Thursday, 16 July 2020

270 To Win

The always excellent The Economist has an article on the US Presidential Election and their model for the Electoral College which: by simulating 20,000 paths for the election, each time varying candidates’ vote shares to account for polling error, changes in turnout or the political environment and the effects of campaigning. The bars below represent the predicted likelihood of every plausible electoral-vote outcome.
Less than three weeks ago, I wrote that:
...the five states to keep a close eye on, and in order of importance, are Florida (29 votes), Pennsylvania (20), North Carolina (15), Arizona (11) and Wisconsin (10).
The Economist has only three states as "toss-ups", namely Ohio, Georgia and Iowa (two of those states having a good number of votes at stake).

Pennsylvania and Wisconsin are moved to the "Likely Democrat" column, and North Carolina, Arizona and Florida move to the "Leans Democrat" category although with a win probability for the Democrats of 0.84, (The Economist's estimate, not mine), I would have Florida in the "Likely" category. Betfair has the Democrats in Florida at 1.62, having traded as low as 1.5.   

With 270 Electoral College votes needed to win, any one of Pennsylvania, Florida, North Carolina or Arizona would be enough for the Democrats, assuming no huge surprises higher up the list of "likely" states. 

With less than four months to go until the election, the Trump response to COVID-19 as poor as it has ever been, and with no sign of getting any better any time soon, the 1.66 on Biden to be Next President or 1.57 on Democrats to be the Winning Party look great value. Barring a miracle cure for COVID-19 (and his suggestion to stop testing so that no new cases are found does have a slight flaw) it's hard to see where he will add to his base to be close to competitive.

The fight for control of the Senate looks likely to be close, with "No Majority" currently the favourite at 2.5 and "Democrat Majority" second favourite at 3.15.

A tie would probably suit the Democrats (assuming they win the Presidency) since the deciding vote is cast by the Vice President. Who that might be isn't yet known, but Kamala Harris of California is currently the favourite. 


Matthew Trenhaile said...

"The always excellent The Economist"

This is presumably an exaggeration and I don't believe even The Economist themselves would claim to be always excellent. In fact I would suggest that "always" when applied to excellency is a tough standard to meet for any entity.

I was particularly disappointed at the last general election that the publication suggested that all readers who had the same values as it should vote for the Liberal Democrats. When voting in a first past the post environment wilfully throwing away a vote in this fashion is completely meaningless. If general elections were conducted in the not always excellent but more excellent than FPP way of proportional representation and single transferable vote then the advice might of held merit. The most important thing when voting is that your vote carry positive expected value much like in betting. The more optimal way of voting when dissatisfied with the status quo is to vote for the most popular non incumbent party. Even if you do not agree with that party’s policies the most important thing in politics is invariably to, as cricket commentators are want to say about bowling to a batsmen, “keep ‘em honest”. Safe seats become less safe and marginal seats are often lost. Only when the constant threat of change is embraced can we hope to see progress. To use another cricket metaphor politicians must be forced to constantly play balls in the “corridor of uncertainty” otherwise they might develop a taste for leaving too many deliveries altogether or even worse start slogging with no regard for the team effort.

G said...

I get as an investor you need to look at some form of predictive data but how can you confidently rely on polling modelling is beyond me.
A look back at the 2016 election predictions had Clinton as high as 93% and at worse 73% to win.
Covid modelling predictions from Imperial were wildly inaccurate [not linked to investing in an election off course but I say this to highlight the inaccuracies of predictive modelling in general]
I for one won’t be investing in a Biden win at 1.6x based on some polling/modelling.
These predictions may well turn out to be correct and with hindsight 1.6x may look like a steal but not for me.