How Donald Trump could win

At present, I am aware of at least 18 poll based forecast models. All of these models predict the most likely outcome for the election is a win for Joe Biden. So, I think a fair question is - Could Donald Trump win a second term? My answer is a definite yes. Most likely, it would be a very narrow victory. A victory similar to 2016, with 304 electoral votes or fewer votes is entirely possible.

Polling data are inherently in error, since we don't know who will vote. It was been suggested that Trump won in 2016 because he could motivate his supporters to vote in 2016 better than Clinton. Polling data are more reliable when the margins are large and multiple recent surveys are consistent. Trump will win Kansas and Biden will win New York. So, the data tells us what we already know. If we didn't have any polling data, the historical record of voting would tells us pretty much the same.

Poll based predictions are similar because they all use the same polling data. The forecasters may also look at the same historical voting data. These prediction are not statistical samples of a population which tend to improve over time. The predictions may even become more uncertain over time with an increase in swing states. The underlying population (turnout of each party in November) is not certain and their political preferences will change. If 2016 showed us anything, it was the fallibility of poll based models.

A realistic victory scenario for Trump

As a base case, I chose the Niskanen map. Of the 18 forecasts I am following, it is closest to the consensus in terms of approximate EV's.

I assumed Trump wins four toss-up states, Florida, North Carolina, Wisconsin and Arizona. Now, I also assume he wins three lean-Republican states, Texas, Georgia and Iowa. Our simulation model predicts Trump will win in an extremely tight race, with Trump winning 274 electoral votes to Biden's 264 electoral votes. The probability of a Trump win in this case is 58% verses 42% for Biden. I could have easily assumed Trump would win other Republican inclined states such as Utah and Missouri, raising his win probability a bit higher.

Remember in 2016, Trump won all of these states. The website fivethirtyeight.com predicted Hillary Clinton would win Florida, North Carolina, Wisconsin and Michigan. So if 2020 is similar to 2016, this could happen again.

For everyone who thinks that Trump can't possibly win a second term, I would say think again. Trump won all these states in 2016.

Click the map to create your own at 270toWin.com

The Incumbent Advantage factor

Analytical modelers have known about the incumbent factor for a long time. It is a reason why all analytical modelers have favored Trump albeit with a very tight race. They were forecasting Trump's victory before knowing who his opponent would be.

Elected presidents at the end of their first term have a tendency to be re-elected. A second rule is that when the same party has been in office for more two terms or more, voters will elect the candidate from the opposite party. Voters are now tired of the party.

On re-elections: Nixon was re-elected in 1972, Reagan in 1984, Clinton in 1996, Bush in 2004 and Obama in 2008. Five solid cases of the incumbent victories. Some would also consider Johnson in 1964, as a re-election of the JFK/LBJ ticket, so this makes 6 incumbent victories.

Now there is one solid exception: Carter lost his re-election bid in 1976 to Reagan. The others which seem like exceptions are: Ford took over in 1973 following Nixon's resignation and lost his first real election as president in 1976. Also, Bush senior lost his re-election in 1992. But, in both cases, the Republican party had been in office for more than 2 terms, so the second rule applies. Voters had grown tired of the party.

Bush senior's first election in 1988 is an exception to the second rule. It was time for a Democrat to win. Otherwise, these two rules work pretty well. Analytical modelers may consider many other factors in predicting elections.

The basis for analytical equation modeling is to find factors to explain the exceptions in the Carter 1976 and Bush 1988 elections. Other factors include the candidates popularity and economic trends.

I provided on my website a number of analytical models which predicted in 2019 or early 2020 a victory for Trump. Moody's model was very comprehensive and included as a factor, the results of the 2016 election. This was based on the observation that what voters did in 2016 was likely to carry over to 2020. The model has not been updated. This is understandable, as this model uses long term economic trends and it is very difficult to rigorously factor in the disruption from Covid-19.

Recently I have been asked about Professor Helmut Norpoth's Primary Model. His model predicts a 91% chance that Donald Trump will be re-elected with 362 electoral votes. The calculations are not found on his website. I like forecasts outside of the mainstream, but without serious documentation, I must assume these are his opinions, and nothing else. It is interesting, however when someone says an event will take place with 99% certainty, it may be taken just as a statement of confidence, but when someone says there is 91% certainty, it implies a calculated value.

Complacency factor/ Underdog factor

Being behind in polling has an advantage of mobilizing one's base. Being ahead can lead to complacency. Thus in an ironic way, Trump may benefit with so many models suggesting Biden will win. It can tell voters in November that Biden is so far ahead, it is unnecessary to vote for him. So the model predictions themselves can affect the election outcome, contrary to their predictions.

Final Thoughts

I have included more information on the Historical tab in this website on the 2016 upset, where all poll based forecasters were wrong. They were right in that they called correctly almost all states, but wrong on the final outcome.

I consider of the 538 votes, counting only toss ups, lean Republicans and lean Democrats, there are 191 electoral votes still up for grabs. There is certainly sufficient undecided/swing votes for Donald Trump to win in November.

David Lord

July 12, 2020