Election preview: United States

The United States goes to the polls tomorrow morning to choose its next president, with the contest between vice-president Kamala Harris and former incumbent Donald Trump too close to call.

Results will come in during the day tomorrow (Australian time), but they will do so in a way that by the standards of other democracies is quite unusual. Most votes are irrelevant, because most states are safe for one party or the other. Only a few marginal states really matter, and unless there is a major upset none of them will be able to be called (as the safe states will) on the basis of exit polls alone.

Because election administration is so decentralised, actual counting happens in different ways in different states: some start counting postal and pre-poll votes before the polls close and therefore have a large chunk of results to report straight away, while others count the on-the-day votes first. And since Republicans have demonised the idea of postal voting (a stance they have recently tried to walk back from), postal votes – which Americans confusingly refer to as absentee votes – are often quite different in complexion.

FiveThirtyEight has a comprehensive guide to when results can be expected; it’s worth careful study (its times are all given as US Eastern; subtract eight hours and add a day to get eastern Australian summer time, so 10pm Tuesday in New York is 2pm Wednesday in Melbourne). For more about what to watch for, you can read my preview of the 2020 election here, and my four-week look at this one here.

There will most probably be only six states that you need to worry about: Arizona, Georgia, Michigan, North Carolina, Pennsylvania and Wisconsin. Whichever candidate wins the majority of those will, with a high degree of certainty, win the election. In the event that they split three-all – which can’t be ruled out, although (for reasons that we’ll get to shortly) it’s not as likely as the polls would have you think – Nevada may matter as well, although in that case Harris (but not Trump) has winning chances without it.

Of those seven, North Carolina will probably be the first to report meaningful results, which in Melbourne will be from about 11.30am. Georgia and Pennsylvania should follow from about midday, Michigan and Wisconsin from about 1pm, Arizona 2pm and Nevada more like 4pm. But varying numbers of votes will be involved, and in most cases it will take a few hours to get a clear picture; longer if it’s very close.

Other states may be worth watching too, partly because an upset is always possible (a recent poll putting Harris ahead in Iowa has attracted a lot of attention) but also because swings tend to be pretty consistent. If Florida, for example (a state that reports quickly), shows Trump struggling, that would be a good sign for Harris, even on the assumption that he still wins it. Conversely if Harris is having trouble in a state like Virginia.

If the polls are right, the swings – at least in the marginal states – will be very uniform indeed. FiveThirtyEight’s aggregator, for example, shows margins in the six ranging from 0.5% for Harris in Michigan and Wisconsin to 1.0% for Trump in Arizona; in 2020 they went from 1.4% for Biden in Michigan to 0.7% for Trump in North Carolina.* That’s an extraordinary degree of consistency. Moreover, there’s been very little variation between pollsters as well, either in state or national polls.

That’s led to a widespread view that pollsters have been “herding”: either massaging their data or selectively releasing results so as to stay close to the consensus. Here’s Nate Silver on the subject; here’s Australia’s own Peter Brent. Both of them point out that the existence of herding increases the chance of error but doesn’t tell you in which direction the error is likely to go. They also suggest, though, that there is some reason to think the polls are more likely to be understating Harris’s chances than overstating them.

But it’s also possible that the conclusion the polls are herding around is in fact correct, and that this is just an election with extremely small and extremely uniform swings. Certainly America has been going that way for a while. Back in the 1960s and ’70s, swings varied quite a bit; typically about two-thirds of the states would swing by more than two points either side of the national average.

Then things began to change. In 2000, the majority of states swung only within the ±2% band, for only the second time in a century. Swings got smaller still, and by 2020, with three successive elections having moved by less than 2% nationwide, an amazing 41 of the 50 states were in that same narrow band. Among the marginals, only Florida (for Trump) and New Hampshire (for Joe Biden) produced more substantial swings, and they were still modest.

No-one knows exactly why this is, but it’s clearly self-reinforcing and it highlights even more the absurdity of the electoral college. When only a few states are seen to be in play, parties increasingly concentrate their resources there; no-one makes an effort to reel in a long-shot state, so voters there don’t shift much, and campaigning in the marginals is so intense and so targeted that it attracts diminishing returns.

Perhaps, with so much at stake, tomorrow will produce some surprises and show that there is still life in the republic. If not, a deeply and evenly divided country will face a grim future, even if a Harris win should ward off immediate disaster.

.

* I express those margins in the Australian fashion, as the two-party swing required for a change. American usage takes the gap between the two sides, which is double the required swing.