GENDER AND RACE IN THE LABOR MARKET
Sponsored by a Grant TÁMOP-4.1.2-08/2/A/KMR-2009-0041 Course Material Developed by Department of Economics,
Faculty of Social Sciences, Eötvös Loránd University Budapest (ELTE) Department of Economics, Eötvös Loránd University Budapest
Institute of Economics, Hungarian Academy of Sciences Balassi Kiadó, Budapest
2
Author: Anna Lovász Supervised by Anna Lovász
June 2011
Week 7
Measuring discrimination IV:
tests in Hungary and experiments Literature for next week
• Borjas chapter 6.
• Rao et al. (2003)
• Further recommended reading:
• Sakellariou 2004
• Kuhn-Weinberger 2006
Hungary: Sik–Simonovits 2009
• Summary of survey results on the perception of discrimination and discrimination testing
• 2005: Equal Treatment Committee
3
• Many complaints about discrimination existing problem
• But: hard to measure with a single number Goal: measure using 3 different methods
Sik–Simonovits 2009: methodology
• Perception of discrimination based on surveys
• Problem: mixture of 3 effects (extent of discrimination, social sensitivity, functioning of the institutional system)
• Chances of becoming a victim (experiences) based on surveys
• Problem: respondents may hide/exaggerate their experiences over/underestimate
• Experimental testing – based on success of applications to job ads
Results – perception
• EU-MIDIS survey: 27 countries, 45 minorities
• Hungarian roma respondents:
• 90% perceive a high level of discrimination based on ethnicity
• 2nd highest in the EU (1. place: North-Africans in Italy)
• Entire Hungarian population: ”for equally qualified applicants does it hinder employment chances if…”
• 65% skin color (EU average = 42)
• 67% age (EU = 45)
• 50% disability (EU = 41)
• 29% gender (EU = 22)
• Less than the EU average: name, accent, religion, appearance
Not a precise measure of discrimination, but an interesting comparison. How would you interpret these results?
4
Results: perception – EU-MIDIS
• Most disadvantaged group by country: roma
What could this results mean?
5
EU-MIDIS: perception of immigrants
Results: perception – CSO
• 2008 supplementary labor force survey, ages 19–64
• Question: have you experience discrimination during job search or firing based on ethnicity, gender, age, family status, education level, or health?
• Results:
• 16% during job search, 8% during firing
• Mostly: age (6 és 3%), family status (4 és 2%)
• Control: Budapest natives
• Worse perception of immigrants
• Mostly workplace
What does this mean? Empirical problems?
6
• Men felt more often based on ethnicity, gender, age, family status
• As a measure:
• Problem: not only labor market
• Education and health may affect productivity, so not necessarily basis for discrimination
• Measurement error: hidden/exaggerated answers
• Lack of controls
Results: perception – Tárki 2009
Women, young people, low-skilled, and roma perceive higher discrimination
No observations of workplace for young, low- skilled, roma
7
Testing – NEKI 2009
• Methodology:
• Analysis of job ads: discriminatory content?
• Phone-based testing: 86 jobs, Budapest and Szabolcs-Szatmár county
• Successful test: open position, competent respondent, 1 control and 2 minority applicants, outcome measures
• Jobs that do not require specialized training
• Groups: gender, ethnicity, family status, and age combinations
• Trained applicants, reveal their group membership during the conversation – how?
Results – testing (NEKI 2009)
8 Implicit rejection: behavior of employer changed when minority membership was revealed (tone, sigh, pause, …) – objective?
Results – testing, 2006
• Previous results are contradictory:
• Roma received fewer callbacks, women mostly implicit rejections
• Explicit rejection of roma males was the highest
• Explanation?
9
Sik–Simonovits (2009): summary
• Results
• Clear perception of discrimination against roma people
• Testing: female occupations, employers prefer women and young in these jobs
• Problems
• Description of methodology not clear
• Measurement of outcomes: subjective, researchers may influence results
• Absence of reference categories
• Similar test pairs: may be differences during conversations
• Method for revealing group membership may affect response
• Not representative:
• Very narrow occupational categories, female jobs
• Small sample size
10
Experiments: games – Fershtman–Gneezy 2001
• Different games, in which researchers analyze the behavior of participants against opponents of different ethnicities
• Ashkenazi (A: European or American) or Eastern Jews (E: Asian, African, lower living standards, prejudice)
• Ethnicity of opponents: random pairings, clearly ethnic names
• Do the two groups discriminate against each other?
• Is there a difference in strategies based on the ethnicity of the opponent?
• Do they prefer their own group, or does everyone systematically discriminate against one group?
• Is the discrimination taste or statistical based?
• Is the perception correct (stereotype)?
Fershtman–Gneezy 2001: methodology
• Participants:
• 966 BA students in Israel
• Haifa+AC Tel Aviv (2nd player), Tel Av.U. (1st player)
• Choose typical A or E names for men and women (well-known to be recognizable)
• Random pairings
• Game procedure:
• Same referee, rounds by university
• Players’ decisions, payouts are confidential
• Referee does not see them until the end
• Survey: gender, birthplace
11
Fershtman-Gneezy: 1. game – Trust
• Trust is a crucial element of economic transactions
• Procedure:
1. round: Player 1 receives a description of the game, the name of his/her pair, and 20 NIS (money)
Decides and writes down how much to send to Player 2, and his/her own name the referee triples the amount2. round: Player 2 receives the game description, the name and decision of Player 1, and the (tripled) amount sent to him/her
Decides how much to send back to Player 1 The referee views the decision, and pays Player 1 as wellTrust game
• Cooperation maximizes the payoff:
• If Player 1 trusts Player 2, he/she sends the full amount in the first round.
• If Player 2 is trustworthy, he/she shares the amount.
• If Player 1 does not trust Player 2, he/she sends very little/none.
• If Player 2 is offended, he/she keeps the full amount.
• Typical outcome in previous experiments (not based on groups): Player 1 sends a positive amount, Player 2 often sends back even more.
Is there a difference in the strategy of Player 1 based on the ethnicity of Player 2?
12
Results of trust game, men
13
Trust game, men
• Is everyone prejudiced against the other group, or is everyone prejudiced against E-s (systematic prejudice)?
• Is there a real basis for the prejudice? Do E-s send back less?
14
2. Dictator game
Is this based on taste or stereotypes?
• Round 1 same as trust game
• No round 2: Player 1 keeps their part, Player 2 receives the tripled amount sent to them.
• Player 2 has no role/decision.
• Optimal decision: not send anything, but in previous experiments usually send some amount (fairness)
Stereotypes regarding the behavior of Player 2 have no role in Player 1’s decision If they still discriminate against the E-s, it suggests taste-based discrimination, not lack of trust.Dictator game results, men
15
Ultimatum game
• Frequent stereotype: E-s are very proud, and will angrily avenge any injustice.
• Round 1: same as trust game
• Round 2: Player 2 decides whether to accept the offer or not (everyone gets 0).
• Optimum: Player 1 sends a small amount, Player 2 accepts.
• Previous results: Player 1 sends not too small amount, Player 2 sometimes rejects it.
Is there ethnic discrimination? If so, then Player 1 will fear the anger of E-s, and they will receive higher amounts on average.
Ultimatum game results, men
16
Results, women
Antonovics et al. 2003: The Weakest Link
• Based on data from the TV game show:
• Do players discriminate based on gender or race?
• If so, which model can explain their behavior: taste-based, statistical, or strategic (group membership may be the basis for cooperation. Holm (2000):
gender is often the basis in games)
• Procedure:
• Preliminary rounds: players jointly grow the pot, general knowledge questions.
17
• At the end of every round, one player is voted out.
• The last remaining two players then compete for the pot.
The Weakest Link – methodology
• Analyze the voting decisions of players based on conditional logit models.
Can take the characteristics of the voter and the recipient into account.
• Contribution:
• Good measure of individual productivity: the ratio of correct answers
• We know what characteristics are observed by players, so OVB is less of a problem.
• Decisions can be compared to those of other players to filter out further unobservables.
• Can examine the source of discrimination: is there a difference in the performance of different groups?
Problems, issues?
The Weakest Link – results
18
The Weakest Link – summary
• No evidence of discrimination against women by men or against blacks by whites.
• In early rounds, women discriminate against men.
• Significant: for similar productivity players, the chance that a woman votes against a given man is 23.3%, against a women is 15.1%.
• Behavior is only consistent with taste-based discrimination.
• Does this result contradict the general knowledge that there is discrimination against women and blacks?
No: there is no difference in the performance of the groups statistically, so there is no statistical discrimination in this game. This suggests that the labor market discrimination usually observed is statistical.