---

 

Update: September 12, 2020

I look at a very wide variety of forecasts, poll based, gambling sites and non-poll based methods. No one has discovered the perfect crystal ball.

Polling based forecasts:

The poll based forecasts favor Biden. They have for the last 4 months at least. The expected approximate electoral votes for Biden are 331 with a standard deviation of 20 EV's. The simulation model estimates 319 EV's for Biden with a 90% chance of winning.

I decided not to include the Real Clear Politics electoral vote estimate, because they have classified 201 electoral votes as toss-up. This meant giving half to each party could be considered far too subjective. I decided to include Ryan Gest forecast this time as presented on YouTube.

The close agreement of forecasts may be interpreted as a reliability indicator, but this can be false impression. Forecasters look at the same data, the statewide polling data which can only survey potential voters. They do not know who will actually vote. The forecasters know what others are predicting and likely take this into consideration. So, these are not independent assessments.

Ryan Gest, the YouTube forecaster, made an excellent point, that the really "tipping point" state is Florida. A Trump victory in Florida, plus winning a couple states Trump won in 2016, namely Michigan and Pennsylvania, adds 65 electoral votes to his total, and Biden's cushion is suddenly gone. Many of the forecasts predict Biden can take back states won by Trump by a narrow margin: AZ, FL, PA, WI and MI.

Gambling Sites:

I keep an eye on Real Clear Politics gambling summary which shows a lot more change. These people are putting their own money into their forecast. It shows a certain amount of conviction in their forecast. They will be more meaningful when we get closer to the election. Joe Biden odds of winning according to the gamblers is only 53% as opposed to Trump's 47%, so this is a forecast of a close race. Biden's winning odds were 61%, a comfortable lead over Trump in late July.

Non-poll based Models

This group is hard to classify. It ranges from Dr. Litchman check list of 13 questions to statistical analysis of variables which affect elections (Moody's analytical model is a great example). In general, the models are predicting a Biden win.

There are 3 forecasts which predict Trump will win. Moody presents 3 cases, one of which shows Trump winning because of high turnout among Republican voters.

I believe in presenting all forecasts which provide clear documentation, regardless of their predictions. I have included Dr. Norpoth forecast of Trump winning 362 electoral votes. I understand his concept that the incumbent candidate has a historical advantage. Other consider that the voters grow tired of an incumbent party after the second term (not Trump's case). Dr. Norpoth uses autoregression to calculate election cycles. There is a lack of documentation with his forecast, so I will likely remove it from the summary table in the future.

Summary

If you look at just the poll-based forecasts, you could easily conclude that Biden has no chance of losing. But, the gambling sites show that Biden's odds have been diminishing. So, I see some conflicts not agreement in these forecasts. Biden's victory will depend on taking back states that Trump won by a narrow margin including AZ, FL, NC, MI, PA and WI. What appeals to Biden or Trump voters in North Carolina is likely to equally appeal to Ohio, Michigan, Pennsylvania and Wisconsin voters. This is called systemic error because the popular vote margins in each states are dependent.

One new aspect in all this is expected increase in mail-in balloting. I don't know how it will all end up, but people who prefer to send in their ballots, may opt to vote in person. For forecasters, it's just added one more complication to their predictions.

It is also amazing how little the forecasts have changed since the last updates in August 11, 2020. In that time, the Republican and Democrats have had national conventions. About 10 new books have been published, strongly pro and anti-Trump. There are recent reports that Russia is attempting to interfere in our elections.

The first debate will be September 29, 2020. I urge all Americans to keep informed and vote.

You can also comment on this site using the form provided. I enjoy emails and will respond to all of them.

David Lord

September 12, 2020