Generic Congressional Polls: Believe Them . . . or Not?

It's no surprise that the topic attracting the lion's share of the attention this campaign year is Republicans' chances of regaining control of the House after two disastrous election cycles (with the possibility that the Senate may switch hands starting to be taken seriously, as well). And perhaps one of the most-watched measures of how things are going is the generic congressional ballot.

The question on the generic ballot is usually some variation of this: "If the election for the U.S. House of Representatives in November were being held today, would you vote for the Democratic candidate or the Republican candidate in your congressional district?"

The venerable polling organization, Gallup, has been doing the generic ballot since 1950, and it captured headlines at the end of August with a survey that showed Republicans with an "unprecedented" 10-point lead over Democrats among registered voters. That was the largest margin in favor of one party or the other in Gallup's history of tracking midterm elections.

Polling guru Nate Silver wrote, "When a poll produces an 'unusual' result, it simply reflects random noise and the best advice is to wait for the next edition of the poll to come along, when more often than not it will revert to its previous position." Silver, who considers Gallup's generic ballot an important element of the models he uses for predictions, made that observation not as a criticism of the pollster, but as the reason he does not often comment on individual polls.

Silver added, "If you were to take the exact same survey and put it into the field again - but interview 1,450 different registered voters, instead of the ones Gallup surveyed - you would most likely not find the G.O.P. with a 10-point advantage."

And, when Gallup did go into the field again between Aug. 30 and Sept. 5, it found, in its poll released Tuesday, that Republicans and Democrats were tied at 46 percent each.

Between those two Gallup surveys, a Washington Post/ABC News poll had Republicans ahead 47 percent to 45 percent among registered voters, but with a bigger 53-percent-to-40 percent advantage among likely voters; the Wall Street Journal/NBC News poll had a 43 percent split among all adults, but a 49-percent-to-40 percent advantage for the GOP among likely voters; and, CNN put the Republicans ahead 52 percent to 45 percent among registered voters. When CNN included "all Americans," the Republican edge was 49 percent to 45 percent. (For a round-up of those polls and links to the polls themselves, click here).

So, what to make of the generic ballot? After all, people don't vote for "generic" candidates on election day. They also tend to favor their own representatives, as is illustrated by the fact that only a fraction of the 435 House seats are really in play in any given cycle, although this year there is more ferment out there.

(The University of Virginia's Larry Sabato rates 29 seats as true toss-ups; the Cook Political Report puts that number at 45 and CQ Politics at 36. When races that are leaning one way or another but still considered competitive are added in, the total usually comes to something over 70).

Gallup says its generic poll, particularly when election day draws nearer and it shifts to "likely voters," has proven to be a good indicator of how the final result will turn out.

Scott Keeter, director of survey research at the Pew Research Center, agrees: "We think the generic ballot is a useful predictor of the national two-party vote for Congress, which itself is a good predictor of swings in the partisan distribution of House seats. Given that conducting individual House district polls is not feasible on anything like the scale you'd need for a national prediction, the generic ballot is the next best thing."

But how to square the generic numbers, as a predictor, with the fact that only a minority of all the races are in doubt?

Keeter acknowledges that there "are many uncontested elections and others that are lopsided," but said the generic polling is important because "there are potentially many competitive races where a shift in the national mood or preference between the parties can matter."

"That was especially true in 1982, 1994, 1998, and 2006," he said. "In 1994, we saw the end of the old New Deal alignment in which conservative southern Democrats got replaced by even more conservative southern Republicans, due to a strong differential turnout between Democratic and Republican voters. The situation in 2006 was similar, except that the strong Democratic turnout edged out many Republicans in conservative districts. This year, those pickups are perhaps the most vulnerable to switching back due to the change in national mood."

Now that we're past Labor Day, the traditional start of the campaign, the focus of pollsters will be on likely voters -- something that puts even more importance on the measures of voter enthusiasm. (Gallup has found a 2-to-1 enthusiasm gap in its two latest polls in favor of the Republicans).

Nate Silver wrote, "We've found that the gap between registered and likely voter polls this year is about four points in the Republicans' favor -- so a 10-point lead in a registered voter poll is the equivalent of about 14 points on a likely-voter basis."

The shift to the "likely voter" model may produce even more variation in results among different pollsters, given the different ways they use to determine the makeup of that "likely voter" base. That's one reason "likely voter" samples usually have a larger margin of error than polls of registered voters. Pollsters weigh factors like voter interest in the election (enthusiasm), stated intention to vote, past voting history, and even awareness of where to vote.

"One thing that is tough about the current election is that Republicans have been fired up for most of the year, while Democrats have been somewhat depressed," Keeter said. "The normal situation is that most voters don't get engaged in the fall elections until the fall. So, a big question for the Democrats is whether their core voters are going to get engaged once the local campaigns get under way in earnest."