Today's post brings us to a fun intersection.  So far, my colleague,Blue Leader, and I have been writing in parallel.  Our posts appear around the same time, but they don't really intersect with each other.  This is the post that you've been waiting for! The clash of the Titans! The Rumble in the Jungle! Okay, it's nothing like that.  In fact, it's really nerdy.  What we are going to discuss today is making strategic use of your voter file when designing your poll, which brings us to an intersection with Blue Leader's excellent series, Building a Voter File.

In my last post, we started the discussion of knowing what to ask about.  You obviously need to divvy up your time between a certain amount of dependent and independent variables, and that's what we're going to talk about today.  Stepping into your role as campaign manager, you want to commission a message testing poll. You only have the money for 500 respondents at fourteen minutes length.  You don't want to cut out any of the dependent variables you're testing, but at the same time, you don't want to lose that valuable contextualising information.  What are you to do?   This is where having data in which you have high confidence becomes invaluable.

There's more...

To quote from Blue Leader,

Once the data is (yes, is, prescriptivists--I went there) in a standardized format, we move from the realm of "interesting" into "faintly creepy".  The information from Secretaries of State or state parties is generally pretty innocuous--name, address, maybe phone number or age.  The appended consumer data, on the other hand, is more unsettling.  There's nothing on there that would do real damage if anyone knew it--no credit card numbers, nothing that people could use to steal your identity--but it can be kind of strange to think who realizes that you own two dogs and a cat.
 
Most of this consumer data is gathered by for-profit companies, who then retail it to both the state parties and the for-profit companies that are creating these files (if you take a look at our resources page, InfoUSA is one such vendor).  They get their information anywhere they can--state licensing agencies (think it might be worthwhile for the McCain campaign to know who has a gun license?), magazine subscription lists, grocery store value card memberships...basically, if you have to fill out a form for it, somebody wants it, and will get it unless prevented by law.

Let's ignore for a minute just how ugly to the ear it sounds when someone uses data as a singular noun and look at the ideas that he's put forth here.  When pollsters are collecting the lists of whom they're going to call, it's usually from a list like the ones Blue Leader is describing.  If I'm getting a little worried about having the time to ask all the questions I want to ask, it's theoretically possible for me to skip asking the questions in the poll and rely on the data I get from my vendor to supplement it.  For example, let's say that I get a little worried about time, and I decide to see what I can cut from the independent variables section of my poll. If I trust that Blue Leader generally delivers me good data, I can say, "Well, he has race, gender, age and party registration on the file. Maybe I can cut those questions from the poll, rely on the file and that'll open up some room for me."  Depending on the circumstances, this is the right tactical decision to make.

So how do you know when you can get away with this?  There's an easy answer, but it's not one that you're going to like.  One word: validation.  What this means is that you have to look at previous projects and see how accurate the data on the file were compared to what you found when you asked the questions.  Let's say that Blue Leader is able to get vote history, gender, age and income dead on, but for some reason, he's terrible about coding race.  You would know this by checking the answers to your questions for vote history, gender, age, income and race and counting the percentages of how many records were listed correctly.   Each campaign is able to make their own decisions as to what constitutes an acceptable level of validation, but if you're above sixty-five per cent, it's an option you want to keep in mind.

You also want to measure by what degree the listed data were off.  So, for example, let's say that he lists two people as being in one household and has their separated incomes listed on the file.  When you call, however, the respondent gives you the combined household income. If you check the file and find that what Blue Leader gave you added up correctly to the combined household income, you've found a fix to your problem.  Of course, you then have to validate what you've found.  Similarly, if Blue Leader is listing people as Hispanic in a state where the preferred term is Latino, you've found a fix.

Obviously, there is no perfect substitute for field acquired data.  That's why we do the polling in the first place.  If you have a data vendor whom you trust implicitly, however, you have a valuable partner in your undertaking.

Dirty D