Does Power Cause Walks? Test 1: Bonds 2002

In a previous post I proposed that Bonds’s increased walks must be independent of power. Since that time I have come to believe that power might affect walks by causing pitchers to avoid pitching to Bonds. So, I decided to test it.

Thanks to a thoughtful reader I was able to acquire game-by-game stats for the 2002 season on ESPN.com. Unfortunately, I cannot find the game logs for 2001 anywhere online. I have tried hard, too.

To see if power precedes walks I examined the following model:

Walk-Rate = Constant + B (Power in the past 10 games) + V(Controls) + e

I used three measures of power in the previous 10 games: Slugging, Iso-Power, and HR.
The Controls were score differential, total score, and a dummy if the Giants won.

The results for coefficient estimates power are as follows:

Var.	Coef	T-stat	Elasticity	
SLG	0.099	0.9	0.28	
Iso-P	0.128	0.93	0.21	
HR	0.07	0.6	0.11

Although all of the estimates are positive, none are statistically significant. I corrected for heteroskedasticity and tested for autocorrelation (none found and correcting does not affect the results). If you don’t know how to interpret the elasticity estimates, a 1% change in SLG is associated with a 0.28% change in the walk-rate, at the average. As best I can tell, there is no evidence that pitchers were shying away from Bonds when he was hot; however, Bonds was pretty hot over the entire season. I would prefer to see the estimates for 2001.

Other notes: The R-squared is about .04 for all estimates. The sample included 149 games. I tried several possible controls, these were the best and the power estimates were never statistically significant. I also tried some non-OLS estimates such as multinomial logit and poisson, but found the results were similar to OLS. I am not as familiar with these techniques so it is possible I fouled them up.

UPDATE: I just got the 2001 numbers. I have run them, and there is a difference from 2002. SLG does seem to lead the walk-rate by a larger amount, and it is statistically significant. Iso-Power is also larger and borderline significant. HRs do not seem to matter. Unfortunately, I will not be able to post until tomorrow. Thanks to Doug and Alec for the data.

Comments are closed