Admin Note: We apologize for the tardiness of this post. Dirty D got swamped in his day job and forgot to send me his contribution for editing and posting.  We'll try to be on time in the future.

In my last post, we began a conversation about using your list strategically.  This was the beginning of the intersection with Blue Leader's awesome series, Building a Voter File. I'd like to continue that conversation today.  When we left off, we were talking about using variables on a list in lieu of questions in a poll, and how you could trust what's on there.  I'd like to continue that conversation a bit today.

There's more...

Blue Leader riffed in with this

A good voter file, along with quality polling, are complements, not competitors, and when used together can be much more powerful than if they're isolated from one another.

I didn't mean to imply that polling data were a competitor to what's on a file.  Here's what I will say, though. It's probably more accurate and suitable to your needs.   Given that most of what's on a listed file comes from models that are built on purchased data, what's on the file is layers and layers away from self-reporting, which is important.  Each level of modeling and derivation introduces more and more possibilities of inaccuracy (what we call error), and that's what makes field acquired data so much more valuable.

Now, there are times when you have to rely on recorded data, like vote history lists, because people have a tendency not to remember when and if they voted, and, more importantly, will often over-report their participation.  Also, people tend to alter their memories to remember voting for the victor.  This is why, if you have a good data vendor, you should always trust what they have on file from the Secretary of State when it comes to vote history.

I hope that the distinction here is clear: for data that rely on reproducing data that are listed and acquired from the field, like vote history, it's pretty safe to rely on your voter file.  For things that are modeled, though, your preference should be to validate unless you are completely comfortable with the modeling techniques that your vendor uses.

Now, we discussed validating the data on your voter file...This leads to an obvious conclusion: never, never, never trust modeled data the first time that you use a data vendor. It shouldn't matter what other people have told you or what kind of relationship you personally have with the vendor. In this case, it's all about the data.  The first time that you use a vendor, it's worth the extra money to ask all the questions you can and then compare them back to the sample file to discover where the holes in coverage are.

That's all for now, but we're going to continue this in my next post, too.

Dirty D