Discrimination in Labor MarketsA. Introduction1. Disparities in earnings across groupsa. annual earnings of black men and women and whitewomen are less than those of white menb. some convergence in recent years (due mostly to dropin earnings for white men)c. pre-1967 differences were even more pronounced2. How do we account for these differences? To what extentdo they reflect discrimination?3. Concepts (Cain 1986)a. Discrimination – unequal treatment for members ofgroups who differ on some observable characteristicsuch as gender or ethnicity but do not vary in theirunderlying productivity; unequal treatment ofotherwise equal individualsb. Individual discrimination – unequal treatment of anindividual or by an individual firmc. Market discrimination – unequal outcomes for allmembers of a particular group; in labor markets, wetypically focus on wage outcomesd. Prejudice – psychic disutility associated with aworker’s demographic or other characteristics whichare unrelated to the worker’s productivity (motivationfor individual discrimination)e. Segregation – unequal representation of differentgroups with equal productivity and motivation indifferent markets or occupationsB. Theoretical Models of Discrimination (Cain 1986)2. Discrimination by consumersa. price of discrimination1. assume that consumers are prejudiced (have tastesfor discrimination) and are willing to pay adifferent price to associate only majority workers2. price of a majority worker, p3. price of a minority worker, p! d4. d captures the value of the taste for discriminationb. advantages of measure: continuous, potentiallymeasurable, cost interpretationc. disadvantages: does not measure stigma felt by victim,requires interaction between minority workers andconsumersd. leads to segregation but not necessarily wagedifferentials; hence, market level of discriminationmight well be zeroe. provides an insight into the difference betweenindividual- and market-level discriminationf. more intriguing insight: segregation eliminates wagediscrimination2. Discrimination by workersa. majority worker’s wage demand with other majorityworkers, wb. demand with minority workers, w + dc. again leads to segregation but not necessarily marketdiscrimination; if there are equal skills across workers,competition among employers, and mobility acrossmarkets, minority workers can still earn w elsewhere3. Discrimination by employersa. assume all employers discriminate and set wages forminority workers at w! D; this does lead to asustainable equilibriumb. if instead employers discriminate to varying degrees(w! di), market differential will (if there are fewenough minority workers) simply be the lowest levelof di (0 if some employers do not discriminate)c. again, the model leads to segregation but notnecessarily wage discrimination4. Product monopolya. two advantages of assuming a monopoly:1) definitional uniformity in tastes2) above-competitive profits to support dib. disadvantage: does not necessarily translate tomonopoly power in labor marketc. thus, the firm would still have to pay w to minorityworkers, though it could segregate workers or payhigher wages to majority workers5. Monopsony – single buyer of labora. in this case the firm has power in the labor marketb. permits exploitation ([VMP ! w] / w); greaterexploitation allowed for group with more inelasticlabor supply curvec. genuine monopsonies (e.g., one company towns) arerare; although its not hard to imagine that some firmsat some times have a limited amount of market powerd. need to consider the elasticity of labor supplycarefully; for instance, the model does not explaindiscrimination against women (who presumably havemore elastic labor supply curves than men)6. Labor unions as monopoliesa. to secure monopoly rents, some method of restrictingentry is a necessary first stepb. if discrimination is consistent with members’ tastes,could help to prevent members from defectingc. not clear that union story matches up with time seriesevidence on discriminationd. also is not consistent with recent trends in unionmembership7. Statistical discriminationa. let q be worker’s true innate productivityb. employer observes y = q + u; y provides employerswith a signal of qc. employers pay w = E ( q | y )d. leads to individual discrimination but not groupdiscrimination (in fact some minority workers aremade better off)C. Empirical tests2. Regression methodsa. Standard model, regress earnings on a set of observedproductivity characteristics, X, and a group(minority = 1) identifier, ZW = XN# + AZ + eestimates that A < 0 are taken to indicate thatdiscrimination is presentb. first issue is whether the productivity characteristicsare exogenous or reflect past or pre-marketdiscrimination1) e.g., previous employers could have discriminatedin deciding whether to provide training2) there may be segregation or discrimination inaccess to quality schooling3) no simple rules as to what variables to include c. second issue is the possibility that a productivitycharacteristic that is correlated with group status isomitted 1) leads to omitted variable bias2) effects of productivity and group status will beconfoundedd. findings1) using “standard” controls for productivity such aslevel of schooling and potential work experiencetypically reduces but does not eliminate thedifferences in wages2) including more elaborate productivity/abilitycontrols such as AFQT (see, e.g., O’Neill 1990)can eliminate the difference in wages3. Audit methodsa. methodology1) send two equally (or artificially) matched jobcandidates to apply for jobs2) record the reactions that each encounters3) look for systematic differences in treatmentb. advantages:1) presumably controls for productivity differences2) can examine reactions at all stages of theapplication process3) is useful in detecting subtle differences in behaviorc. disadvantages:1) hard to truly control for all relevant differences;Heckman (1998) shows how particulardistributions of relevant differences can lead tofalse conclusions of discrimination2) relatively expensive (have to train and compensatetesters)3) informed testers may look too hard for differences(more generally, there may be differences inperceptions between majority and minority testers)4) possibly unethical (people are applying for jobsthey don’t intend to fill and taking up firms’interviewing resources)5) only indicates individual discrimination, notmarket discriminationd. results from