There is a groundswell of attention being paid to voter registrations in key swing states, driven in part by a widely circulated report by well-known JPMorgan quant analyst Marko Kolanovic. I first wrote about the importance of registrations last Fall, using Michigan as the example. I am making four arguments to underpin my view – they are debatable:

  1. The shift in voter preferences in 2016, particularly in swing states was driven by trends that have only been reinforced by the pandemic – from globalization to China tensions to culture wars
  2. Where you live tends to influence how you vote, so county level registrations matter
  3. Turnout percentages tend to remain within a steady band
  4. Other factors that get a lot of attention such as party energy and crossing party lines, cancel each other out with both sides working hard to secure an advantage

If you agree, what it boils down to is how many voters each party registers in each county – remember, you can’t vote if you don’t register.

By this measure, Republicans have an advantage.

I have completed a model of 11 swing states (Florida, Pennsylvania, Ohio, Michigan, Georgia, North Carolina, Arizona, Wisconsin, Minnesota, Iowa, and Nevada) based on currently available voter registration data. Only five of these states have closed voter registration. I will keep updating the forecast as states report final or near-final voter registration statistics (and project Election Day registrations). For now, the data implies that President Trump garners 288 electoral votes and wins re-election.

Electoral VotesBidenTrumpRegistration Deadline
Grand Total250288
Michigan16Up to day of election
North Carolina15October 31
Wisconsin10Up to day of election
Minnesota10Up to day of election
Iowa6Up to day of election
Nevada6Up to day of election

Why the polls may be wrong is not the goal of this post – for some views on that, see here. For a couple of additional views on voter registrations, read here and here.

The View from Fall 2019

Underlying my argument was the assumption that party preference and turnout at the county level in key swing states had shifted from 2008/2012 to 2016 as a result of underlying trends and that those patterns would persist.

  • I partially made this point by showing that in the 12 counties in Michigan that flipped from Obama in 2012 to Trump in 2016, in 8 of them, President Trump received more votes than President Obama did.
  • I also argued that county flip-flopping was rare by showing that of the 207 U.S. counties that flipped during a Presidential election from one party to the other from 2008 to 2012, only 23 reversed back in 2016. (Source:, Guardian)

Thus, what matters was how many voters each party could register in counties, and preference and turnout would take care of the result. At that time, the U.S. was experiencing record low employment, healthy economic growth, and relative peace. I also showed that key economic indicators in Michigan there were healthier than the country as a whole. I further showed that even if Democrats were to drive greater registrations and turnout in important counties such as Wayne County (home to Detroit), that increased registrations in Rebpulican leaning counties were still going to be difficult to overcome. There simply were not enough voters available.

What About Now?

Despite the horrid events of 2020, I argue that the underlying trends support consistency in likely preference and turnout in counties in the key swing states. This is the core of our current thesis, and the most debatable. To account for this debate, I have produced a set of voter preference and turnout projections that take into account a shift in favor of Democrats. Here is my methodology:

Source Data: I sourced county level presidential voting records for the 2008 – 2016 elections for Florida, Pennsylvania, Ohio, Michigan, Georgia, North Carolina, Arizona, Wisconsin, Minnesota, Iowa, and Nevada (note: Georgia data quality was suspect). This data is widely available. I then sourced voter registration data for the same states from respective state websites.

Step 1 – Final Voter Registrations: The first step was identifying an expected base of registered voters. In some states, registrations are closed and numbers published. In those states that remain open, I estimated final voter registration data based on prior elections. For example, on average Michigan adds 2% to its final registration base from July to October. I grossed Michigan registrations up by that percent. This was done at the individual county level in each state.

Step 2 – Base Case Looks Like 2016: I created a base case that looks like 2016, in terms of turnout (votes/final registrations) in each county and split of Democrat v. Republican v. Other. The big assumptions here are the high level of votes for the Other candidate and of course Republican leanings. This case, unsurprisingly, shows a decisive Trump victory because in the swing states, far more registrations have occured in counties that voted Republican in 2016.

Step 3 – Alternate Case Looks Like 2012: I created a 2012 case, similiar in methdology to the base case, using turnout and party splits from 2012 mapped to Final Voter Registrations from Step 1. This shows a decisive Biden victory because it assumes a lot fewer votes for the third-party and that many 2016 Republican counties are back to Blue.

Step 4 – Combined Case: Under the assumption that a lot of what is getting attention will cancel itself out as a draw (crossing party lines, GOTV operations, tampering/suppression, good luck/bad luck, party energy), I combined the Base and Alternate Cases to form a Combined Case. To do that, I took the 2016 numbers for any county that voted Republican in 2016 and the 2012 numbers for any county that voted Democrat in 2016, and summed them for each state. This shows a narrrow Trump victory

Step 5 – Adjusted Combined Case: In the Combined Case, Other votes were still 3% of those cast in these swing states (down from 5% in 2016 but above the 1% of 2012). To account for the possibility that this was still too high. I created an Adjusted Combined Case assuming that some percentage (picked 35%) of remaining Other voters from the Combined Case would come back home to either major party. I åssigned those votes back based on the existing Democrat v. Republicans split in the Combined Case.

Projected Combined CaseDemocratRepublicanOtherWinner
North Carolina2,091,3622,619,16170,166R

Any opinions or forecasts contained herein reflect the personal and subjective judgments and assumptions of the author only. There can be no assurance that developments will transpire as forecasted and actual results will be different. The accuracy of data is not guaranteed but represents the author’s best judgment and can be derived from a variety of sources. The information is subject to change at any time without notice.