During a campaign I am avidly attentive to polls and tactical details. But I don’t enjoy Election Night itself. If I could resist the social draw, I’d make an early night of it, then wait until the next day to see the results.
The basic problem is that returns are slow to come in and often misleading. This low quality can foster worries about vote fraud. Yet careful consideration suggests little support for the idea that significant-scale fraud occurred in any major race – including the Alaska Senate contest…
As I wrote a few days ago, pre-election polls had great predictive power this year, as they did in 2004. This is why I advised you to focus on one small state – New Hampshire. It’s small and relatively homogeneous, and could test whether the overall measured spread was correct. It was.
Erratic reporting of results can feed inaccurate thinking about the dynamics of races. Some people seem inclined to overestimate the likelihood of technology problems, suppression, and fraud in voting. This style of thinking may be primed by the 2000 election, a scarring event for progressives and liberals. It is true that Republican officials like to act in ways that reduce the other side’s vote. But it is a long way from known tactics to large-scale fraud. Existing known tactics are likely to make a difference only when races are within one or two percentage points.
This year, several mini-dramas have been resolved in ways that are consistent with polls.
President – Indiana, Missouri, North Carolina. Polls predicted that these would be toss-ups, as indeed they turned out to be. A fourth state, ND, appeared to be near-tied, but polling was sparse and dominated by partisan polling organizations.
President – Virginia. There was consternation in the evening’s count, as the race appeared far closer than the pre-election average of Obama +4%. But after late votes and absentee votes were counted, the final result of Obama +6% was a good match.
Senate – Alaska. This is an interesting case that is currently attracting comment: Why is Stevens leading, if polls showed Begich ahead? Why does the turnout appear so low? In both cases, I believe there is less to the story than meets the eye.
Polls were volatile, a fact that would escape attention if you were only watching who was ahead. Indeed, I had failed to notice this. Begich led by 1-2% on 10/14-19; 22% on 10/28-30, soon after Stevens’s conviction; and 7% on 11/2.
What should we make of the fact that Stevens leads in votes counted so far? First, early voting is showing the same pattern as everywhere else: a tilt toward Democratic candidates. Second, despite speculations of suspiciously low turnout, Election-Day votes (224,000) and uncounted ballots (90,000) add up to 314,000, similar to the 313,000 votes recorded in the 2004 Presidential race. So there’s no evidence for skulduggery. Stevens currently leads by only 3,000 votes. I predict that the final margin, which should be announced tomorrow, will end up in the range of the last two polls, a Begich win by 2% to 7%.
To return to my previous post: The main message was that polls do quite well. We should always keep the simplest hypothesis in mind: – polls reflect sentiments, and so does voting – after the votes are all counted. Unexpected events can happen, and it’s a good thing to watch for them. But we should not necessarily be on a hair trigger.