These last few installments in these series have focused heavily on how a pollster makes use of the lists given to her, and it's for a pretty good reason.  That reason is not, by the way, to promote Blue Leader's series on Building a Voter File.  The reason is that polling would really, really, suck if there were no lists. This is why our tag line reads "Analytics, data, modeling..." these three fields are so closely interrelated that I'm tempted to call them areas of specialty in one large area, and not three distinct undertakings.  Without list vendors out there, we pollsters would be severely limited in what we could do.

This brings us to what I'd like to talk about today, which is how you pick a data vendor. 

There's more...

The first step is to ask your colleagues around you for referrals.  While they most likely won't share their actual numbers with you on how the data validate, they can give you a really good sense of where the strengths and weaknesses of a vendor are.  The next step is to start looking at which fields each vendor offers.  You'd be surprised at how much this varies from vendor to vendor, but it's important to keep track of.  As a political pollster, there are a few things that are of immeasurable importance to you. Do not compromise on quality for these on your purchased file, or you'll suffer for it later.

1. Vote History - Even though this comes directly from the Secretary of State (in most cases), it's really important that you ensure that your data vendor has a good way of measuring and tracking vote history.  Some people take the extra steps of getting municipal elections, too, which is always nice.  Blue Leader can tell you: keeping track of vote history can be a nightmare when you're dealing with different levels of government and frequent updates.  This is really important to you. Take the time to talk to your vendor to find out how frequently they get this.
2. Name/Telephone (Phone Match) - No one's list is ever perfect. You will never have a list that has 1005 of the records with the correct name matched to the correct telephone number.  Give up on that right now.  Also, I'd say that giving up on having anything close to 90% is a good idea. Now that you've given up, start investigating what the phone match of the file in question is.  If you can't call someone, you can't ask her questions. This is the single most important thing on the file.
3. Race - If it's on there, you want it to be good.  This is one of the most frequently botched fields out there, so you should take great pains to make sure that you understand exactly how your vendor is coming up with that data.  Also, whether or not they get race correct is a pretty good indicator for how well they get other variables.

Notice something about this list?  Aside from one's name, two out of the three things that I told you to watch out for are two of the three most powerful predictors of voting behavior.  If they botch these things, you're not going to be able to get anywhere with your data vendor.

So what else are we looking for?  Well, there are ways to consider this. What is the breadth of coverage, and what is the depth of the coverage?  Put more concretely, how many different fields do they claim to offer you, and how much coverage do they have in each of those fields?  For example, it does you no good if your data vendor claims to be able to tell you about the petfood purchasing patterns of people, but can only do so for three percent of the file.

So what do you want? As with anything else, it depends on the specifics of the project. I would say that you should always prefer data vendors who get the basics right, so prioritize firms who get demographics and vote history well.  After that, start worrying about the frou-frou.  In other words, it's better to get a cup of coffee with no sprinkles than it is to get a cup that's otherwise empty, but, man, does it have some nice sprinkles.

Dirty D