Custom Search

Tuesday, September 25, 2012

Pollster Defends Oversampling Democrats, But Argument Proves Conservative's Points

By Susan Duclos

Yesterday we saw two Democratic pollsters criticizing polling organiszations for oversampling Democrats in their polls. One called it "utter irresponsibility" and explains "the effect, is it to build an artificial picture, an illusion, of great Obama momentum.”

Schoen pointed out that the Pew poll was based on Democrats sampled for having an 11 percent voters registration edge over Republicans. He further added, “saying that America has gotten more Democratic than 2008, which is a questionable assumption.”

Today, the National Journal talks to Doug Schwartz, the director of the Quinnipiac University Polling Institute, who conducts polls in battleground states for CBS News and The New York Times, as he defends the oversampling of Democrats in the polls.

"A good example for why pollsters shouldn't weight by party ID is if you look at the 2008 presidential election and compared it to the 2004 presidential election, there was a 7-point change in the party ID gap," Schwartz said. Democrats and Republicans represented equal portions of the 2004 electorate, according to exit polls. But, in 2008, the percentage of the electorate identifying as Democrats increased by 2 percentage points, to 39 percent, while Republicans dropped 5 points, to 32 percent.

Unless I am missing something or haven't had enough coffee, (Quite possible since my brain doesn't function without caffeine) he claims that is an example of  why pollsters should not weight by party ID, yet considering that in the 2008 presidential election, Obama beat John McCain by 7 percentage points, (53% to 46 percent -  CNN), it seems to indicate strongly that pollsters should be using correct party ID in their sampling.

 Mr. Schwartz appears to have made our point for us.

Morrissey over at Hot Air, who routinely looks into the sampling of polls and has for quite some time now, makes a couple of points:

Polls are intended to be predictive.  In order to be predictive, the sample has to hew closely to the turnout model of the actual election. The best way to calculate that is to check the trends from the most recent election cycles.  One can get surprised by this when turnout shifts dramatically, as it did in 2008 — but that was in favor of the Democrats for a D+7 result, and it’s unlikely to happen a second time, especially after the all-even turnout model from the 2010 midterms.  That means that D+11 on national samples aren’t going to be terribly predictive of the outcome in November, nor would R+11 samples, and so it’s difficult to take those results seriously.  Furthermore, with just a few weeks before the election, pollsters need to start finding likely voters rather than just registered voters or general-population samples if they expect consumers to rely on them for predictions of voter behavior — again, the entire point of polling.

If there is one valid criticism of conservative poll analysis, it’s that we tend to focus on just party ID rather than a broader range of demographic categories — gender, age, income, and geography.  Most of that data exists in exit polling, too, so it isn’t terribly difficult to check, but it is time consuming.  However, if a national poll features a turnout model of D+11 or R+11, that’s enough to make the results unreliable without checking the rest of the demos, just as one with 60% men would be.

Rasmussen's most recent partisan breakdown shows that 37.6 percent of Americans consider themselves Republicans, 33.3 percent Democrats, and 29.2 percent Independents.

According to the above linked National Journal article,  Lee Miringoff, director of the Marist College Institute for Public Opinion in Poughkeepsie, N.Y., admitted to National Journal that he is now taking note of the party-identification results in the polls he is conducting.

"I look at our party-ID spread because I want to anticipate the reaction," he said. He added: "I guess it makes for good pundit sport at this point."

Jim Geraghty at NRO takes on the media's defense about the oversampling of Democrats in the polling:

First, the basics: Most Democrats are going to vote for the Democratic candidate, most Republicans are going to vote for the Republican candidate, and the independents are usually going to split somewhat evenly. So the proportion of the three groups more or less determines which candidate the polls are going to show ahead.

No one claims to be able to predict, with absolute certainty, what the partisan makeup is going to be on Election Day. But we do have a range from recent history — from even in 2004 and a seven percentage point advantage for Democrats in 2008. If a pollster believes that the electorate will be even more heavily Democratic in 2012 than it was in 2008, I’m willing to hear those arguments, but I think it’s a tough case to make......

Then he nails the media narrative of misrepresenting what Conservatives are asking of pollsters:

Yet our objection keeps getting misstated and twisted by poll defenders time and again. Here’s Chris Cillizza, claiming the complaints are “a series of false assumptions none bigger than that because the country has been virtually evenly divided on partisan lines for the past decade or so that the party identification question should result in something close to a 50-50 split between Democrats and Republicans.”

No. I do not demand an even split, nor do any of the other folks paying attention to this factor. I do think that a split of D+7 or more is excessive, and that a D+3 or D+4 divide — halfway between the GOP peaks of 2004 and 2010, and the Democrat peak of 2008 — is more likely. But more importantly, if a pollster is going to assert that the electorate is going to be more heavily Democratic this cycle than it was the previous historical high, they ought to explain why they think this is the case.

Because for much of this year, in quite a few national polls, we’ve seen Romney winning almost all the Republicans and hold a lead among independents, and still trail Obama, because Obama is winning almost all the Democrats, and Democrats make up such a large share of the sample.
Possible? Sure, anything’s possible. But if a pollster is going to offer a hypothetical electorate that looks different from everything we’ve seen before, I’d like to know why they think this is the case.

 Emphasis mine.

Gallup in August of 2012, found "44% of all U.S. adults have identified as or lean Democratic and 40% are Republican."

Polls should also reflect the drop in Democratic registration in eight battleground states which was nearly 800,000, the drop in Republican registration which was one tenth of that and came in at 79,000 and the rise of Independent registration which came in at nearly half a million. (Source- Politico)