Building A Voter Profile With Big Data

Yesterday we discussed how the “old” way of campaign strategy is to build voter lists based on their past voter habits. The idea being that if a voter casts a vote in the last one, two, three or more elections, then they are likely to cast a vote in the upcoming elections. The lists that the Chavez, Escobar and Fenenbock campaigns are likely using are derived from either the county’s voting records, the Democratic Party VAN, or both. Both Escobar and Fenenbock have likely enhanced their target lists by using additional metrics provided by their political consultants. Escobar likely also enhanced her data from Beto O’Rourke’s data and her previous election data. Either way, their data is limited to voter demographics and contact information along with some issue-related notes. However, Escobar has the added advantage of using O’Rourke’s linked data of families and friends that would be used to drive voters to the polls through peer pressure among the likely voters.

All three campaigns are targeting the same set of voters, concentrating on about 39,200 voters that are likely to cast a vote in March. (The average of all votes cast in the last four primaries) As you can see, the campaigns are fighting over 8% of the electorate, leaving a voter block of about 400,000 voters that are not expected to cast a vote in the primary. (As of February 13, there were 439,377 registered voters in the county.)

Now that the campaigns have their lists, they likely ranked them by affixing a voter score to each voter. For example, if voter A was eligible to vote on the last three elections and cast two votes, their score would be 66%. Obviously, the 100% voters are the prime voters the campaigns want to mobilize. But there is an inherent probability error with this technique. A twenty-something voter may have a 100% score because they were eligible to cast one vote and because of the novelty of that new right, they cast the first vote they could.

Are they likely to cast another vote is the question that needs to be asked.

It is unclear whether the campaigns are taking this potential problem into account, if not, it could skew the target lists.

Now that the voter lists are ranked, the next step is to match email addresses, telephone numbers and, possibly social media channels. Some of the more sophisticated campaigns match voters who have donated to the campaign previously as well as voters who regularly write campaign checks.

For the most part, this is where the campaign mailouts, door-to-door walking lists and robocalls are generated from. That, along with the mass media messaging, i.e. billboards, radio and televisions spots are the strategies being used by the three El Paso campaigns.

The notion is that the likely voter needs to be influenced to vote for the candidate targeting them.

Remember that the strategy targets about 8% of the target voter universe while ignoring, for the most part, the other 92%.

Thanks to big data, the El Paso Votes APP gives us better access to voter metrics that we can use to enhance our voter list.

Let’s pick on Claudia Perez, formally, Claudia Ordaz. Ignore for the moment that she is an elected official, let’s just see what we can glean about Perez via big data as a potential likely voter.

We know by looking at her voter profile that Claudia Perez has a voter score of 100% based on her record of voting in the last five general elections. I am using the November 2008, 2010, 2012, 2014 and 2016 elections as I believe that they offer more insightful data as to the voter’s propensity to vote in March. The campaigns still use the primaries as their dataset.

We also know that she is 32 years old and has been eligible to vote since 2004.

The El Paso Votes APP gives us more information that allows us to create a better profile about her. For example, we know that her criminal record is clean, except for one traffic violation in 2013 for failure to maintain financial responsibility. We also know that the house she lives in is owned by Vincent Perez, her husband.

The deed was transferred to Perez in 2013. We know that her house is worth about $104,000 and she pays about $3,234 annually in property taxes. This gives us a starting point about what her, and her husband’s income is. It also let’s us know if taxes might be a trigger issue for the couple. (Yes, I know both have no problem taxing others, it’s just an example.)

The El Paso Votes APP gives us information about their potential finances. We know that they, or rather Vince Perez, owed $97,375 on the house in 2007 and that it is likely that First National Bank holds the mortgage on the house.

From this information we get a snapshot of what the Perez’ finances may be like. For targeting them as a voter, we can use the property tax issue as one element in our profile. But, let us see what else we can find.

The El Paso Votes APP shows us that Claudia has one household member that is also a voter, Vincent. Vincent is 36 and has been eligible to cast a vote since 2000. Vincent has a 100% voter score, casting a vote in each of the last five elections in our data set. The APP also shows us that Jaime Abeytia is a “friend” of the couple, giving us a data thread to follow.

Jaime Abeytia is 42 years old and, also has a 100% voter score. Jaime cast votes in all five of the elections in our target list and has been eligible to vote since 2008. Abeytia lives in a small apartment complex on Old Country Road. This complex has about 18 units. That tells is that Abeytia’s income is in the lower spectrum of the county. Abeytia’s criminal record is well known so there is no reason to repost it here.

The El Paso APP gives us more information we can use to build our target list. We now know from the APP that Abeytia has an additional friend – Katheryn Rose Lucero. Lucero is 44 years old and has an 80% voter score. She missed casting a vote in 2010. Katheryn has used at least two different last names: Hairston and Lucero.

According to the El Paso APP, Katheryn has three criminal cases on her record. The first was in 2011, followed by 2012 and 2013. All three involve charges of her child not attending school. If our candidate was running for a school board, this information would be useful in crafting a message targeted to Katheryn for her vote.

In Katheryn’s household there are four additional voters. The house in which Katheryn lives is owned by two of the household voters. (As they are not public officials I’m leaving their names off this post, but the information is available in the APP) The 71-year-old female has a voter score of 80% and the other, a 64-year-old male has a voter score of 75%. The house is valued at about $91,628 and they pay about $2,849.00 annually in property taxes. The house was built in 2005 and purchased by the couple in 2011 for about $106,236,000. There are two other voters in this household. One is an 18 year-old-female who just became eligible to vote in 2017 and the other is 24 year-old-female with a lackluster voting record of 50%, casting one vote out of two eligible elections.

We now know that these eight voters represent the average income El Pasoan and that they have mortgages and property taxes to contend with. We also know how to contact them, either via email, telephone or directly at their house.

As you can see from this example, Claudia Perez allowed us to create a likely voter list composed of three excellent voters (100%) and five additional voters that have average records but can be leveraged to vote by their links to Claudia. We now have a total of eight voters we can target.

You’ll note that this information was readily available on the El Paso Votes APP.

But we need more information about the voters to determine what would trigger them to cast a vote for our candidate. This is where data mining comes into play. We’ve only hit the tip of the iceberg. Tomorrow, we’ll investigate building a psychographic profile on our voters so that we can target them individually with our messaging. Psycho?!? What?!? I know, but hang in there, it’ll be extra cool.