Archive for November 4, 2020

US Election Night and Day

Posted in Bad Statistics, Biographical, Politics on November 4, 2020 by telescoper

Before you ask, no I didn’t stay up all night for the US presidential election results. I went to bed at 11pm and woke up as usual at 7am when my radio came on. I had a good night’s sleep. It’s not that I was confident of the outcome – I didn’t share the optimism of many of my friends that a Democrat landslide was imminent – it’s just that I’ve learnt not to get stressed by things that are out of my control.

On the other hand, my mood on waking to discover that the election was favouring the incumbent Orange Buffoon is accurately summed up by this image:

Regardless of who wins, I find it shocking that so many are prepared to vote for Trump a second time. There might have been an excuse first time around that they didn’t know quite how bad he was. Now they do, and there are still 65 million people (and counting) willing to vote for him. That’s frightening.

As I write (at 4pm on November 3rd) it still isn’t clear who will be the next President, but the odds have shortened dramatically on Joe Biden (currently around 1/5) having been short on Donald Trump when the early results came in; Trump’s odds have now drifted out between 3/1 and 4/1. Biden is now clearly favourite, but favourites don’t always win.

What has changed dramatically during the course of the day has been the enormous impact of mail-in and early voting results in key states. In Wisconsin these votes turned around a losing count for Biden into an almost certain victory by being >70% in his favour. A similar thing looks likely to happen in Michigan too. Assuming he wins Wisconsin, Joe Biden needs just two of Michigan, Nevada, Pennsylvania and Georgia to reach the minimum of 270 electoral college votes needed to win the election. He is ahead in two – Michigan and Nevada.

This is by no means certain – the vote in each of these states is very close and they could even all go to Trump. What does seem likely is that Biden will win the popular vote quite comfortably and may even get over 50%. That raises the issue again of why America doesn’t just count the votes and decide on the basis of a simple majority, rather than on the silly electoral college system, but that’s been an open question for years. Trump won on a minority vote last time, against Hillary Clinton, as did Bush in 2000.

It’s also notable that this election has once again seeing the pollsters confounded. Most were predicted a comfortable Biden victory. Part of the problem is the national polls lack sufficient numbers in the swing states to be useful, but even the overall voting tally seems set to be much closer than the ~8% margin in many polls.

Obviously there is a systematic problem of some sort. Perhaps it’s to do with sample selection. Perhaps it’s because Trump supporters are less likely to answer opinion poll questions honestly. Perhaps its due to systematic suppression of the vote in pro-Democrat areas. There are potentially many more explanations, but the main point is that when polls have a systematic bias like this, you can’t treat the polling error statistically as a quantity that varies from positive to negative independently from one state to another, as some of the pundits do, because it is replicated across all States.

As I mentioned in a post last week, I placed a comfort bet on Trump of €50 at 9/5. He might still win but if he doesn’t this is one occasion on which I’d be happy to lose money.

P.S. The US elections often make me think about how many of the States I have actually visited. The answer is (mostly not for very long): Kansas, South Dakota, Colorado, Iowa, Missouri, Arkansas, Louisiana, California, Arizona, New York, New Jersey, Maryland, Massachusetts, New Hampshire, Maine, and Pennsylvania. That’s way less than a majority. I’ve also been to Washington DC but that’s not a State..