Customer Preference and Race in the NBA

Sep 26, 2014; Cleveland, OH, USA; Cleveland Cavaliers forward Kevin Love (0), Cleveland Cavaliers forward LeBron James (23) and guard Kyrie Irving (2) pose for a photo during media day at Cleveland Clinic Courts. Mandatory Credit: Ken Blaze-USA TODAY Sports

Conversations about race are inextricable from the culture and politics of the NBA on all levels. In the last year we’ve seen two different front-office oustings for racist comments. Players have been vocal contributors to public dialogue about the deaths of Mike Brown and Eric Garner, among others. The news this year has been more racially charged than it has been in recent memory (at least in my young lifetime) and NBA players, commentators, and writers have been in the thick of the conversation. Those who have participated have received significantly less vitriol than their peers in other sports, in part due to an audience that is younger, more diverse, and generally more liberal.

This in mind it may seem that the NBA is good at policing itself for racist practices. Issues of race that have popped up recently, like Donald Sterling, have been near uniformly denounced and remedied. But a paper published in the Journal of Applied Economics in 1991 titled, Customer preference, attendance and the racial structure of professional basketball teams, found a link between the proportion of a team that is white and the proportion of the surrounding area that is white. The authors’ thesis, based on customer-based discrimination, is that if fans prefer to see players of their own race then basketball teams will be influenced in their selection of players by the racial composition of the surrounding area.

The first model in their paper, which finds the link between the team and the surrounding area, covered the 1980-81 through 1985-86 seasons. They use two models, a linear probability regression and a probit regression, to estimate the effect of a Metropolitan Statistical Area’s (MSA, as defined by the US Census Bureau) racial composition on the racial composition of its associated NBA team.

After controlling for income and workforce variables, both models found a positive correlation that is significant at the .01 confidence level. They found in their linear probability model that a 1% increase in the proportion of the metro area that is white is associated with a 0.916% increase in the proportion of the team that is white, or nearly a 1:1 relationship. Interestingly this equation has an R^2 value of 0.40, meaning that for all the complex factors that go into making a team, approximately 40% of the variability in the racial composition of a team is explained by demographic variables of the MSA. When all other variables are removed except for the proportion white in the MSA that R^2 value drops to 0.32, so the bulk of the variability in team racial composition is determined by the MSA’s racial composition.

I updated these models to see if innate racism is still fueled by revealed preference for NBA teams. Two complications of my models are the influx of foreign born players since the late 1980s and the qualitative nature of determining a player’s race. Race is a social construct, so it’s not quantifiable in the same way as other demographic variables, I do not have access to anything where players have been surveyed and asked to identify their race. To do this I went player by player on each team for the 2009-10 through 2013-14 seasons and tried to determine if a white ticket purchaser would feel a player “looks like them”. This method is far from ideal. Foreign born players that “look white” further complicate updating this study. To alleviate, this I iterate each model twice, once for only including white, US-born players in the dependent variable and once including all white players regardless of birth place. Finally, I did not include the Toronto Raptors because of data availability.

Full results are included below, but the big takeaway is that the effect observed in the 1980s still exists, but has been significantly reduced. The racial composition of the area around an NBA team appears to affect personnel decisions on that team. The “% Pop. White” variable is the proportion of a MSA that identifies as white in the most recent US Census. The first column shows a coefficient of 0.14 meaning that a 1% increase in the proportion of the population that is white increases the proportion of the team that is white and born in the US by 0.14%, slightly more than 84% reduction in the effect seen in the original paper. Interestingly, the effect is seen to increase when we include white players that are not born in the United States. The coefficients in the probit regression are z-scores, which are associated with a Q probability. This means that the effect in the third column with a 3.06 coefficient, associated with a Q probability of 0.0011, means a 1% increase in the proportion of the population that is white increases the proportion of the team that is white (US born) by 0.11%. Also interesting is that in the probit model including the foreign born white players actually decreases the strength of the effect. The R^2 value in the linear probability model is also more than halved, so MSA demographics explain less of the variability in the racial composition of the team than they did in the 80s.

There are obvious limitations for what this can tell us. The original authors did further analysis that showed that if the demographics of the team matched the demographics of the area attendance would increase. In doing so they split players by whether or not they were starters during that season, and this effect was statistically insignificant. In my analysis there’s no differentiation by playing time, theoretically a player who never sees the floor may have a different effect on attendance than a starter. Also I can’t argue that this really proves anything, all I can do is defend the controls used and hypothesize why we see this effect. I can’t say whether or not GMs and owners are more or less racist, or if they ever were, whether they may just be profit maximizing.

Kevin Arnovitz’s recent piece on the rebranding of the Atlanta Hawks presents the idea that pursuing affluent, white males may not make sense anymore as an NBA business strategy. If fans of the NBA are truly responsive to teams looking like them we may be seeing players receive inflated or deflated contracts based on their ability to draw fans, which is not a new concept. A clickbait version of this analysis might say something about empirical proof of a race problem in the NBA, which is clearly misleading, but this does show, along with much recent history, that the NBA is not free from discrimination at its base definition.