According to the Marquette Poll, the races for President and Senate are essentially tied with Tammy Baldwin down by a point and the President up by a point among likely voters.
Meanwhile the NBC-Marist Poll shows the President up by six (51-45) and Baldwin up by four (49-45) among likely voters.
Why the difference? Since both polls of registered voters are nearly identical-- Marist has the presidential race at 50-44 and Marquette has the race 50 to 45-- something happened significantly different in each poll's likely voter screen.
Marquette says their screen for likely voters is based on poll respondents likelyhood of voting: If they answered "absolutely certain" they are considered a likely voter, while if they answered "very likely," "50-50" or "not voting" they didn't make the cut. Meanwhile, Marist bases their likely voter designation on three things: likelyhood of voting, interest in the election, and past election participation.
Marist's likely voter universe consists of 77% of registered voters, while Marquette's likely voter universe consists of 87% or registered voters.
So, Marquette cut-out fewer voters, but the voters they did cut out were overwhealmingly Democratic voters: Of the 13% they cut-out, Obama leads by a 59% to 27% margin.
Of course, this is the same "likely voter" screen that Marquette used when Obama was up by 14 points a few weeks ago-- the difference is that the first debate energized republicans and depressed democrats and that had a big effect on Marquette's likely voter screen.
Since the Marquette Poll was taken before the second debate, I expect that more Democrats will return to Marquette's "likely voter" screen when the next poll is taken.