Page 1 of 12 May 26, 2015 Theories of Labor Market Discrimination • Differences between groups in labor market outcomes do not necessarily imply discrimination. • There are two main theories of labor market discrimination: – Preference-based discrimination – Statistical discrimination Preference-Based Discrimination Employer Discrimination • Start from the supposition that men and women are equally productive. Labor is the only factor of production. • However, employers do not like hiring women and act as if there is an extra cost to hiring women. VMPE=WF(1+d) – Where d is the discrimination coefficient. – d may vary across firms ECON 139 SP ‘15 Antonovics 9 5-26-15 1Page 2 of 12 Firms that Do Not Discriminate (d=0) • Suppose WM>WF. • A firm that does not discriminate, will hire only women since they are cheaper. • The firm will hire women up to the point where the value of the marginal product of labor equals the wage. • VMPE=WF Firms that Discriminate ( d>0 ) Psychological (non-mandatory) cost associated with hiring women: WF (1+d) • Hire only men if: WF (1+d)>WM • Hire only women if: WF (1+d) < WM As d increases, firms switched from hiring women only to men only. Discrimination Lowers Profits On the previous slide, note that discrimination lowers firms’ profits for two distinct reasons. 1. Employers who hire women are hiring the “wrong” number of women (if d>0). E0 and E1 2. Employers who hire men are paying higher wages.Page 3 of 12 dc solves WF (1+dc) = WM Discrimination should not survive if markets are competitive. Pi max means the highest possible profit Labor Market Equilibrium Nd=0 means number of woman that can be hired by non – discriminating firm.Page 4 of 12 Predictions 1. The wage differential between men and women depends on the number of women in the labor force. 2. The wage differential between men and women depends on the number of firms that discriminate. Preference-Based Discrimination Employee Discrimination • Whites do not like working with blacks and blacks are indifferent between working with whites and members of other racial groups. • This implies that employers must pay white workers a compensating wage differential to work with blacks. • Segregated Workforce. Firms can avoid these higher labor costs by hiring either only blacks or only whites.Page 5 of 12 • No Wage Differential. Employers will hire whichever group is the cheapest. Thus, competition will eliminate any racial wage differential. Preference-Based Discrimination Customer Discrimination • Consumers don’t like buying goods made or sold by blacks. • Purchasing decisions based on utility-adjusted price, p+d. • Consumer discrimination reduces the demand for goods and services made and sold by minorities. • Segregated workforce. Firms have an incentive to place minority workers in jobs with little customer contact. • No wage differential. No wage differential between blacks and whites as long as firms can “hide” black workers. Statistical Discrimination • Basic Logic – Two people, identical resumes. – One of the applicants is a man, and the other is a woman. – Statistical evidence indicates that women are more likely to quit their jobs than men. – Quits hurt employer because disrupt teamwork, etc. – Thus, the employer will hire the man. • Statistical discrimination arises because firms do not have complete information about worker productivity. • Same logic as racial profiling. The Impact of Statistical Discrimination on Wages • Employers observe T, a noisy productivity measure. Can be thought of as a test score. • Ta is the average productivity/test score for group. • Wages based on what you observe about individual and what you know to be true about their group. W = pT + (1-p) Ta • p measures the correlation between the test score and productivity. Two Ways in Which Statistical Discrimination Leads to Wage Inequality CASE 1: The average productivity varies across groups, but p is the same. In this picture, blacks are less productive than whites on average (Tw>Tb). Even if blacks and whites have the same T, blacks paid lower wages.Page 6 of 12 CASE 2: The average test score is the same across groups, but p differs. In this picture, firms have less accurate information about blacks than about whites (pb<pw) High scoring blacks earn less than high score whites. Low scoring blacks earn more than low scoring whites.Page 7 of 12 Should Employers Use Group Averages? • From the standpoint of profit maximization: yes. • From the standpoint of efficiency: yes (maybe). • From a moral/legal standpoint: no. • And, the very process of using the group average to infer productivity can lead to self-fulfilling prophecies . . . – Example: women and the decision to leave the labor force to have children. Measuring Discrimination How do you measure or quantify discrimination? • One possible definition of discrimination is to look at the difference in the mean wage between two groups, say men and women. • This measure is unappealing because there are many reasons why men earn may more than women (differences in education, experience, occupation). • A more appropriate measure would compare the wages of equally skilled workers. Oaxaca Decomposition • To simplify the exposition, let’s suppose that only one variable, schooling, affects earnings. Male earnings function: WM = αM + βMS Female earnings function: WF = αF + βFS • Can estimate these parameters by running a regression. • β tells how much an individual’s wage increases as the result of an additional year of schooling. • α tells the level of the earnings profile for each group. • So that the difference in the average wage of men and women becomes:Page 8 of 12 Diagram of the Oaxaca DecompositionPage 9 of 12 May 28, 2015 Critique of the Oaxaca Decomposition n The Oaxaca decomposition may overstate the extent of discrimination if – We cannot control for all of the relevant observable differences between men and women. – In this case, part of the wage difference