2020 US Election Night — A Guide for Brits
Should I stay up on Tuesday night?
Despite the importance of this election, this may be a year to consider skipping the excitement of election night. As always, the earliest the polls close will be 7PM ET, which is midnight our time. However, only a handful of states are important in deciding the results of the election. When you filter out the dead races, you’ve got:
Midnight: Georgia
12.30am: Ohio
1.00am: Florida, Pennsylvania
2.00am: Arizona, Michigan, Minnesota, Texas and Wisconsin.
Just because the polls close at these times does not necessarily mean that we’ll be getting winners. Even in a normal year, results can take quite a while to come though. The last 3 elections have been called at the following times (2016–7.47am; 2012–4.18am; 2008–4am). However, this is not a normal year, there have an been unprecedented 92 million early votes (as of Sunday), and it’s unknown how this will affect counting. Pennsylvania, which is most likely to be the tipping point state, has said counting there may take place until Friday because they can’t start counting until 7am on election morning.
In short, a better strategy may be getting up early and watching the madness with some waffles.
Can Trump win?
Yes, obviously Trump can win. However, I’ve been quite surprised for the last 2–3 years that people in my social circles have been quite bullish on Trump’s prospects — “he’s going to win again, isn’t he?” when there haven’t really been any data to support it. Before 2020, most betting markets had the race at a coinflip between an unnamed Democrat and Trump and Biden has been repeatedly ahead by a sizable margin in the polls since being named the nominee. Moreover, the models have Biden at 90%+ to win right now (more on these later) and the betting markets have him at about 66%. I suspect that the confidence in Trump may have more to do with managing expectations than being grounded by data.
What about those pesky shy Trump voters?!
Shy Trump voters do not exist. For that matter, shy Tories don’t exist either. While there is long-standing academic study into how people may answer some questions dishonestly — social desirability bias — there is little reason to think that this phenomenon extends to mainstream politics. I have a feeling that this is a liberal delusion — we can think of nothing worse than a Trump or Tory voter and therefore assume that others must be embarrassed about it. In reality, people are usually pretty forthcoming about their political views and voting intentions. Whether in the context of the US or the UK, this theory has been largely discredited.
What if the polls are wrong like they were in 2016?
It’s entirely possible that the polls are wrong in some other systematic way. However, it’s worth noting that the polls weren’t actually that far wrong in 2016. The average of the national polls predicted that Clinton would win by 3.9% in 2016 and she ended up winning by 2.1%. As I’m sure you all remember, the story of this election was Trump doing unexpectedly well in a small number of states. The same thing could absolutely happen again, but given the 8.5% polling lead that Biden currently has nationally, they’d have to be wrong on a much greater scale. Moreover, Biden is quite far ahead in a number of the key states that Trump unexpectedly won in 2016.
The models have Biden at 90%+ and the markets only 66%?!
For me, this is where things get interesting. Full disclosure: I’ve emptied the clip on Biden because I think there’s a good spot of value here.
Individuals like Nate Silver create probabilistic models based on aggregating all the polls, as well as also taking into account other factors like turnout, the economy etc. Silver rose to prominence after “correctly” calling 49 out of the 50 states in the 2008 election. Since then, his data journalism company fivethirtyeight and become the leading player in this field. However, some have criticised Silver’s models — most notably Nicholas Nassim Taleb — for misunderstanding how to deal with different types of uncertainty. Some uncertainty can be contextualised within a system of rules, whereas for other types, the rules themselves are uncertain. Taleb argues that elections are largely the latter and is also displeased the Silver does not have “skin in the game” by hedging his bets with probabilistic forecasts. You read a really nice summary of this debate here or read the full Taleb paper here (I fully 100% understand the latter and do not skip over the mathsy bits).
So we may have a reason to think that the markets are underestimating uncertainty, but there’s another factor at play here too. Betting markets are reactive to… well… betting. A number of bookies are saying that a huge chunk of the money is coming in on Trump which naturally pushes the market to be closer to even money. This could be due to some clever gamblers exploiting overconfident idiots like me, or it could just be a lot of punters with no real clue inflating the market (See: Mayweather vs McGregor).
So the question is, which is more likely to be systemically flawed. It’s entirely possible they both are and the whole thing is so uncertain that we can’t make a good estimate. However, I’m a degenerate so I’m going with the models.
I’m watching on election night and there are some serious weirdly unexpected results coming in
One way in which the US differs from the UK is that counties report individually as they get results. This is compared to the UK, which waits for a whole constituency to get their results. This might mean that you see something like: “Kentucky has reported 21% of its counties and Biden is winning by 10%”. Don’t let this fool you, wait until whichever channel you’re watching on officially “projects” the race. Anything before that is coloured bubbles. The channels decision desks will still do this based on a projection, but they’ll wait until they’re 99%+ confident and have a very good track record.
