Written by by Brion Hurley
In June 2020, former University of Iowa football players raised concerns about the mistreatment of Black players in the program. As a former player at Iowa, this news impacted me personally, and I’ve been following this story since it first came out. Just to be clear, I am white, and was born and raised in Iowa City my whole life.
To tackle a complicated problem like this, Six Sigma can be useful to provide a data-driven approach to problems, to better clarify, understand and ultimately solve the issue.
One of the complaints brought up by former players was the amount of Black student-athletes that had transferred or left the program as a result of the complaints and mistreatment.
On June 30, 2020, HawkeyeNation.com writer Rob Howe collected and shared the data of player retention from 2009 through 2019, broken down into two categories: “Black” vs “White/Non-Black.”
The charts below show the data by year comparing the two groups, color-coded into three colors: Signed (blue), Finished Career (red) and Still in Program (Yellow).
These charts are very helpful for visualizing a problem, and it appears that the red bars are much lower than the relative blue bars for Black student-athletes, which suggests that Black student-athletes have more difficulty finishing the program than White/Non-Black student-athletes.
However, in Six Sigma, we want to balance what we see visually in a chart with statistical analysis and discipline, so we don’t incorrectly interpret or misread the charts. We also want to ensure that the data used to collect this information is valid and correct, called Measurement System Analysis (MSA). In other words, if other people replicated this data collection exercise, would they have come up with the same data results? For now, I’ll assume that the categorization of the student-athletes into Black and White/Non-Black was done correctly, as well as whether they were marked correctly as “Finished Career.”
One statistical test we can use to compare the two groups over this time period is a hypothesis test called Chi-square Test of Independence. It helps determine whether there is a statistically significant relationship between categorical variables (Black vs Non-Black). For this data, we can compare the two groups to see if the difference in the percentage of student-athletes that finish their career is statistically valid.
For example, if 70% of Blacks complete the program, and 73% of Non-Blacks complete the program, we could easily conclude that the difference seems very minor, and would probably not be statistically valid.
If the data shows that 25% of Blacks complete the program and 95% of Non-Blacks complete the program, then it would be obvious to conclude that the difference is statistically valid (not just due to random variation in the data).
I took the data from the charts, and put it into a spreadsheet.
I decided to keep the data as clean as possible, so instead of including data from 2016-2019 with student-athletes marked “Still in Program,” I chose to analyze the data from 2009-2015 only. Here are the results over that time period.
Overall, about 50% of the student-athletes finish the program, regardless of race. However, 58% of White/Non-Black players finished, compared to 42% of Black players. That seems like a statistically significant difference, but let’s verify that difference with confidence intervals and the Chi-Square test.
Confidence intervals estimate the likely difference in percentage between both groups, at the 95% level (5% chance that it could fall outside of that interval range). The interval is the likely true difference between groups. The true difference is somewhere between 0% and 32%, and it is most likely to be around 16%. This supports the idea that it’s likely that there is a difference between Black and White/Non-Black. But there is a wide spread in the interval due to the small sample of players (about 150 players total).
For the chi-square test, the way the test works (in layman’s terms) is the assumption that there is no difference in the groups (both finish the program about 50% of the time). However, if the actual data from each group differs too far away from 50%, then the probability that it is just coincidence goes down, and the likelihood that it’s due to the group designation (race) goes up. The probability that the difference is just coincidence (random variation in the data) is called the P-value.
If you haven’t taken any Six Sigma classes, you might be confused by the chart above, so let me explain the results. Let’s focus on the very bottom number, the P-value.
When the P-value is less than 5% or 0.05 (which is true in this case, P-value = 0.042), then we can conclude that there is a statistical difference between Blacks and White/Non-Blacks. Black players have a statistically lower chance of finishing the program than Non-Blacks. There IS a real problem in the Iowa football program.
However, that’s all we can conclude from this data. We can’t explain WHY there is a difference between the two groups. Further investigation and analysis will need to be done to determine what common factors are present with players that did not finish, such as past experiences, specific incidences that occurred in the program, success on the field, success in the classroom, and many other contributors.
If you’d like to learn more about the statistics, check out this article to learn more about Chi-Square tests.