مدل عرضه و تقاضای همکاری کارگر، کارفرما و تبعیض مشتری

When racial wage differentials reflect customer, co-worker and employer discrimination, how can these sources of discrimination be theoretically and empirically distinguished? We develop a wage model fashioned around professional basketball that includes customer, employer and co-worker discrimination when there is racial integration. We find that the model is capable of: (1) predicting unexpected effects of the various types of discrimination on wages; (2) the three types of discrimination will interact nonlinearly; and (3) compensation attributable to co-worker prejudice is endogenous. Using data from the National Basketball Association, we find evidence consistent with co-worker discrimination by white players and customer discrimination by nonwhite fans.

In his pioneering work on the economics of discrimination, Becker (1971) suggested three sources of discrimination in the labor market—employer discrimination, employee (or co-worker) discrimination and customer discrimination.1 While the literature abounds with studies of racial and gender discrimination in wages, no study has considered all three types of discrimination in a wage model theoretically and empirically. Many of these studies have confirmed the presence of racial discrimination in various industries, but they have generally not been concerned with attributing estimated, cet. par. racial wage differentials to a specific source(s) of discrimination. There are three important reasons why separation of Becker's three sources of discrimination has eluded researchers. First, the traditional methodology used to estimate racial discrimination in wages has precluded the identification of a specific discrimination type in racial wage differentials. This methodology involves estimating the coefficient on a dummy variable indicating the race of a worker in a wage regression.2 Unfortunately, the estimated coefficient on the worker race dummy cannot pinpoint the source(s) of discrimination because the coefficient could be measuring a combination of all three.3 Second, no study has tried to empirically distinguish between Becker's three sources of discrimination in a wage regression because little is known about how they might be theoretically distinguished. That is, without formally deriving an equilibrium wage equation from worker and firm optimization, most researchers have only estimated some version of a Mincerian wage regression. Third, when researchers have done tests of specific types of discrimination, they have typically used models that assume only one type of discrimination and the tests have been indirect in nature. In fact, a majority of the studies attributing estimated wage differentials to a specific source of discrimination have been in professional sports.4 Overall, drawing any conclusions about specific sources of discrimination is problematic because no theoretical model exists that directly allows for separation or even considers whether separation is possible. In this paper, we focus on the problem of theoretically and empirically distinguishing between the specific sources of discrimination in a, cet. par. racial wage gap. We believe that it is very important to examine the distinction between Becker's three types of discrimination for several reasons. First, a theoretical model which considers all three types of discrimination would be helpful in understanding how the market, cet. par. wage differential (Becker's Market Discrimination Coefficient) is influenced by each discrimination type and their interactions. A theoretical model would be useful in understanding how each type of discrimination influences production, cost and labor supply decisions and, in turn, how those decisions affect the level of market discrimination. Second, finding a method that considers Becker's types of discrimination can help shape public policy focussed on the eradication of discrimination. For example, if we know employer prejudice is the primary source of discrimination in the labor market, policymakers could develop policies specifically targeted at influencing the behavior of employers. On the other hand, if customer discrimination were found to be the primary source of discrimination, then policymakers might develop different policies. In what follows, an equilibrium wage model is developed, which includes Becker's three sources of discrimination. We extend Becker's model of discrimination when complete segregation is impractical. Becker's model primarily assumes perfect substitution between minority and majority labor. One implication of perfect substitution is complete racial segregation when labor cost differs between the two groups. However, racial integration is the rule, not the exception, in most industries. Therefore, we develop a model consistent with racial integration by assuming imperfect substitutability between white and nonwhite workers. Hashimoto and Kochin (1980) and Goldberger (1984) suggest that many discrimination studies have used imperfect proxies for productivity. Use of such proxies can lead to omitted variables bias. Consequently, we carefully chose an industry on which to base our theoretical model where the risk of omitted variables bias was smallest. Because of the high quality of worker-specific data in professional athletics, we have fashioned our model around professional basketball. While our results are from the NBA, our theoretical model is sufficiently general to be applicable to many other industries. We assume that team profits depend on attendance and we allow for the possibility that owners indulge their tastes for discrimination and cater to the prejudicial tastes of fans and coplayers. Our model has several novel features. First, our model is based on “two-way” discrimination, i.e. whites can be prejudiced against nonwhites and nonwhites against whites. With the exception of McCormick and Tollison (1997), this approach differs from the “one-way” discrimination of previous studies. Second, we introduce customer discrimination in the firm's production function, employer discrimination in the firm's cost function and co-worker discrimination in the reservation wage equations. Third, we explicitly derive reduced form demand and player wage equations separately for white and nonwhite labor. Player wage equations are estimated using a sample of 489 NBA players from the 1985/1986 and 1990/1991 seasons. Our salary regressions include proxies for the three sources of discrimination and illustrate how our model can be applied to other industries.

This paper has demonstrated that Becker's three sources of discrimination can be incorporated into a single model, both theoretically and empirically. However, unlike the clear predictions of standard discrimination models, a model of a racially integrated firm in which all three types of discrimination may coexist yields ambiguous predictions. The ambiguity results from relaxing the perfect substitution assumption and allowing for two-way co-worker discrimination effects on wages. Our wage model is the first we know of that has all three types of discrimination in a wage equation, where the equation is derived from a model of labor demand and supply and where imperfect substitution is allowed between the two classes of labor in production. One insight of the model is that the firm's racial composition can influence wages. In the literature, racial composition of the firm has received little attention. We estimated a short-run counterpart to the long-run theoretical model because institutional constraints in the NBA limit the ability of the team to alter its racial mix in a given year. We found no evidence of employer discrimination that had a significant effect on wages. We found evidence consistent with conventional white co-worker discrimination, which suggests that co-worker discrimination should receive more attention in future research. We believe that the absence of black co-worker discrimination in our sample may reflect the dominance of blacks in the NBA labor pool. Finally, we found evidence consistent with black fans engaging in customer discrimination, but not in the case of white fans. Regardless, we caution that these results should be very carefully interpreted given the possibility that our proxies for discrimination could be given a different explanation. For example, the model in Section 2 implies that discrimination may have unexpected effects on wages. However, this study represents the first effort to identify all three forms of discrimination and illustrates some potential avenues for future researchers to examine all three types of discrimination. Finally, this study has raised another important question for future investigation: in the case where whites and nonwhites are perfect substitutes and there is labor market discrimination, why don't we see more occupational or firm level segregation? Our study has demonstrated that the production function and reservation wages are key to understanding racial integration in American industry. Other factors driving integration unrelated to production, e.g. search frictions, imperfect information, and other sources of transaction costs, must also be examined.