Perhaps it was our article from last Friday which showed that contrary to most popular polls predicting a Blue Wave sweep on Nov 3, a recent analysis by the Trafalgar Group (which correctly called the outcome of the 2016 election) found that a Trump victory and Republicans holding the Senate were the two most likely outcomes on election day. Or perhaps it was just general nerves over the upcoming election, which will be contested according to the latest BofA Fund Manager Survey …
… but whatever the reason, JPMorgan’s top quant Marko Kolanovic wrote and interesting note yesterday in which he said that clients have been asking him “what data in addition to polls are available to assess possible US election outcomes in various states.”
To answer this question, Kolanovic said that he looked at the historical changes in voter registration data and back-tested their relevance to election outcomes.
Why? Because according to Marko, it is intuitive that “if more voters register for a party, that they will also vote for that party.” However, being a quant, Kolanovic immediately focused on the causative, not correlative aspects of this observation, and as he puts it, “the question is how significant the relationship is and what is the “beta” of additional registered voters to actual votes.” The chart below shows data for seven battleground states that provide these data and presidential elections over the past 20 years.
Looking at the chart above, Kolanovic finds that changes of voter registration (change in registered D, R, D-R), is a significant variable in predicting the voting outcomes, a critical observation which virtually no existing polls takes into account since it would indicate that Republicans have a high likelihood of winning virtually all battleground states!
The regression above indicates that the change in D-R (Democrats less Republicans) registrations highly correlates with the subsequent change in D-R voting outcomes. Kolanovic also notes that historically, the “beta” was averaging 1.1 across states, a relatively high number which can be explained by large pools of independent voters (in fact, there is positive correlation between the registration “beta” and fraction of independent voters in a state).
The JPM strategist next looks at an example of this analysis in the case of arguably the most important swing state for the upcoming election: Pennsylvania (PA). PA has a particularly strong relationship between voter registration and election outcome, according to Kolanovic, and in Figure 2 below he shows the relationship between changes in voter registration (D-R) and subsequent changes in election outcome (D-R vote count).
Three months ago, the Philadelphia Inquirer conducted a similar analysis when it found that “since the 2016 primary election, Republicans have added about 165,000 net voters, while Democrats added only about 30,000. Democrats still maintain an 800,000-voter edge over Republicans. But that’s down from 936,000 in 2016, when Trump still won the state by less than 1%.”
“Look, the president won our state by 44,000-plus votes in 2016,” said Lawrence Tabas, chair of the Pennsylvania Republican Party. “We have since picked up and narrowed the gap between us and the Democrats [by 135,000]. So we were already ahead 44,000, and look what we’ve picked up. I predict we’re going to narrow the gap further between now and November.”
While the other swing states with available data typically show a weaker relationship, in most cases it is significant from a statistical standpoint.
Finally, the punchline: what is the current change in voter registration over the past four years in swing states? This is shown in the table below.
Needless to say, if Kolanovic’s assumption is accurate, the change in voter registration data shown above would immediately invalidate all polls such as this one from Real Clear Politics showing Biden sweeping across the Battleground states. In fact, while he does not say it, the implication from the Kolanovic analysis is that Trump may well end up winning the critical trio of Pennsylvania (20 Electoral votes), Florida (29 votes) and North Carolina (15 votes).
Even more remarkable, the Kolanovic analysis means Trump could win Pennsylvania, something which not even the Trafalgar analysis of tossup states assumed.
Of course, lest he be seen as predicting Trump victory based on these data (something which last month prompted Nate Silver to have a hilarious meltdown on twitter), Kolanovic caveats that voter registration is only one variable in determining the election outcome, and “these results should not be taken as a prediction of state election outcomes.”