As there is a lot of confusion about the nature of challenged ballots in Minnesota, a field guide of sorts is in order. From what I can gather, a campaign has four principal reasons to challenge a ballot. For purposes of this exercise, everything will be written from the standpoint of the Coleman campaign -- the four reasons that Coleman might want to challenge a ballot. If you want to think of things from the Franken campaign's point of view instead, just substitute Franken every time you see Coleman, and vice versa.
Type 1: LL (Legal Vote/Legal Vote).
Description: Ballot initially counted as a legal vote for Franken. Elections judge rules it a legal vote for Franken. Coleman challenges, argues that it shouldn't count.
Disposition: Franken loses one vote from his count. If challenge is rejected, Franken gets that vote back.
Rationale: Usually, a campaign will argue to disqualify a ballot for one of three reasons; (i) they claim it is an undervote, e.g. the voter did not fill in the oval all the way; (ii) they claim it is an overvote, e.g. there appear to be markings beside more than one candidate; (iii) they claim that the voter marked the ballot -- e.g. revealed his identity through a signature or some other distinguishing means (under Minnesota law, marked ballots are considered void).
Frequency: This is by far the most common type of challenged ballot, probably representing something like 75-80% of all challenges (or perhaps more). This is also the type of ballot most prone to gamesmanship, e.g. you can deduct a vote (temporarily) from the opponent's score even if the challenge has a low chance of being upheld. There is some evidence -- although it is debatable -- that a higher percentage of Coleman's challenges fall into this category. If so, that is good news for Franken, since these ballots have already been deducted from his total, but are likely to be added back once the challenge is rejected.
Prognosis: I would surmise that very rarely will this type of challenge prevail -- probably not more than 5% of the time, and perhaps not more than 1-2%. Minnesota has a fairly liberal (lower-case 'l') voter intent law, and if a ballot has been accepted by both the machine scanner and by the elections judge, good luck getting it rejected. About the only successful challenges, I would guess, will revolve around cases when the voter did appear to unambiguously mark the ballot, such as by signing it or providing his Social Security number.
Type 2: LX (Legal Vote/No Vote).
Description: Ballot initially counted as a legal vote for Coleman. Election judge rules that the vote isn't legal. Coleman campaign challenges, arguing that the vote should count after all.
Disposition: Coleman loses one vote from his count. If challenge is accepted, Coleman gets that vote back.
Rationale: If a machine scanner initially counted a ballot, and the elections judge then rejects it, the campaign will usually have a self-explanatory reason to appeal and will probably do so some large fraction of the time, as human beings will usually count more ballots than machines. A fair number of these cases probably involve marked ballots (see above).
Frequency: Probably about 5-10% of all challenges. May be slightly more common from the Franken campaign, if Democrats tend to be responsible for a higher fraction of sketchy ballots.
Prognosis: This is probably the type of challenge with the best chance of prevailing, since (i) the elections judge and machine scanner disagree, and (ii) Minnesota's voter intent law is in the challenger's favor, and will tend to be interpreted to count more, rather than fewer votes.
Type 3: XX (No Vote/No Vote).
Description: Ballot determined by machine to be undervote or overvote. Coleman argues it is a legal ballot and should count. Elections judge disagrees with Coleman (but agrees with machine), rules that the vote is illegal. Colmean challenges, arguing that the vote should count.
Disposition: Since the vote wasn't counted in the first place, nothing changes in the state's count. If challenge is accepted, Coleman gains a vote.
Rationale: By definition, the campaign will argue that voter intent is discernible. Among the more controversial cases will be those where voter appears to have crossed the name of one (or more) candidates out before voting for another one; these will usually be read as overvotes by the machine but are allowable if voter intent is clear.
Frequency: Likely the next most common case after Type 1, probably accounting for about 10% of challenges. My inference is that this type of challenge is more common from the Franken campaign, as it appears to occur more often in blue precincts.
Prognosis: The challenge may have a fighting chance, but with both the machine scanner and the elections judge having ruled against the ballot, it's an uphill battle.
Type 4: XL (No Vote/Legal Vote).
Description: Ballot determined by machine to be undervote or overvote. Elections judge disagrees, rules that the vote shows intent for Franken. Colmean challenges, arguing that the vote should not count after all.
Disposition: Since the vote wasn't counted in the initial count, nothing changes immediately. However, Coleman has to win his challenge, or else Franken gains a vote. For that reason, you don't so much mind Type 4 challenges, since most will be rejected and you'll probably be gaining a vote.
Rationale: If the machine rejected the ballot the first time around, it will usually have some type of imperfection, as instances of true machine error are relatively rare among optical scanners.
Frequency: Somewhat self-limiting, since a prerequisite is that machine and elections official have to disagree ... unlikely to be more than 5% of challenges. May be slightly more common from the Coleman campaign, if indeed there are more marginal Franken ballots.
Prognosis: The campaign is on the wrong side of the voter intent presumption, although the fact that there was a disagreement between the machine and the elections judge provides hope. You'd certainly prefer to take your chances with this type of challenge than a Type 1.
That should account for almost all challenges, although there are a few oddball cases. Apparently, for instance, both campaigns may challenge the same undervoted (and rejected) ballot if they each think it shows intent for their candidate. Theoretically, a campaign could also argue that a vote has been counted for the wrong candidate (e.g. Coleman argues that a Franken vote should be a Coleman vote, rather than a no-vote), but I'd guess that these cases are exceptionally rare. One exception may be votes for third-party candidates (e.g. a Barkley vote that Coleman wants counted as a Coleman vote), where a campaign might have a slightly easier time closing the sale since the third-party won't have representation in the room; these instances would be essentially analagous to Type 3 challenges.
Takeaways and Extensions:
- Type 1 challenges are by far the most common type, and are probably becoming even more common as challenge wars escalate, since there's an essentially unlimited supply of legal ballots. However, casual observers seem to assume that *all* challenges are Type 1; they are not.
- Type 1 challenges are also the only type that result in an immediate deduction from the opponent's pre-recount total.
- There is more of a one-to-one correspondence between the number of challenges and deductions from the opponent's total in red (Coleman) precincts, suggesting that a higher fraction of challenges are Type 1 in such precincts. (n.b. Franken has challenged a higher percentage of ballots in such precincts).
- Type 1 challenges almost certainly have the least chance of being accepted.
- Type 2 challenges probably have the best chance of being accepted.
- Type 1 and Type 4 challenges are "no win" challenges. The vote is already being excluded from your opponent's total, which is the best you can do; if the challenge is rejected, the opponent nets a vote. You would like for as many of your opponent's challenges as possible to be Type 1 and Type 4 (especially Type 1).
- Type 2 and Type 3 challenges are "no lose" challenges. The vote hasn't been counted for you yet, but could be counted if the challenge is accepted. You would like for as many of your challenges as possible to be Type 2 and Type 3 (especially Type 2).
- Bayesian probability is largely responsible for dictating the relative frequencies of different types of challenged ballots. There are far more legal ballots than undervotes, and the elections judge will agree with the machine scanner more often than he disagrees with it. Therefore, Type 1 challenges are the most common (legal ballots where the judge agrees with the machine), and Type 4 are probably the least common (undervotes where the judge overrules the machine).