CHAPTER 9: LABOR MARKET DISCRIMINATION — EXAM-

READY SUMMARY
Borjas, Labor Economics (McGraw-Hill, 2016)

CHAPTER OVERVIEW
Chapter 9 analyzes how labor market discrimination affects the earnings and employment
opportunities of minorities and women. The chapter begins by documenting the raw wage and
employment gaps across race, gender, and ethnicity, then builds a series of theoretical models to
explain how discrimination arises. Becker's framework of taste discrimination (with the
discrimination coefficient) provides the backbone for analyzing employer, employee, and
customer discrimination. The chapter also covers statistical discrimination (no prejudice
required), experimental evidence, the Oaxaca decomposition for measuring discrimination, and
policy applications examining the Black–White wage ratio, the Female–Male wage ratio,
affirmative action, and discrimination against other groups.
Chapter Structure:
9-1: Race and Gender in the Labor Market (raw facts)
9-2: The Discrimination Coefficient (Becker's framework)
9-3: Employer Discrimination (equilibrium analysis with Figures 9-2 to 9-5)
9-4: Employee Discrimination
9-5: Customer Discrimination
9-6: Statistical Discrimination (with Equation 9-4 and Figure 9-6)
9-7: Experimental Evidence on Discrimination
9-8: Measuring Discrimination (Oaxaca Decomposition — Equations 9-5 to 9-8, Figure 9-7)
9-9: Policy Application: Determinants of the Black–White Wage Ratio
9-10: Discrimination against Other Groups (Hispanics, Asians)
9-11: Policy Application: Determinants of the Female–Male Wage Ratio

SECTION 9-1: RACE AND GENDER IN THE LABOR MARKET

Short Summary
The chapter opens by documenting the raw gaps in earnings and labor market outcomes across
race and gender in the U.S. labor market (2009–2010 data). White men earn the most ($55,800
annually); Hispanic women earn the least ($28,100). However, these raw gaps partly reflect
differences in labor supply, education, and other observable characteristics — not discrimination
per se. Racial and gender wage gaps exist in virtually every country, not just the United States.
Bullet-Point Notes
Annual Earnings by Race and Gender (Table 9-1, 2009–2010):

Group Annual Earnings Full-Time Earnings

White Men $55,800 $65,900

White Women $37,000 $47,000

Black Men $41,200 $48,400

Black Women $32,500 $39,500

Hispanic Men $35,200 $42,800

Hispanic Women $28,100 $35,300

Asian Men $66,600 $76,100

Asian Women $45,900 $55,200

Important Qualifications:
The typical white man earns 51% more than the typical white woman in total annual
earnings, but only 40% more among full-time workers — part of the gap reflects differing
labor supply
Part of the racial wage gap stems from education differences: only 12.5% of Hispanic men
are college graduates vs. 30.6% of white men
Gaps also exist internationally: Malaysia (Malay/Chinese wage ratio = 0.57); Canada (blacks
earn 18% less than whites); UK (nonwhite immigrants earn 10–20% less)

The International Gender Wage Gap (Figure 9-1):

A negative correlation exists across countries between the gender wage gap and the
gender employment gap
Countries with smaller employment gaps between men and women (more women
working) tend to have larger wage gaps
Interpretation: When more women enter the labor market, the marginal female worker has
lower potential earnings, widening the observed wage gap

MUST-REMEMBER POINTS
Raw wage gaps reflect differences in labor supply, education, AND potential
discrimination — raw numbers overstate discrimination
U.S. white men earn 40% more than white women on a full-time basis (not 51%)
 Countries with higher female employment rates tend to have larger gender wage gaps
— sample composition matters

SECTION 9-2: THE DISCRIMINATION COEFFICIENT

Short Summary
Gary Becker (Nobel Laureate) formalized labor market discrimination in his 1957 doctoral
dissertation. His model translates racial or gender prejudice into economic language using the
discrimination coefficient (d). A prejudiced employer acts as if hiring a black worker costs more
than the actual wage, treating it as a "utility-adjusted" price. This same concept applies to worker
and customer prejudice as well.

Bullet-Point Notes
Taste Discrimination:
Employer has disutility from hiring black workers
Actual cost of hiring a black worker: wB dollars
Employer's perceived (utility-adjusted) cost: wB(1 + d) dollars
d = discrimination coefficient = the percentage markup in the employer's perceived cost
due to prejudice
The greater the prejudice, the greater d
Example: wB = $10/hour, d = 0.5 → employer acts as if the black worker costs $15/hour

Three Sources of Taste Discrimination:

Employer discrimination: Employer gets disutility from hiring blacks → acts as if cost is
wB(1 + d)
Employee discrimination: White workers dislike working alongside blacks → white worker
acts as if wage is wW(1 - d) in an integrated firm
Customer discrimination: White customers dislike buying from black sellers → customer
acts as if price is p(1 + d)

Nepotism (reverse of discrimination):

Some employers prefer to hire blacks (nepotism coefficient n)
Utility-adjusted cost of hiring favored worker = wB(1 − n) → employer acts as if black labor
is cheaper than it actually is

Connection to Compensating Differentials:

Becker's discrimination fits within the compensating differentials framework: prejudiced
parties incorporate the race/gender of the other party in their "whole of advantages and
disadvantages," requiring compensating differentials for their utility loss or gain
MUST-REMEMBER POINTS
Discrimination coefficient d: perceived cost = wB(1 + d); greater d = more prejudice
Three types of taste discrimination: employer, employee, customer
Becker's book: The Economics of Discrimination (1957)

SECTION 9-3: EMPLOYER DISCRIMINATION

Short Summary
Employer discrimination is analyzed in a competitive model where black and white workers are
perfect substitutes in production. The key results are: (1) discrimination produces a segregated
workforce, (2) discrimination does not pay (it reduces profits), and (3) in equilibrium, employer
discrimination generates a wage gap between equally skilled black and white workers, acting
like a compensating differential that blacks must pay to soften employer resistance. Competition
will eventually erode employer discrimination.

Bullet-Point Notes
Setup:
Production function: q = f(EW + EB) — black and white workers are perfect substitutes
Both groups have the same VMP; differences in outcomes can only arise from
discrimination
Assume wB < wW initially (black wage is lower)

The Non-Discriminatory Firm (Figure 9-2):

Hires only black workers (the cheaper input)
Hires E*B workers where: wB = VMPE (Equation 9-2)

The Discriminatory Firm's Decision Rule (Equation 9-3):

Hire only blacks if: wB(1 + d) < wW
Hire only whites if: wB(1 + d) > wW
Implication: Firms have completely segregated workforces — either all-black or all-white
Firms with small d → "black firms"; firms with large d → "white firms"

Figures 9-3a and 9-3b — The Hiring Decisions:

Figure 9-3a (White firms): hire E*W whites at the higher white wage → hiring the wrong
workers at the wrong price → lower profits
Figure 9-3b (Black firms): hire fewer than E*B blacks because the utility-adjusted price
wB(1 + d) > wB → fewer workers than optimal → lower profits
Discrimination Does Not Pay (Figure 9-4):
Most profitable firm: discrimination coefficient d = 0 → hires E*B blacks at the lowest wage
→ profits = πmax
As d increases, black firms hire fewer workers → profits fall
At threshold dW: firm switches to all-white → profits drop dramatically to πW (paying the
higher white wage)
All-white firms earn the same profits regardless of their d value (since they all hire E*W
whites)
Key implication: Competition is a minority's best friend — color-blind firms can out-
compete discriminatory firms and eventually drive them out

Labor Market Equilibrium (Figure 9-5):

As the black–white wage ratio falls (black wages fall relative to white wages):
Point R: The least-prejudiced firm crosses a threshold and begins hiring blacks
More firms become black firms as the black wage falls further
The demand curve for black workers is downward sloping
Supply of black workers assumed perfectly inelastic (N workers regardless of relative
wage)
Equilibrium at (wB/wW)* — below 1 (black workers earn less than white workers)
The wage gap acts as a compensating differential: blacks must compensate employers for
the disutility of hiring them
Nepotism firms: If some firms prefer blacks, the demand curve shifts UP to D′, potentially
allowing (wB/wW) > 1 if blacks are few

Enclave Economies:
The typical black lives in a neighborhood that is 57% black
Geographic clustering creates opportunities for minority-owned firms to hire minority
workers
Enclave economies may allow minorities to escape the adverse impact of majority
discrimination

MUST-REMEMBER POINTS
Perfect substitute assumption → segregated workforces (all-black or all-white firms)
Decision rule: hire blacks if wB(1 + d) < wW; hire whites if wB(1 + d) > wW
Discrimination does not pay: color-blind firm maximizes profits (πmax)
Equilibrium generates wB < wW (black wage < white wage)
In competitive markets, employer discrimination will "wither away"
 Use the Becker model to explain why discriminating firms are less profitable than non-
discriminating firms.
Draw Figure 9-5 and explain how the equilibrium black–white wage ratio is determined.
Under what condition could the equilibrium black wage exceed the white wage?

SECTION 9-4: EMPLOYEE DISCRIMINATION

Short Summary
Employee discrimination arises when white workers dislike working alongside black workers.
The key results differ importantly from employer discrimination: employee discrimination still
produces segregation, but does not generate a wage differential between equally skilled black
and white workers. There are also no market forces to erode employee discrimination over time.

Bullet-Point Notes
Setup:
White workers have discrimination coefficient d: they perceive their wage in an integrated
firm as wW(1 − d) instead of wW
Black workers are indifferent about coworker race → their actual and utility-adjusted wage
= wB

Employment Decision:
A color-blind profit-maximizing employer would never voluntarily have an integrated
workforce:
If blacks are cheaper → hire only blacks
If whites are cheaper → hire only whites
Mixed workforce requires paying a compensating differential to white workers with
no productivity gain
Result: Segregated workforce (same as employer discrimination)

No Wage Differential:
Competition equalizes wages of the two groups:
If blacks are cheaper → employers increase demand for black labor → black wages rise
If whites are cheaper → employers increase demand for white workers → white wages
rise
In equilibrium: wB = wW for equally productive workers

No Self-Correcting Market Forces:

 Employee discrimination does NOT reduce firm profits (all firms pay the same market
wage)
There is therefore no competitive pressure to eliminate employee discrimination over time
Unlike employer discrimination, it does not wither away under competition

MUST-REMEMBER POINTS
Employee discrimination: segregation YES, wage differential NO
Employer discrimination: segregation YES, wage differential YES
No profit motive to eliminate employee discrimination → it is more persistent than
employer discrimination

SECTION 9-5: CUSTOMER DISCRIMINATION

Short Summary
Customer discrimination arises when customers dislike buying from minority workers, treating
the price as p(1 + d). If firms can hide black workers from customer contact (placing them in
back-office or manufacturing roles), customer discrimination may not affect wages. But when
face-to-face contact is unavoidable, customer discrimination reduces the demand for black
workers in contact positions and can generate a wage differential.

Bullet-Point Notes
When Customer Discrimination Doesn't Affect Wages:
If the firm can allocate black workers to no-contact positions (manufacturing, back office),
competition for these cheap workers equalizes wages
No profit loss from catering to customer tastes in this case

When Customer Discrimination Does Affect Wages:

When black workers are in customer-facing positions: firm must lower the product price to
compensate white buyers → black worker's wage falls to compensate for the firm's lost
profits

Empirical Evidence (Table 9-2) — Difference-in-Differences:

Survey of employers in Atlanta, Boston, Detroit, Los Angeles — classified by customer
contact and customer racial composition:
 Firm Type Mostly Black Customers Mostly White Customers Difference

Contact firms 58% black hires 9% black hires 49.0 pp

No-contact firms 46.6% 12.2% 34.4 pp

Difference-in-Differences 14.6 pp

The 34.4 pp gap in no-contact firms is NOT due to customer discrimination (no face-to-face
contact) — it reflects neighborhood/applicant pool effects
The difference-in-differences estimate = 14.6 percentage points → this is the causal effect
of customer discrimination on black hiring

Baseball Card Evidence:

Cards of white players cost 10–13% more than cards of black players with equivalent career
statistics
This is pure customer discrimination in the memorabilia market

MUST-REMEMBER POINTS
Customer discrimination matters only when workers cannot be hidden from customer
contact
Difference-in-differences estimate = 14.6 pp reduction in black hiring due to customer
discrimination
Baseball card premium: white players' cards cost 10–13% more even after controlling for
performance stats

SECTION 9-6: STATISTICAL DISCRIMINATION

Short Summary
Statistical discrimination arises without any prejudice. When employers cannot perfectly predict
a worker's productivity, they use group membership as a statistical signal, basing wages partly
on the average characteristics of the worker's group. This produces unequal treatment of equally
skilled workers belonging to different groups — even in the complete absence of bigotry.

Bullet-Point Notes
The Mechanism:
Employer faces information problem: résumé + interview only imperfectly predict true
productivity
 When individual signals are noisy, the employer rationally uses the group average as a
supplementary signal
Result: two workers with the same test score but different group membership receive
different wages
The Wage-Setting Formula (Equation 9-4):

w = αT + (1 − α)Tˉ

T = applicant's own test score

Tˉ = mean test score of the applicant's group
α = weight on individual signal = correlation between test score and true productivity
α = 1 → perfect individual prediction; employer ignores group average
α = 0 → test is completely uninformative; employer relies entirely on group average

Two Ways Statistical Discrimination Affects Wages (Figure 9-6):

Case (a): Groups have different mean test scores (Figure 9-6a):
If blacks have a lower mean test score (TˉB < TˉW ), both the white and black wage-test
​ ​

curves have the same slope (same α), but the white curve lies above the black curve
A black worker with the same score T* as a white worker earns less because the group
average lowers the employer's assessment
Case (b): Test predicts less accurately for blacks (Figure 9-6b, "cultural bias"):
If αB < αW (test is a worse predictor for blacks), the black wage-test curve is flatter
Low-scoring blacks earn more than low-scoring whites (employer doesn't trust low
scores for blacks — pulls wages toward the group mean)
High-scoring blacks earn less than high-scoring whites (employer discounts high
scores for blacks)

Policy Debate — Race Norming:

Should employers use group averages in wage-setting?
Race norming = adjusting test scores to assign the same mean to all groups
Borjas's argument: even if race norming is mandated, employers will search for other
productivity signals. If those signals are correlated with race/gender, statistical
discrimination persists

Statistical Discrimination vs. Signaling:

In the signaling model: workers invest in education to separate themselves from others
(immutable traits not used)
In statistical discrimination: employers use immutable traits (race, gender) as signals →
workers cannot escape it by investment
MUST-REMEMBER POINTS
Statistical discrimination: wage = αT + (1 − α)T̄ — a weighted average of own score and
group average
α = 1: no statistical discrimination; α = 0: full statistical discrimination (employer ignores
individual info)
Cultural bias (lower α for blacks): low-scoring blacks benefit, high-scoring blacks are
penalized
Statistical discrimination requires no prejudice — it arises from imperfect information
alone

Common Exam Questions:

Explain statistical discrimination using Equation 9-4. What does α represent?
Draw Figure 9-6 and explain the two channels through which statistical discrimination
affects wages.
How does statistical discrimination differ from taste discrimination?

SECTION 9-7: EXPERIMENTAL EVIDENCE ON DISCRIMINATION

Short Summary
Because it is illegal to discriminate, employers won't reveal their prejudicial behavior directly.
Researchers have therefore designed controlled experiments to measure discrimination. Two
main approaches: (1) sending fake résumés with racially suggestive names and (2) sending
matched pairs of actual job applicants ("hiring audits"). Both methods provide compelling direct
evidence of discrimination.

Bullet-Point Notes
The Name Experiment (Bertrand & Mullainathan, 2004):
~5,000 fake résumés sent to ~1,300 actual job ads in Boston and Chicago
Race was signaled only through the applicant's name (e.g., Emily Walsh/Greg Baker =
white-sounding; Lakisha Washington/Jamal Jones = black-sounding)
Results:
White-sounding names: 1 callback per 10 résumés sent
Black-sounding names: 1 callback per 15 résumés sent
A black applicant would need 8 more years of work experience to close the callback
gap
40% of black girls born in California between 1961–2000 received names not given to any
white girl in that period → racial naming patterns are measurable and distinct
Hiring Audits:
Two matched applicants (identical in all respects except race/gender) sent to the same
firms
Chicago/San Diego audit (1989) — Hispanic vs. non-Hispanic white applicants for entry-
level jobs:
White applicant was 33% more likely to be interviewed and 52% more likely to
receive a job offer
Restaurant audit (Philadelphia):
Low-priced restaurants: 8 of 10 job offers went to women
High-priced restaurants: 11 of 13 job offers went to men (men directed to better-
paying positions)

NBA Referees Study:

Each game has 3 referees; racial composition of referee team varies naturally
All-white referee team vs. all-black referee team effect:
White players receive ~0.25 fewer fouls from all-white referees
Black players receive similar fouls regardless of referee team composition
Conclusion: White referees engage in nepotism (going easier on white players) rather
than animus toward black players
This discrimination matters in close games: ~4–5% fewer fouls for same-race players
→ ~2–3% more points per game

MUST-REMEMBER POINTS
Name experiment: 1 callback per 10 (white names) vs. 1 per 15 (black names) = 50%
more callbacks for white-sounding names
8 extra years of experience needed for black applicants to equalize callback rates
NBA referee study: nepotism by white referees toward white players, not animus
toward black players

SECTION 9-8: MEASURING DISCRIMINATION — THE OAXACA

DECOMPOSITION
Short Summary
Comparing raw average wages between groups is misleading because differences in observable
skills (education, experience) also generate wage gaps. The Oaxaca decomposition separates
the raw wage differential into (1) a portion explained by skill differences and (2) a residual
portion attributable to discrimination. It is the dominant empirical tool in discrimination
research, but it has important limitations and is frequently used (and abused) in litigation.
Bullet-Point Notes
The Raw Wage Differential (Equation 9-5):

Δw
ˉ=w
ˉM − w
ˉF ​ ​

Problem: Men and women differ not just in their treatment but also in their skills →
comparing raw wages conflates skill differences with discrimination

The Earnings Functions (Equation 9-6):

Male earnings: wM = αM + βM sM
​ ​ ​ ​

Female earnings: wF = αF + βF sF
​ ​ ​ ​

βM and βF = returns to schooling for men and women respectively

αM and αF = intercepts (wages with zero schooling)

The Oaxaca Decomposition (Equation 9-8):

Δw
ˉ = (αM − αF ) + (βM − βF )sˉF + βM (sˉM − sˉF )
​ ​ ​ ​ ​ ​ ​ ​

Component What It Measures

(αM − αF ) +
​ ​ Due to discrimination — the wage gap that arises because employers value
(βM − βF )sˉF
​ ​ ​ men's characteristics more than women's at the same skill level

βM (sˉM − sˉF )
​ ​ ​ Due to skill differences — the wage gap that arises because men and women
have different average schooling levels

Figure 9-7 — Graphical Illustration:

Men's earnings function has a higher intercept (αM > αF) and steeper slope (βM > βF)
Average woman has s̄ F schooling, earns w̄ F
If the average woman were "paid like a man": she would earn w*F (reading off the men's
earnings function at s̄ F)
Measure of discrimination = w*F − w̄ F (the gap the average woman faces due to
differential treatment)
Skill differential = w̄ M − w*F (the gap due to men having more schooling)

Limitations of the Oaxaca Decomposition:

Requires complete information on all skill-determining variables; omitting variables biases
the estimate
 If blacks attend lower-quality schools, a simple years-of-schooling control doesn't fully
adjust for skills → the "discrimination" residual is overestimated
On the other hand: if pre-market discrimination (school funding disparities, occupational
channeling) generates the skill gaps, then the discrimination component is underestimated
by netting out skills
No consensus on the "right" set of controls: adding occupation and industry controls
reduces measured discrimination but raises the question of whether occupational sorting is
itself caused by discrimination
Used extensively in class-action litigation; subject to strategic manipulation

MUST-REMEMBER POINTS
Oaxaca decomposition: Raw gap = Discrimination component + Skill component
Discrimination component = difference in intercepts and returns to skills
Skill component = βM × (s̄ M − s̄ F)
Adding more controls (occupation, industry) reduces measured discrimination but may
absorb discrimination effects
Key limitation: omitted variable bias; quality of controls determines reliability

SECTION 9-9: POLICY APPLICATION — DETERMINANTS OF THE BLACK–

WHITE WAGE RATIO
Short Summary
The black–white wage ratio has improved significantly over the past 60+ years (from ~0.40 in
1940 to ~0.76 for men in 2012), driven by rising black human capital and the impact of civil rights
legislation and affirmative action programs. However, a significant and persistent wage gap
remains, and the declining labor force participation rate of black men is a troubling countertrend.

Bullet-Point Notes
Oaxaca Decomposition of Black–White Differential (Table 9-3, 1995):
Raw black–white log wage differential: −0.211 (approximately 21% gap)

Controls Used Due to Skill Differences Due to Discrimination

Education, age, sex, region −8.2% −13.4%

+ Occupation and industry −11.4% −9.8%

The measured discrimination component is sensitive to which controls are included

 Key question: Is occupational sorting itself a product of discrimination? If yes, controlling
for it underestimates discrimination
Trend in the Black–White Wage Ratio (Figure 9-8):
1940: Black–white ratio ≈ 0.40
1967: ≈ 0.65 (for men)
1980: ≈ 0.71 (for men)
2012: ≈ 0.76 (for men)
Black women: Ratio rose rapidly 1967–1975 (0.75 → 0.96); now stands at ~0.87

Explanations for the Rising Black Wage:

Human capital improvements: In 1940, the typical 30-year-old black man had 6.0 years of
schooling vs. 9.9 for whites. By 1980: 12.2 vs. 13.6 years (gap narrowed to 1.4 years)
Quality gap in schooling nearly closed: Pupil–teacher ratios in southern Black schools were
50% higher than white schools in the 1920s → essentially disappeared by the late 1950s
Rate of return to schooling: Black workers who entered in 1940 had a 4.7% return vs. 9.8%
for whites; by the late 1970s, blacks had a higher rate of return (9.6% vs. 8.5%)
At least half the increase in the black–white wage ratio can be attributed to rising black
human capital

Affirmative Action:
1964 Civil Rights Act → prohibited employment discrimination → established EEOC
Executive Orders 11246 and 11375 → prohibited discrimination by federal contractors;
required affirmative action plans with goals and timetables
Compliance: Contractors with $50,000+ contracts and 50+ employees must file annual
employment reports by occupation, race, and gender
In 1966: black men were 10% less likely than white men to work for federal contractors; by
1980: 25% more likely
South Carolina textiles: Black employment share rose from ~4–5% (1910–1964) to ~20% by
1970 after being covered by executive orders
Impact on wages: Weak direct evidence; "back-door" effect via increased representation in
high-wage large firms accounts for ~15% of the rise in the black–white wage ratio

The Decline in Black Labor Force Participation:

Mid-1950s: Both black and white men had ~85% labor force participation rates
2010: Gap widened to 7 percentage points between white and black men
The declining black LFPR may partly explain the rising observed black wage (Figure 9-10):
Those who left the labor market were likely lower-wage workers → the sample of black
workers in the labor market became positively selected → the "average" observed black
wage rose even without real wage gains for all black workers
 Implication: The Oaxaca measure of discrimination may overstate discrimination because
the black wage comparison is made on a selected sample

MUST-REMEMBER POINTS
Black–white wage ratio: 0.40 (1940) → 0.65 (1967) → 0.76 (2012) for men
At least half the wage ratio improvement attributed to increased black human capital
Measured discrimination: 13.4% (basic controls) to 9.8% (with occupation/industry
controls)
South Carolina textiles: Black share from ~5% to ~20% after executive orders took
effect
Declining black LFPR introduces selection bias into observed black wages

SECTION 9-10: DISCRIMINATION AGAINST OTHER GROUPS

Short Summary
The chapter briefly examines wage differentials experienced by Hispanics and Asians. Hispanics
(especially Mexicans) face large wage gaps primarily due to skill differences, not discrimination.
Asians, by contrast, earn a slight wage premium over whites, largely due to their high average
skill levels. Both groups have grown rapidly as shares of the U.S. population.

Bullet-Point Notes
Hispanic Wage Gap (Figure 9-11, Table 9-4):
Hispanic–white wage ratio declined in the 1980s and has been roughly stable since
Hispanic men earn about 20–25% less than white men
Education gap is severe (Table 9-4):
37.1% of all Hispanics are high school dropouts (vs. 7.9% of non-Hispanic whites)
Only 13.9% of all Hispanics are college graduates (vs. 33.2% of non-Hispanic whites)
Mexicans: 42.6% dropouts, 10.6% college grads
Cubans: 18.6% dropouts, 26.2% college grads (considerably better outcomes)
Over three-quarters of the Mexican–white wage gap can be attributed to differences in
observable skill measures → Mexicans earn less mainly because they are less skilled, not
because they are Hispanic per se

Self-Selection Bias:
The measured wage performance of ethnic groups is contaminated by which individuals
choose to self-identify with their ethnic group in Census surveys
Among Mexican-Americans with mixed heritage (high intermarriage rates), those who
self-identify as Mexican-American are negatively selected — lower economic performers
 This tends to understate the true socioeconomic improvement of Mexican-Americans over
time
Asian Wage Premium:
Asian–white wage ratios: 1.0 to 1.2 for both men and women (Asians earn slightly more
than whites)
Much of the advantage is explained by high average skill levels of Asian workers
Asian population grew from 1.5% (1980) to 4.6% (2009) of the U.S. population

Post-9/11 Discrimination Against Arabs and Muslims:

Weekly earnings of Arab and Muslim men fell by ~10% after 9/11
The wage reduction was not long-lasting; much of it evaporated by 2005
Affected practically all Arab and Muslim men regardless of education, immigration status,
or country of origin
Earnings reduction was larger in areas with greater increases in hate crimes

MUST-REMEMBER POINTS
Over 3/4 of the Mexican–white wage gap is explained by skill differences, not
discrimination
Asians earn a wage premium of 0–20% over whites, mainly due to high skill levels
Self-selection bias: Mexican-Americans who identify as Mexican are negatively
selected → observed wages understate true group improvement
9/11 caused a ~10% temporary earnings decline for Arab/Muslim men

SECTION 9-11: POLICY APPLICATION — DETERMINANTS OF THE FEMALE–

MALE WAGE RATIO
Short Summary
The gender wage gap is large and persistent, but much of it can be explained by differences in
labor market experience rather than outright discrimination. Women's intermittent labor market
attachment (due to child-rearing) reduces their human capital accumulation and causes
depreciation of existing skills during absences. The female–male wage ratio hovered at ~0.60
from 1960–1980, then rose rapidly to ~0.77 by 2012. The rise is largely attributed to increasing
work attachment of women. Occupational crowding (women channeled into lower-paying
occupations) is also discussed.

Bullet-Point Notes
Oaxaca Decomposition of Gender Wage Gap (Table 9-5, 1995):
Raw log wage differential: −0.286 (men earn ~28.6% more than women)
 Controls Due to Skill Differences Due to Discrimination

Education, age, region −0.8% −27.9%

+ Occupation and industry −7.6% −21.1%

Men and women have roughly similar education, age, and location → these variables
explain almost none of the gap
Once occupation/industry are controlled, skill differences explain 7.6% and discrimination-
like residual = 21.1%

Labor Market Experience and the Gender Wage Gap:

Standard decompositions ignore that men and women have very different labor market
histories
Women worked only 71% of their potential years of labor market experience (late 1980s) vs.
men at 93%
Intermittent female labor supply reduces wages through two channels:
1. Women invest less in human capital (shorter payoff period → lower returns to
investment)
2. Human capital depreciates during years away from the labor market (skills forgotten
or obsolete)

Law School Study (Michigan, 1973–1975 classes):

15 years after graduation: Male attorneys earned $141,000 vs. female attorneys $86,000
Two-thirds of the wage gap explained by differences in work histories (e.g., 3 years of part-
time work reduced female earnings by 17% permanently)

The Chicken-and-Egg Problem:

Did weaker work attachment cause lower wages? (Human capital channel)
Or did lower wages (from discrimination) cause weaker work attachment?
This feedback problem is difficult to resolve statistically

Occupational Crowding Hypothesis:

Women are intentionally (or culturally) segregated into particular occupations →
oversupply of workers in "female jobs" → lower wages in those occupations
Evidence (Table 9-6): Under 5% of aircraft mechanics are women; over 95% of kindergarten
teachers are women
Empirical finding: A woman in an occupation that is 75%+ female earns 14% less than a
comparable woman in an occupation that is 75%+ male
 This wage penalty also applies to men in predominantly female occupations — it is the
"femaleness" of the job, not the gender of the worker, that matters
Marriage bars (19th century–1950): Prohibited married women from teaching/clerical work
→ crowded educated women into lower-paying jobs
Occupational Choice (Human Capital Alternative):
Mincer and Polachek (1974): Women with expected interruptions rationally choose
occupations where human capital depreciates less
Some occupations have inherently more or less depreciation during absence
Women's college major choices are partly driven by mathematical ability scores — not just
discrimination
Human capital story and crowding hypothesis are not mutually exclusive

Trend in the Female–Male Wage Ratio (Figure 9-12):

Hovered at ~0.60 from 1960 to 1980
Began rising in the early 1980s
Stood at ~0.77 by 2012
Pre-1980 improvement was masked: increasing female labor force participation brought in
lower-wage women → sample composition effect
Controlling for cohort effects: female wages grew 20% faster than male wages even before
1980
Post-1980 rise in women's wages: ~50% attributable to increasing work
experience/attachment of American women

Blind Auditions in Symphony Orchestras:

Major orchestras adopted blind auditions (applicants play behind a screen) in the 1980s–
1990s
The use of a screen increased the probability that a female musician advanced past
preliminary rounds by 50%
By the 1990s: over 20% of major orchestra musicians were women; ~half of this increase
directly attributable to blind auditions

Affirmative Action and Women:

Weak impact on white women: federal contractors employed 28% of working white women
in 1970 vs. 30% in 1980
Large impact on black women: 35% employed by federal contractors (1970) → nearly half
(1980)
Black women "fill two quotas for the price of one"
MUST-REMEMBER POINTS
Raw gender gap: men earn 28.6% more than women (1995); controlling for occupation
reduces discrimination estimate to 21.1%
Women worked 71% of potential experience; men worked 93%
Two reasons intermittent work reduces female wages: less human capital accumulated
+ depreciation during absence
Michigan law school: 2/3 of the wage gap explained by different work histories
Female–male wage ratio: ~0.60 (1960–1980) → 0.77 (2012)
~50% of the post-1980 narrowing attributed to women's increasing work experience
Occupational crowding: "femaleness" of job penalizes wages; a worker in a 75%+ female
occupation earns 14% less
Blind auditions raised women's odds of advancing in orchestra auditions by 50%

Common Exam Questions:

What is the occupational crowding hypothesis? How does it differ from the human capital
explanation of gender wage differentials?
Why might the Oaxaca decomposition in Table 9-5 underestimate true discrimination
against women?
Explain the "chicken-and-egg" problem in gender wage research.
What evidence exists for the role of work experience in explaining the gender wage gap?

CHAPTER 9 MASTER SUMMARY TABLE

Topic Key Fact / Formula

Discrimination coefficient Perceived cost of black labor = wB(1 + d); d > 0 implies prejudice

Nepotism coefficient Perceived cost = wB(1 − n); n > 0 implies preference for that group

Employer discrimination decision Hire blacks if wB(1 + d) < wW; hire whites if wB(1 + d) > wW
rule

Result of employer discrimination Segregated workforce; wage differential; reduced profits

Result of employee Segregated workforce; NO wage differential; no market correction

discrimination

Result of customer discrimination Wage differential only if workers can't be hidden from customers

Diff-in-diff: customer 14.6 pp reduction in black hiring at contact firms (Table 9-2)
discrimination
Topic Key Fact / Formula

Baseball cards White player cards cost 10–13% more, controlling for career stats

Statistical discrimination formula w = αT + (1 − α)T̄; α = predictive validity of individual test

α=1 Perfect prediction; no statistical discrimination

α=0 Complete reliance on group average

Cultural bias (low α for blacks) Flat wage-score curve for blacks: low scorers benefit, high scorers
penalized

Name experiment White-sounding names: 1 callback per 10 résumés; black names: 1

per 15

Hiring audit (Hispanic) White applicants 52% more likely to receive a job offer

Oaxaca decomposition (formula) Δw̄ = [Discrimination: (αM − αF) + (βM − βF)s̄ F] + [Skills: βM(s̄ M −
s̄ F)]

Black–white raw gap (1995) −21.1%

Discrimination after basic −13.4%; after adding occupation/industry: −9.8%

controls

Black–white wage ratio trend 0.40 (1940) → 0.65 (1967) → 0.76 (2012), for men

Cause of ratio improvement ≥ 50% from rising black human capital; Civil Rights Act + affirmative
action

South Carolina textiles Black employment: ~5% (pre-1964) → ~20% (1970) after executive
orders

Gender raw gap (1995) Men earn 28.6% more; controlling for occ./industry: 21.1%
discrimination

Women's potential experience Women worked 71% of potential experience; men 93%

Michigan law school 2/3 of wage gap explained by work history differences

Female–male wage ratio ~0.60 (1960–1980) → 0.77 (2012)

Cause of narrowing ~50% from women's increasing work experience

Occupational crowding 75%+ female occupation → worker earns 14% less, regardless of
gender

Blind auditions 50% higher probability of female musicians advancing past

preliminary rounds
 Topic Key Fact / Formula

9/11 wage effect Arab/Muslim men's earnings fell ~10%, mostly reversed by 2005

Mexican–white wage gap Over 3/4 explained by observable skill differences

KEY TERMS (Chapter 9)

Discrimination coefficient (d): The percentage markup in a prejudiced employer's
perceived cost of hiring a black worker
Taste discrimination: Becker's framework; prejudice incorporated into economic decision-
making
Employer discrimination: Discrimination arising from the employer's own prejudice;
reduces profits; competitive markets tend to erode it
Employee discrimination: Discrimination arising from white coworkers' preferences;
causes segregation but no wage gap; does NOT wither away under competition
Customer discrimination: Discrimination arising from customers' preferences; affects
wages only when workers have direct customer contact
Statistical discrimination: Differential treatment of equally skilled workers based on group
membership, arising from imperfect information rather than prejudice
Nepotism: Preference for hiring a particular group (positive discrimination coefficient
equivalent)
Oaxaca decomposition: A method to decompose a raw wage differential into a skill-
differences component and a discrimination component

End of Chapter 9 Summary