The Taylor Swift Effect
Do Travis Kelce and the Kansas City Chiefs play better when Taylor Swift is at the game?
An introductory lesson for data science and statistics.
Background
Taylor Swift is undeniably talented and a pop culture icon. In her music career, she has had 11 number one hit songs and sold over 114 million albums. Her Eras tour is selling out stadiums around the world and has already grossed more than $1,000,000,000 (USD) worldwide! Since she began dating Kansas City Chiefs star tight end Travis Kelce and attending games to cheer him on, economic analysis has estimated that she has added more than $330 million in brand value to the Chiefs and the NFL simply by showing up to games. Viewership of Chiefs games is up and sales of Travis Kelce jerseys increased by 400% almost overnight when Taylor started appearing at games and their love story became a storyline for the NFL. This gigantic economic boost from her presence at NFL games has been called “The Taylor Swift Effect” many times by the press.
At DataClassroom we wondered, does this Taylor Swift effect extend onto the field of play. Does her presence at a game influence the success that the Kansas City Chiefs have on the football field? Does Taylor’s presence have a measurable effect on the individual play of Travis Kelce? We know all too well how to approach questions like these, and that is with data.
Examine this dataset to test for The Taylor Swift Effect on the performance of Travis Kelce and the Kansas City Chiefs. Are you…ready for it? Ok, we’ll stop now ;)
Dataset
This dataset contains performance and outcome data from every Kansas City Chiefs game from the 2014 season through Super Bowl LVIII — 185 games, regular season and playoffs. Travis Kelce was drafted in 2013, but he spent almost all of his rookie year on injured reserve and took just one snap, so the data begin with 2014, his first full season.
Each game is flagged twice: whether Kelce was in a relationship with Taylor Swift at the time, and whether Swift was at that particular game. That lets you compare his 2023 performance both against the 2023 games she missed and against nine earlier seasons of his own career. We have also included data for the 2024 season if you want to extend the time occuring during their relationship.
We did not include the 2025 season, as Taylor intentionally tried to not allow her presence be known. The data of “was Taylor at the game” would not be accurate.
Five rows have blank receiving statistics. These are games Kelce did not play.
The data were compiled by Smoliga and Sawyer for their 2025 study of the "Swift effect" in PLOS ONE and are used here under a CC-BY license. The game statistics come from Pro-Football-Reference; Swift's game attendance was coded from news reports.
Here’s the article for further reading: Smoliga, J. M., & Sawyer, K. E. (2025). The folklore of the "Swift" effect. PLOS ONE, 20(9), e0315560
Variables
Season - This info variable references the year the fall season started. For example, the 2015 season ran from Sept 2015 through January 2016. Values include 2014-2024.
Date - This info variable describes the date of the game played. It is listed as month / day / year. It is not in a format to be included in any analysis
Game # - This numeric variable indicates which game was being played. Games are listed chronologically with the first game of the season as 1 and the most recent game (AFC Championship game) listed as 19.
OPP- This categorical variable records which team the Kansas City Chiefs were playing for that game.
location - This categorical variable indicates if the team was playing at Arrowhead Stadium (home) or at an away location.
KC Win or Loss? - This categorical variable records if the Kansas City Chiefs won or lost the game.
KC Points - This numeric variable is how many points were scored by the Kansas City Chiefs in that game.
Opp Points - This numeric variable is how many points were scored by the Chief’s opponent team in that game.
Was Kelce in a relationship with Taylor? - this categorical variable indicates if the game occurred before or after they were in a public relationship. Values include Yes or No.
Was Taylor Swift at the game? - This categorical variable records if Taylor Swift was in attendance at the game or not. It can have the value of Yes or No.
Kelce Receptions - This numeric variable describes how many passes Travis Kelce caught in the game. (A reception is when an offensive player catches a forward pass from a teammate who throws the ball from behind the line of scrimmage.)
Kelce Receiving Targets - This numeric variable indicates how many times the ball was thrown to Travis Kelce during a game.
Kelce Receiving Yards - This numeric variable measures the total number of yards gained on all Travis Kelce receptions during a single game.
Kelce Yards per Reception - This numerical variable describes the average number of yards per reception. Calculated by taking total receiving yards and divided by number of receptions.
Kelce Touchdown - This categorical variable indicates if Travis Kelce scored a touchdown during the game or not.
Number of Kelce Touchdowns - This numeric variable describes how many touchdowns made by Travis Kelce. (A receiving touchdown is scored when a player catches the ball in the field of play and advances it into the end zone, or catches it while already being within the boundaries of the end zone.)
Activity
Main questions:
You will use this dataset to explore two main questions in this exercise.
Q1: Do the Chiefs play better when Taylor Swift is at the game?
Q2: Does Travis Kelce play better when Taylor is at the game?
Get to know the dataset
1) Which variables in the dataset seem most relevant to Q1: Do the Chiefs play better when Taylor Swift is at the game?
2) Which variables in the dataset seem most relevant to Q2: Does Travis Kelce play better when Taylor is at the game?
3) Which variable is likely to be used as your independent (predictor) variable as you do analysis and make graphs to use as evidence for your answers to the main questions above?
4) How many observations (rows) are in this dataset and why?
5) What explains the missing data in row 1 of the dataset? Feel free to use Google for help if you aren’t a Kansas City Chiefs fan and don’t know the answer already.
Build graphs to visually examine the evidence
Investigating Q1: Do the Chiefs play better when Taylor Swift is at the game?
6) Create three different graphs, each with a different response (dependent) variable that can be used as evidence for answering the question of, Do the Chiefs play better when Taylor Swift is at the game? For each graph the variable called Was Taylor Swift at the game? Should be on the X-axis. Paste your graphs in the blank space below.
7) Using your three graphs in the question above as evidence what conclusion would you come to?
Investigating Q2: Does Travis Kelce play better when Taylor Swift is at the game?
8) Create three graphs, each with a different response (dependent) variable that can be used as evidence for answering the question of Does Travis Kelce play better when Taylor is at the game?
9) Using your three graphs in the question above as evidence what conclusion would you come to?
Data Classroom Note: Got extra time, or a fast finisher?
We’ve included all data for Travis Kelsie from 2014 - 2022 (before his relationship with Taylor began). If students feel like the dataset is too small and wish to view his statistics from years before, they can simply undo the exclusion of previous years by clicking the setting button under years, and deselect each year. You can read more about that here in our user guide.
Statistics Extension
Evaluate the strength of evidence
10) Make a graph with Was Taylor Swift at the game? on the X-axis and Chiefs Win or Loss? on the Y-axis. Run a Graph Driven Statistical test to run the Chi-Square test of association. Display your graph and statistical test results below.
11) What conclusion would you draw from the statistical test you ran above? How confident are you in that conclusion?
12) Make a graph with Was Taylor Swift at the game? on the X-axis and Kelce Long Reception on the Y-axis. Run a Graph Driven Statistical test to run the t- test (or Mann-Whitney U test*) . Display your graph and statistical test results below.
14) What conclusion would you draw from the statistical test you ran above? How confident are you in that conclusion?
15) No one would say that this dataset was collected through a controlled scientific study. However, if you were a fan and wanted the Kansas City Chiefs to win the Super Bowl would you want Taylor Swift to be at the game? Use evidence from the dataset to defend your answer.