Wage distribution and firm size: The case of the United States

• Published date: 01 September 2018
• DOI: http://doi.org/10.1111/ilr.12109
• Date: 01 September 2018

International Labour Review, Vol. 157 (2018), No. 3

Copyright © The author 2018

Journal compilation © International Labour Organization 2018

Wage distribution and rm size:

The case of the United States

Damir COSIC*

Abstract. Substantial literature has been produced on the increasing wage gap

in the United States, invoking various possible factors, but largely ignoring the

relationship between rm size and wage distribution. In this study, the author

decomposes wage differences over time between large, medium and small rms,

identifying the effects of observed characteristics (and their returns) along with re-

sidual inequality, i.e. inequality among workers with the same observed character-

istics. From 19 92 to 2012 , trends at small, medium and large rms became more

uniform, while wage inequality rose across the board. Signicantly, it increased

more quickly in the upper half of the wage distribution and at large rms, where

residual inequality was highest.

A

large increase in wage inequality observed in the United States of

America in the 1980s and 1990s has generated substantial literature

on wage distribution. Most of this research has focused on changes in labour

supply (as the composition of the workforce with respect to education and ex-

perience has evolved) and demand (due to technological progress), and on how

labour market institutions have changed (for example, trends in unionization

rates and adaptation of regulatory frameworks).1 However, much of the in-

crease in wage inequality, referred to as residual wage inequality (or inequality

among workers with the same observed characteristics), remains unexplained.

Some of the early studies, such as those of Juhn, Murphy and Pierce (1993)

and Katz and Autor (199 9), estimate that residual wage inequality accounts for

more than half of the overall increase in wage inequality. More recent stud-

ies by Lemieux (2006) and Melly (2005) nd that changes in the composition

of the workforce have more signicant effects. According to their assessment,

which uses an improved methodology and accounts for measurement errors,

the increase attributable to residual inequality is around 20 per cent of the

total increase in inequality, which is still substantial.

* Urban Institute, Washington, DC, email: dcosic@urban.org.

Responsibility for opinions expressed in signed articles rests solely with their authors, and

publication does not constitute an endorsement by the ILO.

1 Studies that examine the effects of increased trade volume on wage inequality, such as

Borjas, Freeman and Katz (199 7), nd them relatively insignicant.

International Labour Review358

One factor that has been largely absent from the literature on wage

inequality is the relationship between wages and rm size. It is well estab-

lished that, on average, large employers pay higher wages than small employ-

ers. Brown, Hamilton and Medoff (1990) nd that in the 1980s, large rms

in the United States paid signicantly higher wages than small rms for the

same positions, occupations and observed characteristics of employees. In add-

ition to paying higher wages, large rms provided better working conditions

and more generous benets. Multiple explanations for this wage premium at

large rms have been proposed. According to some, the higher wages result

from the use of more advanced technology, higher productivity, lower costs of

other inputs or more efcient monitoring. No conclusive evidence has been

produced, though. The literature on wages and rm size (for a survey, see Idson

and Oi, 1999) falls short of discussing the broader issue of wage inequality in

relation with rm size.

This study aims to ll that void. I estimate wage distributions at small

and large rms2 in 1992 and 2012 and decompose the differences between the

two into effects of observed characteristics, returns to observed characteristics

and residual inequality. The wage distribution for each category of rm size is

estimated from the wages of workers employed by rms in that category. I nd

that wages have been distributed more unequally among workers employed

by small rms than among workers employed by large rms. The difference in

wage inequality between small and large rms was particularly pronounced in

1992, when the Gini coefcient of wages at rms with less than 100 employ-

ees was 0.399. At the same time, at rms with more than 1,000 employees, the

Gini coefcient was lower by more than 0.05, at 0.343 (see table 1). For com-

parison, the Gini coefcient of overall pre-tax income in the United States

increased by about the same amount, from 0.415 to 0.469, between 1984 and

2005, when there was rapid growth in income inequality.3 Between 1992

and 2012, overall wage inequality increased, but it grew faster at large rms. By

2012, the end of the observed period, the Gini coefcient had risen by 0.041,

to 0.440 at small rms, and by 0.072, to 0.415 at large rms.

Percentile ratios – a distributional statistic that, unlike the Gini coef-

cient, contains information about the shape of a distribution – reveal that the

difference in wage inequality between small and large rms originates mostly

in the upper half of the wage distribution. This is also where the most dramatic

changes took place over time. Even though the 50/10 percentile ratio4 at large

rms (0.896) was higher than at small rms (0.811) in 1992, indicating higher

2 For the purposes of this study, rms with 1–99 employees are designated as small, those

with 100-99 9 employees as medium-sized, and those with over 1,000 as large.

3 Source: United States Census Bureau, Current Population Survey, Annual Social and

Economic Supplements, 2017, available at: https://www2.census.gov/programs-surveys/cps/tables/

time-series/historical-income-households/h04.xls [accessed 21 June 2018].

4 The 50/10 percentile ratio is the wage or salaryincome earned by individuals at the

50th percentile (those earning more than 50 per cent of otherworkers) compared with the earnings

ofworkersat the 10th percentile (those earning higher than the bottom 10 per cent).