Prognosticators lose the plot in 2024

With a close election up and down the ticket, prediction isn't what matters

Last week my article (gift link) in the Washington Post showed how individuals can use opinion polls to become more effective. Soon after, Ezra Klein wrote about ignoring polls. On Bluesky, Jamelle Bouie points out that there’s no reliable evidence that the shape of the Presidential race has changed.

I have two thoughts today: (1) focus on the whole picture, up and down ticket; and (2) convert your news consumption into action, whether it be canvassing or donation. For example, your activism can affect multiple levels of government in Pennsylvania, Wisconsin, Ohio, Nebraska, and Arizona.

Prediction narcissists

A few days ago, Rick Perlstein committed an epic takedown of pollsters and what they claim to do, namely measure opinion and likely future election outcomes. Two pundits claiming to be able to predict the Presidential election, Nate Silver and Allan Lichtman, have formed a little mutual firing squad. They squabble while ignoring what Harris and Trump might do as President. And the press follows (for a start on what the press could cover instead, see Margaret Sullivan).

For years Allan Lichtman has claimed to predict the outcome of presidential elections using 13 "keys" ranging from economic growth to intangibles such as candidate charisma. He claims a perfect track record. However, he has committed statistical malpractice by coming up with 13 factors that can predict 11 elections. 13 is more than 11, leading to a procedure known as overfitting which will always retrospectively produce a positive track record. Even at this level, his claim holds up poorly. His method purports to predict not the electoral vote, but the popular vote. In 2016, he predicted that Donald Trump would win the popular vote. So in his own terms, that was the wrong prediction.

And then there are the polling analysts. I do not have much to say, except that I see their activity as being excessively oriented to detail. I follow the Princeton Election Consortium tracker, which uses the FiveThirtyEight data feed to produce statistically minimalist snapshots of Presidential, Senate, and House races. And as I argued last week in the Post, at least polling data shows where to put resources. Lichtman can’t do that.

Out of touch - but you don’t have to be

Neither of these gentlemen looks particularly attuned to the moment. Political prediction might have been fun in 2008 or 2012, but in 2024 such activity is a sideshow. The news media is already having trouble focusing on the stakes of the election.

It is worth putting time and resources into any state where a race appears to be within 5 points or less. If what you care about is the Presidency, the places to focus are, from west to east, Arizona, Nevada, the Nebraska Second Congressional District, Wisconsin, Michigan, Pennsylvania, Georgia, and North Carolina (data here). Volunteers would be well advised to go to all of those places – though in the case of North Carolina, disaster relief is still urgent.

The same logic applies to Senate, House, and even legislative races. Wherever the outcome is close, moving a few votes can make a difference. Knowing this kind of information can give you impact well beyond the vote you cast in the place where you live. Use the nonpartisan resource Vote Maximizer to see where you can canvass and donate in an effective manner