As Ed said yesterday, there is still scant factual evidence for various theories of why the polls were so wrong about the Obama and Clinton votes in New Hampshire. There are, however, plenty of different explanations being discussed in various blogs and traditional news sources. Mark Blumenthal, for example, examines the statistical underpinnings of eight theories of Clinton’s upset victory at Pollster.com.
Ken Dilanian’s USA Today article “Pollsters struggle to explain Clinton win,” takes an interesting look at three of the more frequently-cited explanations. Of HRC’s ‘Muskie moment,’ Dilanian quotes Gallup’s Editor-in-chief Frank Newport, pointing out “a lot of last-minute movement in this hothouse environment” and “the intriguing potential impact of the ‘verge of tears’ video,” also noted by Senator Diane Feinstein, GOP strategist Karl Rove and Clinton herself.
Dilanian notes that exit polls indicate that 17 percent of NH voters made up their minds on election day and quotes Zogby on the “havoc” such late-deciders can cause for pollsters. But ABC News Polling Director Gary Langer counters that Clinton had a 2 percent advantage among those who made their decision before election day, according to Dilanian.
Dilanian cites the argument that “the record-shattering turnout” resulted in “a different electorate” than the one used in polling samples. But Newport points out that his sample’s demographics, including the percentage of older women believed to lean strongly toward Clinton, were “very close to those of the actual voters.”
In his Open Left article, “Obama Lost Because Of The Angry With Bush Vote,” Chris Bowers notes that Clinton had a 39 to 34 percent edge over Obama with the 62 percent of voters describing themselves as “angry” with the Bush Administration. Bowers believes that Obama’s “message of conciliatory unity” hurt him and helped Clinton. But if NH voters wanted more anger, I have to wonder why Edwards didn’t do better, especially since he was the guy who tagged Obama for being overly-conciliatory.
In addition to the gender gap favoring Clinton, the AP‘s Charles Babington notes an even more dramatic gap — exit poll data showing that Clinton outpolled Obama and Edwards 14-1 among voters who identified “experience” as the top qualification, possibly offsetting Obama’s 2-1 advantage among voters citing “change” as their top concern.
Ed wonders whether absentee ballots cast before Iowa may account in substantial part for HRC’s win. Seniors do cast a disproportionately large chunk of absentee ballots, and seniors are mostly women. Charles Franklin is skeptical because NH is one of the states that restricts absentee ballots. Still, it would be interesting to see if there is a substantial difference between NH voters before vs. after the IA caucuses.
Absentee ballots are an increasingly-important strategic consideration in many states. The ‘absentee’ (early) voter campaign is certainly huge in California, where almost half of ballots are expected to be cast by voters well before election day, and where the Clinton campaign is already heavily engaged in reaching them.
“Ed wonders whether absentee ballots cast before Iowa may account in substantial part for HRC’s win. … it would be interesting to see if there is a substantial difference between NH voters before vs. after the IA caucuses.”
In WA state where I’ve done some absentee voter polling work in the past, the Sec of State releases tabulations of results (and turnout) of absentee vs day-of-election. But while there’s some disclosure of absentee ballot volume by time period (of receipt), to my knowledge there’s no official tabulation of *results* by time period.
If this were true in New Hampshire as well, there would be no way of knowing ‘officially’ whether pre- and post-IA absentee ballot preferences differed much.
Furthermore, in WA state absentee ballots are accepted after the date of polling as long as they’re postmarked on time, and as last-minute ballots are received, updated counts and results trickle out in a few iterations after the election. This sometimes allows an observant analyst to watch for late trends. However, in NH, absentee ballots must be *RECEIVED BY THE TOWN CLERK* by EOD on the date of election, and I believe all are processed then. So even unoffically, there is likely no way to determine pre-/post-IA from official absentee ballot results.
Possible ways to get at this anyway, I suppose:
– As the pre-election opinion/intention polls get more scrutiny (as appears likely), we should be looking for any large polls which not only encountered a substantial number of absentee voters, but also asked them whether and how they did vote. (It’s not an absolute given that this happened.) If available, these should give some insights.
– Crazy idea: Examine specific geographies known to have a large number of military families, whose absentee ballots would have likely been mailed earlier. Compare difference between absentee/in-person results there versus that difference in other geographies.
But look, as you’ve pointed out, NH restricts absentee voters. People have to request an absentee ballot for each election, and have to have a reason other than convenience for doing so. In 2004, ~5% of the voters went absentee. By my reckoning, there’s simply not enough votes there for a pre-/post-IA effect to influence the outcome as dramatically as it would need to in order to explain the poor predictions of the pre-election polls.