Polls, swings, and margins

So I could totally geek out on you and try to argue that polls are unit roots with drift or other fancy time-series buzzwords to explain why you're wrong Joe, but I have forgotten what most of those words mean since it's been two years since taking Prof. Moody's class. The point is that the polls do swing, but they have a stable mean even if the variance is rather high. In the end what matters is that mean. I'd guess (but have no actual proof) that the variance is actually shrinking at this point compared to October.

Wild swings are expected in polls. People respond to some stimuli that day that fades away in about a week (see Zaller's recent work on the decay factor on campaign ads). This means that variance in the polls can be expected. Besides, there's questions about how well-matched the samples always are, what the current events are, who was on the news more that day... in other words, a fair number of trite external events are driving how people respond to a poll. This is what generates that variance.

Polls involve knee-jerk reactions from members as well, whereas voting does not. Voting you sit in a booth, isolated, and can think without the noise from the outside at that moment. Polls demand instantaneous repsonses. There will be more fluctuation in the polls because of the reasons outlined above.

Plus, we're talking a number of different polling houses. Who knows what goes on under the hood in each one with sampling, timing, frequency, etc? Come on Joe -- we both took Rapoport's class, we should know this! (And hell, I took a class that talked about it in the spring, I should definitely know this, especially since I have a test on it in the fall.) Comparing different poll estimates to one another is a futile effort. That's why most serious polling analysts take averages using some algorithm to handle the differences in techniques used for questions, sampling, etc.

In other words, expect Obama to win. Once again I'm going to impersonate Paul Begala and say it's because of the economy. That's what drives people's voting decisions consistently, and should be especially true this year. The incumbent President's party (because people blame the President more than the Congress), will suffer for this.

Oh, and Joe, don't try to cite evidence of people blaming Congress. Those are strong Republicans and will view it that way anyway. What matters, as always, is that 10 to 20 percent of the population that acts truly independently. Those types do not get nearly as much coverage in the popular press (I hypothesize) because they are far less likely to attend political rallies and end up encountering the media types. So what the media shows is biased against giving us a picture of the independents who vote on the economy and vote against the current President's party when things are going badly.

And hey, I'm not in the wrong field. Oddly enough, most of the big academic work on political polls has come out of UNC -- check out one of the faculty member's websites, which tracks the polls and explains them. It was cited as a good source for our office pool because one of our tie-breakers is the popular vote breakdown (except we included Barr). Currently he's got Obama at about 54 percent and has had him there for the past week. And frankly, I'd trust Stimson's means of averaging the polls more than fivethirtyeight or Pollster -- he's been doing it for decades with another faculty member, MacKuen, and neither have messed up too badly. (Also, MacKuen is good friends with Rapoport, so you should read these guys anyway.)