CHAPTER 9

9-1 Race and Gender in the Labor

Market
9-2 The Discrimination Coefficient
9-3 Employer Discrimination
9-4 Employee Discrimination
9-5 Customer Discrimination
9-6 Statistical Discrimination
9-7 Experimental Evidence on
Discrimination
9-8 Measuring Discrimination
9-9 PA: Determinants of the Black-
White Wage Ratio
9-10 Discrimination against Other
Groups
9-11 Determinants of the Female-
Male Wage Ratio
McGraw-Hill/Irwin Copyright © 2013 by The McGraw-Hill Companies, Inc. All rights reserved.
 Introduction
• Differences in earnings and employment opportunities may
be caused by economic factors such as differences in job
characteristics and workers’ skills.

• However, there should not be differences in labor market

outcomes among equally skilled workers in the same job

• Discrimination occurs when the marketplace takes into

account such factors as race and sex when making
economic exchanges.

9-2
 Introduction
In this chapter, we will cover topics such as:

 Empirical evidence on earnings differences across race

and gender
 Economic framework for analyzing discrimination
 Types of discrimination and how to measure them
 Impact of public policy on the well-being of discriminated
groups

9-3
9-1 Race & Gender in the Labor Market

• Men earn more than women, and whites usually earn

more than non-whites (except Asians).
• Differences in annual earnings are partially explained by
labour supply and educational attainment differentials.
(51% gap in Annual earnings vs 40% gap in FT workers)
9-4
9-1 Race & Gender in the Labor Market

1) Considerable wage gap between men and women in most

developed countries.
2) Differences not only in the size of the gender wage gap & but
also the employment rates of men and women.
3) Gender wage gap is higher in countries where employment
gap across gender is smaller (-ve correlation). 9-5
 9-2 The Discrimination Coefficient
Gary Becker’s PhD Dissertation (1957):
The Economics of Discrimination
• Taste Discrimination: translates the notion of racial
prejudice into the language of economics
• Two types of workers: Whites (W) and Blacks (B)
• Wage rates: WW and WB respectively (competitive
employer)
• Employer: If prejudiced against blacks, gets disutility
from hiring black workers, perceive & act as if the cost
of hiring a black worker is:
WB(1+d) rather than WB
where d is called the “discrimination coefficient”
9-6
 9-2 The Discrimination Coefficient
• Employer: If prejudiced against blacks, perceive & act as
if the cost of hiring a black worker is:
WB(1+d) where d is discrimination coefficient
Example:d=0.5 and WB=$10
If prejudiced, perceive the cost of hiring a black worker
as $15
• Similarly, if an employer prefer to hire blacks (nepotism),
cost of hiring would be perceived as:
WB(1 – n) where n is nepotism coefficent
• 3 types of discrimination: Employer, Employee, Statistical

9-7
 9-3 Employer Discrimination
• Main assumption:
Blacks and whites are perfect substitutes
q=f(LW+LB) (B and W are equally productive)

Hence, same output with either of the combinations below:

LW=0 LB=100, LW=50 LB=50 or LW=100 LB=0

• Competitive employer: Constant input prices (WB & WW)

• A color-blind (not prejudiced) employer will hire whichever

group is cheaper.

 Hence, hire only blacks if WB<WW and vice versa.

 Hire up to the point where WB=VMPL. 9-8
 The Employment Decision for a
Firm That Does Not Discriminate
Dollars
 If the market-determined black
wage is less than the white
wage (WB<WW), a firm that
does not discriminate will hire
only blacks.

wB  It hires black workers up to the

point where the black wage
VMP E
equals the value of marginal
product of labor, L*B.
LB
Employment

9-9
 9-3 Employer Discrimination
• A prejudiced (against blacks) employer perceives the
cost of a black worker as:
WB(1+d) where d>0

• Hiring decision of a prejudiced employer:

 Hire only blacks if WB(1+d) < WW
 Hire only whites if WB(1+d) > WW

• Hence, assuming that employers have mixed tastes (color-

blind and prejudiced):

9-10
 9-3 Employer Discrimination
Implications of the Becker model
1) If blacks and whites are perfect substitutes, employers
have a segregated work force.

2) Level of employment depends on the discrimination

coefficient. Larger d, smaller is the number of hires.
Prejudiced employers hire the wrong type of worker
and/or they hire the wrong number of workers.
9-11
 9-3 Employer Discrimination
Implications of the Becker model
3) Crime does NOT pay: πColor-Blind > πPrejudiced

a) White firms: “High level of prejudice”, high costs (WW)

and sub-optimal levels of output (LW < L*)
b) Black firms: Prejudiced firms (low or medium) hire
wrong (smaller – suboptimal) number of workers!
9-12
 Determination of Black/White Wage Ratio
in the Labor Market
Black-White Wage Ratio  If the black-white wage ratio is
very high, no firm wants to hire
blacks.
S
 As the black-white wage ratio
falls, more and more firms are
(wB/wW) compensated for their disutility
1 and the demand for black
R D workers rises.
 The equilibrium black-white
(wB/wW)*
wage ratio is given by the
D intersection of supply and
demand, and equals (wB/wW)*.
0 N
Black Employment

9-13
 Determination of Black/White Wage Ratio
in the Labor Market
Black-White Wage Ratio
 If some firms prefer to hire blacks,
they would be willing to hire
S
blacks even if the black-white
wage ratio exceeds 1, shifting the
demand curve up to D.
(wB/wW)
1
D
 If the supply of blacks is
R sufficiently small, it is then
possible for the black-white wage
(wB/wW)*
ratio to exceed 1.
D
0 N
Black Employment

9-14
 The Black-White Wage Ratio
in the Labor Market
• Employer discrimination generates a wage gap
between equally skilled black and white workers.

• The quantity demanded for black labor increases

as the black-while wage ratio falls.

9-15
 9-4 Employee Discrimination
• Suppose that the source of discrimination is the
employees (rather than the employer)
• Moreover, assume that blacks are indifferent to working
with whites, however, white workers dislike working with
black workers.
• Hence, if a firm adopts a mixed racial composition in its
workforce, white workers perceive their income to be
lower than it actually is:

WW(1-d)
where d is the discrimination coefficient

9-16
 9-4 Employee Discrimination
• Hence, a color-blind firm would find a multi-racial workforce
suboptimal, since it would have to pay a wage premium to
white workers, which would lower its profits.
• Therefore, a color-blind firm would hire black or white
workers only! (whichever is cheaper). Employee
discrimination implies a completely segregated workforce!
• There will be no wage gap. Some hire only Blacks and
others hire only Whites, but wages will be equal in the
equilibrium as long as skills are identical (same VMP):
WW=WB
• Consequently, employer discrimination does not affect
profitability.
9-17
 9-5 Customer Discrimination
• Customers may have a taste for discrimination.
• In particular, customers may derive disutility from
purchasing goods and services sold by minorities.
• Hence, they may perceive the prices of these goods and
services to be higher than they actually are:
p(1+d) where d>0
• Profit maximizing firms will react to this via allocating black
workers to positions that require little/no customer contact.

9-18
 9-5 Customer Discrimination
• In effect, the employer segregates the workforce so that
white workers fill sensitive positions such as sales and black
workers fill non-sensitive positions such as production.
• Equally skilled black and white workers would receive the
same wage.
• Catering to customer tastes doesn’t reduce the firm’s profits.
• However, if a (competitive) firm cannot hide black workers,
then it would have to lower the price, which would lower the
profits.
• Therefore, customer discrimination can have an adverse
effect on black wages.

9-19
 9-5 Customer Discrimination

• Looking at contact firms, firms with predominantly black

customers hire significantly more black workers than firms
with predominantly white customers.
• However, to assess the true difference, we need to compare
the difference in the hiring behavior of firms with “no
contact” firms.
• Difference-in-Differences estimate of the impact of customer
discrimination is much smaller (14.6%).
9-20
 9-6 Statistical Discrimination
• Statistical discrimination is based on treating an individual
on the basis of membership in a group and knowledge of
that group’s history.
• Example: 2 candidates for a job (Software Development)
 Identical resumes
 Same interview performance (flying colors)
 Same education (college graduates with same field)
 Enrollment in same courses with similar class rankings
 Both perceived to be bright, motivated and knowledgeable
 Both assert that they intend to stay at the firm for the next
years
 Only difference: One man and the other a woman

9-21
 9-6 Statistical Discrimination
• To make an informed decision, the employer will evaluate
the employment histories of similarly situated men and
women that this firm hired at the past.
• Suppose that statistical evidence from the employment
records show that many women leave the firm when they
reach their late twenties.
• The employer has no way of knowing whether the female
candidate intends to leave the firms in a similar manner.
• Nevertheless, the employer infers from the statistical data
that the woman has a higher probability of quitting her job
prior to the completion of the software program.
• The profit-maximizing employer offers the job to the man.
9-22
 9-6 Statistical Discrimination
• Statistical discrimination arises because of the underlying
uncertainty.
• The employer is encouraged to use statistics about the
average performance of the group to predict a particular
applicant’s productivity.
• Statistical discrimination arises not only in the labour market
but in many other markets as well.
• Insurance companies:
 Premiums and Life expectancy (Males vs Females)
 Premiums and Auto Insurance (Teenagers vs Adults)

9-23
The Impact of Statistical Discrimination on Wages
Dollars Dollars

White White

Black Black

T* Test Score T Test Score

(a) Whites have higher average score (b) Test is better predictor for white workers

 Suppose all the information gathered for a candidate is summarized with

a test score (T).
 Uncertainty of the test score’s accuracy to predict personal productivity
 Employers may link productivity to the individual’s test score (T) and the
average group score (Tave):

W = α T + (1- α) Tave
9-24
The Impact of Statistical Discrimination on Wages
Dollars Dollars

White White

Black Black

T* Test Score Test Score

T
(a) Whites have higher average (b) Test is better predictor for white workers
score
The worker’s wage depends not only on his own test score, but also on the
mean test score of workers in his racial group.
(a) If black workers, on average, score lower than white workers, a white
worker who gets a score of T* earns more than a black worker with the
same score.
(b) If the test is a better predictor of productivity for white workers, high-
scoring whites earn more than high-scoring blacks, and low-scoring
whites earn less than low-scoring blacks.
9-25
 9-7 Experimental Evidence on Discrimination
• Difficult to measure a particular employer’s discrimination
coefficient since it is illegal to discriminate.
• A number of studies conducted labor market experiments
to bypass this measurement problem.
 Employers are contacted at random.
 5,000 fake resumes sent to 1,300 real job ads at Boston and
Chicago newspapers.
 Resumes did not specify the applicant’s race but included
hints (i.e. name: Lakisha Washington vs Jamal Jones).
 Holding the skills constant, the applicants with black-
sounding names got 1 callback for every 15 resumes sent.
 A black applicant needed 8 more years of work experience
to even out the gap.
9-26
 9-7 Experimental Evidence on Discrimination
• Experimental approach extended beyond fake resumes to
experimental human beings sent to actual job interviews.
• Summer 1989, hiring audit conducted in Chicago and San
Diego areas and 360 firms were audited.
• Average job applicant was…
 a neatly dressed 22 year old man,
 with a high school diploma,
 did not have a criminal record,
 had some college credits,
 some work experience as a stockperson or waiter
 Only difference is racial and visual indicators (Hispanic
looking with an accent vs White)
9-27
 9-7 Experimental Evidence on Discrimination
• Systematic differences were found in the way that
employers responded.
 White job applicants was 33 percent more likely to be
interviewed,
 And 52 percent more likely to receive a job offer.

9-28
 9-8 Measuring Discrimination
• One possible measure of discrimination is the difference
in mean wages:

• The definition is unappealing because it is comparing

apples and oranges.
• Many factors, other than discrimination, generate wage
differentials between men and women.
• A better measure would compare the wages of equally
skilled workers.
9-29
 9-8 Measuring Discrimination
• Suppose that only one variable, schooling affects earnings

Male earnings function: wM = αM + βM SM

Female earnings function: wF = αF + βF SF

• “β” measures how much a wage increases if an individual

gets one more year of schooling
• If employers value the education acquired by both gender
equally, βM = βF

• Raw wage differential can be written as follow:

9-30
 9-8 Measuring Discrimination
Oaxaca decomposition (Ronald Oaxaca):
• A technique that decomposes the raw wage differential into
two components:
i. a portion related to a difference in skills.
ii. and a portion attributable to labor market discrimination.
• Consider the raw wage differential:

• Lets add & subtract the term to the right hand side:

9-31
 9-8 Measuring Discrimination
Oaxaca decomposition (Ronald Oaxaca):

• A technique that decomposes the raw wage differential into

two components:
1) a portion related to a difference in skills.

2) and a portion attributable to labor market discrimination.

9-32
 Measuring the Impact of Gender
Discrimination on the Wage
 The average woman has years
Differential of schooling and earns
attributable to
skill differential  The average man has years of
schooling and earns
 Part of the wage differential arises
because men have more
Should schooling than women.
be  If the average woman was paid as
if she were a man, she would earn
w*F.
Actual
 A measure of discrimination is
then given by

9-33
Does the Oaxaca Decomposition Really
Measure Discrimination?
 Validity of this decomposition depends on whether all the
dimensions in which the skills of the groups differ have
been controlled.
 If there are some skills characteristics (ability, effort,
motivation) that affect earnings but are left out, there will
be an incorrect measure of labor market discrimination.
 For example, if men & women attend systematically attend
institutions that vary in quality, the Oaxaca decomposition
generates a biased measure of discrimination.
 If blacks attend lower-quality schooling than whites, it
would be incorrect to label wage differences between
workers with same level of schooling as discrimination.
9-34
 9-9 PA: Determinants of the
White-Black Wage Ratio
The Oaxaca Decomposition of
the Black-White Wage Differential, 1995
Controls for Differences Controls for Differences
in Education, Age, Sex, in Education, Age, Sex,
and Region of Region, and Occupation
Residence and Industry
Raw log wage differential -0.211 -0.211
Due to differences in skills -0.082 -0.144
Due to discrimination -0.134 -0.098

• Magnitude of the differentials attributed to discrimination

(13.4% vs 9.8%) depends on the the list of controls used.
• On the other hand, part of the differences may be due to
employment barriers that prevent blacks into moving certain
type of jobs? 9-35
• Overall, long-run trends are clear: The relative wages of
black men and women are substantially higher today than
they were in the late 1960s.
 In 1967: Ratio of 0.65
 In 1980: Ratio of 0.71
 In 2009: Ratio of 0.77
9-36
 9-9 PA: Determinants of the
White-Black Wage Ratio
• There is a number of hypotheses proposed to explain
the improving economic status of African Americans:

1) Increases in the quality and quantity of black schooling.

2) Impact of Affirmative Action
3) The Decline in the LFPR of Black
4) Unobserved Skill Differences

9-37
 9-9 PA: Determinants of the
White-Black Wage Ratio
Increases in the quality and quantity of black schooling
Quantity of schooling:
Years of schooling obtained by a typical 30 yr old man
In 1940: white man 9.9 yrs vs black man  6.6 yrs
In 1980: white man 13.6 yrs vs black man  12.2 yrs
Quality of schooling:
Rate of return to school for a worker who entered the labor
market
In 1940: white man 9.8% vs black man  4.7%
In 1970: white man 8.5% vs black man  9.6%

9-38
 9-9 PA: Determinants of the
White-Black Wage Ratio
Impact of Affirmative Action
 Enactment of 1964 Civil Rights Act: Prohibits employment
discrimination on the basis of race and sex.
 The federal civil rights program prohibited discrimination
(by race and sex) among government contractors.
 Executive orders brought by this program compel federal
contractors to not discriminate and take affirmative action
to ensure they don’t.
 Response: Detailed affirmative action plans with
employment goals (such as quotas) for affected groups
and timetables for meeting these goals.
9-39
 9-9 PA: Determinants of the
White-Black Wage Ratio
Impact of Affirmative Action
 Likelihood of employment at federal contractors
 In 1966: Black men were 10% less likely to work in
federal contractors
 In 1980: Black men were 25% more likely to work in
federal contractors
 Fraction of black employment at federal programs: Textile
industry in South Carolina – main manufacturing employer
 Between 1910&1964: The fraction of black employment
4-5%
 By 1970: Nearly 20% of the workers in the industry were
black
9-40
 9-9 PA: Determinants of the
White-Black Wage Ratio
The Decline in Black LFP

 In 1950: 85% of the black and white participated in the LF

 By 2009: Gap between LFPR of white and black workers
were over 7%.
9-41
The Decline in Black LFP
The increase in the
relative wage of black
men may be an illusion
created by sample
selection bias
(i.e. not an
improvement in
employment
opportunities but due
to low income blacks
no longer participating
in the Labor Force.)

 Suppose that the decline in the LFPR is due to an increase

in the reservation wage (due to large scale public
assistance programs in 1960s)
 If the portion of blacks that drop out of the LF are relatively
low skill, this would to a higher average wage.
9-42
 9-9 PA: Determinants of the
White-Black Wage Ratio
Unobserved Skill Differences & the B-W Wage Differential
 There may be other unobserved skill differences between
two groups that may account for the wage differential
labeled as discrimination.
 Some studies use a particular measure of skills:
 Armed Forces Qualification Test (AFQT)
 Standard test given to all recruits in the U.S. military
 Substantial racial differences in the AFQT scores
 Blacks tend to have lower scores than whites
 Racial differences in AFQT scores practically account for
the entire wage differential between black and white
9-43
 9-9 PA: Determinants of the
White-Black Wage Ratio
Unobserved Skill Differences & the B-W Wage Differential
 Given the same AFQT score, a black worker only earns 5%
less than a white worker
 What exactly does the AFQT measure?
 Convincing evidence that it is not a straightforward measure
of innate ability
 Persons who have more schooling or go to better schools
have higher scores
 Partly measures skills acquired prior to labor market entry
 Hence, these studies indicate that much of the
discrimination is attributable to skills differentials between
these groups (observed and unobserved).
9-44
9-10 Discrimination Against Other Groups
• Astounding growth of the Hispanic population in the U.S.:
 U.S. Population (1980) Hispanics 6.4% vs Blacks 11.7%
 U.S. Population (2002) Hispanics 13.4% vs Blacks 12.7%

 Declining earnings ratio for Hispanic men and women.

9-45
9-10 Discrimination Against Other Groups

• Differences in wages can be linked to varying educational

attainment: Less skilled workers earn less
• Hispanics have a higher % of HS dropouts and lower % of
college graduates relative to white workers
• Therefore, adverse changes in earnings ratio of Hispanics
are due to observed skills differences
9-46
 9-11 PA: Determinants of the Male-Female Wage Ratio
The F-M Wage Gap and Labor Market Experience

• Decompositions conducted above ignore a key determinant of

female earnings: Differences in labor market histories
• Women drop out of the labor market during child-raising years.
• By late 1980s, % of potential years worked:
Typical man 93% vs Typical woman 71%
• Argument: The value of woman’s human capital is reduced by
her intermittent labor market attachment.

9-47
• Hypothesis: Discontinuity in female LS over the life cycle
generates a gender wage gap for two reasons:
1) Relatively, males acquire more human capital
2) The child-raising years increase the wage gap because
woman’s skills tend to depreciate during that period.
• Evidence supports the hypothesis but disagreement of
the magnitude explained by labor market histories:
• U of Michigan law school classes of 1973 and 1975.
Annual earnings of attorneys 15 years after graduation:
 Males: $141,000 vs Females: $86,000
 2/3 of the gap can be explained by labor market history
 Females worked PT for 3 years during child-raising years
 Earnings permanently reduced by 17%!!!
9-48
Occupational crowding

Occupational crowding hypothesis: Women are intentionally

segregated into particular occupations.
9-49
Occupational crowding
• Crowding may simply the result of a social climate in which
young women are taught that the some occupations are “not
for girls” and channeled into “appropriate” jobs.
• The crowding of women into a relatively smaller number of
occupations inevitably reduces the wage of so-called female
jobs and generate a gender wage gap.
• Studies typically find that “female jobs pay lower wages” even
after holding constant the worker’s human capital and other
factors:
 Typical man/woman in a “female job” earn 14% less than a
typical man/woman in a “male dominant” job
 It is the femaleness of the job that leads to lower wages
rather than the gender of the person in that job.
9-50
Trend in Female –Male Earnings Ratio

 F-M wage ratio hovered around 60% between 1960-1980s.

 F-M wage ratio increased rapidly in the beginning of 1980s.
 By 2009, it stood at 77%.
9-51
Trend in Female–Male Earnings Ratio
• Constant F-M wage ratio between 1960-1980 period masks the
improvement in the economic status of women due to
comparison of different samples of working women.
• Cohort effects: Samples toward 1980s had a disproportionate
number of newer labor market entrants with lower wages.
• In fact, growth of female wages were 20% higher than the
growth of male wages prior to 1980.
• Also explains the negative correlation between the magnitude of
the gender wage gap and the differences in the employment
rates observed in Figure 9-1 (Smallest employment differences
paired with largest wage gap.)
• Finally, 50% of the increase in the F-M wage ratio after 1980s
can be attributed to the increasing work attachment of women.
9-52
9-53