By Eric Brescia
To win elections, we need data. Lost in the recent primary-versus-convention debate is an issue often relegated to a footnote: the benefit of identifying hundreds of thousands of Republicans statewide via the primary process.
Presidential primary voter data is extremely important and is produced on a scale that conventions simply cannot match. As we all know, Virginia does not have party registration, so the main method campaigns use to identify partisans is via their primary election participation history. While whom a voter casts a ballot for is private, their participation in the primary is public information.
This is the type of data that is used when volunteers make phone calls for GOTV efforts or are given a walk book to knock on doors. It is used to send direct mail pieces to households and to do analysis in media buys. Its importance cannot be overstated.
For my day job, I’m an economist who conducts predictive analytics and often builds statistical models. I’m also on the Arlington GOP Executive Committee and have been active in campaigns in Northern Virginia since 2007, including running myself in 2009 for the House of Delegates. Most recently, I was the campaign manager for both of John Vihstadt’s successful back-to-back County Board elections, a rare feat in deep blue Arlington.
The cross between my political activism and professional career provides a useful perspective on this issue. In recent years, the application of predictive analytics and statistical modelling to voter outreach is becoming increasingly sophisticated and important. Just as retailers and websites offer customized messaging and advertisement targeting, so are political campaigns.
For example, commonly used voter databases now contain “support propensity scores” that indicate the likelihood of individual voter support for a given issue. These values can be used to micro target messaging and to focus campaign resources.
For Vihstadt’s campaign, I was able to build predictive models to identify likely support and opposition regarding pertinent local issues. I used polling information, demographics, and crucially, primary data, to predict likelihoods of individual voters’ propensity to agree with our issues and to vote for our candidate. These models helped us decide which Republican voters needed extra nudging to get out to the polls, and which Independents and Democrats were most likely to cross over.
Based on this, we sent different mailers to different sub populations based on predicted likelihoods associated with different issues; we created walk lists and call sheets; it gave us insights into Facebook ad targeting and the purchasing of cable ads.
We employed a much more sophisticated data strategy than most local campaigns currently do. However, I expect techniques to become even more advanced and more widely used. And the Democrats are surely going to employ them even if we do not.
None of this would have been possible without having primary election voter data. The last time we had a widely attended presidential primary was in 2008, and our data is becoming poorer as time passes. The amount of known Republicans diminishes every year.
This has particular consequence for places like Northern Virginia, where the population is transient. The median start of residency for an Arlington voter is in 2007, which means that close to half of Arlington’s registered voters were likely not in the County the last time good data was collected.
No other method for ID’ing voters is as effective. Polling completion rates are falling over time. Robo polls net only a 2-3% completion rate, and are useless for the growing number of cellphone-only households. In my experience, live polls only achieve about 7-8% completions, and are costly. Further, it is impractical in most cases to canvass for party ID purposes (absent primary data) on the needed scale.
There is no better way to identify who is who than a widely attended presidential primary.
To the extent that there are concerns regarding “mischief voters,” my analysis has shown that few voters engage in this type of cross over. In Arlington, only about 10% of those with a primary voting history have voted in both party’s primaries, and for those that have, the vast majority were one-off occurrences. The amount of people who will cross over in any given year is minimal, especially when both parties are holding them.
But for those who do, the analysis showed that a mixed primary voter was more likely to be open to vote for Vihstadt, as compared to a person with an exclusively Dem primary history. In addition, when you find a voter like this, it has predictive value for others in the household as well. I found that cross over voters have, on average, an independent streak; these are the types of voters that we need to win statewide, and it is worthwhile to identify them.
I am joining others in asking that the State Central Committee support a primary for 2016. Not doing so would be unilateral disarmament of one of the most important campaign tools there is – data.
I doubt the presidential campaigns this year, or our next round of Senate and governor candidates, want to do without it. And we may not have another opportunity for 8 years, a time when 2008-based data will be mostly worthless. Not collecting this now will have negative consequences for elections up and down the ballot for years to come.
A primary is greatly needed.
Eric Brescia is an Arlington GOP Executive Committee Member, a former House of Delegates candidate, and former campaign manager for Arlington County Board Member John Vihstadt.
