Reassessing customer discrimination in local market National Basketball Association television viewership

ABSTRACT

This study revisits the question of whether evidence of race-based customer discrimination exists in local market television viewership. We assess how consumers in 27 of the largest U.S. media markets respond to the racial composition of playing talent using modern local market television viewership data from the National Basketball Association. An advantage of using viewership data from this context is the ability to control for the perceived and actual quality of the product, allowing for the ability to isolate the effects of player race. We first uncover statistically significant increases in viewership when Non-White players occupy more roster spots and receive more playing time relative to White players. Individual player race models produce similar results and indicate consumers respond favourably to Black and Hispanic players. The results oppose previous research in the same setting and suggest viewership preferences related to race have changed substantially over time.

KEYWORDS:

• Consumer discrimination
• racial preferences
• television viewership
• National Basketball Association
• consumption

JEL CLASSIFICATION:

• J15
• L83
• Z20

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ The active roster size limit for NBA teams is 15 players, but a team must dress a minimum of eight and a maximum of 13 for a given game. Therefore, the maximum number of players in a single game is 26.

² There are ten players on the court in a professional basketball game (five per team) and each player plays 48 minutes (12 minutes per quarter). Regardless of substitution frequency, the total amount of playing minutes is 480 (240 per team), unless the game goes into overtime.

³ The Elo rating is a commonly used individual team quality measure which updates at the conclusion of every league contest. Please see Hvattum and Arntzen (2010) and Mills et al. (2016) for further details.

⁴ A typical NBA regular season runs from mid-October through mid-April. Thus, the league calendar coincides with MLB through the end of October, NFL through the first week of February, and NCAA football through January.

⁵ Alternatively, we re-estimated all models controlling for the minority (Non-White) population percentage in each MSA and the results are practically similar to what is presented.

⁶ More information on the metrics utilized can be found at: https://www.basketball-reference.com/about/glossary.html. In the limited cases where there are no Hispanic players or players of other races in the top five of team playing minutes, we impute the replacement player value for those player races in that contest (WS/48 = 0.000; VORP = −2.0). We also re-estimate all models using advanced metrics specific to the top ten players on each team (i.e. to account for all players who regularly see game action) and again by all rostered players and the results do not change in a meaningful way.

⁷ All-Star players of other races is the omitted category as no players fit this criteria during the sample. All metrics are calculated specific to season t as season-to-season player performance varies. Player quality metrics are not strongly correlated with team quality measures or the absolute contest quality measure and therefore, collinearity is not a concern.