Update on Election Predictions October 5, 2020

As of October 5, 2020, Joe Biden is ahead of Trump in all poll based predictions. Note the average of all these forecasts is 337 electoral votes, just 6 electoral votes above the September average. The swing states are very well known at this point with Ohio, North Carolina and Georgia to be the most difficult to forecast. Some include Iowa, Texas and Florida as toss-ups. Yes, I said Texas with 38 electoral votes. I have added YouGov website to our list of sites, and removed Predictit.com. Predictit.com is really a gambling site.

The gambling sites have shown in the last week, more conviction to a Biden win, with odds of 61% for Biden verses 37.5% for Trump according to the Real Clear Politics. The Predictit website suggests Biden will win with 335 electoral votes, incredibly close to our average of 337 electoral votes.

The non-poll based methods show much more divergence. A new addition, Dr. Campbell's Political Economic Model, considers the election to be very close with Joe Biden just barely winning the popular vote 50.8 to Trump's 49.2%. I fear if he is right, we may end up in the courts, like the 2000 election. On the other hand, Don Luskin has updated his model in August, showing a strong Trump victory with 388 electoral votes.

Another new forecast based on linear regression and economic variables, has been developed by Lewis-Beck and Tien, that predicts Joe Biden will win by 470 electoral votes. I believe implicit in their forecast, is that recessions in the first two quarters of an election year hurt an incumbent, but this was the result of a lockdown supported by both parties. This is in sharp contrast to Helmut Northop's model that predicts Trump will win by 362 electoral votes. He uses auto-regression techinque to capture the cyclical nature of our elections.

I will be updating our simulation model next week, and add it to the list of predictions. Our September simulation showed Biden would win with 331 electoral votes.

Although many predictions point to a Biden victory, those predicting a Trump victory may in the end be right. The toss-up states total over 100 electoral votes in a number of models, which really are just to close to call. Polls never know how many of the respondents will actually vote, so it is the candidate that can motivate his base to actually vote in high numbers that will win.

David Lord

Oct 5. 2020.

I