NFL Football, Free Agency, the Salary Cap, and Parity
Here at StatsExamples our favorite sport is NFL football. your preferences may vary, but if you're still reading then let's assume you're interested in NFL football.
What is free agency and what were people afraid of?
The NFL has been around since the 1920s and there may been a ton of changes since then. Here are just a few examples. (1) When football started forward passes were illegal, the NFL introduced this concept. (2) In the beginning the league champion was decided by which team had the best record, it wasn't until the 1940s that playoffs and championship games were introduced. (3) in 1962 the league finally made grabbing a facemask into a penalty. (4) Seasons had between 10 and 12 games until 1961 when it increased to 14 games followed by an increase to 16 in 1978 and in 2021 it increased to 17 games.
For this page we want to look at a major change that happened in 1992, free agency. The ability of players to move between teams was long restricted and limited - allowing teams that got good players to keep them for their entire careers. Throughout the 1970s, 1980s, and 1990s various lawsuits gave players a bit more freedom to switch teams. Finally, in 1992 there was a lawsuit called "White vs NFL" which was resolved by the creation of true free agency beginning in the 1993 season.
At the time, many people claimed that this would lead to very bad outcomes. The lawyer for the NFL stated that the new policies "would be the destruction of the National Football League that we know today." This new policy, along with a salary cap that meant every team spent the same amount of salaries, led many pundits to claim that this would lead to parity in the NFL. All the teams woudl become equal because good players would follow higher salaries and all the teams would end up with the same number of the best players. They saw a dystopian NFL in which all the teams would be equal and the game wouldn't be fun to watch anymore.
So, did this happen?
Did parity increase after the creation of free agency and the salary cap?
To examine what happened, we here at StatsExamples decide to look at the data. We collected all the season records from 1970 up to 2020, 51 years in total. These years are roughly balanced, with about half before and half after the 1992/1993 introduction of free agency.
What's the best way to measure parity? We could look at the best record in each season, but that's based on just the best team and prone to lots of chance effects. What we decided instead was to look at the variation in the records of the teams. If the naysayers were correct, then we should see that the variation in season records decrease after free agency and the salary cap as all the teams become more equal. The true nightmare scenario would be a league full of 7-7 or 8-8 teams and the variation would be zero. On the other hand, if the variation stays the same then their fears were unjustified.
There are several ways to measure variation in a data set, our page on summary statistics mentions a few, but here we'll focus on three:
The STANDARD DEVIATION of the records will give us a value that, if doubled, is about as wide as the region that contains the middle 2/3 of the data. This value is commonly used to measure variation because it is related to the variance which is used in many statistical tests. It is prone to being inflated by the presence of single teams that are much better or worse than the others.
The INTERQUARTILE RANGE of the records gives us the width of the middle 50% of the values. This value is the most resistant to outliers (i.e., super good or bad teams) and will give us a good sense of the variation of the intermediate teams.
The RANGE gives us the difference between the best and worst records. This value will be more variable and prone to noise because it depends on the presence or absence or super lucky and unlucky teams in each season.
For all three of these metrics, increased parity would be reflected by reductions in the value of the metric. If free agency and the salary cap caused parity, then we would expect these three values to be lower after their start in 1992/93.
Plugging in all the values and looking at the STANDARD DEVIATION values we get a figure that looks like this:
The first thing to notice is that it looks like the league was headed towards parity before the introduction of free agency, the downward slope of the standard deviation values suggests this. The slope is not statistically significant however (p=0.157), so may well be just a random pattern. The slope after the change is upwards and statistically significant (p=0.007) however, which indicates that parity has been decreasing ever since the introduction of free agency.
We can see from this figure that while the variation may have initially decreased, the overall trend since the 1993 introduction of free agency has been an increase in the variation, the opposite of what the naysayers were afraid of.
Plugging in all the values and looking at the INTERQUARTILE RANGE values we get a figure that looks like this:
This figure shows the same pattern of increasing parity changing into decreasing parity. The slopes are not statistically significant (p=0.198 and 0.084), but the pattern is consistent with free agency reversing a trend towards parity. This pattern certainly is strong evidence against the claim that free agency caused more parity.
Plugging in all the values and looking at the RANGE values we get a figure that looks like this:
This figure also shows the same pattern of increasing parity changing into decreasing parity, albeit with nonsignificant slopes (p=0.799 and 0.102). Again, the pattern is consistent with free agency reversing a trend towards parity and while these results aren't significant, they are strong evidence against the claim that free agency caused more parity.
A possible objection to this analysis is note that it looks like there is a drop in the variation after the 1992/93 transition. If we ignore the yearly pattern over time, we could compare the overall average variation before and after the 1992/93 division.
It's true that the average values are lower after the transition (SD: 0.197 vs 0.189, IQR: 0.297 vs 0.265, Range: 0.754 vs 0.734), when we compare these values using a heteroscedastic t test, the differences are nonsignificant (p-values of 0.151, 0.081, and 0.425 respectively). At best, this could be used to refute a claim the transition increased the variation in season records.
Free agency and the salary cap seem to have reduced parity
Taken together, the data shown above makes a very strong argument against a claim that free agency and the salary cap have increased parity in season records.
While the data is weak that free agency and the salary cap specifically increases parity, the pattern is very good evidence against a claim that free agency and the salary cap increased parity - the observed pattern is the exact opposite of what we would expect.
While there may have been an initial increase in parity, the long term trend and overall before and after differences don't show significant evidence of an increase in the parity of NFL teams as measured by their season records.
The fears of the people worried about free agency and the salary cap don't seem to have come true.
tl:dr, "Data undermines the claim that free agency and the salary cap increased parity in the NFL and may even suggest that it is having the opposite effect in the long term."
