Home Address:
  111 Park Street, Apt. 17B
  New Haven, CT 06511

Telephone: (203) 215-8500

Labor Economics
Public Economics
Economics of Discrimination

Labor Economics
Public Economics
Microeconomics
Applied Econometrics

October 2002 (Oral): Labor Economics
May 2002 (Oral): Public Finance
May 2001 (Written): Microeconomic and Macroeconomic Theory

"The Role of Statistical and Taste Discrimination in Racial Disparities"

Professor Hanming Fang
Professor Joseph Altonji
Professor Patrick Bayer

May 2006

M.Phil. (2003), Economics, Yale University
M.A. (2002), Economics, Yale University
B.A. (2000), Economics, University of California, Berkeley (with departmental honors)

Yale University Dissertation Fellowship, Fall 2004
John F. Enders Award, Summer 2003
Summer Fellowship, Yale University, 2001 and 2002
Yale University Graduate Fellowship, 2000-2004

Teaching Assistant, Public Finance, Spring 2005
Teaching Assistant, Microeconomics with Environmental Applications, Fall 2002 and Fall 2003
Head Teaching Assistant, Introductory Microeconomics, Spring 2003

Research Intern, Economic Research Service, U.S. Department of Agriculture, Summer 1999

Professor Hanming Fang
Department of Economics
Yale University
PO Box 208264
New Haven, CT 06520-8264
Phone: (203) 432-3547
Fax: (203) 432-6323
Email: hanming.fang@yale.edu

Professor Patrick Bayer
Department of Economics
Yale University
PO Box 208264
New Haven, CT 06520-8264
Phone: (203) 432-6292
Fax: (203) 432-6323
patrick.bayer@yale.edu

The theoretical literature on discrimination has proposed two main reasons why differential outcomes can arise for observationally equivalent individuals of different races. Taste-based theories postulate that differences in outcomes can develop if economic agents have a distaste or prejudice towards a particular racial group. On the other hand, statistical discrimination models attribute these outcomes to decision-making evaluators having incomplete information about the individuals under consideration. If evaluators believe that an individual’s unobservable skill level is correlated with their racial background, they will have an economic incentive to take an individual’s racial group into account when assessing their skill level.

Although there is an extensive empirical literature studying discrimination, most studies have focused on quantifying how much of the outcome differentials between racial groups can be attributed to discriminatory behavior. Very few studies have attempted to empirically determine whether these practices have arisen from taste or statistical discrimination. This is an important distinction because the effectiveness of policies to reduce discriminatory behavior depends upon the type of discrimination that is present. My dissertation provides evidence on why discrimination arises and what factors affect it. The first essay investigates whether people statistically discriminate when evaluating the skills of others, and whether the presence and degree of statistical discrimination depend on the demographic background of the evaluator. The second essay develops an empirical test that determines whether racial differences in motor vehicle searches are due to taste or statistical discrimination. The third essay evaluates whether taste discrimination on the part of customers might be responsible for the wage differential between black and white professional basketball players.

I. Testing for Cultural Affinity and its Implications on Statistical Discrimination Using Data from the Game Show Street Smarts (job market paper)

This essay tests the validity of an assumption made in some statistical discrimination models which requires that every evaluator has the same perception of the unobservable skill distribution of a given racial or gender (cultural) group. A more realistic view is that this perception depends upon the cultural background of the evaluator, a notion that has been termed cultural affinity. Past research studies that have assumed away cultural affinity ignore the possibility that the amount of statistical discrimination a cultural group faces might depend upon the cultural background of the group evaluating them.

Although ideally one would like to determine whether cultural affinity is present in labor market hiring situations, there is no existing data set that contains detailed information at the individual level about both applicants and employers. To circumvent this problem I use data from the television game show Street Smarts. During this game contestants earn money by predicting how random people, known as street savants, answered basic trivia questions. These predictions will depend on both a savant’s observable skill, as well as a contestant’s perception of their unobservable skill. In order to use contestant predictions to estimate these perceptions, one must control for taste discrimination on the part of contestants, which can also influence their predictions. Due to the unique dynamic structure of this game, the effect a savant’s unobservable skill has on a contestant’s predictions will depend on the savant’s prior performance. The effect of taste discrimination towards a cultural group of savants will not, however. Estimating a structural model of contestant behavior allows me to use this difference to empirically distinguish between the effects of statistical and taste-based discrimination.

I find evidence that cultural affinity is an important factor in how the unobservable skill distribution of a particular cultural group is perceived. More importantly, the results show that while black contestants view the average level of unobservable skill for blacks to be lower than that of whites, white contestants view it to be the same for both groups. In terms of gender, male contestants view average unobservable skill to be lower for females than for males, while female contestants view men and women equally. Although admittedly these perceptions are identified from a case study and thus may not be directly applicable to real world settings, they imply that the statistical discrimination blacks and females encounter in the labor market is likely to depend upon the cultural background of the employer. With slight improvements to current labor market hiring data, the methodology proposed here could be used to identify the perceptions employers have when making hiring decisions.

II. An Alternative Test of Racial Prejudice in Motor Vehicle Searches: Theory and Evidence (joint with Hanming Fang)

This essay proposes a theoretical model of trooper behavior that implies a simple empirical test to determine whether troopers exhibit relative racial prejudice in their decisions to search motor vehicles. We use a unique data set which contains demographic information about both the motorists searched on Florida highways and the troopers that conducted each search during a two-year period. The data also records the outcome of each search (i.e., whether or not the search was successful). Exploiting the information we have about trooper race allows us to develop an outcome test for determining relative racial prejudice that does not rely on comparing the search outcomes of the marginal motorists searched in each race. While all previous outcome tests have relied on this, they cannot be applied to the data because researchers cannot empirically identify the marginal motorist.

Our test relies on comparing the average search outcomes between trooper races for a given racial group of motorists. The key idea is that if troopers are not using racial prejudice in their search decisions, the ranking of average search success rates between white and minority troopers should be independent of the racial group of motorists. Because our test relies on comparing the search outcomes of the average motorists searched (which are observable) instead of the marginal motorists searched, it provides a partial resolution to the infra-marginality problem that has plagued the empirical application of outcome tests in the past. Applying our test to the data, we find that although black and Hispanic motorists are searched more than whites, we cannot reject the hypothesis that troopers do not exhibit relative racial prejudice.

III. Wage and Customer Discrimination in the Professional Basketball Labor Market

This essay estimates the mean salary differential between equivalent black and white basketball players in the National Basketball Association (NBA) for the 2000–01 season, and tests whether fan preference for white basketball players (customer discrimination) might be a factor in this wage differential. The NBA provides a unique environment to estimate wage differentials because, unlike standard labor market situations, the productivity statistics of players are observable to the researcher. I find that, among players who are in at least their second contract, blacks make 17% less than equivalent white players.

I estimate fan preference for white players by looking at the effect the racial composition of the road team has on the home team’s attendance. Because the home team’s preferences are unlikely to affect the racial composition of the road team, this identification strategy alleviates the endogeneity problem that is present in previous studies which measure fan preference as the effect the home team’s racial composition has on the home team’s attendance. The results show that increasing the percent of the minutes played in a game by white players on the visiting team by 1% would have led to an increase in per-game fan attendance of the home team by .11%. Although one cannot show that the relationship between customer discrimination and wage discrimination is causal, finding evidence of both of these types of discrimination against black players is consistent with the notion that customer discrimination will cause wage differentials to appear between equally qualified NBA players.