Tuesday, May 31, 2011

The Evils of the Cook PVI

In 1997, Charlie Cook came up with a method of analyzing congressional districts voting patterns that'd enable people to compare districts across the country, the Cook PVI. The people rejoiced. The number used only Presidential data, thus taking out local candidates who might have partisan followings that'd skew a district in a way that wouldn't continue after he or she left office.

This was a good way to compare congressional districts across the country. Like the RBI in baseball the PVI tells a part of the story but not always the best story. So when Republicans in Illinois win districts with Democratic PVIs but Republicans in Tennessee don't, people assume it's because Republican congressmen in Illinois overachieve.

The problem is that the PVI's perceived strength, not letting one local politician skew the number, is also a great weakness. Chances are that Republicans in Illinois aren't the outlier. It's Barack Obama and John McCain in Illinois. McCain underachieved what other Republican statewide candidates usually get. The Republicans in Illinois aren't overachieving at all. Obama-McCain numbers don't accurately reflect local voting patterns.

A better way to judge how a party will do statewide is to look at how candidates for statewide office do. A better way to judge how a party will do in a congressional district is to look at how the party performs in that congressional district. Sure, one good candidate or a bad one can skew the numbers but chances are that this candidate is the one running again next year. Barack Obama won't be running for a congressional seat any time soon. Yet people cling to his performance in a district as being indicative.

Cook PVI isn't a bad stat. A Democrat is more likely to win a D+6 seat than an R+5. When looking at a district or a state you need to look at how Obama and McCain did compared to statewide candidates and how the congressional candidates do. As noted below, every incumbent Republican beat Barack Obama by 9.6 points or more in 2008. They can't all be good candidates running against bad ones. Chances are the number you're comparing their performance to, Obama-McCain 2008, is inaccurate.

No comments:

Post a Comment