New evidence of salary discrimination in major league baseball ☆ Paul Holmes ⁎ State University of New York at Fredonia, United States abstract article info Article history: Received 25 March 2010 Received in revised form 23 November 2010 Accepted 30 November 2010 Available online 23 December 2010 JEL classification: J71 L83 Keywords: Salary discrimination Baseball MLB Quantile regression Salary discrimination in MLB has largely been discarded as a research topic. However traditional quantitative methods (particularly least squares regression) have concentrated on the effect of race for the average player. If only a particular salary class of players are subject to discrimination, or if the size of the discrimination is small at the average, then least-squares techniques will struggle to identify discrimination. I use quantile regression to uncover salary discrimination against black players in the lower half of the salary distribution. Not only are the premia for white and Hispanic players statistically significant, but they are large: up to 25% of salary for the bottom quintile of players. © 2010 Elsevier B.V. All rights reserved. 1. Introduction Professional baseball in America has acted as a microcosm for examining racial and ethnic discrimination. Baseball has followed, and in some cases led, the American trend toward integration in the twentieth century. Jackie Robinson's breaking of baseball's color line in 1947 set the precedent for integration, though it hardly removed all discrimination. Baseball (and professional sport in general) offers an excellent testing ground for labor economics theories. Sport (and especially baseball) is replete with performance measures, an integral part of many labor studies. Studies of discrimination patterns in professional sports can be especially fruitful, with long periods of detailed observations of individuals and organizations, a wide variety of situations to examine, and major rule changes leading to natural experiments. If salary discrimination can exist in such a highly scrutinized and visible industry, the lessons we learn should be helpful in understanding and uncovering discrimination in wider contexts. The clearest empirical examples of discrimination in Major League Baseball would be violations of the “equal pay for equal work” ideal, as set forth by the Equal Pay Act (1963) and the Civil Rights Act (1964). Recent examples have not generally been forthcoming, with most studies finding no significant difference (see Kahn, 1991, 2000). However, Fizel (1996b) finds that salary arbitrators seem to favor White players over their Black and Hispanic counterparts. The typical response to not finding discrimination in earnings has been to look elsewhere. Bellemore (2001) found that blacks and Hispanics were less likely to be promoted to the major leagues than whites in the 1960s and 1970s, with blacks still facing this discrimination in the 1990s. Jiobu (1988) found that blacks in the 1970s and 1980s tended to have shorter careers than whites, once performance variables were taken into consideration, a finding which he interpreted as discrimination. Other researchers have considered various forms of customer-based discrimination and positional segregation; for a detailed overview, see Kahn (2000). This essay reconsiders the question of “equal pay for equal work”.I use a longer time frame than is common – nine years compared with the more typical one or two – and restrict attention to free agents (players who are free to sign a contract with any team they choose). This is a sensible restriction. Some other researchers simply estimate salary equations for all players in a given season (e.g. Kahn, 1993; Palmer and King, 2006). A concern with this approach is that player performance can change markedly within a long contract, and multi- year contracts are common. The link between salary in the last year of, say, a seven year contract, and the performance variables in the sixth year, is likely to be strained. Much better is to consider only players who have just signed a contract, since the salary should account for all performance characteristics observed up to that date. While this results in a smaller data set than would otherwise be available, the data quality is likely to be superior. I make up for the reduced number Labour Economics 18 (2011) 320–331 ☆ This essay formed one chapter of my dissertation at the University of Illinois at Urbana-Champaign. I wish to acknowledge the helpful comments of Dan Bernhardt, Brad Humphreys, Anthony Krautmann, several anonymous reviewers, and seminar participants at the Illinois Economic Association meetings in October 2007, and at the University of Illinois. A data appendix with additional results, and copies of the computer programs used to generate the results presented in the paper, are available from the author upon request. ⁎ Tel.: +1 716 673 3829; fax: +1 716 673 3506. E-mail address: holmes@fredonia.edu. 0927-5371/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.labeco.2010.11.009 Contents lists available at ScienceDirect Labour Economics journal homepage: www.elsevier.com/locate/labeco