Insiders: Weight by Party ID
A bi-partisan consensus of political Insiders say pollsters should weight their surveys by Party ID instead of adjusting their sample only by demographics like race and gender, according to the latest National Journal Political Insiders Poll.
Sixty-four percent of Democratic and Republican Insiders, including majorities in both parties, said polls should weight by party ID. Republicans on the whole felt stronger about siding with party identification, with 76 percent of them answering that way. Just over half, 51 percent, of Democrats agreed.
The controversy over whether polls should account for party ID has raged for weeks, particularly among conservatives who thought many surveys of the presidential race sampled too many Democrats. Many of those polls reported President Obama held a comfortable lead over Mitt Romney, but in an electorate that also reported included more Democrats than in recent presidential elections.
Pollsters countered that party identification is an attitude, much like their choice for president, and should be measured. Artificially weighting a poll to correlate with party ID, they contend, would risk missing important shifts in the mood of voters.
Many Insiders, however, disagreed.
Should pollsters weight their surveys by party identification, or should they measure only by adjusting their sample to race, gender, and other demographics and allow for changes in party-ID attitudes?
|Weight by party identification||51%||76%|
|Sample only by demographics||49%||24%|
"Have been doing this for 20 years," said one Republican. "You have to make certain assumptions based on historical and likely voters. If you make the right assumption, far more accurate than just demographic sampling."
Added another GOP operative, "Party ID accounts for more than two-thirds of votes cast. If you get that wrong, then you get the ballot wrong. Simple as that."
That Republican Insiders were more likely to support party sampling than their Democratic counterparts isn't a surprise: Their complaints about party ID in polls have been far louder this election. But many Democrats agreed with the criticism.
"Allowing party ID to wildly fluctuate is as silly as thinking that party ID actually fluctuates that wildly," according to one Democrat. "It doesn't."
Still, some Insiders, including a few Republicans, advocated for paying closer attention to a survey's demographic breakdown.
"Party ID is as fluid as candidate preference," said a Democratic Insider. "We have repeatedly seen the same individuals identify as both Democrats and Republicans over the course of an election. Re-weighting by party ID makes no more sense than re-weighting by candidate preference."
Added a Republican, "Party ID is a fluid measurement. Let the market speak and be transparent about results."
Some Insiders also believed that a middle ground existed between both perspectives.
"I'm a pollster, and the true answer lies somewhat in the middle," said one Democrat. "But given the alternative, more emphasis ought to be put into demographics, and party ID should be adjusted only when it seems to have deviated too much from a previous norm."
Added another Democrat: "An ideal poll weights by demographics as party ID often changes based on enthusiasm. Sampling techniques are more sophisticated than simply weighting however, as they need to be concerned with cell phones, who is likely to be home on particular days, and screening bias."