2010 State PVI County Maps and Bellwethers for States With Senate Elections

[cinyc]

I would recommend you rerun Oklahoma. Your map is heavily biased by Obama's performance in the state, a performance was quite markedly different from almost any other Democratic performance in the state's recent history. For example, Pushmataha County, which you have as a bellweather, is typically one of the more Democratic counties in the state in local elections. It is only because of Obama's presence that the state, and a number of counties, swung wildly to the Republicans in 2008.

Rerun it with what?  The math is correct.  The OK PVI numbers are what they are.  I haven't adjusted ANY of the maps because of a non-uniform Obama swing.

I meant with gubernatorial numbers like SC and VT. The 2008 numbers are rather out of line with Oklahoma's usual voting record (I suppose it is in a number of southern and Appalachian states). I would think that if you are basing your projections on a model that is comprised with half (what I would consider) abnormal results, that it would not serve as useful when it comes to comparing to results we will get back on Tuesday. Not that it will probably matter, I suppose.

I only pointed out Oklahoma because it presented an especially egregious case of this: the 2008 OK Democratic strength map looks literally like none other in Oklahoma's history.

I don't know.  We'll know after Tuesday, but as long as Obama's the President, I wouldn't be surprised to see similar voting patterns in the Appalachian/Southern Senate races, particularly open ones.  To the extent these elections will be a referendum on Obama, the same areas of those regions that didn't vote for him probably don't like him much and may vote against the Democrat despite their more distant past.  If I remember correctly, MA PVI was a relatively good, but not perfect, predictor in the Massachusetts special election.

I wouldn't even attempt to translate federal PVI to gubernatorial races, though, particularly in the South.

By the way, one problem with using other statewide races to try to compute anything is that there is a home county or region effect that potentially skews results.  If last cycle's opponent was from Stuebenville and this year's opponent is from Lima, we'd have to adjust for it anyway.  I could use partisan voter registration in the states that have it, but that probably masks some trends, too.

The Oklahoma State Election Board doesn't have the 2002 OK gubernatorial results on their website.  So I can't compute the OK GPVI without getting it from another source.

[RI]

I would recommend you rerun Oklahoma. Your map is heavily biased by Obama's performance in the state, a performance was quite markedly different from almost any other Democratic performance in the state's recent history. For example, Pushmataha County, which you have as a bellweather, is typically one of the more Democratic counties in the state in local elections. It is only because of Obama's presence that the state, and a number of counties, swung wildly to the Republicans in 2008.

Rerun it with what?  The math is correct.  The OK PVI numbers are what they are.  I haven't adjusted ANY of the maps because of a non-uniform Obama swing.

I meant with gubernatorial numbers like SC and VT. The 2008 numbers are rather out of line with Oklahoma's usual voting record (I suppose it is in a number of southern and Appalachian states). I would think that if you are basing your projections on a model that is comprised with half (what I would consider) abnormal results, that it would not serve as useful when it comes to comparing to results we will get back on Tuesday. Not that it will probably matter, I suppose.

I only pointed out Oklahoma because it presented an especially egregious case of this: the 2008 OK Democratic strength map looks literally like none other in Oklahoma's history.

I don't know.  We'll know after Tuesday, but as long as Obama's the President, I wouldn't be surprised to see similar voting patterns in the Appalachian/Southern Senate races, particularly open ones.  To the extent these elections will be a referendum on Obama, the same areas of those regions that didn't vote for him probably don't like him much and may vote against the Democrat despite their more distant past.  If I remember correctly, MA PVI was a relatively good, but not perfect, predictor in the Massachusetts special election.

I wouldn't even attempt to translate federal PVI to gubernatorial races, though, particularly in the South.

By the way, one problem with using other statewide races to try to compute anything is that there is a home county or region effect that potentially skews results.  If last cycle's opponent was from Stuebenville and this year's opponent is from Lima, we'd have to adjust for it anyway.  I could use partisan voter registration in the states that have it, but that probably masks some trends, too.

The Oklahoma State Election Board doesn't have the 2002 OK gubernatorial results on their website.  So I can't compute the OK GPVI without getting it from another source.

It might still skew that way in the Senate race, maybe. The governor's race, which should be the closer of the races, probably won't. I could be wrong though.

If you want more information, the 2002 Governor's results are here, and the latest voter registration is here.

[cinyc]

It might still skew that way in the Senate race, maybe. The governor's race, which should be the closer of the races, probably won't. I could be wrong though.

If you want more information, the 2002 Governor's results are here, and the latest voter registration is here.

Agreed.

Thanks for the link to the 2002 results.  I added an OK GPVI map based on the 2002 and 2006 gubernatorial results.  Payne County appears to be a reliable, not-too-swingy PVI bellwether in federal and state races.

I actually compiled the most recent registration data for the 36 states when putting together the original maps.  Many of the states don't have party registration at all, though.