Will the Republicans take control of the House of Representatives this fall?
Elections for president lend themselves to simple horse-race polling questions, both nationally and in a dozen or so key battleground states. The same can be said for the battle to control the Senate, which will ultimately rest on a handful of contests that pollsters are already beginning to swarm over.
But public polling within individual House races is harder to come by. Our colleagues at Cook Political Report currently rate 65 House seats as "competitive," but right now we have seen independent polls in, at most, a few dozen districts. If the 2010 cycle is like the last two, we will not be ready use district-level polling to assess the overall seat count until well after Labor Day.
So for the next several months, the attention of political junkies will focus on a poll question usually referred to as the "generic" House ballot. This is the Gallup version: "If the elections for Congress were being held today, which party's candidate would you vote for in your congressional district, the Democratic Party's candidate or the Republican Party's candidate?"
When asked of a sample of likely voters a few days before the election, the question produces a reasonably accurate prediction of the national vote for Congress -- that is, the total votes cast nationally for Democratic and Republican candidates. The relative accuracy of the final-week result makes sense: By then, most voters in most districts have their minds made up and should know the party affiliation of their preferred candidate.
But what about now? Voters are far less likely to know the names of the candidates, and most pollsters are not yet applying their likely voter models.
Still, many are tracking the generic House ballot: The most recent Gallup tracking poll shows a dead-even result, with each party preferred by 46 percent of registered voters. Polls by other organizations show a wide range of results. As of this writing, our Pollster.com trend estimate shows a 1.5 percentage-point Republican lead, slightly less than the 2.3 percent GOP lead on the RealClearPolitics average.
Republicans appear poised to make major gains in November, but how many seats would that translate into "if the election were held today?" That's a tough question to answer.
To explain, let's go back to the relatively easy scenario. It's the week before the election. How would the generic ballot result translate into seats in the House? While the total national vote for Congress does not translate neatly into each party's share of House seats (due mostly to the advantages of incumbency), Emory University political scientist Alan Abramowitz created a relatively straightforward statistical model eight years ago that helps translate the final poll results into seats.
The Abramowitz model uses the final Gallup result for the generic House ballot, the number of seats held by the Democratic Party and the party of the current president to project the number of House seats with reasonable accuracy. When applied to the midterm elections between 1950 and 2006, it has predicted the outcome to a mean standard error of 5 seats.
Applied to 2010, the model tells us that a dead-even contest on the generic ballot just before the election will translate into a 32-seat Republican gain. To pick up the 40 seats necessary to win the House, the Republicans would need a lead of at least 3 percentage points on the final Gallup poll. If their margin grows to 7 or 8 percentage points, Republicans would see something on the order of the 54-seat gain they experienced in 1994.
But these seat projection statistics come with a huge caveat: They only apply to the final Gallup poll.
Complicating matters further is that the generic ballot typically overstates the Democrats' standing in the months before the general election. Political scientists suggest different mechanisms for this phenomenon, but the implications for 2010 are consistent: If past trends repeat, Republicans will be doing better on the generic House ballot in October than they are now.
It is important to remember that the various statistical models are typically based on no more than 16 cases -- the number of midterm elections held since pollsters started asking the generic ballot question. My Pollster.com colleague, University of Wisconsin political scientist Charles Franklin, created charts for each election that can help us reach our own conclusions. These show Democrats losing ground on the generic vote in the last 200 days of 6 of 7 midterm elections held with a Democrat in the White House. But notice that the one exception to the pattern was the most recent: 1998.
Some potential exists for movement in the generic ballot result in the coming months: "Over the last two decades," Franklin writes via e-mail, "we've seen movement of as much as 8 points over the last 200 days, but most years show as little as 2 or 3 points over time."
Either way, Democrats are facing a tough year. "If the Republicans gain another 2 or 3 points, that should give them a significant advantage," Franklin writes, "while if the Democrats recoup ONLY 2 or 3 points, they are still looking at significant losses, at least based on the history of the generic ballot and midterm seat change."