Former Bush speechwriter Michael Gerson has a column up at The Washington Post arguing that Obama isn’t delivering on his promise of bipartisanship—that he is, in fact, incredibly polarizing. “Who has been the most polarizing new president of recent times? Richard Nixon? Ronald Reagan? George W. Bush?” Gerson asks in a series of non sequitur queries. “No,” he writes, “that honor belongs to Barack Obama.”

The basis for his argument is a survey the Pew Research Center released last week of Obama’s job approval ratings. Obama’s overall approval rating was 59 percent. (Bush’s rating in 2001 was 55 percent.) But while 88 percent of Democrats had a favorable opinion of Obama’s performance, only 27 percent of Republicans did. That indicates a partisan gap of sixty-one points, a number that the report compares to Bush’s fifty-one point partisan gap in 2001. After citing the numbers, Gerson states: “Obama has been a unifier, of sorts. He has united Democrats and united Republicans—against each other.” And again, later: “Obama’s polarizing approach challenges and changes the core of his political identity.”

Gerson, though, skips over an important point in parsing these numbers. Yes, the gaps seem to speak for themselves—there’s a bigger difference between partisan-based opinions of Obama than there was for Bush a few months into his first term.

But those numbers don’t account for the gradual change in respective sizes of the two groups—as in, what slices of the entire pie they make up. Nate Silver has a smart post about it over at FiveThirtyEight.com, noting: “measurements of the partisan split in support for the President…are not quite as straightforward as they might seem.”

Silver points out that the percentage of Republican voters is shrinking—just 24 percent of voters identified themselves as Republican when Pew conducted this survey in March, which Silver writes “is roughly as low as that total has ever gotten.” Look at last week’s job approval ratings with that knowledge, and the numbers assume some additional meaning, more nuanced than the one Gerson affords them. Here’s Silver:

We see some evidence of these effects in the comparison of Obama’s numbers to those of George W. Bush’s at a comparable point in his presidency. Obama and Bush had roughly the same level of support among members of their own party (88 percent for Obama, 87 percent for Bush) and roughly the same level of support among unaffiliated voters (57 percent for Obama, 56 for Bush). Bush, however, had more support from the opposition party (36 percent of Democrats versus 27 percent of Republicans). And yet Obama, not Bush, had the higher overall approval rating, because Democrats are a significantly larger constituency than Republicans.

So yes, Obama has a smaller percentage of Republican voters’ approval than Bush had of Democrats’, but that smaller percentage also comes from a smaller slice of the voter pie.

Gerson bypasses that numeracy nuance in his column. He does, to his credit, attempt to elucidate the wider partisan gap by explaining (with some help from Ron Brownstein) that the political parties have become “more ideologically uniform,” that media outlets have become more partisan, that ideological interest groups have become more common. But he puts the onus of the sixty-one point partisan gap on Obama. “Obama was supposed to be the antidote to the poison of partisanship,” he writes.

And that’s the problem—not the conclusion he draws, but the logic he uses to get there. It’s true that Obama has found bipartisanship a harder sell (and harder to execute) than he or his followers might have expected (cue: stimulus bill). But the partisan gap figure doesn’t necessarily, or exclusively, prove that point; rather, it illustrates a bigger and more systemic story—of Republican voters becoming a smaller and more ideologically distilled group, and of changes in the nature of partisan identification—than the one Gerson chooses to tell.

This isn’t entirely on Gerson. Silver thinks the survey itself could have been better constructed. Calling partisan identification too fluid, he suggests “a more telling measure might be to see a breakdown in support by voters who identify themselves as conservative, moderate or liberal.” That might produce a clearer and less hang-up-filled picture of the population’s opinions, but it won’t deter the uncritical usage of its results. That requires columnists to take a step back and think about how numbers are constituted, and what they do or don’t actually prove.

Jane Kim is a writer in New York.