Vulnerability Index for House Elections

Over the holidays, SSP readers seemed to have a lot of fun with the vulnerable House Republicans and vulnerable House Democrats threads. This left me wondering, as so many things seem to do, “is there a way to quantify that?” In other words, is there a data-driven way to approach the question instead of just relying on perceptions (and also to make sure that potentially overlooked races don’t fall through the cracks)?

Here’s what I tried. It’s actually a bit reminiscent of my PVI/Vote Index, in that it measures representative performance against the district’s lean, except here performance is measured by the rep’s margin in the last election. (The data for many of the 2008 electoral margins is available in the recent “How’d We Do?” post, conveniently arranged in order from closest to least close.)

Look at the top 20 most vulnerable Republicans to see how it works. As pretty much everyone would expect, Anh Cao in LA-02 is the most vulnerable GOPer. He had the 5th weakest margin of any Republican who survived 2008 (beating Bill Jefferson by 2.7%, behind only Fleming (0.4%), McClintock (0.6%), Calvert (2.4%), and Luetkemeyer (2.5%). Needless to say, he’s in the GOP-held district with the least favorable PVI (D+28, using “old,” i.e. 00-04, PVI). At #2 is Jim Gerlach in PA-06; he had the 9th worst margin at 4.2%, and he’s in the 6th worst district for a GOPer at D+2. And so on…

District Rep. Margin
rating		 PVI
rating		 Total
LA-02 Cao 5 1 6
PA-06 Gerlach 9 6 15
IL-10 Kirk 13 4 17
WA-08 Reichert 16 5 21
MI-11 McCotter 17 16 33
MN-03 Paulsen 22 12 34
NJ-07 Lance 24 13 37
OH-12 Tiberi 34 14 48
CA-50 Bilbray 11 40 51
MN-06 Bachmann 6 46.5 52.5
FL-25 Diaz-Balart 18 37 55
CA-44 Calvert 3 55 58
AL-03 Rogers 25 34 59
LA-04 Fleming 1 60 61
FL-15 Posey 31 30.5 61.5
MN-02 Kline 39 23 62
CA-26 Dreier 33 30.5 63.5
MO-09 Luetkemeyer 4 60 64
NY-26 Lee 38 27 65
PA-15 Dent 58 8 66

Is this much different from SSP readers’ predictions? No, not much; it’s the wisdom of crowds at work. Still, I see a few names on there that didn’t get much of any mention in our prediction thread: especially Pat Tiberi in OH-12 (34th worst margin at 12.6%, 14th worst district at R+1) who seems to fly under the radar every single freakin’ election. Other names revealed by this list that wouldn’t necessarily be intuitive picks include Thad McCotter, John Kline, Mike Rogers (AL), and Bill Posey, who benefited from our big-time recruitment failure in the FL-15 open seat.

Here’s the flipside: the Democratic seats that seem likeliest to flip, based on 2008 numbers. Some of these may not be much cause for alarm; Chet Edwards, for instance, is probably not in any imminent danger except in case of a 1994-sized event, but he’s probably doomed to uncomfortable margins for all eternity. On the other hand, time will tell whether Walt Minnick can quickly fortify himself, or if we’re only renting ID-01 for a couple years.

District Rep. Margin
rating		 PVI
rating		 Total
ID-01 Minnick 5 1 6
AL-02 Bright 2 5 7
MD-01 Kratovil 4 10 14
TX-17 Edwards 19 2 21
VA-05 Perriello 1 26.5 27.5
AL-05 Griffith 10 20 30
MS-01 Childers 25.5 8.5 34
NY-29 Massa 6 29.5 35.5
VA-02 Nye 15.5 22 37.5
CO-04 Markey 34 11.5 45.5
PA-10 Carney 35 14 49
GA-08 Marshall 39 13 52
FL-08 Grayson 12.5 44 56.5
MI-07 Schauer 7 49.5 56.5
NM-02 Teague 33 23.5 56.5
WI-08 Kagen 20 38.5 58.5
OH-15 Kilroy 3 58 61
AZ-05 Mitchell 23 38.5 61.5
PA-03 Dahlkemper 8 54 62
OH-16 Boccieri 27.5 40 67.5

More over the flip…

In describing this method to DavidNYC, he quite rightly asked “Wait, does this thing actually work?” So, after a lot more data entry and some testing based on how well the 2006 numbers would have predicted the 2008 results, I can conclude it does work fairly well. Here is what the 2006 numbers would have predicted for GOP held seats in 2008.

District Rep. Margin
rating		 PVI
rating		 Total
NM-01 Wilson 3 7 10
NY-25 Walsh 9 5 14
PA-06 Gerlach 7 9 16
CT-04 Shays 16 2 18
WA-08 Reichert 14.5 8 22.5
NV-03 Porter 10 13 23
IL-10 Kirk 24 4 28
NJ-07 Ferguson 8 20.5 28.5
OH-15 Pryce 4.5 24.5 29
MI-09 Knollenberg 22 16 38
OH-01 Chabot 20 18.5 38.5
NC-08 Hayes 1.5 38.5 40
PA-15 Dent 33 11 44
FL-13 Buchanan 1.5 46.5 48
IL-06 Roskam 12.5 36.5 49
MI-07 Walberg 41 10 51
NY-03 King 17 34 51
AZ-01 Renzi 28 30.5 58.5
IL-11 Weller 34 24.5 58.5
NY-13 Fossella 45 14 59

One problem leapt out at me: the role of open seats, and the accompanying loss of the benefits of incumbency. So, I performed a tweak that took open seats into account (by taking out the margin, and just leaving the open seat’s strength based only on its PVI rating). That takes it a little closer to the way things actually shook out. 13 out of the top 20 were pickups, which seems like a good but not amazing rate of prediction.

Without doing a lot of putting your thumbs on the scales of individual races, I don’t know how you’d build a model that somehow predicted, say, Tom Feeney’s implosion, or the fizzle in the open seat in NM-02, or Dave Reichert’s confounding staying power, or Bob Roggio’s amazing lack of name recognition… or that Bill Sali was vulnerable (he was #106) if only because of sheer malice and stupidity. Any good prognostication has to include at least some kind of qualitative analysis of candidates’ levels of, well, suckiness.

By the way, in case you’re wondering what this formulation means would happen to Peter King’s seat if he bails out to run for NY-Sen, it would vault up to #2 on the list if it were open. (It’s the 7th most Dem PVI of any GOP-held seat, so for 2010 the score of 7 would slot an open NY-03 right before LA-02.) So, a year from now, once we have a sense of where seats will open up, I’ll have to revisit this project.

District Rep. Margin
rating		 PVI
rating		 Total
NY-25 Open 0 5 5
NJ-03 Open 0 6 6
NM-01 Open 0 7 7
NY-13 Open 0 14 14
PA-06 Gerlach 7 9 16
CT-04 Shays 16 2 18
MN-03 Open 0 18.5 18.5
NJ-07 Open 0 20.5 20.5
VA-11 Open 0 20.5 20.5
WA-08 Reichert 14.5 18 22.5
NV-03 Porter 10 13 23
IL-11 Open 0 24.5 24.5
OH-15 Open 0 24.5 24.5
IL-10 Kirk 24 4 28
AZ-01 Open 0 30.5 30.5
MI-09 Knollenberg 22 16 38
OH-01 Chabot 20 18.5 38.5
NC-08 Hayes 1.5 38.5 40
NY-26 Open 0 42 42
PA-15 Dent 33 11 44

Finally, here’s what the 2006 numbers would have predicted for the Democratic-held seats in 2008, including the tweak for open seats (of which we didn’t have many). Three of the top 10 did, in fact, fall. Plus, LA-06 isn’t on the list because it changed hands during a special election. However, my back-of-the-envelope calculation for Cazayoux based on his 3% margin in the special election and an R+6.5 would’ve given him a score around 24, good for 4th place. On the other hand, the fifth Dem seat to fall, LA-02, clocks in at #187!

District Rep. Margin
rating		 PVI
rating		 Total
GA-08 Marshall 4 9 13
AL-05 Open 0 17 17
KS-02 Boyda 11 11 22
IN-09 Hill 15 12.5 27.5
PA-10 Carney 18 10 28
TX-22 Lampson 29 4 33
NC-11 Shuler 23.5 12.5 36
WI-08 Kagen 6 30.5 36.5
TX-17 Edwards 39 1 40
FL-16 Mahoney 5 38.5 43.5
IL-08 Bean 21 23 44
AZ-05 Mitchell 14 30.5 44.5
UT-02 Matheson 43 2 45
NY-19 Hall 12.5 36 48.5
PA-04 Altmire 7.5 41 48.5
IN-02 Donnelly 25 26 51
IN-08 Ellsworth 44.5 8 52.5
CA-11 McNerney 20 33 53
TX-23 Rodriguez 26 27.5 53.5
OR-05 Open 0 55 55

58 thoughts on “Vulnerability Index for House Elections”

  2. Over 50 seats in two elections, it’s only natural to see the target list get tougher.  However, I’m impressed that our defense list doesn’t look much more vulnerable then theirs.  Yeah, Minnick and Kratovil are in trouble, but the others all faced top-tier opponents and prevailed.  Of course if things really go south guys like Perriello and Massa could find themselves swept out, but even in a neutral environment I think they hold on.

  4. Are the most endangered and also appear to be well on their way to being the worst members of Congress on our team this session.  Some seats you can’t win with Democrats.

  5. I for one don’t see Tom Perriello going anywhere. He is one of the more energetic, inspiring, and impressive candidates I’ve ever seen. He has done nothing but impress people in this district with what he’s done once he won (he has already introduced legislation to increase access to higher education.)

    He is also an organizer at heart. His focus will remain almost entirely on the rural southern part of the district.

    He’s only 34, and I think that this seat is Tom’s for as long as he wants it.

  6. Just b/c it’s Alabama?!?  The district is 33% african american, it was drawn by the dem legislature to be a democratic district.  Segall is 29 years old and a newcomer, made Red-to-Blue but didn’t get a dime, and was outspent 2-1.  The dem performance is in part based on the fact that presidential candidates haven’t run campaigns in Alabama and we didn’t run strong candidates in 06 and 04.  The guy has done NOTHING in congress.  

  7. Politics1.com is reporting that Todd Tiahrt (KS-04) is planning to join the melee for the open Senate seat in that state.  It’s hardly fertile territory for Democrats (59-40 for McCain) but the right Democrat, in an open seat situation, could make it a race.

  8. One possible concern is open seats may have too much of an advantage when the margin is set to 0.  To adjust for this, the margin could be considered missing data for open seats, with the total based solely on PVI (which in effect is the way it is calculated now).  But, for non-open seats, the margin and PVI would be averaged instead of summed.

    On the other hand, open seats are always the most likely to switch, so the way it is calculated now could be a better predictor of vulnerability.

  9. Cao’s district is the most Democratic Republican-held district by twenty points. Granted, you were only trying to rank by vulneariblity, but I’d like to see this tool try to predict more than relative order. After all, relative order means pretty much nothing excepts for us stats geeks – everybody else just notes who won and lost.

    Is there any kind of clever regressional analysis or similar you could do to take into account PVI (rather than just relative PVI) and size of margins over several cycles (rather than just order of margins over one cycle?)

Comments are closed.