Three professors (two from Dartmouth, one from UCLA) have published a paper (PDF) showing why many pundits believe that the recount favors Al Franken.

It mostly comes down to what are often called "undervotes" (although these three professors complicate the terminology and call them "residual ballots." Read the full paper if you need to know why.)

If you are just tuning in, "undervotes" is a term that became familiar during the Florida recount of 2000. It refers to a ballot that, when read by the vote-counting machine, registers a valid vote in some races but not others. Did the voter intend to skip a race, or simply make a mark that the machine couldn't read?

In the current case, the relevant ballots were mostly those that registered a preference for president but none for senator. Apparently there were more than 30,000 such ballots that registered no preference in the Senate race.

The vast majority of these will turn out to be exactly what the machines first thought they were — ballots cast by citizens who intentionally left the Senate portion blank because they didn't want to vote for any of the candidates. It's no secret, for example, that Barack Obama was more popular than Franken (or else we wouldn't be having this recount, now would we?) Some Obama voters simply didn't vote in the Senate race. An additional clue, according to the three professors, is that the concentration of such undervotes was high in precincts with a lot of African-American voters.

But some of voters who cast those approximately 34,000 undervotes surely tried to vote for a Senate candidate (this is confirmed based on the first night of the recount) but made a mark on the ballot in the Senate race but the machine didn't read, presumably because they didn't fill in the oval thoroughly enough.

In many of these cases, the visual examination of the ballot will determine the voters' intent and someone will gain a vote. If 10 percent of the undervotes turn out to show discernible voter intent, that would be plenty of votes to change the result, depending on the breakdown of the newly counted votes.

The next step in the analysis is knowing where these undervotes were concentrated. The three professors analyzed those precincts and found that they were mostly areas of DFL strength in the last two elections. Therefore they concluded that the majority of the undervotes that turn out to be countable will be votes for Franken. They imply but do not state that this will be enough to change the outcome. And this is only an educated surmise.

Bear in mind, in addition to the undervotes, there are those absentee ballots that were rejected. We don't know yet whether the Canvassing Board will look at those ballots to see whether some of them should be counted (as the Franken campaign has asked) or whether they will be subject to court action that follows the recount.