Likely Voters and Unlikely Scenarios

As I've been telling people all week here in Pittsburgh, there's ample reason for Democrats to be worried -- perhaps deeply so -- about 2010. Without major intervening events like 9/11, the party that wins the White House almost always loses seats at the midterm elections -- since World War II, an average of 17 seats in the House after the White House changes parties. Democrats have substantially more seats to defend than Republicans, particularly in the House. They appear to face a significant enthusiasm gap after having dominated virtually all close elections in 2006 and 2008. And the economy and health care are contingencies that could work either way, but which probably present more downside risk to Democrats than upside over the next 12-18 months, particularly if some version of health care reform fails to pass. While the Democrats are not extraordinary likely to lose the House, such an outcome is certainly well within the realm of possibility (I'd put the chance at somewhere between 1-in-4 and 1-in-3). The Senate picture is a bit brighter for them, but they are probably more likely now to lose seats in the chamber than to add to their majority, in spite of the spate of Republican retirements in Ohio, Missouri and other states. In a wave-type election, a net loss of as many as 4-6 seats is conceivable.

With all that said, I would reserve some healthy skepticism for polls that apply aggressive "likely voter" models to elections like the midterms that won't occur for another 16 months. In Pennsylvania, for example, Rasmussen now finds Arlen Specter a 12-point underdog to Pat Toomey among what they define as likely voters. Toomey also leads a more "generic" Democrat, Specter's primary rival Joe Sestak, by 8 points in Rasmussen's polling. By contrast, Research 2000, which in its polling for Daily Kos also uses a likely voter model (but evidently a less aggressive one), puts Specter 5 points ahead of Toomey and Sestak one point ahead of the Republican. These numbers represent big downward shifts for the Democrats, particularly in Specter's case, since Research 2000 last polled the race in May. But obviously, there is a big difference between Specter's -12 number under Rasmussen's likely voter model and his +5 under Research 2000's.

We can learn a little bit about these likely voter models by evaluating other polls that these firms conduct. Rasmussen's likely voter universe, for instance, trusts Republicans more not just on hot-button issues like the economy and health care, but also on traditional Democratic strengths like Social Security (by 4 points) and education (by 3 points).

If the electorate that goes to the polls next November is in fact one which trusts Republicans more than Democrats on education and social security, then Democrats will lose the Senate seat in Pennsylvania and undoubtedly almost every other competitive race -- it will be really, really ugly for them. But I just have a little bit of trouble accepting that as a likely scenario. In 2004 exit polling, voters who listed education as their top priority went to John Kerry over George W. Bush by a 3:1 margin. As of pre-Katrina 2005, when Social Security was being polled frequently in what was not a particularly great time for the Democratic party, Democrats led Republicans by an average of about 15 points on the issue -- and that was long before the market collapse that would seem to have undermined Republicans' calls to partially privatize the system.

Is it possible that the electorate which is voting in November 2010 will be so down on the Democrats that they trust Republicans more on issues like these? Sure, it is possible -- if the enthusiasm gap is wide enough, if Obama's approval is low enough, if the health care debate has been bungled enough, and if the economy is still hemorrhaging jobs. But I'd consider it something of a worst-case scenario. That's probably the best way to regard these Rasmussen polls for the time being.