Does Dating a Celebrity Affect Athletic Performance? A Closer Look at the “Kardashian Effect”

By Matty Cheng

Some believe that dating a celebrity could affect athletic performance, whether due to the actual relationship or the media and paparazzi that come with dating a celebrity.  Given the explosion of athletes dating celebrities in the news, I set out to re-examine the effect of celebrity dating.  To investigate the matter, I examined athletes from the NBA, NFL, MLB, and NHL.

I defined a relationship with “celebrity” as when an athlete and their partner were mentioned in one of the top gossip news sites (including sites such as TMZ, People, E Online). I first scraped all the athletes from the four major U.S. sports leagues in the last 20 years. Then, I ran a search algorithm utilizing Google News to determine whether each athlete was mentioned in one of the top gossip news sites in a dating context. Afterwards, I manually searched gossip sites for each potential candidate to determine whether they dated a celebrity, what type of celebrity, and the timeline of their relationship.  This gave me a dataset with 104 athlete-celebrity relationships (some were not included in the final dataset due to injuries such as Julian Edelman who dated model Adriana Lima, but was injured the entire season that he dated her), including some athletes who dated multiple celebrities such as NBA player Jordan Clarkson who dated Kendall Jenner and Bella Hadid.

I measured an athlete’s athletic performance with various advanced statistics, specific to each sport.  For the NBA, I used Win Shares per 48 minutes (WS/48), which is a statistic that attempts to calculate a player’s value for winning NBA games.  Next for the NFL, I used Approximate Value (AV), a metric to measure the seasonal value of a NFL player. I used Wins Above Replacement (WAR), a baseball sabermetric to measure the value for team wins, for the MLB.  Finally, for NHL, I used Point Shares (PS), which also attempts to measure the number of wins created by a hockey player.

To examine the impact of dating a celebrity, we need to compare an athlete’s performance before and after dating.  We must also take into account that there are natural trends as players get older.

To address this, we can do a difference-in-difference analysis, which compares the difference between two groups before treatment and then looks at the same two groups after treatment. In this case, the “treatment” is beginning to date a celebrity.  The first “group” is a particular athlete who dates a celebrity while the second group is the average value for that athlete’s sport. Using the difference-in-difference analysis is crucial because consider a scenario in which an NBA starts dating a celebrity in his second season, and subsequently sees an increase in player value.  But if the natural trend is for NBA players to make strides in their second year, then the jump in his performance may not be correlated with dating a celebrity at all. So, we use the difference-in-difference technique to determine the true effect of dating a celebrity.

As an example, below is a graph displaying the effect of dating Kendall Jenner had on Blake Griffin’s value.

The blue line plots Blake Griffin’s WS/48 over time while the orange line plots the average WS/48 of a NBA player as he gets older. The green line is the time of Blake Griffin beginning to date Kendall Jenner and the red line is the time of the breakup.

The difference between Blake Griffin’s blue line in 2017 and the average NBA player’s orange line is 0.0909, which is the first difference and means that Blake Griffin was 0.0909 WS/48 better than league average before dating Kendall Jenner. The difference between the Blake Griffin’s blue line in 2018 and the orange line is 0.0334, which is the second difference and means that Blake Griffin was only 0.0334 WS/48 worse than league average after beginning to date Kendall Jenner. Then we take the difference of these two differences to get -0.0575, which means that Blake Griffin became -0.0575 worse after beginning to date Kendall Jenner relative to how a typical NBA player progress through his career. Comparing this to Blake Griffin’s original WS/48 before dating, we find that his performance has dropped by 32% relative to the trajectory of a typical NBA player.  Using the same technique, we find that after breaking up with Kendall Jenner (red line), Blake Griffin’s performance amazingly increases by 28%.

I used this process for the entire dataset, finding that athletic performance dropped by an average of 49.25% after beginning to date a celebrity. After a break up, athletic performance on average increased by 84.15%.

We can break this down by sport as seen below:

For fun we can break it down by occupation of the celebrity:

We can also look at the effect of dating a Kardashian in particular:

**Note the “Kardashian Effect” was actually statistically significant at the 10% level using the paired t-test as described below


We can also take a closer look at the Kardashian effect by specific athlete and relationship. (Note: Kendall Jenner is part of the Kardashian family)   

To examine the statistical significance of these changes, I z-scored the value statistics (WS/48, AV, WAR, PS) of each sport separately for every athlete, and then ran the same difference-in-difference analysis as before. Finally, I ran a paired t-test to compare the value of athletes before and after beginning to date a celebrity as well as from dating a celebrity to after breaking up.  The p-value for comparing dating a celebrity to post break up was not statistically significant at 0.383 despite each sport showing gains in athletic performance, likely due to the smaller sample size of players who break up during their playing careers (many broke up after retiring) and the huge variance in this smaller dataset. From the breakdown by celebrity type, the decline when dating an actress was statistically significant (p-value of 0.02).  Dating a Kardashian had a statistically significant(at the 10% level) negative effect on athletic performance. Finally, overall the p-value for comparing performance pre-dating and during dating was 0.00169, decisively concluding that beginning to date a celebrity negatively affects athletic performance.

If you have any questions for Matty about this article, please feel free to reach out to him at chengm@college.harvard.edu

About the author

harvardsports

View all posts

Leave a Reply

Your email address will not be published. Required fields are marked *