Well , I’m happy to finally introduce the Windjammer Partisan Voting Index for US State Houses. Like the Windjammer Partisan Voting Index for US State Senates, a D+1 district means that if the presidential democrat candidate gets 50%, in this house district, he would get 51%.
For those who like mathematics:
PVI in the House District n: [(Obama2012 score- Romney2012 score)/2 -1.95 +(Obama2008score-Mccain2008score)/2-3.6]/2
Here we go!
Great work, but the math is a little off. You didn't divide to get the two party vote, which makes the gap smaller than it would be with the correct weighting. Using differences, the formula is
PVI(D+) = 50%*[(D2012-R2012)/2*(D2012+R2012) - 0.0197 + (D2008-R2008)/2*(D2008+R2008) - 0.0369]
Thanks for pointing out.
Well, I understand the formula. I guess the problem with my formula is that it doesn't take into account third parties? Does that change a lot of things in the end considering third parties in the US<3%?
It would typically change the results by 0.1 or 0.2. BTW I corrected my formula, since I forgot the extra factor of two due to the fact you used the difference between the two major candidates instead of the share of the two-party vote.
However, I was looking at your IL results, and some districts seemed a bit off so I dug into the data for one district to check. I have all the PVIs for the new 2012 districts calculated with the 04-08 numbers, which were derived when the new maps were announced. Yours were a bit different, so I thought I should crosscheck one district with 08-12 numbers to confirm the discrepancy.
You list IL HD49 as PVI -1, and my 2004-2008 PVI of that district with the new boundaries was PVI -3.1. The 2012 result in that district was Obama 22.0K - Romney 23.5K or a difference of -3.3% which if I divide by 2 and subtract 1.97% gives a 2012 factor of D-3.6. Using the 2012 precincts for the new district (which has almost nothing in common with the old HD49) the 2008 vote in HD49 was Obama 23.7K - McCain 20.0K or a difference of +8.5% which gives a 2008 factor of D+0.5. When I average these two I got a net PVI(D)-1.6. It's close to yours, but it would round off to D-2.
Is it just that you have a number of HDs that are slightly more Dem than mine and rounding shifts them consistently up a point or so? If you could share your details for that HD I can confirm my suspicion.