Last month, I printed an analysis of the Democratic judicial primaries by University of Houston political science professor Brandon Rottinghaus, who along with grad student Chris Nicholson built a mathematical model to predict the number of votes a candidate would receive based of a variety of factors such as money raised, ballot position, and specific endorsements. At the time, I said Dr. Rottinghaus was working on a similar analysis for the Republican primaries and that I’d share it here when he completed it. I now have his analysis, which is as follows:
Our friend Charles Kuffner was kind enough to post our earlier analysis of the Democratic Primary. To follow up, we gathered data on the Republican Primary election in Harris County for all county-wide elected offices. As with the Democrats, we collected all the relevant available candidate-specific data we could think of, including:
• Endorsements by influential groups (Conservative Republicans of Harris County, United Republicans of Harris County, Houston Realty Business Coalition, Houston Police Officers Union, and the Harris County Deputies Organization)
• Endorsement by the Houston Chronicle
• “Well Qualified” ratings from the Houston Bar Association Poll
• Order of listing on the ballot
• Campaign funds raised and spent
While some data is missing (principally Contribution & Expenditure Reports not submitted by candidates), we have the majority of data for each candidate and contested race. This gives us a total of 42 candidates in contested primary races for analysis.
Model 1: Total Votes Received
We ran one statistical model, given that the relatively small number of races (15) caused multicollinearity among the independent variables in the logistic regression model. This model explains the factors that predict the total number of votes received by each candidate using an ordinary least squares regression model with a continuous dependent variable (goodness of fit diagnostics available upon request). Specifically, we control for gender, race, having a Hispanic surname, endorsements by the organizations listed above, listing order on the ballot, “Well Qualified” Bar Poll votes, and the amount of money raised on the January 15, 2010 and 30-Day Contribution & Expenditure reports. While each of these variables exerts some effect on the number of votes received by each candidate, we are interested in those relationships that reach statistical significance ― that is, relationships that do not occur by chance.
Results: Total Votes Received
The model results indicate that two factors were particularly influential in predicting Republican Primary vote totals (number of additional votes for each effect in parentheses): endorsement by the Conservative Republicans of Harris County (+39,530) and endorsement by the Houston Police Officers Union (+12,014). Both effects are strong and the model is “fit” well from a statistical standpoint. We can be equally sure these results are robust: almost regardless of which sets of variables were tested in alternative models, the same results held. None of the other variables we gathered attained statistical significance.
Summary: The 2010 Republican Primary in a Nutshell
The results of this analysis seem clear: akin to the recent Democratic Primary, endorsements appear to be central to the likelihood of victory in a low salience, low turnout election when holding all other factors constant. For Republican candidates, backing from the Conservative Republicans of Harris County and\or the Houston Police Officers Union were the most important factors in this cycle. In a stripped-down logistic regression model (not described here) that used only gender, a few endorsements, the January 15 Report of funds raised, and ballot position, only endorsement by the Conservative Republicans of Harris County achieved statistical significance.
It is interesting that such endorsements appear to be relatively more important than ratings from the Bar Poll, endorsements from the Chronicle, or ballot order listing. This suggests that targeting one’s voters carefully and narrowly is critical to primary electoral success. Spending more money or being well-known among one’s legal peers appears less important in predicting additional votes than having the support of committed voter groups.
University of Houston
A spreadsheet with all of the data is here. My thanks to Dr. Rottinghaus for sharing these analyses with me.