About Those Polls
Posted on June 25th, 2008
by Daniel Larison |
|
If this McCain spin is even close to being right and the L.A. Times/Bloomberg poll had something like 22% Republican party ID, there is something seriously wrong with it and it almost certainly underrepresents McCain’s strength. Republican party ID has declined, but there is no reason to think it has dropped by 14 points since 2006. That would help to account for McCain’s fairly anemic showing in that poll. The Newsweek poll also uses a pool of respondents with just 22% Republican party ID (23% among RVs). So the model of both the LAT and Newsweek polls is questionable, and both seem to be underrepresenting Republicans–hence the large leads for Obama. I have seen several people suggest that the LAT poll might mean that the Newsweek result was more reliable, but both are based on what seems to be a seriously flawed model of the electorate. Polling registered voters is potentially misleading enough as it is, but these results require you to believe that fewer than one in four voters in America identifies as a Republican. While not unimaginable (GOP party ID sank to these levels after Watergate), it is pretty unlikely.
P.S. The surveys that rely on a pool of adults or RVs tend to have the lowest Republican party ID, while polls of LVs tend to have considerably higher ones. The party ID flaw in these polls derives from polling registered voters rather than likely ones. I should add, however, that the Gallup daily tracking poll is also a survey of registered voters, and shows the election tied at 45.
Filed under: politics
There seems to be an awful lot of sample smoothing all around. If you remember, undersampling and oversampling independents threw off some of the estimates for the early primary states. I think the pollsters are having a lot of trouble knowing what a random sample should look like right now.
I recently interviewed a pollster named George Barna, who mostly polls for religious organizations, but came out three weeks ago with a national poll that reached almost exactly the same conclusions as Newsweek’s. He agrees that the crucial distinction is between registered voters and likely voters, but argues that because far fewer McCain voters believe their candidate will win (31%) far fewer of them are likely to go to the polls than Obama voters, who are confident of their candidates chances (53%). You can see the same kind of “enthusiasm gap” between supporters of McCain and Obama in the Times/Bloomberg poll, where 81% of Obama voters are enthusiastic, and only 45% say the same of McCain.
http://www.barna.org/FlexPage.aspx?Page=BarnaUpdate&BarnaUpdateID=300
As for McCain spin, isn’t his campaign also arguing that Obama’s plan to staff campaign offices and spend money in every state is a sign of weakness?
Polls? Call me in October.
I’m convinced that Obama’s true poll numbers are and will always be at least 5% below the technical numbers, and perhaps more. Its simply a fact that many people who have serious doubts about voting for a black man will keep them secret to avoid ridicule. I’m no political strategist, but if the GOP machers don’t have a plan to unleash a flurry of doubt-seeding quasi-racist offensives right before the elction, they ain’t worth their salt. Pathetic though it may be, their only hope, all things remaining equal between now and election day, is to try and facillitate a “Bradley Effect” that will show up in the results but not in the polls.
Then again, It might be a wash, assuming unprecedented turnout among African-Americans and young people.