I will admit, I was
very skeptical of the
Constituent Dynamics "Majority Watch" U.S. House polls. The methodology was certainly credible, the sample sizes were pleasingly large, and the margins of error pleasingly small, but too many of the results did not, in my pre-election opinion, pass the sniff test. But the results were far more accurate than conventional wisdom, or at least my personal opinion, in a ton of these races.
More commentary and a race-by-race analysis after the jump.
Not to give Constituent Dynamics too many kudos, it still
is the case that their polls
were innaccurately biased in favor of Democrats. But that bias was small, not overwhelming and ridiculous as it seemed to me before the elections.
And another more important caveat: voter opinion is fluid, so that just because an outcome is different from what MW polled a week or a month earlier doesn't mean MW was wrong. MW could have been right at the time, but the status of the race changed. Conversely, just because an MW poll a month out provided the correct margin or even vote shares in the actual election doesn't automatically mean the MW poll was correct when conducted. But certainly MW nevertheless deserves the benefit of doubt in that case absent evidence to the contrary. All told, the MW polls correctly showed the ultimate winner leading in an overwhelming majority of these close battleground races.
Now let's take a look race-by-race. Warning: this is loooooong as MW polled a lot of races! Below the race-by-race analysis is my summary analysis.
AZ-01, Renzi (R) vs. Simon: MW showed a dead heat, first with Simon by 5 and then Renzi by 2, 48-46, in late October. Renzi won 51-44, barely within the margin of error for the final MW poll although not the dead heat MW suggested. MW correctly picked the winner.
AZ-08, pre-primary: MW's only poll here, in a race in which Democrat Giffords beat Gooper Graf, showed an unnamed Dem winning 50-46 over an unnamed GOPer. I'm guessing the MW folks didn't follow up here because they bought the conventional wisdom that the Dems would take this seat. Giffords won 54-42, and I'll give MW credit for picking the winner.
CA-04, Doolittle (R) vs. Brown (D): MW gave Doolittle a 49-46 lead a couple weeks out, following a 51-44 advantage a couple weeks before that. Doolittle won 49-46, MW nailed it exactly.
CA-11, Pombo (R) vs. McNerney (D): MW's one poll 2 weeks out showed McNerney with a 48-46 advantage, and he won 53-47. Credit MW for picking the winner and nailing Pombo's number within a point.
CO-04, Musgrave (R) vs. Paccione (D): MW polled twice here, showing Musgrave up 47-41 in August but giving Paccione a 48-45 edge 2 weeks out. Musgrave won 46-43 with an indy getting 11%. MW never named Eidness, the indy, in its polling, but "other" still got 5%. Perhaps if they named Eidness that would have showed stronger support for him? Either way, MW missed on this one, but to their credit still not by much.
CO-07, Perlmutter (D) vs. O'Donnell (R): they polled here 3 times, the first 2 times showing a dead heat before giving Perlmutter a 51-46 edge 2 weeks out. Perlmutter won 55-42, living up to conventional wisdom that he was way ahead all along. MW missed on this one, although still gets credit for picking the winner.
CT-02, Simmons (R) vs. Courtney (D): MW polled twice here, both times showing Courtney up 51-45 including once 2 weeks out. Courtney is clinging to a 167-vote lead in a 50-50 finish that is subject to recount, and while the winner is still undetermined, MW deserves credit for picking the winner if Courtney's edge holds up. But MW still appears to have missed in the larger sense of having giving Courtney a more comfortable edge than he had. Indeed, conventional wisdom the final weeks said that Simmons held a narrow edge.
CT-04, Shays (R) vs. Farrell (D): MW polled twice here and gave Shays a comfortable margin both times, the last time a 52-43 lead a couple weeks out. Shays, to my great surprise, won 51-48. MW missed on the closeness of this race but still deserves credit for picking what I thought was a surprise winner.
CT-05, Johnson (R) vs. Chris Murphy (D): MW polled twice here, first showing Johnson up 5 in mid-October but then giving Murphy a 51-43 lead 2 weeks out. That seemed too big at the time, but Murphy won 56-44. I credit MW both for picking with winner and seeing Murphy catching up and then pulling away, which was consistent with conventional wisdom.
FL-13, Jennings (D) vs. Buchanan (R): MW polled 3 times here, once pre-primary showing a big lead for an unnamed Gooper but then twice with a named ballot showing Jennings up. Jennings "lost" by a couple hundred votes in an outcome that is being contested, after MW showed her up 49-47 2 weeks out. Given what I've read about dramatically and abnormally high undervotes due to the House race having been hard to find on the ballot, I'm going to credit MW here for possibly getting it right. But for the time being on the MW "win-loss" count I'll count it a loss, just as I'm tentatively counting Courtney a "win" for MW predicting the winner.
FL-16, Mahoney (D) vs. Foley/Negron (R): MW did something interesting here, polling on two tracks at the same time: they tested Foley alone on one track, and Foley with an explanation that it's really for a then-unnamed other Gooper (Negron had not yet been selected) on the second track. The Foley-alone track produced a 50-43 Mahoney advantage, but that shrank to 49-47 with an unnamed Gooper. Mahoney won 49-48, close enough for me to say MW nailed it.
FL-22, Shaw (R) vs. Klein (D): MW polled twice here, giving Shaw a 52-44 lead in mid-October but Klein a 50-48 edge a couple weeks out. Klein won 51-47, close enough for me to say MW nailed it.
IA-01, Braley (D) vs. Whalen (R): MW polled only once here, and it was waaaaay out in August, but the 54-41 Braley lead was mostly in line with conventional wisdom and compares quite favorably with the actual 55-43 Braley win. Kudos again to MW.
IA-02, Leach (R) vs. Loebsack (D): this was a strange choice at the time since no one thought Leach was in trouble, but MW polled twice here and shockingly found a dead heat both times: Loebsack up 48-47 in mid-October and down 50-48 a couple weeks out. Loebsack won 51-49, and I'm crediting MW for an accurate call even though I can't give them a "win" on the "scorecard". This was one of MW's marquee accomplishments.
IA-03, Boswell (D) vs. Lamberti (R): MW polled only once in August and gave Boswell a 54-43 edge. Boswell won 52-46. Credit MW for picking the winner.
ID-01, Grant (D) vs. Sali (R): MW polled once in mid-October and gave Sali a 49-42 edge; Sali won 50-45. MW deserves credit here for picking the winner and coming real close to the margin.
IL-06, Duckworth (D) vs. Roskam (R): MW polled 3 times here, all 3 times with one candidate or the other with a one-point lead in what clearly was a dead heat. The last poll gave Duckworth a 48-47 edge, but she lost 51-49. Like the Loebsack race, I can't give them credit for picking the winner, but they nailed the closeness of the race.
IL-08, Bean (D) vs. McSweeney (R): 3 MW polls here, the last giving Bean a 50-45 edge a couple weeks out. The actual result was Bean 51-44. MW was close enough to say they nailed this one.
IL-10, Kirk (R) vs. Seals (D): MW polled twice here in October, first giving Kirk a 46-44 lead before giving Seals a 48-46 edge. Kirk won 53-47, and MW deserves credit for seeing the closeness even though they were off a bit.
IL-14, Hastert (R) vs. Laesch (D): MW polled once in mid-October right after the Foley debacle exploded and showed Hastert up 52-42. Hastert won 60-40, but it's very possible it was closer a few weeks earlier when MW actually polled the race. Credit MW for, in my opinion, probably having caught a Laesch rise that later fizzled out--not to mention they picked the winner.
IL-19, Shimkus (R) vs. Stover (D): MW polled and found Shimkus up big at 53-36 in mid-October, and Shimkus won 61-39. I think it's fair to say MW nailed it pretty well.
IN-02, Chocola (R) vs. Donnelly (D): MW always had Donnelly holding small leads here, the last time 48-45 a couple weeks out. Donnelly won 53-47, and I credit MW for mostly getting it right even if the numbers were slightly off.
IN-08, Hostetler (R) vs. Ellsworth (D): MW polled twice in September and then October, showing Ellsworth up comfortably but not by a blowout margin both times. The last time was 53-43 a couple weeks out. Credit MW for picking the winner.
IN-09, Sodrel (R) vs. Hill (D): 2 MW polls here in September and October, the 2nd time a couple weeks out showing Hill up 51-43. The margin was high as Sodrel had closed by then, but Hill won 49-46. Credit MW for picking the winner.
KY-03, Northup (R) vs. Yarmuth (D): MW polled twice here in October and produced results that seemed out of whack with conventional wisdom at the time: first a 48-48 tie, then Yarmouth up 52-46 a couple weeks out. Like the Loebsack race, MW caught something here some others didnt, as Yarmuth won 51-48.
KY-04, Davis (R) vs. Lucas (D): MW polled 3 times here, twice showing Davis up 49-46 before giving Lucas a 50-46 lead a couple weeks out. MW misfired on this one, as Davis, not Lucas, had momentum in October and pulled away to a 51-44 win.
MN-01, Gutknecht (R) vs. Walz (D): MW polled twice in October, showing Walz down both times at 48-47 and then 50-47 a couple weeks out. Walz won 53-47 and MW missed on the winner, but credit them again for seeing the closeness of it.
MN-06, Wetterling (D) vs. Bachmann (R): MW polled 3 times here including twice in October, showing Wetterling up 49-45 before giving Bachmann a 48-47 lead a couple weeks out. A pre-primary MW poll gave an unnamed Gooper a 53-42 edge over an unnamed Dem. Bachmann won 50-42, and credit MW for picking the winner, correctly noticing her October momentum, and correctly pegging the district's partisan leaning.
NC-08, Hayes (R) vs. Kissell (D): MW polled twice here and produced shocking results just like IA-02; Kissell led in both polls, 51-44 and 48-44 both in October. Hayes "won" in a race that's not really decided yet and might go to recount, but tentatively strike MW for missing the winner while giving them huge kudos for catching that Kissell was the real deal.
NC-11, Taylor (R) vs. Shuler (D): 3 MW polls here all showing Shuler winning, the last a 53-44 edge a couple weeks out. Shuler won 54-46, and credit MW for nailing the winner and coming close enough to say they nailed the vote shares and margin.
NH-02, Bass (R) vs. Hodes (D): MW polled once 2 weeks out and found Hodes up 50-47. Hodes won 53-45. MW picked the winner and might have nailed the margin at the time of polling, as Hodes was still gaining momentum right through to election day.
NJ-07, Ferguson (R) vs. Stender (D): MW polled twice i October and found it close, giving Ferguson a 3-point lead both times including 46-43 a couple weeks out. Ferguson won 49-48, and credit MW for picking the winner and recognizing the closeness.
NM-01, Wilson (R) vs. Madrid (D): MW polled twice but the 2nd time was in mid-October with Madrid up 52-44. Wilson "won" in another 50-50 outcome that might not be quite over due to closeness, although her lead looks big enough to likely hold up. Can't give MW credit for picking the winner, but they may have nailed it correctly at the time of that last poll since it did match conventional wisdom. It was that debate in late October that killed Madrid.
NV-03, Porter (R) vs. Hafen (D): MW polled twice in October and gave Porter 51-43 and then 51-44 advantages. Porter won 48-47 in a race that everyone agreed tightened late and in which private polling a week or two out actually might have given Hafen the edge. Credit MW for picking the winner, but they may have missed the closeness.
NY-03, King (R) vs. Mejias (D): MW polled twice in October and gave King a 51-44 edge 2 weeks out. King won 56-44, and credit MW for picking the winner, seeing a comfortable margin, and nailing Mejias' final vote share.
NY-19, Kelly (R) vs. Hall (D): one of several races where MW polls that at the time appeared to have jumped the shark, MW gave Hall a big 49-40 lead in mid-October and a slim 49-47 lead late. Hall won 51-49, and credit MW for nailing this one.
NY-20, Sweeney (R) vs. Gillibrand (D): MW showed Gillibrand with big, double-digit leads twice in October that clearly weren't right, but late momentum turned what was a small Sweeney lead into a dead heat into a Gillibrand edge and a 53-47 Gillibrand win. Credit MW for picking the winner.
NY-24, Arcuri (D) vs. Meier (R): MW showed Arcuri up 52-43 in mid-October, and Arcuri won 54-45. Credit MW for nailing it.
NY-25, Walsh (R) vs. Maffei (D): MW really did jump the shark here by showing big Maffei leads twice, when Walsh was always really ahead and eeked out a 51-49 win. I just can't give MW any credit here, they were waaaaaay off.
NY-26, Reynolds (R) vs. Davis (D): MW showed Davis up 56-39 in mid-October hard on the heels of Reynolds reeling from the Foley scandal. Another indy poll showed a similar Davis margin to seemingly corroborate MW, and Survey USA showed a smaller Davis lead. But MW missed on the winner in a race where their month-out polling can't be blamed since it was more-or-less on target at the time.
NY-29, Kuhl (R) vs. Massa (D): more shark-jumping, as MW twice in October showed Massa up double-digits. Kuhl eeked out a 52-48 win; MW deserves no credit here.
OH-01, Chabot (R) vs. Cranley (D): MW showed Cranley up 48-46 a couple weeks out, but Chabot won 53-47. MW missed on this one, even though the dead heat they showed seemed consistent with conventional wisdom at the time.
OH-02, Schmidt (R) vs. Wulsin (D): MW showed Wulsin up in mid-October but gave Schmidt a 51-46 edge a couple weeks out. Schmidt apparently has won 51-49 although Wulsin thinks she still has an outside chance based on some uncounted provisionals, but I think Schmidt has won this one.
OH-06, Wilson (D) vs. Blasdel (R): MW showed Wilson up 56-40 in August and didn't poll again. Wilson won 62-38; credit MW for nailing the blowout Dem win.
OH-12, Tiberi (R) vs. Shamansky (D): MW gave Tiberi a shakey 51-46 edge a couple weeks out, but Tiberi won in a 61-39 blowout. MW gets credit for picking the winner but really was way off.
OH-15, Pryce (R) vs. Kilroy (D): MW gave Kilroy a commanding 53-41 lead in mid-October that wasn't inconsistent with conventional wisdom at the time, but Pryce rallied and probably has won 51-49 in another race whose outcome isn't settled. I think Pryce's lead holds up and MW misfired on the winner even though they might have had it right at the time.
OH-18, Space (D) vs. Padgett (R): Space was up 51-42 in mid-October and ended up winning big, 62-38. I'll credit MW for picking the winner and also a comfortable margin; no one saw a 24-point spread coming.
PA-04, Hart (R) vs. Altmire (D): MW gave Hart a 51-47 edge in its only poll, a couple weeks out. Altmire won 52-48. MW missed on the winner but might have nailed it at the time of polling, as Altmire still was just riding momentum that did not crest until election day.
PA-06, Gerlach (R) vs. Lois Murphy (D): MW missed here, giving Murphy 5-6-point leads in 3 polls including 51-46 a couple weeks out. Gerlach won 51-49.
PA-07, Weldon (R) vs. Sestak (D): MW gave Sestak a 51-44 lead in its only poll a month out. Then it came out Weldon was a FBI target and Sestak won 56-44. Credit MW for nailing it in my opinion.
PA-08, Fitzpatrick (R) vs. Patrick Murphy (D): MW gave Murphy a 50-47 lead a couple weeks out after having had Fitzpatrick up 53-45 in August. Murphy won by just over 1,500 votes in a 50-50 race; credit MW for nailing it in my opinion.
PA-10, Sherwood (R) vs. Carney (D): MW gave Carney a 50-43 edge way back in August in their only poll. Carney won 53-47. I'll give MW credit on this one, they picked the winner and got the margin within a point.
TX-17, Chet Edwards (D) vs. Van Taylor (R): MW gave Edwards a huge 55-38 lead a month out, and Edwards won 58-40. Credit MW for nailing it, I say.
VA-02, Drake (R) vs. Kellam (D): MW polled 3 times and showed big swings: Kellam 51-43 in August; Drake 49-45 in mid-October; Kellam 50-45 a couple weeks out. Drake won 51-49, and MW missed it.
VA-10, Wolf (R) vs. Feder (D): MW showed Wolf up by a scary-close 47-42 margin a month out. But Wolf won 57-41, and I doubt it was ever as close as MW polled. MW missed here.
VT-AL, Welch (D) vs. unnamed R: MW polled only in August pre-primary and gave Welch a 54-40 edge. Welch won 53-45. I give MW credit here for correctly seeing a comfortable Welch win and getting his vote share within a point.
WA-05, McMorris (R) vs. Goldmark (D): MW gave McMorris a 51-46 lead a couple weeks out, which might have been right or close to right at the time given some information that McMorris was sweating. McMorris won 55-45, and I credit MW for picking the winner but seeing it was a race.
WA-08, Reichert (R) vs. Burner (D): the votes aren't even all counted yet, but Reichert has held a 51-49 edge the whole time that analysts expect to hold firm. MW went see-saw on this and ended up giving Burner a 49-47 edge a couple weeks out. Conventional wisdom said Reichert held a slim lead all along, so I think MW missed here.
WI-08, Kagen (D) vs. Gard (R): MW polled 3 times including a pre-primary and then two named ballots; the named ballots both showed Kagen up, the 2nd time by a 51-45 edge a couple weeks out. Kagen won 51-49. Credit MW for picking the winner. Also interestingly, MW's unnamed ballot in August showed a Dem winning 48-44, defying conventional wisdom that this was a GOP-advantaged seat.
WV-01, Mollohan (D) vs. Wakim (R): MW gave Mollohan a 52-42 edge in August, and he went on to win in a 64-36 blowout. Credit MW for picking the winner and seeing a comfortable Mollohan advantage, and it might very well have been closer way back in August.
**
So, what's the bottom line here?
Assuming thin leads hold up in the yet-undecided contests, MW correctly picked the winner in 42 of the 60 races it polled, going by the final poll in races where it polled multiple times. It missed on the winner 18 times. But those 18 times included correctly predicting dead heats a bunch of times, such as in CO-04 and IA-02 as only a couple examples. Indeed of the 18 races where MW missed on the winner, I think there were only 2, NY-25 (Walsh/Maffei) and NY-29 (Kuhl/Massa) where they really were waaaaaay off. The other polls either were validated by actual election results or were pretty close to conventional wisdom at least at the time the polling was done. Now, there were races where MW correctly picked the winner but still probably jumped the shark, like in NY-20 in showing Gillibrand with consistent double-digit leads when that clearly was wrong. But elections are a winner-take-all deal, so I might as well credit a pollster, too, for picking the winner.
Majority Watch polls ultimately were stronger than I thought they were. I hope they continue to try to identify their flaws as there clearly are a few, but equally I hope they keep polling because they were more right than wrong including on some of their surprises. After all, Robert Novak insisted on Fox News on election night, as he shook his head in disbelief, that no one thought Jim Leach was ever in trouble. Well, Bob, Majority Watch sure did know!