Don Luskin's Model

I listened to Don Luskin's short video in which he presented the wild swings in his model, dating back to pre-Covid-19 period in February 2020, until now. The model went from very positive for the incumbent (Donald Trump), to extremely bullish to his opponent Joe Biden, then back to a a record breaking victory for Donald Trump. His model predicts the the election margin = EVrep - EVdem. A margin of 100 EV's would mean Trump wins by 329 to Biden's 219. The prediction is based on six economic variables and one variable on incumbency. Trump automatically receives 81 electoral votes because he is the incumbent candidate.

As of September 1, 2020, the model predicts Trump wins 418 electoral votes verses Biden's 120 electoral votes. The values of the six variables has varied over time, as presented in the table below. A nominal round-off error was added to the addition correct. The economic values are issued monthly, with the exception of oil prices. So, the Covid-19 lockdown impact is apparent in the payroll employment values, but not the impact of the stimulus program is not seen until the July 30 forecast. When both impacts are accounted for, the Covid-19 pandemic becomes a net positive in terms of electoral votes. Trump's chances of being re-elected improved as oil prices drop to historical low levels but this did not last.

The table below includes only about half of all the forecasts that Don Luskin shows in his video. The video ends with the July 30, 2020 forecast. He has issued an update to the forecast on September 1, 2020.

A margin is easily converted into actual electoral votes for both candidates. It is interesting his forecasts have been far outside of the consensus, both in support of Trump and then in support of Biden. In fact, he notes that in his video, he has calculated margins in above 538 and below minus 538, which are impossible. The 344 margin is inline with his earlier predictions, pre-covid-19.

The table shows the impact of the economic variables on the incumbent's electoral vote margin. For example, Trump EV went from 49 to 143, a gain of 94 electoral votes, when oil prices suddenly collapsed during the early lockdown. Don Luskin has apparently performed a linear transform, EVoil = A - M * oil price to obtain proper relationships. I calculate the coefficients A and M to be 235.75 and -4.43, respectively. I do not know the other linear transformations involved, but they could likely be calculated from a comparison with the published economic data.

While I readily admit to an affinity to numbers rather than concepts, I think both are very important in Luskin's analysis. He has been right on the money in his prior predictions. His central idea in the video is that a contraction in consumer consumption occurred during the initial shock of the lockdown, then everybody got stimulus checks, so it was like Christmas in July and August, and the consumer is making up for lost time. He's suggesting this is the impetus for the rising stock market, amidst terrible unemployment numbers. I also believe it is a common conviction that a vaccine is coming soon.

Thus, his modeling shows really two effects, payroll employment taking a nosedive then slowly coming back and the improving GDP and personal income, thanks to the stimulus package. These effects result in his very bullish forecast for Trump in February (margin = 290) and even a more bullish case in the Sept 4 forecast (margin = 355). With stimulus II on it's way, Luskin's model looked pretty good.

This held up alright, until 2:00 pm today (Oct 6, 2020), when Trump through a tweet, said he was through with negotiating with House Speaker Nancy Pelosi on the next stimulus package, and this would have to be put off until after the elections. Airline stocks were particularly hard hit, falling 3 to 4% on the news.

One area which I wish I could avoid, was Luskin's aside about Covid-19 being less deadly than the 1957 H2N2 flu attack. Covid-19 is extremely contagious, given that it spread in just 3 months to nearly every country on our planet and is continuing to take lives at a rate of 5,000 per day. I would like Covid-19 to be history like the 1957 flu is today, and then comparisons are appropriate.

I generally will not post links with political messages, but I have done it in this case, because his insight in forecasting has excellent value and offers a different perspective.

David Lord

October 6, 2020

Links:

Don Luskin's Video of August 3, 2020

Don Luskin's forecast of September 1, 2020

Don Luskin's forecast of September 4, 2020