There's been some new polling on statewide races from Franklin & Marshall College. As the P-G has already noted, overall the results suggest that not much has changed: For one thing, massive numbers of people haven't picked a candidate for governor yet. 

But I'd like to hone in on one weird little finding in the race for U.S. Senate.  

That race will almost certainly feature Republican Pat Toomey facing the winner of this May's Democratic primary -- either incumbent Arlen Specter or challenger Joe Sestak. If you look at results among registered voters, things don't look so bad for the Dems. Here's how Franklin & Marshall breaks down the possible November match-ups: 

33% Specter, 29% Toomey
25% Toomey, 22% Sestak

But when F&M considers "likely voters," the result is much different

44% Toomey, 34% Specter
38% Toomey, 20% Sestak

See what happened there? The Dems did about the same among both groups of voters ... but when you focus on likely voters, Toomey got a double-digit percentage point increase. 

Others noticed the disparity as well. The DailyKos observed that "there is usually some gap between registered voters and likely voters, but this strikes me as more than a little absurd."

What's going on here? I put a call into Terry Madonna, F&M's polling guru. Madonna acknowledges there's "a big swing in those numbers," but that they reflect a race where a lot is in flux. "Polls are all over the place now," he says. But he adds that in any case, the numbers show an uphill climb for Dems. 

As noted here previously, there's no fixed definition of a "likely" voter -- different pollsters determine who is likely based on a variety of different questions and procedures. In this case, Madonna says he determined a voter was "likely" to vote based on previous voting history and how closely the voter claimed to be watching the campaign. 

Madonna then used a very narrow cohort of voters: The ones whose answers suggested they would vote in an election where turnout was only in the mid-30-percent range. That's "probably on the low end" of what turnout has been in similar contests, he says -- a somewhat less pessimistic forecast on turnout would have improved numbers for the Dems. But "To be candid, I don't have a clue about what turnout will be in November. So many things can happen between now and then." 

What the numbers reflect now, he says, is a serious "enthusiasm deficit" among Democratic voters. The voters signalling a strong interest in this race now are disproportionately Republican, which explains why Toomey does so much better among "likely" voters. 

"I think Democrats will lose lots of seats this year," Madonna predicts. "There are cycles in politics, and you can't usually reverse these trends completely. But if Democrats can increase their turnout, they can maybe mitigate some of the damage." 

I've raised the enthusiasm gap before. But I've cited it as a reason for why Sestak could be a larger threat to Toomey than Specter. Clearly, this poll only supports the first part of my argument. I mean, if Sestak were lighting people up, you'd think that he would fare better in a poll of likely voters as well.

There's no question the guy is trying -- he's been out here in western PA a lot more than Specter has, and his campaign staff is, well, highly dedicated. But I just don't see sign that he's getting much traction. 

If he's lucky, that's just because of the icy road conditions of recent weeks -- people might be too busy digging out their cars to dig into the issues. But the way these numbers are shaping up, there may be a bigger storm in the offing.