Writings and observations

(Less than) 3 months to go


One of the online places political junkies get their fix – it’s really hard to stay away for long – is the fivethirtyeight.com web site and especially its election forecast section.

There are other polling analysis sites around the web, but 538, led by the remarkable statistician Nate Silver, is the most sophisticated. Most prominently it has a section showing, based on current information, how the candidates for president are doing. It updates the information whenever a new data point becomes available, which may be several times in a day, or even several times in an hour. Every time I check back in, it seems to have changed. And there’s more: The site offers three rounds of current estimates, the “now cast,” which estimates who probably would win and by how much if the election were held now; the “poll only,” which analyzes polls and nothing else; and the “polls-plus,” which adds in economic, historic and other factors.

As I write this, 538 estimates Democrat Hillary Clinton has an 86.3% chance of beating Republican Donald Trump, according to “polls only.” The number will change, up or down, by the time you read this.

538 also breaks down the probability estimates by state. As I write this, the odds Trump will win Idaho have been calculated – polls-only – at 96.3%. It is the third highest probability of a Trump win in the country, behind only West Virginia and (in first place) Oklahoma (at 98.7%). The polls-plus probability of a Trump win in in Idaho in November hit 99.1%, which is almost as close to a certainty as 538 gets, while the now-cast (if the election were held today) is at 98.2%. The now-cast estimates that in Idaho, Trump would get 54.6% of the vote, Clinton 35.7% and Libertarian Gary Johnson 7.8%.

You can see a consistent pattern here.

Some states, especially many of the battlegrounds, are polled frequently, but Idaho isn’t, which creates an obstacle for analysts like 538. They’re relying in large part on three polls from Dan Jones & Associates.

Polling analysts put a lot of attention into not so much the snapshots that individual polls can generate, but the trend lines – are numbers rising or falling over time – and comparisons between pollsters, when those are available. In Idaho, those numbers have been mostly stable all year.

Idaho’s neighboring states have been a little more variable, swinging around significantly during July (the month of conventions) in blue Washington and Oregon, red Utah and Montana (though not red Wyoming, which stayed stable) and purple Nevada. In the first couple of weeks of August, however, all have begun to settle into familiar patterns.

The most interesting of the neighbors – in the possibility it might break from familiar patterns – is Utah. Utah actually has been polled with some regularity this year, and by several pollsters. Trump is given an 80% probability (polls only) of winning it, but that’s far less than Idaho or Wyoming. At 80% probability, you have an operating assumption that Trump will take the state, but the chance of an upset is not completely off the charts. Put another way, the chance Trump may lose Utah is greater than the chance that he wins the November election. If he did lose Utah, might that affect the Idaho percentages in reflection of how the large LDS vote might turn?

Utah is one of several western states of interest, in having polling numbers that force both parties to keep a wary eye on them. Nevada and Arizona are near-battlegrounds, Colorado is in the gray area for battleground status, and party activists might be wise to keep an eye on Montana, where Trump has a probability of winning now sitting at 76%, which is less than secure.

For the time being, though, after all the post-convention talk about changes in the races, Idaho still looks pretty well locked down.

Share on Facebook