Jonathan Alter’s most recent Newsweek column rightly takes Hillary Clinton to task for her insistence that more people have voted for her than Obama.

That claim needs to be aggressively footnoted, not only because Clinton seeks to include votes cast in Michigan and Florida’s rogue primaries. In Michigan, Obama removed his name from the ballot after the major candidates agreed not to campaign in the state. (And as I’ll get to in a second, in the absence of his name, many Obama supporters voted for “Uncommitted.”) That means Clinton suggests we tally some 300,000 votes from the state in her column, and none to Obama.

The party’s rules never foresaw the popular vote being a meaningful metric in the nominating contest, and considering the kaleidoscope of rules and systems—open primaries! Semi-open primaries! Closed-primaries! Non-binding primaries! Territorial conventions! Total support reporting caucuses! Non-total support reporting caucuses! The Texas two-step!—it’s downright impossible to have a consistent, agreed upon metric for how to count this number that was never meant to be counted. Until Clinton came to rely on it as one of her few remaining talking points, the candidates weren’t in a race to win the popular vote; they were, as both remaining campaigns once acknowledged, in a race for delegates.

So good for Alter for taking a stab. But he goofs when he argues that any fair accounting should give all of Michigan’s “Uncommitted” voters to Obama:

Beyond not being official numbers, there’s another problem with counting Michigan in these totals. Obama wasn’t on the ballot there. You can say this was his own choice, but that doesn’t change the fact that had he been on the Michigan ballot he would have received a lot of popular votes. How many?

Try 238,168. That’s the number of Michiganders who voted for “uncommitted.” Were they possibly genuinely abstaining? Maybe a few hundred of them at most. The rest were clearly Obama supporters who launched a grass-roots campaign. Everyone in Michigan knew on January 15 that a vote for “uncommitted” was a vote for Obama.

Wait. The 238,000 or so were “clearly Obama supporters”? That would come as news to the nearly 40,000 of that number who, judging from exit polls, were Edwards supporters. On January 15, he was still an active candidate gearing up for the Nevada caucus. And his name, like Obama’s, was absent from the Michigan ballot.

The exit polls suggest that 12 percent of voters in the primary favored Edwards, and of them, 30 percent ended up voting for Clinton, 11 percent for Kucinich, and 57 percent voted “Uncommitted.” So here’s the math: 57 percent of 12 percent is 6.84, and 6.84 percent of the total votes cast that day gives us, again, just over 40,000 Edwards voters backing “Uncommitted.”

Am I entirely comfortable with using an exit poll to guess at votes that were never cast? No. Reasonable people could disagree on the validity of that step. But it’s a far better idea than just signing over the total uncommitted vote to Obama, as Alter proposes.

The question is just another illustration of the impossibility of counting the so-called popular vote. Here’s another: as I noted above, the exit poll also suggests that significant portions of Clinton’s vote total (4 percent, by the same process) came from people who would have voted for Edwards had his name been on the ballot. Well, by the same process, the exit poll suggests that 6.3 percent of Clinton voters would have favored Obama if he’d been on the ballot. Those voters certainly undermine Alter’s claim that “Everyone in Michigan knew on January 15 that a vote for ‘uncommitted’ was a vote for Obama.” It looks like 20,000 or so people missed the message. And that leads to another question: Should exit-poll estimates of those who said they only voted for Clinton because their candidate was missing from the ballot be docked from her tally? Maybe so.

Clint Hendler is the managing editor of Mother Jones, and a former deputy editor of CJR.