Exit discrimination in Major League Baseball: 1990-2004.

• Author: Groothuis, Peter A.

1. Introduction

   The topic of discrimination in labor markets has attracted a great deal of research by economists over time. From the theoretical constructs developed by Becker (1971) to the development of the residual method by Oaxaca (1973), economists have struggled to determine the cause and effects of prejudice on minority workers. Professional sports have provided a fertile field for researchers of this topic because of the abundance of performance data and salary information available. In Kahn's (1991) article on discrimination in professional sports, he reviews empirical findings from numerous articles published between 1972 and 1989. He divides his summaries under four major headings: salary discrimination, positional discrimination, customer discrimination, and hiring discrimination. Many of the articles offered findings on more than one of the topic headings.

   Salary discrimination drew the most attention. Twenty-four articles with empirical work on this topic are critiqued by Kahn (1991): 12 on baseball, eight on basketball, and two each on football and hockey. The strongest evidence of salary discrimination against black athletes comes from studies of the National Basketball Association (NBA) based on data from 1984-1986. However, more recent research (Bodvarsson and Brastow 1999; Hill 2004) has failed to find salary discrimination in the NBA using data from the 1990s. Twelve articles involving research on positional segregation were reviewed (Kahn 1991): five from baseball, three from football, two from basketball, and two from hockey. Evidence suggested positional segregation or stacking existed for blacks in baseball and football; evidence also suggested that blacks were underrepresented at the center and forward positions in the NBA. Customer discrimination was the main or auxiliary focus in 15 of the papers (Kahn 1991): nine from basketball and six from baseball. Results are mixed for attendance at NBA games; a recent article (Kanazawa and Funk 2001) finds a negative correlation between the percentage of blacks on the roster and cable TV ratings. In baseball, evidence from mid-1970s revenue data fails to find customer discrimination, but studies using earlier data did. Customer discrimination also appears to exist in the baseball trading card market (Nardinelli and Simon 1990).

   Of the 13 pieces of empirical work focused on hiring discrimination reviewed by Kahn (1991), six deal with data from baseball, four with basketball, two with hockey, and one with football. Findings failed to confirm hiring barriers for black athletes in basketball and baseball; whites with inferior performances were not hired over superior black athletes. However, Johnson and Marple (1973) found evidence from 1970-1971 NBA data that marginal white players had longer careers than marginal black players. Hoang and Rascher (1999) explored the concept of racially based retention barriers in the NBA in more detail using data from the 1980s. They, too, found evidence that, holding performance constant, there was "exit discrimination" in the NBA. Groothuis and Hill (2004) failed to confirm Hoang and Rascher's results using more recent data, height as an added explanatory variable, and a duration model that allowed for both stock and flow samples.

   Jiobu (1988) found evidence that race decreased career length, ceteris paribus, for black players but not Hispanics using Major League Baseball (MLB) data from 1971-1985. The purpose of this article is to determine if Jiobu's findings can be replicated using more recent baseball data and a richer duration model that has season-variant data; Jiobu used only unchanging, career performance variables in his model. Exit discrimination is more subtle than positional segregation or pay discrimination. If exit discrimination in MLB is no longer present, it would seem to affirm Becker's theoretical implications that market competition will eliminate discrimination over time.

2. Theory

   Discrimination in the labor market implies that certain individuals or groups of workers are somehow treated differently than others unrelated to ability or performance; in the literature on professional sports, the focus has been on differential treatment of black, Hispanic, and French-Canadian players. The differential treatment can occur in a variety of formats. When pay discrimination occurs, workers in the group receiving the unequal treatment receive lower pay than others for the same performance level. Most researchers believe pay discrimination against black players in professional baseball was eventually eliminated by free agency and salary arbitration in the mid-1970s (Hill and Spellman 1984; Cymrot 1985). Research by Johnson (1992) even suggests that white players earn less than black players of similar ability on teams that have fewer black players.

   Hoang and Rascher (1999) define exit discrimination as "the involuntary dismissal of workers based on the preferences of employers, coworkers, or customers" (pp. 69-70). Jiobu (1988) and Hoang and Rascher (1999) concluded that career lengths for black players in MLB and the NBA, respectively, were lower than those of their white counterparts, ceteris paribus. While Jiobu does not make any calculations on the impact of exit discrimination on career earnings, Hoang and Rascher conclude that this form of discrimination led to an almost two and a half times greater decrease in black career pay compared with the more heavily analyzed form of pay discrimination.

   Becker (1971) suggests that the source of the personal prejudice that leads to various forms of labor market discrimination may be employers, coworkers, or customers. Recent research by Kahn and Sherer (1988) on pay discrimination in the NBA and Hoang and Rascher (1999) on exit discrimination in the NBA has focused on customers as the source of the prejudice; both studies found empirical evidence to support the existence of customer discrimination. The pay premium for white players was explained by the higher value of their performances compared with black players because of the prejudiced preferences of white, majority fans.

   Hoang and Rascher (1999, p. 74) hypothesized: "To satisfy the fans, there is a minimum number of white players on a team. The second assumption, that the pool of quality available talent is becoming increasingly black, causes annual replacement of players with rookies to occur mostly among black players. The white players have longer careers simply because there are fewer qualified white rookies to replace them ..."

   In his study of exit discrimination in MLB, Jiobu (1988, p. 532) does not specifically test for customer discrimination, but he does state: "Perhaps, motivated by the concern that white fans will not support a predominantly black team, management has silently placed an 'invisible ceiling' on the black percentage. When coupled with the desire to have a winning team, this ceiling would generate strong pressures to (a) employ as many black players as possible in order to capitalize on their performance, but (b) in order to remain under the ceiling, to eliminate black players as soon as their performance declined, and (c) to retain white players of declining but similar ability."

   Research on this topic assumes that all turnover is involuntary; Kahn (1991) argues that the high salaries paid in sports make voluntary quits unlikely. Thus, these studies are essentially survival models. If white players have longer careers than black players with similar performance statistics, then exit discrimination is said to exist.

3. Data

   Our data include all individuals, both hitters and pitchers, who participated in MLB from 1990 through 2004 for a 15-year panel consisting of 3185 players. (1) Jiobu's (1988) study on exit discrimination in baseball only looked at hitters. To capture the overall length of players' baseball careers, our data consists of both stock and flow samples. A stock sample consists of all ongoing careers at the start of the panel in 1990. These left-censored data are easily included because we know how many years each player had played in the major league prior to 1990. Our stock sample consists of 690 players who had an average tenure of six years as they entered the 1990 season. Including a stock sample captures information on players whose careers are longer than the panel data set. Using only stock data, however, would underrepresent short-career players, so we also include flow data.

   A flow sample includes all careers that start between 1990 and 2004. This sample captures many short careers in MLB. Including only flow data, however, would allow for no careers longer than 15 years, which is the length of our panel. As with most panels, our data are also right-censored where many careers were ongoing when our sample ended in 2004. Our right-censored data include both stock and flow observations. To estimate a duration model of stock and flow data, we use a technique developed by Berger and Black (1998).

   The variables in our data include both season-variant and season-invariant data. Jiobu (1988) used only season-invariant data for his analysis. We report the means in Table 1A for hitters and Table 1B for pitchers. Our season-invariant data include dummy variables for player's race; a dummy variable for infielders, first basemen, and catchers; and a dummy variable for pitcher's throwing hand. (2) Our season-variant data for hitters are age, age squared, a dummy variable for each team's primary designated hitter, and performance data that include games played as well as slugging average, home runs per game, stolen base percentage, runs batted in per game, walks per game, strikeouts per game, hit by pitch per game, sacrifice flies per game, and runs scored per game. Our season-variant data for pitchers are age, age squared, and performance data that include games played as well as earned run average, wins, losses, saves, strikeout-to-walk ratio, and...