By Daniel Adler
We’ve hit the big-time…an article in the Wall Street Journal (the real deal–print edition!). The article looks at whether increasing payroll leads to more wins. The short answer: no.
The following graph uses a 9.37% rate of inflation so we can compare change in spending across the past 19 seasons. Wins for the strike shortened 1994 and 1995 are adjusted to full season levels. The r2 level is very low and the p-value is insignificant. Payroll data is courtesy of the USA Today salary database.
More analysis after the jump…
Looking at last season, the ten teams with the largest increases in opening day payrolls fared a total of 77 games worse than in 2008. Of the ten largest payroll gainers, only the Marlins and Phillies improved and it was by a total of four games. The Yankees did win 14 games more after committing $423.5 million to CC Sabathia, Mark Teixeira, and A.J. Burnett last off-season, but the team’s opening day payroll actually dropped since the high priced contracts of Bobby Abreu, Jason Giambi, Carl Pavano, and Mike Mussina came off the books.
It is still true that a larger payroll does predict a higher winning percentage (more on this to come), but since the teams that increase payroll are often coming off a stronger record, it may appear that spending more is not actually that helpful. In some cases, teams that increase payroll (like the 2009 Rays) may not actually improve their team, but rather “tread water” and pay a lot of money to retain their players.
A future post will look at whether spending more is helpful when you control for previous wins (i.e. if two teams both have 85 wins and team A increases payroll by X and team B decreases payroll by Y, how does this impact their chances?). My guess is that when controlling for previous wins, spending more will actually be a predictor of winning more.
Special thanks to David Biderman at the Journal (or Michael Scott likes to refer to it, “The Wall”) for working with HSAC on this article.
I think you have to be very careful when looking at one-year shifts in payroll and then drawing conclusions about whether increasing payroll produces more wins.
One problem is back-loaded contracts. Many players sign multi-year contracts that pay them much more at the end than the beginning. This may be for reasons of taxes, team payroll, whatever. But it means that the final year of Jason Giambi’s contract, for example, in no way reflects the value the team thought he would have that year — the Yankees were still paying him in part for his earlier, more productive seasons. Taking it off the books has a huge impact on payroll, but smaller effect on team talent. So you MUST work with the annualized value of these contracts, to get more realistic sense of what a player was really being paid in that year.
You also have the problem of younger pre-arb players becoming arbitration-eligible. Their salary skyrockets at that point, increasing payroll, but the player’s improvement in talent is small.
And you have to consider WHY a team would suddenly change it’s salary budget. Perhaps the team won a lot of games the prior year, creating a revenue surge, so they increased payroll. Well, that team is likely to win fewer games next year just due to regression to the mean.
On balance, I doubt you will find one-year changes in payroll to be a helpful way to study the payroll-performance link. This is an area where you want big, multi-year samples to see the actual underlying relationship.
FYI: For good discussion of why annual change in payroll doesn’t tell us much, see this post by Tango:
http://www.insidethebook.com/ee/index.php/site/comments/wins_and_payroll/.