RALEIGH – A while back, I discussed the use of generic-ballot polls for predicting the outcomes of congressional and legislative races. Here are a few more thoughts about the subject.

Right now, opinions about the generic polling tend to line up fairly well with the political preferences of those expressing opinions. Given that Republicans enjoy a sizable lead in the generic polling for Congress – an average edge of seven percentage points in both the Pollster.com and RealClearPolitics.com averages – many GOP-leaning analysts are extolling the predictive power of such polling while Democratic-leaning analysts are downplaying it.

It’s fine to have political preferences. They ought not to cloud one’s judgment of the data, however.

As an empirical matter, the generic-ballot test has been a fairly good predictor of congressional outcomes in the past. It doesn’t translate directly into the percentage of seats won, of course, but the general direction of the electorate is usually similar to the latest genetic-ballot polling before the election.

As I previously wrote, if anything generic polling tends to underestimate Republican support in midterm congressional elections, so there’s nothing odd or outlandish about assuming that a seven-point edge in the generic ballot will translate into major GOP gains in Congress.

When it comes to generic polling for legislative elections, however, the picture gets murkier. The simple fact of the matter is that there hasn’t been nearly as much publicly available polling of legislative races, in North Carolina or elsewhere, so it’s hard to draw firm conclusions from such a limited set of data.

For what’s it worth, here’s my own experience with legislative polling and outcomes.

Starting in 1996, the John Locke Foundation began a biannual statewide poll in conjunction with our Agenda series of briefing books and candidate briefings. While most of the polling was devoted to policy questions, on issues ranging from fiscal policy to school choice, we always included some political and horserace questions.

For our first poll, we contracted with Scott Rasmussen’s then-new polling company. From 1998 to 2004, we employed the services of Tel Opinion Research, a polling outfit in Virginia. After the founding of the Civitas Institute, which included monthly polls as part of its regular programming, we ended our own polling program (Tel Opinion is still one of Civitas’ polling providers).

On our Agenda polls for 1996, 1998, and 2002, we included generic-ballot questions for both the N.C. House and Senate. I went back and compared the findings of those polls to what actually happened in the subsequent legislative elections. I also did the same for a Civitas poll in 2008.

Here’s what I found:

• Overall, the generic polling proved to be a fairly good predictor of House races. In 1996, for example, 51 percent of the North Carolina voters expressing a preference for the House in the JLF poll said they would vote Republican. GOP House candidates actually got 51 percent of the statewide vote and won 51 percent of the House seats. Something similar happened in 2008. In the other two years, there was a significant difference between the GOP popular vote and the number of seats won in the House, so the generic polling essentially predicted one but not the other.

• In the Senate, generic polling has been a fairly good predictor of the share of votes cast for Republican candidates – but has tended to overestimate the number of seats won by the GOP. That’s because, at least until recently, the Senate districts were horribly gerrymandered. In 1998, for example, Republican candidates won a majority of the votes for Senate but only 30 percent of the seats (15 out of 50).

At this point, perhaps the safest conclusion to draw is that if generic polling is showing a sizable Republican lead in legislative races, they’re going to have a good year. But translating the margin into a specific prediction of seats won would be tricky.

Hood is president of the John Locke Foundation.

p.s. If you want to read up on the key races that define the 2010 election cycle in North Carolina, there’s no better place to go than Carolina Journal’s Exclusive Series Covering the 2010 Elections.