Wednesday, June 13, 2012


Generally, Rasmussen is a decent pollster. Nate Silver says Rasmussen has a Republican lean but that is largely because of the likely voter model they use. So long as they are somewhat consistent in what they do and open about their methodology and assumptions, we can assume their results tell us something meaningful.

Today, Rasmussen released a poll of Wisconsin that had Romney up by 3. Those who read here know that is a result that raises my eyebrows so I wanted to take a closer look. I am not a Rasmussen subscriber so I tried to read between the lines of the press release and I did find a few things that jumped out at me.

First, the press release says this:
Though a plurality (46%) of Wisconsin voters recognizes that the government lost money in the bailouts of General Motors and Chrysler, 45% consider the auto industry bailouts a success. Thirty-six percent (36%) regard the bailouts as a failure. However, when told that the government is expected to lose billions of dollars on those bailouts, just 33% consider them a success, and 56% see them as a failure.
Now, there are a couple of issues here. One is that it really isn't quite a sure thing that the federal government has lost money on the auto bailout. To really measure that, you'd have to consider what the counter-factual would be. What would revenues look like if the bailout hadn't happened for instance? There's also another problem with this claim. The federal government still owns 500 million shares of GM. If the government sold them all today, the government would have lost close to $15 billion. But we're not selling today. Not only that, the price of the stock is artificially depressed because voters know Treasury will need to exit. The bottom line is we don't know when Treasury will exit and at what price. So it is not quite right to say this money is lost quite yet.

But there's a potentially bigger problem here. I don't know where in the poll this question was asked relative to the topline question. If the pollster points out that the auto bailout has lost billions of dollars and then asks whether the respondent supports Obama or Romney, that's really problematic.

A week ago, voters in the recall election were asked (in exit polls) whether they support Obama or Romney and Obama came out ahead 51-44. The turnout in November is likely to be a little better for Obama than it was last week. So, either there has been about a 10-13 point shift in Wisconsin in a week or the exit polls were very wrong or the Rasmussen poll is very wrong or some combination of these. The least likely of these explanations is a 10-13 point shift in the last week. I've provided some reasons the Rasmussen poll may be suspect above. That's probably the best explanation but we'll see what other pollsters find in the coming days.

No comments: