Likely Voters, Elections, and "Plebiscites"
This item by Ed Kilgore was first published on March 10, 2010.
One of the oldest and hoariest debates among pollsters and political scientists is the measurement of public opinion according to likelihood to vote in a particular election. Some polls show results for "all adults," some for "registered voters," and some for "likely voters." This last category is especially useful, if perilous, in projecting election results. It's useful for the obvious reason that the views of people who don't wind up voting are irrelevant to actual election results. It's perilous because determining likelihood to vote is not an exact science, and moreover, can produce some serious distortions. Pollsters typically use two different methods for measuring likelihood to vote: some are subjective, mainly involving poll respondents' own expressed interest in an election, and some are objective, including past voting behavior, and most controversial, post-survey "adjustments" of raw data to reflect the expected composition of the electorate. "Adjustments," in fact, are one of those factors (others include question language and question order) the biases of pollsters or their clients can become pretty important, but in general, "tight" likely-voter screens have recently produced results more favorable to Republicans.
Aside from measurement factors, there are two important reasons why going into the November elections, "likely voters" are more likely to lean Republican than "registered voters." The first is that historically, midterm elections attract an older and whiter electorate than presidential elections; given the weakness of Barack Obama among old white voters even in his 2008 victory, that's significant. The second is that likelihood to vote measures intensity of political engagement, and right now, there's little question Republicans are more "energized" than Democrats. So I'm certainly in full agreement that Democrats have what Jonathan Chait recently called (after examining the latest Democracy Corps/Third Way data on "drop-off" voters) a "turnout emergency" in 2010
But it's a very different matter altogether to use public opinion surveys sifted for likelihood to vote in the next election to measure the current "mood" of the American people on this or that issue--in other words, to treat polls as a sort of plebiscite on the wishes of the electorate as a whole. You see this every day when conservatives argue that "the people" or "America" has rejected health reform because likely 2010 voters in a poll tilt heavily against some formulation of health reform legislation. Such polls may well indicate a possibility that voters in November will react poorly to the enactment of health reform, but do not present a fair representation of public opinion on the subject. No one would seriously argue that only those voting-eligible adults who get through a pollster's LV screen are "people" or "Americans." So no one should use LV data to construct some sort of plebiscite. LV's will have their say in November. Let all Americans have their say when they are asked to express it.