Primary Model

Updated on August 24, 2020

My selection of forecasts on this site has nothing to do with who is predicted to win the election. However, I must admit that when so many forecasters are predicting a Biden win, I am very happy to post a view to the contrary. Yet, forecasters should provide the basis for their predictions. I am perfectly happy to post any forecast regardless of their professional affiliation.

Right now, there are two forecasts using quantitative analysis predicting a Trump victory, the Primary Model by Professor Helmut Norpoth and Vinod Bakthavachalam's model, which has not been updated since October 2019. Vinod B. has recently posted demographic statistics in the swing states which shows that Donald Trump could win because of the demographics in the swing states and the "all or nothing" rules of the Electoral College. I have added a link to this posting at the end of this blog.

I note that Moody's analytical model includes a case where Trump wins in a close race, but it is dependent on a strong turnout of Trump voters. Since this will not be known until after the election, I consider this case just an example of how their base prediction could be wrong.

The Primary Model website provides the forecast of Trump gaining 362 electoral votes and 91% chance of winning. Since the values in the model are already determined, this prediction will likely stay the same until the election. Dr. Norpoth states that this time, the electoral vote prediction is not from the popular vote estimate, but comes straight from his model.

The prediction has been noticed by others, and I suspect it will be popular among Trump supporters. The website does not provide details on the estimation of expected electoral votes and the chance of winning.

Prior Equations used by Dr. Norpoth

The 2020 prediction is different from his prior work which focused on the prediction of the percentage of the popular vote for each candidate. There have been only five elections where candidates won the popular vote, and lost the election - in 1824, 1876, 1888, 2000 and 2016. So, whoever wins the popular vote is highly likely to win the electoral vote. Obviously the elections of 2000 and 2016 are recent exceptions.

An equation provided in a 2014 paper by Dr. Norpoth is:

V_t = 49.5 + 0.525 V_t-1 - 0.474 V_t-2

where, if applied for the 2020 election, V = percent popular vote of Republicans, subscripts t-1 = 2016 election, and t-2 = 2008 election. A high Republican popular vote in 2008 decreases Trump's popular vote in 2020, which is likely the result of identified election cycles from Norpoth and others research. The calculated result using V₂₀₁₆ = 46.1% and V₂₀₁₂ = 42.7% equals 53.5% popular vote for Trump or a 7% vote margin over Biden, if we neglect other third party candidates.

Dr. Norpoth also has provided a second equation, which incorporates the result of two early primaries, New Hampshire and South Carolina for both candidate as they applied to his 2016 prediction. I believe the correct form of the equation for the 2020 election is:

V_t = 50.6 + 0.429(PS-Rep) - 0.170(PS-Dem) + 0.361 V_t-1 - 0.377 V_t-2

Where PS-Rep and PS-Dem are the primary "scores" based on the Republican and Democratic primaries in New Hampshire and South Carolina and the popular vote variables are evaluated for 2016 and 2012 elections as in the prior equation.

The equation as provided on the website for 2016 for Democrat and Republican primary scores are reversed, because the incumbent was Democrat in 2016. A high primary score for the Republican candidate should help his popular vote. I have kept the primary score as positive values, and changed the sign on the 0.170 coefficient to negative.

Other analytical forecasters have noted the cyclical nature of our elections, and the advantage of the incumbent president in re-election. They have also noted that the electorate can feel less inclined to elect the same incumbent party for more than two terms. This is called the "shopworn" effect by Don Luskin or the fatigue factor by Moody's Analytical model team.

As more information becomes available, it will be posted to my website.

David Lord

Updated, August 20, 2020

Links:

Primary Model

Norpoth, Helmut. 2014. "The Electoral Cycle.” PS: Political Science & Politics 47: 332-335.

Vinod Bakthavachalam, Even With Huge Polling Leads for Biden, The Electoral College Still Favors Trump

Vinod B. Recent Postings on Medium.com