Still counting, still close

The other week I tried to demythologise the United States election result a bit, with a promise to return to the topic when counting was complete. It is still not quite finished, but close enough to be able to say some definitive things.

On the latest numbers from AP, Donald Trump has 50.9% of the two-party vote, a lead of just under 2.7 million over Kamala Harris. The uncounted votes come mostly from California, so that gap will narrow a little more (it has already come down from about five million on the night); the final figure will be close to 50.8%, with a lead perhaps just under 2.5 million votes.

That makes this the closest election in terms of the popular vote since 2000’s Gore vs Bush. Trump’s vote is a little worse than that of Hillary Clinton in 2016, who had 51.1%. Taking the raw rather than the two-party vote (as American sources usually do), Trump currently has 50.1% of the vote to Harris’s 48.3%; the late counting will probably bring that fractionally below 50%, which some might take to have psychological significance but which is fundamentally irrelevant.

It was a poor election for the minor parties. Rogue Green Jill Stein placed third with 0.5%, or about three-quarters of a million votes, just ahead of Robert Kennedy – who had withdrawn but was still on the ballot in the majority of states – also on 0.5%. They were followed by Libertarian Chase Oliver on 0.4%, his party’s worst result since 2008. In two states, Michigan and Wisconsin, the third-party total was greater than Trump’s winning margin, but it’s most unlikely that their presence made any net difference.

Trump’s two-party swing of about 3.0% is the largest since 2008 but is still well below the historical average. He carried 31 states plus a single district in Maine for 312 electoral college votes; Harris won 19 states plus the District of Columbia and a single district in Nebraska, giving her 226. That’s exactly the result that the pendulum predicted for a uniform swing of that magnitude, with six states (Arizona, Georgia, Michigan, Nevada, Pennsylvania and Wisconsin) changing hands.

Unlike the last two elections, no state was exceptionally close. The closest was Wisconsin, which Trump won with 50.44% two-party, a margin of about 30,000 votes. Harris’s narrowest win was in New Hampshire, where she had 51.4%. The uniformity of swing was again quite remarkable: every state swung Republican (the first time that has happened since 1952), and all within a narrow band, from Washington at 0.4% to New York at 5.6%.*

I pointed out in my preview how extraordinary the last election was in its uniformity, and it’s happened again. All but eleven states swung within two points either side of the national average (last time it was all but nine), and the range from largest to smallest swing of 5.2 percentage points is, as far as I can tell, the smallest ever recorded.

Nor is there anything much in the way of a regional pattern; swings were a bit smaller in the midwest and bigger in the north, but the difference is hardly noticeable. Swings were, however, generally bigger in Democrat states. Four of the five largest (Florida is the exception) were in safe Democrat states – although one of those four, New Jersey, is no longer safe, its margin having been cut by five points to 2.9%.

Apart from this strange ossification of voting behavior, what other lessons can we learn from these numbers? Three things seem worth mentioning.

First, the bias in the electoral college has disappeared, at least for now. In 2028, the Democrats will need a uniform swing of 0.9% to win back the presidency, almost exactly what they will need to win a majority of the two-party vote. In other words, the result in the median state, Pennsylvania, was about the same as the result in the country as a whole, in contrast to 2020 when it was about two points better for the Republicans.

That’s because Harris performed somewhat better than average in the marginal states (except for Arizona). As I’ve said a number of times, while there is a small structural bias in the electoral college towards the Republicans (because of the over-representation of smaller states), most of its effect is random: a party that loses out one year may gain the advantage four years later.

Secondly, demographics didn’t really matter. A lot of ink has been spilled in trying to pinpoint just where Harris lost the election, but the truth is she lost it across the board. As always, there are differences in the candidates’ performance among particular demographic groups, but they’re minor; a swing as uniform as this one means that national issues predominated. The Democrats can stop flagellating themselves as to how they should have done a better job with this or that specific group – what they need is more votes nationwide.

Finally, nobody knew that this would happen. Many pundits predicted the result more or less correctly, but any who were certain about it were wrong. A result that close is never certain; it could easily have gone the other way. In addition to learning to be more numerate, a lot of commentators need to learn some humility.

.

* Some of those numbers might change slightly with late counting, but not so as to make any substantial difference.