Being a Republican pollster can bring grim tasks these days. More often than not, you have to deliver some measure of bad news to clients.
The responses of clients given bad news varies widely. Some nod in a knowing way; they expected it. Others get melancholy, while still others steel themselves for the fight of their life. These are reasonable responses to unfavorable polling.
But some clients can take a different tack altogether. These candidates act as if they are surprised by the bad news. “Yes, the Republican brand is damaged,” they acknowledge, “but I’m sure I’m weathering the storm,” they reason. “Everyone says I’m safe.” They exhibit incredible denial of reality and may even turn against the messenger, their loyal pollster. Or, to save their pollster’s credibility somewhat, they speculate, “It must be a bad sample.”
Of course, it could be a bad sample. Sampling theory allows that one in 20 surveys will be what a client once referred to as a “wild hair sample,” one that has a higher margin of error than the 95 percentile figure traditionally cited by pollsters. But it seems like there are more “wild hairs” sprouting these days than could be reasonably expected if sampling theory is really a science.
The doubtful client is also set off when the sample doesn’t seem to have enough Republicans. “A-ha!” they exclaim, “you undersampled Republicans! That’s why I did so poorly.” In states with permanent party registration, this could never occur. We can establish quotas by party registration and ensure that we get the proper number of registrants from each party.
But in those states without permanent registration by party, like my Texas, it can be messy. In these states, pollsters have traditionally relied upon measures of partisanship, such as “Do you think of yourself as a Republican, Democrat or independent?” This sort of partisanship is too soft, we have learned in the past, to use when weighting samples to the proper percentages of partisans. Because voters tend to adjust their partisanship to achieve consistency with their anticipated voting behavior, people who would have told me four years ago, or even two, that they think of themselves as Republicans may now say they are independents, or in a few cases, Democrats, because they plan to vote for some non-Republicans.
We pollsters adjusted our questions to avoid the “floating voter” problem of partisanship in transition. So we started asking about something firmer than their current partisan identification. We ask about past voting behavior. “Regardless of how you plan to vote in the next election, in the past have you mostly voted for Republicans or Democrats or have you split your ticket?” The problem is that voters seem to change their reports of past generic voting behavior, too. So it’s hard to know how to weight your samples using partisanship.
I recently explored this conundrum in a Midwestern state that doesn’t have party registration. Of voters polled statewide who said they have voted “only” or “mostly” for Republicans in the past, 19 percent admitted migrating toward independent status, and 6 percent acknowledged a full-tilt shift to the dark side and voting for Democrats. So should I count these voters, one in four one-time loyal Republicans, as GOP voters or not?
It is undeniable that the Republican “base” is shrinking. If GOP pollsters boost the percentage of Republican identifiers in their poll samples to match what we saw four years ago, there are going to be lots of unhappily surprised clients come Election Day. Republicans must acknowledge that some of our one-time followers are now camped out as independents. If we know where they really are, and how many there are of them, it will be simpler to win them back.
Hill is director of Hill Research Consultants, a Texas-based firm that has polled for GOP candidates and causes since 1988.