Be wary of late New Hampshire polls
View the New Hampshire polls, you see in the final hours before the voting occurs with a large grain of salt.
It’s not because the pollsters doing the surveys aren’t good at their jobs.
It doesn’t mean they won’t turn out to be correct.
But the very short window between the Iowa caucuses and the New Hampshire primary makes it almost impossible to do the kind of quality polling that professionals would like, if given their druthers.
Iowa voted on Thursday, and New Hampshire goes to the polls Tuesday.
Historically, the Iowa results have had an effect—sometimes major, sometimes less so—on the New Hampshire returns. Generally candidates who do well in Iowa see their fortunes improve in New Hampshire and vice versa.
But that was when there was a minimum of eight days between the two contests.
This time, the interval is half that. For practical purposes, the time for polling will be very narrow, at most three nights.
Polls that came out Sunday evening or in the Monday morning newspapers reflect at most three days of polling. Those that come out Tuesday morning—the day of the actual voting—could reflect four full nights.
Pollsters like to have larger periods to poll. Part of the discipline of the field is that once a random sample is drawn, good pollsters make every effort to call back the telephone numbers that did not answer, rather than call extra ones, in order to preserve the randomness and integrity of the original sample.
Fewer nights of calling means fewer opportunities to get the people who were not home the first or second times they were called.
Because it would be worthless to poll before the Iowa results were available, Friday was the first day pollsters could be in the field.
Moreover, historically Friday is the least efficient evening to poll. Overall, fewer people are at home, and certain demographic groups are far less likely to be available to answer their telephones. Sunday, however, is the most efficient night to poll.
All polls are snapshots in time—glimpses of public opinion during a specific period. The Iowa results will inevitably change New Hampshire voters’ perceptions of the race and perhaps some of their votes. How much information the public has when it is polled matters greatly.
For instance, it is clearly true that a New Hampshire voter who is polled on Sunday night will have more current information about what happened in Iowa and its effect on the race than if that same voter is polled on Friday.
Yet, reality dictates that those polled on Friday will be roughly half of a two-day sample and a third of a three-day sample.
That could be especially crucial to Massachusetts Gov. Mitt Romney’s poll numbers. His loss in Iowa will almost certainly help shape his fate in New Hampshire. Before Iowa voted, Romney had already lost the lead he had held for months in the New Hampshire polls to Sen. John McCain. History suggests that a disappointing Romney showing in Iowa makes his candidacy less attractive to New Hampshire voters.
On the Democratic side, of course, the same is true. History suggests Sen. Barack Obama’s victory over Sen. Hillary Clinton and former Sen. John Edwards will likely benefit him in New Hampshire.
In any case, how voters who had been supporting candidates who fared poorly in Iowa is the great unknown in New Hampshire. And, this might not become clear until the actual voting on Tuesday—too late to be picked up by the polls.
None of this is to suggest that the poll numbers that begin coming out in New Hampshire in the final day before the primary will necessarily be inaccurate.
But the compressed time period for such surveys should make everyone a bit wary.