant ECO2012-34581.

Received October 13, 2014; Accepted October 15, 2014.

C2014 Wiley Periodicals, Inc.

Journal of Public Economic Theory, 18 (3), 2016, pp. 363–384.

363

364 Journal of Public Economic Theory

States, the rise in the college premium is primarily driven by a reduction in low-skill

wages.2In the left-hand graph of Figure 1, we can observe the simultaneous increase in

the college premium and the proportion of higher education students that took place

in the United States in the period 1978–2008. Remarkably, as we can observe in the

right-hand graph of the figure, the difference between the median earnings of high-

skill (bachelor’s degree or higher) and low-skill workers (some college, high school, or

less than high school) has essentially widened as a consequence of the reduction in the

latter.3

We construct a signaling model with wealth heterogeneity to provide a supply-side

explanation for the facts presented in Figure 1: (i) the increase in the supply of college

graduates, (ii) the rise in returns to higher education and (iii) the decline in the

wages of low-skilled workers. Our explanation focuses on the ability composition effects

triggered by improved incentives to attend higher education. General conditions on

the joint distribution of wealth and ability are sufficient to guarantee the reduction in

low-skill wages.

Our model is a modified version of Spence’s (1973) signaling model that includes

wealth heterogeneity as in Hendel, Shapiro, and Willen (2005). There is a labor market

with competitive firms and heterogeneous workers. Workers are heterogeneous in two

unobservable aspects: ability and wealth. In contrast, an individual’s level of education

is observable. Workers’ productivity at firm level depends on their unobserved ability.

As higher education is more difficult for low-ability individuals to obtain, a college cre-

dential (partially) reveals the unobservable ability of individuals.

The signaling effect captured by the model operates as follows. Given the positive

correlation between education and unobserved productivity, wages are equal to the

average productivity conditional on educational attainment. Therefore, the wage at

a particular level of education depends on the proportion of high-skill and low-skill

workers that attain that education level. Nevertheless, if there are imperfect credit

markets and wealth heterogeneity, higher education is not only a signal of ability

but also of individuals’ (parents’) wealth. The presence of this additional source of

heterogeneity may distort the signaling consequences of increased enrollment in

higher education. Consider a case in which individuals can only be wealthy or poor

and of high or low ability. If higher education becomes more attractive, then both

high-ability but low-wealth individuals and high-wealth but low-ability individuals might

move from the noncollege-educated to the college-educated share of the population.

Consequently, the change in the ability composition of the higher educated workforce

is ambiguous, as is the change in the signaling power of higher education. However, in

contrast, only a minority of low-ability individuals with low wealth remains low skilled.

This makes the absence of higher education a more evident signal of low ability, driving

noncollege wages down toward the low-ability productivity level.

As we show, a monotone likelihood ratio and log-concavity of the cumulative dis-

tribution of wealth are sufficient conditions to realize the intuition described in the

2We use the terms “college” and “high-skill” interchangeably, as well as “noncollege” and “low-skill,”

throughout the paper.

3The college wage premium is calculated as the logarithm of the ratio between college and high school

median earnings. According to the theory that the signaling effect is more relevant for younger grad-

uates (Altonji and Pierret 2001), and because changes in women’s earnings can be highly influenced

by their progressive participation in the labor market, we have considered median earnings for males

aged 25–34. The enrollment rate has been computed as the ratio between college students and the pop-

ulation aged 18–24. All data were obtained from the US National Center for Education and Statistics

(NCES).