We keep track of so many different elections here at DKE that sometimes it's important to step back and try to look at national trends. Take the 2012 House Elections. How "predictable" were they? How much did they fall into a pattern? Where were the real upsets? How important was incumbency?
We'll never be able to definitively answer those questions, but we can use the information we have now--Obama's 2008 performance in each district and the preliminary vote counts--to take a look. Who were the best and worse candidates each party fielded relative to the national pattern? Which candidates did better than expected? And how badly did Michele Bachmann do this time? Let's find out more.
Introduction/Graph:
Our own twohundertseventy came up with an idea for evaluating candidate quality relative to national trends using Presidential performance by district and incumbency.
Here's the idea: It's important to have a baseline in mind, and the best baseline is to compare similar situations in similar districts.
This graph shows Obama's 2008 performance in the 2012 House seats plotted against the Democratic performance last Tuesday (both using two-party vote share). I used preliminary House results courtesy of David Nir, who suggested I share them copied-and-pasted, here. Let me know via PM if you want to see the full spreadsheet I used. I also used DKE's Presidential Results by Congressional District for the Presidential numbers and for which seats were open, although they missed a few primary losses (which I'm counting as open seats). I'm always terrified of typos and small mistakes, but with this many elections, they shouldn't matter too much.
I limited myself to races where a Democrat and a Republican faced off in the general election, and eliminated Louisiana for its jungle primary weirdness. Red dots represent districts with Republican incumbents, blue dots represent districts with Democratic incumbents, yellow dots are open seats, and the green dots are the two districts where incumbents from both parties faced off thanks to redistricting.
As you can see, all 387 of these elections mostly line up into a predictable pattern. If you squint, you can sort of see the graph as three parallel lines, with the blue line above the yellow line, and the yellow line a bit above the blue line. In other words--Democratic incumbents, on average, did a bit better than Democrats running in seats, who did a bit better than Democrats running against Republicans.
But the difference is pretty small on average.
Also, note how much crazier the graph gets around where seats tend to be actively contested--45% Obama to 60% Obama or so.
We can also easily see a few outliers. For example, that red dot floating above the trend at about 35% Obama? That's TN-04, where Eric Stewart held Scott Desjarlais to about 55% of the two-party vote. No Democratic challenger did quite so well in quite so red a district.
A Regression Equation:
We can quantify the above pattern with a regression equation, again, using twohundertseventy's idea.
The result (thanks to Wessa)?
On average, Democratic House performance was highly predicted by Obama's 2008 performance, although Democrats did about 5 points better when they were incumbents, and about 5 points worse when they were challengers.
Or, to put that into an equation rounded to the nearest tenth:
2012HouseDem= 2.9+ 0.9*Obama2008+ 4.9(in races with D incumbents) -5.4 (in races with R incumbents).
That's a pretty weak incumbency advantage for what's generally not considered to be a wave election.
In 2004, a similar calculation gave incumbency effects that were roughly twice as large.
The median error was about 2.9 points. You get a much weaker fit if you look at the most heavily-contested elections, but I like this way, because you can compare the predicted results of elections where there was a lot of campaigning to the predicted results of elections with minimal or nonexistent campaigning.
Best and Worst:
The 15 Democratic candidates who out-performed their predictions by the most were in the following races. For each race, I've listed the Democratic candidate, the Republican candidate, the Democrat's actual two-party vote share, and then the one predicted by the national pattern.
MA 8 Lynch Selvaggi 76.26% (actual) 62.21% (predicted)
TN 4 Stewart DesJarlais 44.24% 30.92%
FL 26 Garcia Rivera 55.49% 43.46%
MN 6 Graves Bachmann 49.41% 37.83%
MN 7 Peterson Byberg 63.39% 52.35%
MD 6 Delaney Bartlett 60.61% 50.02%
HI 2 Gabbard Crowley 80.57% 71.41%
TX 20 Castro Rosa 65.68% 56.79%
TX 23 Gallego Canseco 52.51% 43.92%
CT 3 DeLauro Winsley 74.55% 66.30%
AL 6 Bailey Bachus 28.64% 20.71%
TX 28 Cuellar Hayward 69.57% 61.66%
NJ 10 Payne Jr. Kelemen 88.95% 81.07%
NY 26 Higgins Madigan 74.64% 66.90%
TX 33 Veasey Bradley 73.80% 66.36%
I think most of these elections fall into a few broad categories.
Some involved popular Blue Dogs and moderates who either do well in hostile districts or do really well in neutral/light blue districts, like Stephen Lynch, Collin Peterson, and Henry Cuellar.
A few others had Republican incumbents facing scandals, like Scott DesJarlais or David Rivera, or who were just generally bad politicians, like Michele Bachmann (perhaps the most consistently weak Republican incumbent in the country).
Plus there are a few elections where Republicans didn't really try at all, and where the Democratic candidate is particularly popular or appealing: Tulsi Gabbard's general election opponent was a "homeless handyman".
The presence of three districts on the border of Texas (plus one in Dallas) is interesting. It's possible that, once we can properly account for Obama's 2012 performance in this region, it'll turn out that some of those Representatives didn't really over-perform by as much as 2008 would indicate. But it certainly suggests that South Texas Democrats had a good night in Presidential performance, in House performance, or both.
The 15 worst Democratic candidates ran in the following races:
HI 1 Hanabusa Djou 54.60% 73.46%
RI 1 Cicilline Doherty 55.79% 70.02%
IL 18 Waterworth Schock 25.79% 38.77%
MI 10 Stadler Miller 30.25% 42.52%
CA 21 Hernandez Valadao 39.97% 51.71%
MA 6 Tierney Tisei 50.51% 61.27%
IL 2 Jackson Woodworth 72.87% 83.01%
MI 4 Freidell Wirth Camp 34.70% 43.92%
OR 2 Segers Walden 29.59% 38.24%
IL 1 Rush Peloquin 73.79% 82.26%
TN 2 Goodale Duncan 21.62% 29.99%
PA 4 Perkinson Perry 36.57% 44.72%
WV 2 Swint Capito 30.22% 38.35%
NJ 4 Froelich Smith 31.31% 39.28%
MO 8 Rushin Emerson 25.45% 33.14%
Once again, we have a few potential (overlapping) explanations. Some of these elections had the kind of moderate Republicans who tend to do well, like Chris Smith, Jo Ann Emerson, Charles Djou, Richard Tisei, and Shelley Moore Capito (plus Jimmy Duncan, who isn't really a moderate, but who is an unusual paleoconservative Republican). Others had scandal-tarred Democrats like David Cicilline, John Tierney, and Jesse Jackson. There are also a few races where Democrats didn't really try against particularly competent Republicans. And it's certainly possible that Obama's disproportionate drops in Michigan, Missouri, downstate Illinois, and Hawaii help to explain a few of these.
The Wrong Calls:
Another way to look at this is: What elections didn't go the way they "should" have?
Here are the races Democrats lost where they had the highest predicted vote share. If the last column is over 50%, it means the Democrat "should have" won:
IL 13 Gill Davis 49.76% 54.00%
CA 21 Hernandez Valadao 39.97% 51.71%
IA 3 Boswell Latham 45.47% 51.14%
MI 11 Taj Bentivolio 46.64% 49.84%
KY 6 Chandler Barr 48.00% 49.59%
PA 12 Critz Rothfus 48.22% 49.59%
IN 2 Mullen Walorski 49.27% 49.36%
NY 27 Hochul Collins 49.27% 49.08%
MT AL Gillan Daines 44.64% 47.94%
WI 8 Wall Ribble 44.07% 47.63%
NV 3 Oceguera Heck 45.95% 47.63%
CO 6 Miklosi Coffman 48.08% 47.63%
WI 7 Kreitlow Duffy 43.87% 47.21%
MI 6 O'Brien Upton 43.93% 47.21%
NY 19 Schreibman Gibson 46.53% 47.21%
Note that only three of these elections are real misses: IL-13, CA-21, and IA-03. As for the others, Obama's disproportionate drops in Montana, Wisconsin, Michigan, Indiana, and downstate Illinois might account for quite a few of these. Although David Gill was considered a weak candidate anyway, for example.
But--on the other hand--perhaps we shouldn't really regard the losses in KY-06, PA-12, or NY-27 as upsets. This was, in general, a pretty nationalized election, where incumbents had relatively small advantages, and those are red seats. (On the other hand, Chandler, Critz, and Hochul all have histories of over-performing baseline expectations, so perhaps it could be reasonably called an upset when they don't.)
Among the races Democrats won, they had the lowest predicted vote share in the following:
FL 26 Garcia Rivera 55.49% 43.46%
TX 23 Gallego Canseco 52.51% 43.92%
FL 18 Murphy West 50.29% 44.85%
CA 36 Ruiz Bono Mack 51.93% 44.88%
NY 18 Maloney Hayworth 51.69% 45.78%
CA 7 Bera Lungren 50.43% 45.82%
NC 7 McIntyre Rouzer 50.08% 46.42%
UT 4 Matheson Love 50.62% 46.66%
NH 1 Shea-Porter Guinta 51.95% 46.70%
WV 3 Rahall Snuffer 53.95% 47.20%
MN 8 Nolan Cravaack 54.47% 47.21%
AZ 1 Kirkpatrick Paton 51.62% 47.51%
GA 12 Barrow Anderson 53.69% 48.25%
CA 52 Peters Bilbray 50.30% 49.08%
NH 2 Kuster Bass 52.66% 49.49%
NY 24 Maffei Buerkle 52.48% 50.02%
MD 6 Delaney Bartlett 60.61% 50.02%
AZ 9 Sinema Parker 51.60% 50.77%
MN 7 Peterson Byberg 63.39% 52.35%
Some 15 of these elections are real misses (with two more close)--suggesting that Democrats benefited more from campaign-by-campaign effects than Republicans, or that Obama's relative district-by-district performance varied from 2008 to 2012 in a way that helped these Democrats, or both.
Sometimes it probably really is both: South Florida and the Texas border probably trended Democratic from 2008 to 2012, and Democrats were also helped there by such quality challengers as Pete Gallego and Raul Ruiz, along with such lame opponents as Allen West and David Rivera.
Another example: Obama had an especially good performance in Maryland this time around, and Rep. Bartlett probably didn't have much of an incumbency advantage after his district was blown up. A strong challenger, Sean Patrick Maloney, and a strong Obama performance in New York (even relative to 2008) probably came together in the upset defeat of Nan Hayworth.
Democrats also have a few more durable red district Representatives like John Barrow, Jim Matheson, or Mike McIntyre, although there seem to be fewer and fewer of them every two years. Nick Rahall is probably the only Representative who could end up on a list like this with a performance everyone seemed to find surprisingly poor.
And Republicans also had to defend a lot of weak incumbents who were swept in by the 2010 wave, like Chip Cravaack and Ann Marie Buerkle.
Conclusions:
Here are some thoughts I have as a result of this exercise.
1. Incumbency wasn't irrelevant, but this was a nationalized election. The incumbency advantages were pretty small, even in the overall regression equation, although the regression is a fair bit less accurate if you ignore incumbency outright. My guess is some of that will disappear once we have the 2012 Presidential results by Congressional district. As jncca speculated earlier, ticket-splitting might well be declining in general. And it's striking how much you can predict using Obama's performance from 2008--more reason to think that "what is most remarkable about 2012 is not its radical change but instead enduring stability".
2. Moderates are popular. If I remember right, Bobby Bright topped twohundertseventy's analysis of 2010. (Which used PVI, not just Obama's vote share.) Moderate Democrats and moderate Republicans will often--not always--rack up the biggest margins when they're uncontested and be the hardest to defeat when they're contested. This idea that if you "give people a choice between a Republican and a Republican they'll chose the Republican every time" doesn't have a lot of support that I can see.
3. Small errors don't rule out campaign or candidate effects. Although I'm skeptical of a lot of stories about candidate and campaign effects, I should point out that this sort of analysis doesn't rule out the possibility. What it does suggest, I think, is that if a race ended up more or less as predicted, then candidate and campaign effects were likely either national factors or district-specific factors that happened to mostly cancel out.
For example, take PA-12. My suspicion is that Democrat Mark Critz was an unusually strong incumbent who was likely to over-perform, but who was also at a considerable disadvantage in outside spending and in campaign spending against Republican Keith Rothfus, especially when you realize how much of the listed total Critz actually spent on his primary campaign. (And Rothfus might have been a good candidate himself--I have no idea.) Their strengths probably cancelled out, with the election's actual result very close to the predicted one.
Consider these kinds of patterns, but don't take the results at face value. There were still 435 (or 387) distinct House elections last Tuesday, even if the results were quite highly correlated.
Appendix: Over/Under-Performances in Notable Races:
I looked up the 65 races where either the NRCC or DCCC spent money this cycle (some of the districts on this list are almost certainly only on here because of screwed-up redistricting numbers, like NY-09, but anyway).
Limiting ourselves to those races, but keeping the above equation's predictions, here are the races where Republicans out-performed their predictions, and won:
IA 3 Boswell Latham 45.47% 51.14%
NV 2 Koepnick Amodei 38.55% 43.93%
IL 13 Gill Davis 49.76% 54.00%
WI 8 Wall Ribble 44.07% 47.63%
WI 7 Kreitlow Duffy 43.87% 47.21%
NV 3 Oceguera Heck 45.95% 47.63%
KY 6 Chandler Barr 48.00% 49.59%
PA 12 Critz Rothfus 48.22% 49.59%
NC 8 Kissell Hudson 45.86% 46.81%
NY 19 Schreibman Gibson 46.53% 47.21%
IN 2 Mullen Walorski 49.27% 49.36%
Here are races where Republicans out-performed their predictions, but lost:
RI 1 Cicilline Doherty 55.79% 70.02%
MA 6 Tierney Tisei 50.51% 61.27%
CA 9 McNerney Gill 54.09% 60.88%
CA 16 Costa Whelan 55.05% 61.82%
CA 24 Capps Maldonado 54.88% 60.88%
IL 10 Schneider Dold 50.49% 55.99%
NY 25 Slaughter Brooks 57.18% 62.59%
NY 21 Owens Doheny 51.12% 56.09%
CO 7 Perlmutter Coors 56.43% 61.27%
IA 2 Loebsack Archer 56.59% 61.27%
IA 1 Braley Lange 57.70% 62.21%
CA 26 Brownley Strickland 51.87% 56.01%
CT 5 Esty Roraback 51.52% 55.46%
AZ 2 Barber McSally 50.10% 53.30%
NY 1 Bishop Altschuler 52.17% 55.16%
WA 1 DelBene Koster 53.55% 55.46%
NV 4 Horsford Tarkanian 54.29% 56.01%
CA 41 Takano Tavaglione 57.59% 58.85%
IL 17 Bustos Schilling 53.28% 53.77%
IL 8 Duckworth Walsh 54.72% 55.06%
The Democrats who over-performed but lost anyway were in the following races:
FL 10 Demings Webster 48.24% 41.13%
OH 6 Wilson Johnson 46.63% 39.70%
FL 2 Lawson Southerland 47.26% 41.13%
MI 1 McDowell Benishek 49.65% 44.39%
TX 14 Lampson Weber 45.49% 41.94%
IA 4 Vilsack King 45.60% 42.52%
IN 8 Crooks Bucshon 44.67% 42.06%
AZ 8 Scharer Franks 35.34% 32.78%
CA 10 Hernandez Denham 46.27% 44.88%
OH 16 Sutton Renacci 47.77% 46.45%
CO 3 Pace Tipton 43.47% 42.52%
CO 6 Miklosi Coffman 48.08% 47.63%
NY 27 Hochul Collins 49.27% 49.08%
And finally, the Democrats who over-performed on their way to winning were in the following races:
TX 23 Gallego Canseco 52.51% 43.92%
NY 26 Higgins Madigan 74.64% 66.90%
MN 8 Nolan Cravaack 54.47% 47.21%
CA 36 Ruiz Bono Mack 51.93% 44.88%
WV 3 Rahall Snuffer 53.95% 47.20%
NY 18 Maloney Hayworth 51.69% 45.78%
FL 18 Murphy West 50.29% 44.85%
GA 12 Barrow Anderson 53.69% 48.25%
NH 1 Shea-Porter Guinta 51.95% 46.70%
CA 7 Bera Lungren 50.43% 45.82%
AZ 1 Kirkpatrick Paton 51.62% 47.51%
UT 4 Matheson Love 50.62% 46.66%
NC 7 McIntyre Rouzer 50.08% 46.42%
IL 11 Foster Biggert 58.06% 54.71%
NH 2 Kuster Bass 52.66% 49.49%
NY 24 Maffei Buerkle 52.48% 50.02%
NY 9 Clarke Cavanagh 88.26% 85.80%
CA 52 Peters Bilbray 50.30% 49.08%
AZ 9 Sinema Parker 51.60% 50.77%
IL 12 Enyart Plummer 54.60% 54.00%
OR 1 Bonamici Morgan 64.28% 64.09%
I include these lists for two reasons. First of all, I wanted to make sure that people could see how candidates in high-profile races did. And I also think it's important to understand that candidate quality isn't always the same as winning or losing.
Maybe Republicans are making fun of much-hyped but unsuccessful challengers like Val Demmings or Chrisite Vilsack, and I'm sure Democrats love making fun of much-hyped but unsuccessful challengers like Ricky Gill or Maggie Brooks.
But, in fact, quite a lot of those kinds of candidates probably did very well considering circumstances beyond their control: how Democratic the district tends to be, and whether or not there was an incumbent running.