Monday, December 31, 2012
Why an increase in payroll may not lead to sustained success
Posted by Glenn DuPaulLast week, I discussed the issue of teams "going for broke" and how significant increases in payroll may affect overall success (win totals). One of the more interesting findings in that piece dealt with teams who had great success the season before they increased payroll.
I found that from 2001 to 2012 there were 28 teams who won at least 88 games in the year before they increased payroll by at least 20 percent. Of those 28 teams, only 14 teams were able to get back to that 88-win threshold in the subsequent season.
This result sounds surprising when taken at face value. In most cases, the assumption would be that the increase in payroll was an attempt to prevent regression and sustain, or build on, past success.
Based on the results I found in that piece, I concluded that the increase in payroll had no real effect on a team's ability to sustain success.
Over at the Book Blog, Tom Tango pointed out a major flaw in that conclusion, namely the lack of a control group:
Specifically, in one test, he finds a bunch of playoff teams that adds alot of money, and he still finds that a good portions wins fewer games. But, even if they did not add alot of money, they would STILL win fewer games. That’s regression toward the mean at work. If you take a group of teams that win 88+ games, you will find, on average, that they will win less in the next season. This is because teams that win 88+ games are teams that have both more good players than bad players and more good luck than bad luck. The next season, the luck will cancel out, so, all other things equal, are expected to win fewer games.Tango's response brought me not to a different question, but instead a much better way of answering my original question. Instead of comparing successful teams that then increased payroll to themselves, I included every team that won 88 games from 2001-12 in the sample for a more illuminating comparison.
On average since 2001, the 113 teams that won at least 88 games in one season won 6.6 fewer games in the next season. This result was expected because of regression toward the mean that Tango discussed.
Despite this regression, 63 of the 113 teams (55.7 percent) were able to reach the 88-win threshold again in the subsequent season.
Does an increase in payroll have any effect on reducing the regression toward the mean?
As suggested by Tango, I broke the sample into three subsets:
- Teams who increased payroll considerably (15 percent or higher increase)
- Teams who increased payroll somewhat (2 to 15 percent increase)
- Teams who increased by a negligible or negative amount (Less than two percent increase)
| Payroll Change | Num. of Teams | Avg. Change in Wins | Percent >=88 |
|---|---|---|---|
| > 15% | 39 | -7.02 | 56.4% |
| 2-15% | 38 | -7.26 | 52.6% |
| <2 % | 36 | -5.42 | 58.3% |
As you can see, the number of teams in each subset was pretty close to evenly distributed. Quite surprising, though, is the subset of teams that were the most successful in preventing regression toward the mean in the subsequent season.
The control group of teams that did not increase payroll or actually decreased payroll, had the smallest average drop-off in wins from the previous season.
The last column in the table shows the percent of teams in each subset that was able to get back to 88 wins after reaching that plateau in the previous season. The control group also lead in this category with the highest percentage of teams, 58.3 percent, that were able to (sort of) maintain their success.
The results for teams that substantially increased their payroll (by 15 percent or more), surprisingly, were below the average for the entire sample.
Interestingly, improving the structure of the test resulted in the exact same conclusion as the one from the original study. Based on the results found in this study, it seems that increasing payroll has no real effect on a team's ability to sustain success and dodge the regression-toward-the-mean bullet in the subsequent season.
All payroll information comes courtesy of Baseball Prospectus' Compensation Tables.
Glenn is an Economics major at Lehigh University. He works as a Research & Development intern for Baseball Info Solutions. He also writes about sabermetrics for Beyond the Box Score. You can follow him on twitter @Glenn_DuPaul or email him at .(JavaScript must be enabled to view this email address)
Is there any way to consider WHY payroll was increased? For example, if it was due to arbitration or re-signing core players (something the Nats are faced with) or going out to sign “expensive” free agents because the minor-league system is not provided adequate players to promote or trade?
I realize this further splinters your sample sizes, but it might provide insight.