Misconceptions: “The Election is a Dead Heat”

by Benjamin Studebaker

I’m not one to get mired in election cycle coverage often, but then I saw “news” stories like this one claiming that the presidential election is actually close. This is not true. Barack Obama is almost certainly going to win this election, and his margin in the electoral college is probably going to be more than a couple of states large. Here’s why.

There are two key errors that are being made by the people telling you that the election is close:

  1. They are looking at national polls rather than state polls.
  2. They are looking at individual polls rather than averages of large numbers of polls.

Now, before we dissect this further, please be aware that I am not saying this out of political loyalty or bias. I’m the sophiarchist–I’m suspicious of the entire system’s ability to get things done. Full disclosure, I do not like Romney. I think he has misled the public, that his tax plan doesn’t really work, and so on. Search his name on the blog if you want to read about that stuff; that’s not my purpose here today. Furthermore, I do not think that Obama’s re-election is going to be particularly helpful or provide any panacea. I expect that the next four years will be like the last two–divided government and gridlock. When I say that Obama is going to win, I say it not out of any intense desire to see Obama win, but because there is empirical evidence to suggest that it is actually going to happen.

That said, let’s get to why this will happen and why people are telling you different.

Let’s start with the first mistake. In some national polls the election is very close; Romney even sometimes has a slight lead. However, the election is not decided at the national level. Imagine, for example, the following scenario. In Alabama, Romney leads 60% to 40%. In Ohio, Obama leads 52% to 48%. In this scenario, Romney gets more votes nationwide, but Obama wins 18 electoral votes to Romney’s 9. In short, Romney has big leads in the south and the heartland. Where Romney is ahead, he polls in the high fifties, the sixties, even 70% in Utah. In contrast, Obama often has little five or six point leads in states like Oregon, Minnesota, Wisconsin, and so on. Obama has maintained those leads for weeks and months on end–they’re solid, and Obama will win those states, but he won’t win them by a huge majority, and, as a result, sometimes states that have shown a consistent small bent toward Obama get labelled as swing states.

The second mistake is to look at individual polls rather than poll averages. There are dozens of polls taken each week, particularly in the swing states. What many commentators will do is they will take single, individual polls that contradict the majority and hold these up as evidence of a close race. Consider Ohio. Of the last twenty polls listed for Ohio, 17 give Obama a small lead, 2 give Romney a small lead, an 1 shows a tie. The logical fallacy that gets made here is that the 2 Romney lead polls indicate that the 17 Obama polls could be flawed rather than the reverse. The more polls you have saying a given thing, the less likely that all of these polls share a bias of methodological flaw.

For this reason, election modellers who look at the individual state polls and use poll averages rather than isolated examples come down with Obama winning around 300 electoral votes–a victory with multiple states worth of cushion. The one that has most recently been in the news is Nate Silver, whose model takes all the polls, crunches them together, and then spits out a percent chance of victory based upon previous election history concerning what kind of poll lead constitutes a safe poll lead. Silver’s model predicts an average 306.9 to 231.1 Obama victory as of the time of writing. Here are his poll averages:

Based on the poll averages, of the commonly labelled swing states, the model predicts that Obama wins Nevada with 89% certainty, Colorado with 68%, Wisconsin with 83%, Wisconsin with 94%, Ohio with 85%, Pennsylvania with 97%, New Hampshire with 79%, and Virginia with 71%. Romney is predicted to get North Carolina with 78% and Florida with 53%. Bear in mind in case you missed it, these are probabilities of winning the states, not the percentages of the votes that the candidates are going to come away with.

Now, let’s imagine that Silver’s model is biased (it isn’t, unless all or most of the polls are, and that’s a pretty big conspiracy) and say that any state it predicts Obama wins with less than 80% certainty actually goes to Romney. Even if Romney wins all of those states, Obama still wins the election 277 to 261. Silver’s overall percent chance of an Obama victory? 85%. That’s not certainty, but it’s a long way from a dead heat.

I’m sure many people, maybe some of you reading this, do not like this outcome. That’s completely understandable, if you’re a Romney supporter. If your preferred political outcome is unlikely to transpire, you’re understandably upset and angry. I get that. However, being upset about what is happening does not mean that what is happening is not happening. Accusing Silver and all the pollsters of bias, as some people on the right have begun to do, amounts to self-delusion, because in order for Silver and other modellers to be wrong, not only must some of their polls be biased, most of their polls must be biased; they’re using averages. Historically, this has not been the case–poll averages have predicted fairly accurately the outcomes of past elections. So while I certainly do not expect everyone to be happy with this information, that does not justify irrationally undermining the information with baseless accusations. An unpleasant fact is still a fact.

So do not be surprised if Obama wins a comfortable 300 electoral votes or so and we all go to bed early.