The Horner Effect, Part 1

While Minnesota has a rich history of third party successes, the recent era began in 1998 with the election of Jesse Ventura as Governor. Since then there have been two additional statewide elections in which the Independence Party candidate received a significant share of the vote, Tim Penny in the 2002 Governors race and Dean Barkley in the 2008 Senate race. This year certainly looks like it will be another as Tom Horner is over 16% in the current GPI.

This is the first part of a three part series on what effect we can expect Tom Horner to have on the Minnesota Gubernatorial election. In this installment I'm going to look at past statewide elections and the effect the Independence Party candidate had. Part two will attempt to show why 2010 is different than these past elections and part three will take a look at what the current polling shows is happening.

With that then, here is Part 1...


In a three-way, someone always gets screwed

What follows is a chart of the correlations between election results for each candidate by county, so where it says "Dem v IP", that is the correlation between the Democratic candidate and the Independence Party candidate for the corresponding race. The closer to 1 (or -1) the number is the stronger the correlation (or negative correlation) is.



Exciting stuff right? What you see though, is that in the four races without a strong IP candidate (not in bold) the DFL-GOP vote share had a strong negative correlation, meaning when one candidate did well in a county the other candidate did poorly and as one candidate did even better the other candidate did even worse. In a race with only two competitive candidates this is what happens, you can't have a situation where both candidates do well in a county after all as there are a finite number of votes available, so as one candidate does well, the other candidate is going to do worse.

This high level of correlation is best illustrated in a graph, what follows then is a scatter plot of Amy Klobuchar and Mark Kennedy's vote share by county in the 2006 Senate race (the R-squared value is equal to correlation squared).



The trend is unmistakable, a very tight and linear distribution, when one candidates vote share increases the others decreases by about the same amount. Now let's look at the scatterplots of the IP candidate that year, Robert Fitzgerald and the other two candidates.



His vote share fell in a narrow range of between 2% and 5% and had no effect on the results of the other candidates. In a three way race with a strong third party challenger however, this dynamic changes and you can end up with divergent results as seen in the chart above and the graphs that follow.

1998 Governor



In 1998 Skip Humphrey and Norm Coleman had a very small positive correlation, meaning they didn't really have an effect on each others results, it's just a seemingly random distribution of points. This is largely due to the distribution of Ventura's votes, as he did either exceedingly well in a county, which came at the expense of both the DFL and GOP candidates or he did rather poorly, in which case the other two candidates did much better.



The scatterplot of Ventura with both of the other candidates is somewhat similar to the graph from the 2006 Senate race, certainly not as tight, but somewhat linear nonetheless. What you see though is that Humphrey's floor extended further down than Coleman's; Humphrey has results as low as 15% while Coleman didn't go below 27%.

Jesse Ventura was a unique candidate in that he was able to attract voters from the center of the electorate as well as from both of his opponents flanks, libertarians from Coleman and liberals from Humphrey, in short, a populist. I don't think it can be underestimated how unique he was as a candidate and how difficult it is to make comparisons between him and other IP candidates. His performance though can provide us with a best case scenario for a third party candidate in Minnesota.

2002 Governor



In the 2002 Governors race you see a different dynamic, the correlation between the Democratic candidate and the Republican candidate is virtually the same as between the Republican and IP candidate. In the above graph there are a large number of data points that fall into a somewhat tight linear distribution, but then there are a bunch that don't seem to correlate at all, these were the counties that Tim Penny did well in.



Tim Penny ended up with 16.2% of the vote, but he did disproportionately well in the southern Minnesota counties he represented as a congressman for six terms and not so well everywhere else. What you can see from the graphs is that Pawlenty's vote share floor was far higher than Moe's and while Pawlenty lost some vote share as Penny gained, there reached a level at which Pawlenty's support didn't fall below, 27%. Roger Moe however fell into the teens and even 10% at his low point.

2008 Senate



The 2008 Senate race also featured a strong IP candidate in the form of former Senator Dean Barkley. This is the only one of the three races with a strong DFL-GOP correlation and this is somewhat related to Barkley's vote distribution, which was much broader, geographically, than Tim Penny's.



Barkley didn't win any counties like Penny did, but he didn't get routed in any either, his vote share fell within a consistent range statewide not dropping below %10 much and not going above 25%.

Looking at all of the races though, what appears to be happening is that IP candidates have more impact on the Democratic candidates vote share than Republican candidates, even discounting the 2002 race which featured an ex-Democrat as the IP candidate, this relationship is present in 1998 and 2008. In all three races the GOP candidate hit the floor at around 27% while the DFL candidate fell down into the teens occasionally.

Now, of course, correlation is not causation, so we can't say, for instance, that those IP voters, had the IP candidate not been in the race, would have voted for the DFL candidate more so than for the GOP candidate. What we can say though is that in statewide races when an Independence Party candidate does well in a county it tends to be more strongly correlated with the Democratic candidate not doing well than the Republican candidate.

An argument can be made however that the reason IP candidate have tended to hurt the DFL more than the GOP is because of who the candidates themselves are. While the DFL has tended to nominate candidates who were more liberal than moderate, the GOP has nominated candidates who were more moderate than conservative. The effect of this is that the partisan space available for the IP candidate tends to be more tilted left.

This then will be the starting point for analyzing what effect Tom Horner will have on the 2010 Minnesota Governors race, which will be the subject of part two of The Horner Effect.