Cook Releases 2008 PVIs, With a Change SSPers Will Like

Our friends at the Cook Political Report have released an updated Partisan Voting Index that now includes the results of the 2008 presidential election. They’ve also made a small change in the formula used to calculate PVIs, and I think it’ll be instantly recognizable to SSPers:

To determine the national average for these latest ratings, we have taken the average Democratic share of the two-party presidential vote for 2004 and 2008, which is roughly 51.3 percent, and that of Republicans, which is roughly 48.7 percent. So, if John Kerry captured 55 percent of the vote in a district and Barack Obama carried 57 percent in the district four years later, the district would have a PVI score of roughly D+5. (Emphasis added.)

As we discussed at length, the old PVI formula compared district-level results for the past two presidential elections to nationwide results for only the most recent election. This choice sparked plenty of debate, and some folks even suggested we use our own “SVI” that would compare 2004 to 2004 and 2008 to 2008.

Fortunately, the debate has been resolved. As you can discern from the description above (the key part is in bold), Cook has decided to revise its methodology along the lines proposed by people here. Charlie Cook (an SSP reader, as is House editor David Wasserman) told me he wanted something that was “totally apples and apples,” and I agree with the choice. Ultimately, this means that the new PVIs will be about two points bluer than under the old system – e.g., a district that would have been R+10 will now come in at R+8.

You can find the new PVIs by partisan rank in this PDF, as well as by member name and by state/district. There’s also a giant-size map and a cool chart showing trends in the PVI over the last decade. (As you’d expect, the number of “competitive” districts, at least on the presidential level, has been shrinking.) Have fun!

30 thoughts on “Cook Releases 2008 PVIs, With a Change SSPers Will Like”

  1. two-cycle meaning that we use numbers from the two most recent presidential elections.

  2. in order from most Democratic to most Republican.

    Obama-Republican districts are bolded.

    CA-09 (Lee-D): D+37

    CA-08 (Pelosi-D): D+35

    CA-33 (Watson-D): D+35

    CA-35 (Waters-D): D+31

    CA-31 (Becerra-D): D+29

    CA-37 (Richardson-D): D+27

    CA-28 (Berman-D): D+23

    CA-06 (Woolsey-D): D+23

    CA-12 (Speier-D): D+23

    CA-13 (Stark-D): D+22

    CA-34 (Roybal-Allard-D): D+22

    CA-14 (Eshoo-D): D+21

    CA-07 (George Miller-D): D+19

    CA-17 (Farr-D): D+19

    CA-38 (Napolitano-D): D+18

    CA-30 (Waxman-D): D+18

    CA-16 (Lofgren-D): D+16

    CA-15 (Honda-D): D+15

    CA-05 (Matsui-D): D+15

    CA-32 (Open [Solis] -D): D+15

    CA-53 (Davis-D): D+14

    CA-29 (Schiff-D): D+14

    CA-01 (Thompson-D): D+13

    CA-43 (Baca-D): D+13

    CA-27 (Sherman-D): D+13

    CA-39 (Linda Sanchez-D): D+12

    CA-23 (Capps-D): D+12

    CA-36 (Harman-D): D+12

    CA-10 (Tauscher-D): D+11

    CA-51 (Filner-D): D+8

    CA-20 (Costa-D): D+5

    CA-47 (Loretta Sanchez-D): D+4

    CA-18 (Cardoza-D): D+4

    CA-11 (McNerney-D): R+1

    CA-50 (Bilbray-R): R+3

    CA-26 (Dreier-R): R+3

    CA-45 (Bono Mack-R): R+3

    CA-24 (Gallegly-R): R+4

    CA-03 (Lungren-R): R+6

    CA-46 (Rohrabacher-R): R+6

    CA-48 (Campbell-R): R+6

    CA-25 (McKeon-R): R+6

    CA-44 (Calvert-R): R+6

    CA-40 (Royce-R): R+8

    CA-19 (Radanovich-R): R+9

    CA-52 (Hunter Jr.-R): R+9

    CA-42 (Gary Miller-R): R+10

    CA-49 (Issa-R): R+10

    CA-04 (McClintock-R): R+10

    CA-41 (Lewis-R): R+10

    CA-02 (Herger-R): R+11

    CA-21 (Nunes-R): R+14

    CA-22 (McCarthy-R): R+16

  3. tells you what you need to know about how this map tilts. Democrats have some work to draw in a couple of years.  

  4. Both the 2004 and 2008 elections results are the factors to the PVI results, do you think that the race issue during the 2008 presidential race threw off the numbers? Oklahoma, Tennesee, Arkansas and a couple other states actually took a significant turn to the right during the 2008 cycle despite the rest of the country going in the other direction, (presumably due to President Obama’s race), are those PVI’s accurate or should we be more skeptical?

  5. Cook’s cool chart showing the breakdown of districts by PVI really shows how Democrats got screwed in the 2000 redistricting.  In 2000, there were 142 heavily Republican leaning districts (PVI >5) and 134 heavily Democratic leaning districts, only 8 more R than D safe districts.  In 2010, there are 182 heavily Republican leaning districts compared to 149 Democratic ones; 33 more R than D safe districts.  That is a direct result of us losing the 2000 (and post-2000 in the case of TX) redistricting wars.  We can and must do better in 2010!

  6. I’m surprised to read that the methodological change would tend to produce PVIs that are about 2 points bluer.

    The new PVIs for Iowa are:

    IA-01 D+5 (was D+4)

    IA-02 D+7 unchanged

    IA-03 D+1 unchanged

    IA-04 Even (was D+0)

    IA-05 R+9 (was R+8)

    Iowa’s presidential results have been pretty close to the national presidential results in 2000, 2004 and 2008.

    I wonder why the Iowa PVIs mostly did not get bluer. Does anyone on this board have a hypothesis for me?

  7. For the record, DavidNYC should get the credit for our decision to change the formula for calculating the Partisan Voting Index.  

    We started the PVI 12 years ago and frankly, when DavidNYC emailed me asking why we compared the two previous elections nationally with just the most recent presidential result in each CD, I could not remember why we decided to do it that way to begin with.  

    Our House Editor, David Wasserman and I discussed it with Clark Benson from Polidata, the firm that compiles and computes the presidential vote data for us and together decided to make the switch.  As DavidNYC points out, it changes few districts, but is a better way to do it.

    Though the Cook Political Report has been around for 25 years (as of last month), we still try to view things through fresh eyes each election cycle and welcome constructive suggestions.

    Charlie Cook

    1. CA-46’s swing from Bush 2004 to Obama was a little less than the Obama-Republican districts had, as well as McNerney’s since his district is somewhat similar to the O-R districts partisanship-wise, and the swing was not enough to make it another O-R district.

      CA-44 swung from Bush by 19 to Obama by 1, a swing of 20 points to Obama.

      CA-25 also swung from Bush by 19 to Obama by 1, a swing of 20 points.

      CA-48 swung from Bush by 18 to Obama by 1, a swing of 19 points.

      CA-03 swung from Bush by 19 to Obama by 1, a swing of 20 points.

      CA-24 swung from Bush by 13 to Obama by 3, a swing of 16 points.

      CA-45 swung from Bush by 13 to Obama by 5, a swing of 18 points.

      CA-26 swung from Bush by 12 to Obama by 4, a swing of 16 points.

      CA-50 swung from Bush by 11 to Obama by 4, a swing of 15 points.

      CA-11 swung from Bush by 9 to Obama by 9, a swing of 18 points.

      CA-46 swing from Bush by 15 to McCain by 2, a shift of 13 points.

  8. Sure there are plenty of ways to compute some kind of partisan index for CD’s using other races, but they would only be useful for intra-state comparison.  Using different races in different states would not be comparable across state lines, which is what we wanted to do.

    Actually senatorial and gubernatorial races are often not the best races to use to compute partisan voting behavior, as they are affected by the personalities of the candidates and the campaigns.  Better are low visibility, down-ballot statewide races where no one really knows who most of the candidates are, they are just voting on party.  

  9. Yes, Swingstate basically posted the new Cook PVIs for all districts before Cook did!

    If you compare Cook’s numbers to the SPVI numbers posted here, you’ll see that they are mostly the same.  On about 1 in 10 districts, the rounded SPVI for a district would be +1 or – 1 compared to the numbers Cook posted.  I am certain this is because I used the candidates’ percentages in each district as compiled by Swingstate rather than their raw vote totals.  Comparing my numbers to Cook’s, this seems to result in a +/- error of up to 0.3 in the numbers posted on Swingstate, and thus the occasional discrepancy with Cook’s numbers.

    But, if you compare how similar the results are and Cook’s description of his change in formula, it’s clear that he’s using the exact same formula that was proposed by Swingstate.

  10. Seems to me that there is always going to be some skewing of the PVI numbers based on the “fit” between the presidential candidates and the individual states.  Compared to the average Democrat, Obama was not a good fit with OK, AR, and TN (and neither was Kerry).  But, I think he was a really good fit with places like IL, IN, and a lot of western states (more so than Kerry).  McCain did really well in AZ.  Bush overperformed average Republicans in Texas and the South.  Cook tries to minimize this effect by using the results of two presidential but that only goes so far.

  11. There will always be some “skewing.” The TN PVIs, for instance, dropped so precipitously not just because Obama fared poorly in most of the state but because Gore’s 2000 results have dropped out of the equation. Naderites loved to squawk the ridiculous “accusation” that Gore “couldn’t win his home state,” but as subsequent elections have shown, he did remarkably well there – definitely a high-water mark for a national Democrat.

  12. Bill Clinton did even better, carrying TN in 1996 with 48%.  Clinton won 6 southern states that year (KY, TN, WV, AR, LA, FL).  In 1992 he carried the first 5 of those states plus GA (no FL in 1992).  Republicans over invested in TN in 2000 as an ego thing and it nearly cost them the election as Gore pulled off a lot of close states (PA, MI, WI, IA, NM) and really won FL.  He came awfully close in NH.

    Republicans have a history of investing heavily in prestige victories.  Nixon made darn sure that Vietnam vet John Kerry lost his House race in 1972.  House Speaker Tom Foley was edged in 1994 ( a first).  Daschle went down.  If Republican operatives were running the 2008 campaign, Mitch McConnell would have been scalped.  At great cost.

  13. Cook shows how districts perform against the “national average.”  The national average for 2004/2008 is a 3 point Democratic victory.  Sure enough, Democrats hold 7 of 10 R+1 seats and 10 of 14 R+2 seats.  R +3 through R+5 are essentially even as a group (18 R, 17 D with R+3 being the best Republican set).

    In my own state of NJ, the Cook PVI improved for Republicans in all 13 seats because Obama ran even with Gore in NJ (a 15 point win) but 7 points ahead of him nationally.  Some of those districts did improve for Republicans (notably NJ-5)but some did not.

    Btw, with the exception of Reichert all of the Republican targets (even or less) are in the midwest or the Northeast.  Pat Tiberi was a big, big surprise.

  14. Going all out trying to defeat Walter Mondale following the assassination of Senator Wellstone in 2002.

  15. Given that R+3 to R+5 is about the break-even point for Democratic odds with PVI, can someone with some graphical ability show how these maps would look shifted about four points to the left? In other words, rather than comparing every district to the national average (when the average is +3D), show every district in its D vs R affiliation? Since right now most of the seats at +0 PVI are Democratic-held, that allows a case that those aren’t currently swing seats, and shouldn’t be considered as such.

  16. because the nation as a whole shifted blue at about the same magnitude.

    Now I am surprised that a lot of California PVIs didn’t budge much despite California shifting blue at a greater magnitude than the country did from 2004 to 2008. Going from Kerry +10 to Obama +24, a 14-point shift vs. the country, which went from Bush +2 to Obama +6, an 8-point shift.

  17. I really have to give credit to the community here on this one. I think most folks just figured that the old PVI formula combined both sets of presidential election results. When I told them how it actually worked (something Tim Sahd had previously explained to me), people were very surprised and a ton of discussion ensued, prompting my email to you on the subject.

    I’m delighted that the Cook Political Report is so open-minded and flexible, but I’d expect no less from such a top-notch outfit.

  18. “DavidNYC for Governor, Charlie Cook says he’s better than the current guy.”

    First we found out Sen. Schumer was reading this blog, now Charlie Cook is commenting. I love this site!

  19. Do we have any use for Charlie Wheelan, to move to some sort of upscale Chicago suburb that still votes Republican?

  20. My advice for those really into the voting patterns of districts is to go on our website ( and download the pdf of the PVI ratings in ranked order, from the most Republican district to the most Democratic district.

    Then line them all of up one sequence and all cut the top and bottom margins of every page in the middle and the bottom off the first page and the top off the last page and scotch tape it together, it’s about seven feet long.  

    Then go through with a light blue highlighter and outline each line of a currently Dem held district and with a pink highlighter do the GOP-held seats, so that every line is colored either blue or red.  Then hang it on your wall, it is a great too for focusing your attention on the races that are the most swing districts, where the action historically has been in the districts that ranged from R+3 to D+3.  And when you look at blue Democratic-held seats deep in Red Republican country, look into the circumstances that allowed that to happen, the sme from red Republican-held seats in what are otherwise blue districts.

    While the Cook PVI is a flawed and blunt instrument in determining the tilt of a district, it really is the only legitimate way to measure and organize these districts using universal criteria across 50 states.

  21. I’m interested in alternative PVI systems, based on comparisons with other races.  Such as comparisons to gubernatorial, senate, and other statewide races.  Granted, they aren’t as high-profile, and are thus less useful as partisanship indicators, but it’s just something I’m curious about.

Comments are closed.