Gender and Racial Pay Gaps

1
Gender and Racial Pay Gaps
2
Introduction

• Normally put together (as here) e.g.
  Altonji-Blank chapter in HOLE
• There are similarities but there are also
  differences
• Women have babies
• Ethnic minorities often geographically
  concentrated

3
The Gender Pay Gap

• Men, on average, earn more than women, pretty
  much everywhere
• But there is variation
• across countries,
• over time
• within countries at a point in time (e.g.
  differences in earnings profiles we saw earlier)

4
Variation in the Gender Pay Gap Across Countries
(Blau-Kahn, JEP 2000)
5
Questions and Puzzles?

• Is position of women in southern Europe really
  much better than in Northern Europe, the US?
• Problem of sample selection (Petrongolo and
  Olivetti, CEP DP711) proportion of women
  working in southern Europe much lower

6
Olivetti-Petrongolo CEP DP711
7
Variation in Gender Pay Gap Over Time - UK
8
Blau-Kahn, JEP 2000
9
Variation in Gender Pay Gap Over Time - US
10
Most other countries

• Similar trend but timing different
• But some countries have marked change around time
  of Equal Pay Act (e.g. UK), some do not (e.g. US)

11
Accounting for the Gender Pay Gap the Oaxaca
decomposition

• Think of estimating separate earnings functions
  for men and women
• These must go through the means so

12

• Taking differences
• This can be written as
• i.e. gender pay gap is decomposed into
• Part due to differences in characteristics
• Part due to differences in coefficients
• Second part used to be called discrimination
  but perhaps unwise to do so

13

• But decomposition is not unique e.g. could write
• Will generally give different answers only same
  answer if only difference in intercept
• No reason to prefer one over the other and could
  evaluate at other characteristics generally
  report them all and hope tell a similar story

14
A Graphical Representation
15
Insights from Oaxaca decomposition US 1979
(Altonji Blank, HOLE)
16
Insights from Oaxaca decomposition US 1995
(Altonji Blank, HOLE)
17
What does UK look like?
At male chars At male chars At female chars At female chars
raw differential diffs in chars diffs in returns diffs in chars diffs in returns
No controls -0.285 0 -0.285 0 -0.285
Year Region -0.285 -0.001 -0.284 0.001 -0.286
Education -0.285 -0.045 -0.24 -0.029 -0.256
Potential Experience -0.285 -0.028 -0.257 -0.031 -0.254
Race -0.285 -0.027 -0.258 -0.031 -0.254
MaritalKids -0.285 -0.033 -0.252 -0.032 -0.253
Job Tenure -0.285 -0.058 -0.227 -0.055 -0.23
IndustryEmployer Size -0.285 -0.089 -0.196 -0.101 -0.184
Occupation (2-digit) -0.285 -0.145 -0.14 -0.098 -0.187
18
Things to Note

• Hard to find regressors that can explain most of
  gender pay gap
• Problem bigger now than in the past as men and
  women more similar now
• Perhaps do not have the right variable focus
  especially on actual experience as opposed to
  potential

19
Altonji-Blank Table 6
20
Differences in actual experience not huge e.g. UK
BCS 70
21
Blau-Kahn alternative measure of gender pay gap
position in male wage distribution
22
What insights from this?

• Position of women in US looks better high level
  of wage inequality tends to widen gender pay gap
• Gender pay gap influenced by factors which do not
  seem to have gender-specific angle Blau-Kahn
  JOLE 2003 show that collective bargaining and
  minimum wage important

23
Theories of the Gender Pay Gap

• Human Capital Theories
• Women have lower actual experience
• Invest less in human capital
• Discrimination
• Prejudice
• Statistical discrimination
• Search Theories
• Psychological Theories

24
Human Capital Theories

• Among past generations women had less education
  than men now they have slightly more
• But women still spend less time in paid work than
  men but are gaps big enough?
• Can explain some of the gap but not all of it
  see earlier picture

25
Theories of DiscriminationBeckers Theory of
Prejudice

• Becker The Economics of Discrimination proposed a
  theory of discrimination that was based on the
  prejudice of agents in the labour market,
  employers, workers and customers. It is
• sometimes called a tastes theory.
• Basic idea is that some people are prejudiced so
  will be prepared to pay a price to avoid certain
  groups
• This then affects the demand for labour of those
  groups.

26
A Simple Model of Employer Discrimination

• Suppose that some employers are prejudiced
  against black people
• Implies that if black and white wages were the
  same, they would always prefer to employ white
  workers.
• But they do care about profits so that, if the
  black wage is sufficiently below the white, the
  extra money they can save by employing black
  workers is sufficient to overcome their prejudice.

27

• Suppose the wage for white workers is Ww and
  that for black workers is Wb.
• Firm f will choose to employ black workers if the
  relative wage is given by
• where df is the firms discrimination
  coefficient. If df0 the employer is not
  prejudiced and employs whichever group of workers
  is cheapest
• a larger value of df implies the employer is more
  prejudiced

28
Demands for Black and White Workers

• If the wage paid by firm f is W, denote its
  demand for labour by
• If it employs white workers it will employ
• if it employs black workers it will employ

29

• order the F firms in the order of their
  discrimination coefficient so that there will be
  a cut-off f, below which all firms employ black
  workers and above which they employ white
  workers.
• The cut-off must satisfy
• Hence, the overall demand for black workers is
  going to be

30

• And aggragate demand for white workers
• What does equillibrium look like?
• If no prejudice than black and white wages must
  be equal otherwise all firms employ all black or
  all white workers
• But where there is prejudice white wages and
  black wages will diverge prejudice has
  consequences for demand for labour and hence
  relative wages

31
Assessment

• Prejudiced firms make lower profits as will only
  employ more expensive white workers
• Becker predicted competition would drive out
  discrimination
• Not clear this is a good prediction but
  argument does not work for consumer discrimination

32
Statistical Discrimination

• basic idea is that in a world of imperfect
  information, employers will base decisions on
  observable characteristics (like race and gender)
  that do not directly affect productivity because
  they are correlated with unobserved
  characteristics that do affect productivity.
• Many variants of this theory will present one

33
Simple model of statistical discrimination

• true productivity of a worker, p, is not observed
  by employers but do know true distribution in
  population (mean is µ)
• employers do observe an imperfect signal of
  productivity (e.g. education), s,
• Under certain assumptions (that I dont want to
  go into here), the expected productivity of
  someone with signal s is given by

34

• T a measure of how informative is the signal
  about the quality of the worker concerned.
• In a competitive labour market workers will
  receive a wage equal to their expected
  productivity so wE(p).
• Assume T is lower for whites than blacks (why?)
  so that

35

• black and white workers with the same level of
  education will now get different wages so there
  is some discrimination.
• But the average level of wages across will simply
  reflect the average level of productivity so
  there can only be discrimination overall if
  µw?µb.
• Now consider choice of education
• Leading to first-order condition

36

• If ? lower for blacks than whites then lower
  incentives for blacks to get educated
• leads to a lower average level of productivity
  for blacks and lower average wages.
• In equilibrium blacks get paid less than whites
  because they have lower productivity but this is
  the outcome of a rational decision-making process
  in which they have lower returns to skills
  because employers find it harder to evaluate
  their skills.
• key point in models of statistical discrimination
  is that your fate is determined not just by what
  you do but by what people like you do.
• This can lead to self-fulfilling expectations
  that are discriminatory in nature.

37
Vicious Cycle Theories of Discrimination

• Employers think women are more likely than men to
  take time out of paid work so do not invest in
  them, promote them as much
• Within couples becomes rational for the woman to
  take time out because opportunity cost for man is
  larger
• Relies on a form of statistical discrimination
  get judged by your gender not your personal
  characteristics

38
Search Theories

• Search theories suggest
• faster earnings growth if job offers arrive more
  regularly
• Slower earnings growth if leave jobs more
  regularly
• Women report more constraints on job search,
  commute shorter distances etc

39
For example
40
Search Theories Job Mobility Rates for better job
41
Job Mobility Rates not for better job
42
Psychological Theories

• A more recent development
• By the time they enter labour market men and
  women are different in personalities
• Why is suggest of furious nature vs. nurture
  debate
• Some have suggested this can help to explain
  gender pay gap

43
Example of psychological differences
44
Evidence for this view

• Babcock-Laschever Women Dont Ask women dont
  ask for what they want because of lower
  self-esteem, lower locus of control, more concern
  about what others think
• Gneezy et al lab work women try to avoid
  competition and dont do so well when faced with
  it

45
Gneezy, Niederle, and Rustichini, QJE 2003
46
British Teenagers in BHPS
Percentages agreeing with the statement Men Women
I feel I have a number of good qualities 93.0 91.0
I feel that I do not have much to be proud of 13.9 17.5
I certainly feel useless at times 29.1 44.8
I am able to do things as well as most other people 91.0 90.5
I am a likeable person 93.0 93.6
I can usually solve my own problems 90.2 86.0
All in all, I am inclined to feel I am a failure 8.9 12.1
At times I feel I am no good at all 22.6 35.3
I feel left out of things when I am with friends 15.6 25.0
47
How to map this into wages?

• Tried this in Manning-Swaffield
• Not much success at most a few percentage
  points
• Still ended up with a large gap after 10 years in
  the labour market

48
Racial Pay Gaps

• Harder to compare across countries as ethnic mix
  varies so much
• Tend to be smaller than gender pay gaps
• Often show a lot of variation across ethnic
  groups e.g. in UK Pakistanis/Bangladeshis do
  badly, Indians much better
• Can often relate that to social background prior
  to immigration

49
Black-White Earnings Gap In US
50
Interpretation

• Blacks made significant gains in 1960s and 1970s
  as a result of civil rights movement
• But stagnation after 1980

51
Explaining the Black-White Wage Gap

• Oaxaca decomposition suggests can explain more of
  black-white pay gap than gender pay gap
• Sizeable differences in education
• Neal-Johnson claim that can explain nearly
  everything if include test score variable
• Suggests that any discrimination is prior to
  labour market entry

52
Neal-Johnson, JPE, 1996 evidence
53
Audit studies

• However, audit studies do suggest discrimination
  in labour market
• Get matched pairs to apply for jobs, one white,
  one black and see what happens
• Expensive to do, small samples, questions about
  experimental bias
• Betrand-Mullainathan AER 2004 perhaps more
  persuasive

54
Betrand-Mullainathan AER 2004

• Research design
• Send out CVs
• Measure call-back rates
• Convey race through name randomly assigned
• Find call-back rates lower for black-sounding
  names, bigger difference for high-quality CVs
• Criticism does name convey something other than
  race Fryer-Levitt QJE

55
Bertand-Mullainathan Results
56
Discrimination Estimates from Production
FunctionsHellerstein, Neumark, Troske, JOLE 1999
57
Conclusion

• Being able to find the magic X variable that
  can explain gender pay gap remains elusive
• On black-white wage gap, main question is around
  whether labour market differences totally reflect
  pre-market differences