Is the ‘Swift Effect’ fact or folklore? A Quasi-Experimental Investigation into Taylor Swift's Influence on American football performance

James M Smoliga and Kathryn E Sawyer
Additional contact information
James M Smoliga: Tufts University School of Medicine

No ujws7, OSF Preprints from Center for Open Science

Abstract: Taylor Swift’s present at National Football League (NFL) games was reported to have a causal effect on the performance of Travis Kelce and the Kansas City Chiefs. However, this has not been validated through a robust statistical analysis. Therefore, the purpose of this study was to critically assess whether Taylor Swift's game attendance influenced: 1) Travis Kelce’s football performance, and 2) Kansas City Chiefs' game outcomes in the 2023 American football season. METHODS: A quasi-experimental study with propensity score matching was employed. Chiefs’ pregame Elo score was identified as a confounding factor which influence Kelce’s performance during the pre-Swift era (2014-2022 season). Each Chiefs game from the Swift era (2023 season) were matched to five games from the pre-Swift era by Chiefs’ pregame Elo. Linear mixed effects models were then used to determine how Swift’s presence or absence in Swift-era games influence Kelce’s performance, relative to historical data. Kelce’s yards was the dependent variable, random factor was each “matched set” of six games (5 pre-Swift era, 1 Swift era), and Chiefs pre-game Elo was entered as a covariate. Two models were developed: 1) Swift present (n=13 Swift era games, matched to 65 pre-Swift era games); and 2) Swift absent (n=6 Swift era games, 30 pre-Swift era game). Additionally, a binary logistic regression model was developed to determine if Swift’s presence influenced the Chief’s game outcomes, relative to historical averages. RESULTS: Kelce's performance was similar between the Swift (2023 season, 70.5 yards) and pre-Swift (2014-2022 seasons, 73.9 yards) eras. Linear mixed effects models revealed that Kelce achieved an extra mean 7.1 (95% confidence interval: -12.7, 26.9) yards per game when Swift was in attendance (n=13 games in the Swift era), compared to matched games from the pre-Swift era however this was not statistically significant (p=0.476). When Swift was absent from games (n=6 in the Swift era), Kelce’s performance changed by -28.6 (-69.4, 12.3) yard per game, compared to matched games from the pre-Swift era – again not statistically significant (p=0.163). Swift’s attendance did not significantly increase the Chief’s likelihood of winning [odds ratio = 1.32 (0.33, 5.34), p=0.692]. DISCUSSION: The weak statistical evidence that spawned the concept of the “Swift effect” is rooted in a constellation of fallacies common to scientific and medical research –including attribution bias, unjustified mechanisms, inadequate sampling, emphasis on surrogate outcomes, and inattention to comparative effectiveness. Clinicians and researchers must be vigilant to avoid falling victim to the “Swift effect,” since failure to scrutinize available evidence can lead to acceptance of unjustified theories and negatively impact clinical decision-making.

Date: 2024-08-08
New Economics Papers: this item is included in nep-cul and nep-spo
References: Add references at CitEc
Citations:

Downloads: (external link)
https://osf.io/download/66b502511dd0dda20b519a1c/

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:osf:osfxxx:ujws7

DOI: 10.31219/osf.io/ujws7

Access Statistics for this paper

More papers in OSF Preprints from Center for Open Science
Bibliographic data for series maintained by OSF ().