[Editor's note: If you're reading the Building a Voter File series in order, you should be aware that this particular entry is not part of the main sequence; thus, Errata.  I will return to the tutorial in short order.]

As Dirty D points out, I can indeed lay out some cautionary tales regarding the construction of vote history.  This can be an extremely important field, too, so if you're using vote history it behooves you to be an active consumer.  Ask questions of your vendor, and don't be shy about digging into the data.

First of all--what's the use of voter history? In short, past voting behavior is one of the best indicators of future voting performance.  If someone has voted in every election possible, odds are they aren't going to miss the next one.  It can be useful as a tool to identify targets as well--for instance, if you're running in a midterm, people who only vote in presidential generals might good targets to bump up to the next level of voting in all general electiosn.  This can be extremely important in constructing a model of likely voters to conduct a poll as well.

So what do we need to do to build a useful vote history?  As always, the answer comes back to standardization.  Every state will give vote history in a different format, and we need to come up with a way to make useful comparisons. 

One important step is to define what the types of elections are.  For instance, we might say that "general" elections consist of federal races, plus state legislature and governor's races.  We might distinguish presidential primaries from other primary elections.  And so on.  Needless to say, these categories will become more fluid as we get to the lower-turnout elections like bond issues, municipal referenda, and other such; but in general, having a strictly-defined category system for elections is a Good Thing, since it means we can compare vote history over different states. 

The next thing to do is figure out what the state has actually given us.  For instance, look at the Ohio file again.  It only lists general, primary, and special elections; moreover, some of the "generals" it lists occured in years like 2001, despite the fact that Ohio state leg and governors races occur in even years only.  Classifying elections like this is always somewhat of a judgement call; a good voter file team will do a little research and try to come up with the best fit possible for each election.

Moreover, note that coverage over states will necessarily vary.  Some states are very good at keeping track of tiny little elections--500 people voting in a water district election, a few thousand for a bond issue in a rural county, and so on (some, in fact, are too good, keeping track of hundreds of separate elections, which might make life a tad confusing).  Others only keep the big stuff--statewide primaries and generals, for instance. 

Other problems can come in when the state doesn't maintain consistent standards.  For example, different counties might report the same election using different codes.  Or some counties may be better about keeping track of vote history than others (this can be especially true for marginal elections like those I mentioned above).  Vote history is always most reliable for the largest and most prominent elections.

Overall, building an accurate vote history is one of the more involved parts of building a good voter file, so it pays to ask questions.  If you're going to be using vote history--and it can be very useful--then the more you know about what you need and what you're getting, the better.