Do you need an attorney?

By Sean Phillippi

A favorite pastime of politicos right after an election is to try and decipher what the election results mean for the future. More than a little ink has been spilled over the years seeking to project what Primary Election results will mean for upcoming General Elections.

One prime example is using the Primary Election turnout by the major political parties to predict what will happen in November.

It turns out that you can use Primary Election turnout, along with one other factor, to project results for a November Gubernatorial Election here in Florida, but the results of those projections are not what most people might expect.

The likelihood of success in winning the Governor’s Mansion over the last 20 years actually goes down, not up, with higher Primary Election turnout.

For every increase of one GOP voter in Primary Election turnout margin (my methodology is explained below), their win margin when facing off against the Democratic nominees in November goes down by 1.1 votes. Additionally, when there is an incumbent Republican Governor running for re-election, which is the other factor I used, the marginal electoral benefit has gone to the Democrat (by 254,474 votes) and not the incumbent Republican when they faced off.

This all may seem counterintuitive. I freely admit that these were not the results I was expecting to find as I set out to prove high Primary turnout led to better results in the General Election. I ended up disproving that theory instead of proving it. When looking at the raw data though it made me see why the results of this analysis came out the way that they did.

Only one time in the last 20 years have more Democrats voted in a Primary Election than Republicans (by 390k votes no less). That was 2002, which was also the year where Republicans enjoyed their largest victory in a Gubernatorial Election over that same time period. The largest Republican turnout margin in a Primary Election was 2010, which was also the year in which Republicans enjoyed their closest November victory in the last two decades.

Running a multiple regression analysis not only quantifies how big of an effect one factor has on another, but also calculates how strong that relationship is. 95% is the standard measure used to say whether a factor is statistically significant. In this analysis the significance levels were 99.65% for the intercept (what to expect when all other factors are zero), 99.36% for Primary Election turnout margin, and 96.99% for Republican Incumbency. This regression analysis explains 98.78% of the variability in the overall margin of the Gubernatorial General Election results.

The phrase correlation does not equal causation, meaning just because one thing can predict something else doesn’t mean that it caused that something else to happen, has become a cliche. It is a cliche that absolutely applies here though. Just because strong Primary Election turnout can be used to predict a lower margin of victory in Gubernatorial General Elections does not mean that it causes it, and I strongly believe that there is no causal link.

Data analytics and statistics are sciences, but there is also an art in knowing which analyses to trust and which ones may not be as useful in a particular circumstance.

Using this analysis to project forward to this November after the upcoming August Primary Election is not something I would advise. The reason being that we live in a totally different world politically than we did during the last Gubernatorial Election, much less any election in the last 20 years.

That being said, the avalanche of evidence that this analysis provides is overwhelming. So, while I question predictive value of this multiple regression analysis the picture that it paints as to the state of Florida Gubernatorial Elections over the last 20 years is crystal clear: There is no other option other than to reject the hypothesis that strong Primary Election turnout leads to Gubernatorial Election success.

Methodology: I computed Primary Election turnout margin by taking the most votes cast in a GOP Statewide Primary Election and subtracting the highest number of votes cast in a Democratic Statewide Primary Election. I used dummy variables to account for Republican incumbency.

For anyone who wants to dive deeper into the methodology: Click Here to see the Multiple Regression Output in a chart that expands on my methodology.  

XXXXX

(Sean Phillippi, the Managing Member of TLE Analytics LLC, the political data and consulting firm he founded in 2012, is based in Broward County. He has worked campaigns at the federal, state, and local levels, including the Nan Rich for Governor Campaign, the pro-Hillary Clinton Super Pac “For Florida’s Future”, and the Florida Democratic Party, where he served as Acting Data Director.  Phillippi is also pursuing a master’s degree in Business Analytics at American University.) 

Closed.