Recently, I received an email from Brandon Rottinghaus, who is a political science professor at UH. He and a graduate student of his had put together a mathematical model of the 2010 Democratic primary for the countywide judicial races. I thought it was interesting and worth sharing, so I present to you here what he sent me:
One of my graduate students (Chris Nicholson) and I were interested in the outcome of the 2010 primaries, in particular with respect to judicial races. To appease our curiosity, we collected data pertaining to the Democratic Primary election in Harris County for all judicial county-wide seats. We only focused on county specific races, but we also included candidates for those District Courts which had substantial portions of their district in Harris County. We collected all the relevant candidate-specific data we could think of, including race, gender, endorsements by influential groups (HBAD, the AFL-CIO, GLBT, Harris County Coalition of Elected Officials), ballot order and campaign funds raised and spent. Although we have some missing data (especially on the campaign funds variables) we have most all of the data for each candidate and each race. This yielded 72 total races.
Model 1: Total Votes Received
We ran two statistical models. In the first, we explain the factors that predict the total number of votes received by each candidate. (If anyone is interested, this is an ordinary least squares regression model with a continuous dependent variable; goodness of fit diagnostics are available upon request). Specifically, we control for gender, having a Hispanic surname, endorsements by the AFL-CIO, GLBT, the Harris County Elected Officials, being first on the ballot and the amount of money raised on the January 15, 2010 report and the 30-Day report. Although each variable (referenced above) has some effect on the number of votes received by each candidate, we are interested in those relationships that reach statistical significance ― that is, the relationship occurs not by chance.
Results: Total Votes Received
The results from this first model indicate that the following factors were influential in predicting primary vote totals (number of additional votes for each effect in parentheses): being first on the ballot (+8,680 votes), being endorsed by the Harris County Coalition of Elected Officials (+8,302 votes), being endorsed by the Houston Black American Democrats (+7,089 votes), being a female (+5,725 votes) and more money raised (and reported) in the January 15 filing (+.25 votes for every 1% increase in funds raised). All of these effects are strong, except the latter, which is statically significant but substantively very small.
Model 2: Probability of Winning
In the second model, we truncate the outcome to explaining the probability of those who won (including those who gained enough votes to enter a runoff election) . (If anyone is interested, this is logit model with a dichotomous dependent variable; goodness of fit diagnostics are available upon request and we used marginal effects to calculate the effect of each independent variable on the dependent variable). The variables we included were the same as in Model 1. Again we are looking for statistically significant relationships.
Results: Probability of Winning
The results from this model show that gender and being endorsed by the Houston Black American Democrats were the key factors predicting victory. Specifically, women were 35% more likely to win than men and HBAD endorsees were 50% more likely to win than those not endorsed by HBAD. None of the other variables reached standard levels of statistical significance.
Summary: The 2010 Democratic Primary in a Nutshell
Of course, these results are broad trends, not specific to every judicial race in the County, but the results give us some sense of the most influential factors from the 2010 Democratic primary.
In summing up the results, endorsements were important, but some endorsements mattered than others, and campaign funds raised mattered very little. The Houston Black American Democrats endorsement seemed to be the most influential in terms of prospects for victory. Their organizational push to get the vote out seems to have paid off. In addition, being a female candidate was a good predictor of victory, suggesting that Democratic primary voters were more interested in electing women to the bench than men. This could be the result of a concerted effort by organizations devoted to making the judiciary more diverse, or this could be a demographic effect where more women voting in the primary translated into more votes for female candidates.
University of Houston
The accompanying spreadsheet for their analysis is here. Let me know what you think. They’re working on a similar model for the Republican results, and when they have it I will share that as well. My thanks to Dr. Rottinghaus for sending this to me.