Unlikely Voter Screens

Today Reuters released polling commissioned from Ipsos on the Kentucky Senate race showing the contest tied at 40 percent apiece among registered voters, but with Republican Rand Paul leading Democrat Jack Conway by a 45 percent to 40 percent margin among likely voters. But just how likely is this likely voter screen?

According to Ipsos, 435 of the 600 registered voters polled are "likely voters," which would peg estimated voter turnout at 72.5 percent of registered voters. Interesting. Here are the actual turnout figures among registered voters in Kentucky for the last six federal elections:

• 2008: 64.0 percent turnout among registered voters
  
• 2006: 49.5 percent turnout among registered voters
  
• 2004: 64.7 percent turnout among registered voters
  
• 2002: 47.7 percent turnout among registered voters
  
• 2000: 61.3 percent turnout among registered voters
  
• 1998: 47.8 percent turnout among registered voters

Notice a trend there? You should. Turnout among registered voters hasn't topped 65 percent in any recent presidential election, and has actually been shy of the 50 percent mark during the last three midterm elections. It should be noted that these midterm turnout numbers include 1998, when Kentucky hosted a closely fought Senate election that was ultimately decided by less than 0.6 percentage points. And yet Ipsos believes that turnout in this midterm election will amount to 72.5 percent -- nearly 7 percentage points higher than any recent presidential election and 20 percentage points higher than recent midterm elections. Perhaps I'm using the wrong yardstick. Perhaps it would be more fair to look at the midterms of 1994 (.pdf), when Republican performance was at its highest in decades nationwide. In that midterm election just 38.8 percent of registered voters headed to the polls in Kentucky -- a rate more than 30 percentage points lower than Ipsos' projection.

Maybe it's the case that pollsters are catching an unprecedented trend, one that will see 20 to 30 percent more of the electorate vote this November than have voted in midterms past. Or maybe, just maybe, motivated Republicans are answering pollsters' calls at a higher rate than in previous years, skewing the results of "likely voter" polls...