Looking at President Obama’s popularity in Iowa, you might think Democrat Bruce Braley would have little chance of winning the U.S. Senate seat there. Obama’s average approval rating in the state, according to live telephone polls, is just 39 percent. That’s lower than his national approval rating (which is terrible).

But Braley is still in the game, partly because he’s had some success disentangling himself from the White House with a crucial voting bloc in the Hawkeye State: He’s held on to non-college-educated white voters even as they’ve abandoned Obama.

Obama’s standing with working-class white voters has been poor nationally for some time. Iowa was an exception. Obama won 52 percent of non-college whites there in 2008 and 52 percent again in 2012. Winning that group helped win him the state; working-class whites make up between 55 and 65 percent of the state’s voters, depending on the survey.

But that was then. By early summer 2014, the president’s approval rating among non-college whites in Iowa slipped to just 34 percent, according to an average of Marist and Quinnipiac polls. (A note: The cumulative size of these subsamples is well above 1,000 voters.) His average approval rating among college-educated voters in these same polls hovered around 50 percent — roughly equal to Obama’s performance with this group in 2012.

In other words, Obama’s drop in the Iowa polls was almost entirely because non-college whites started to feel the same about Obama as non-college whites in the country as a whole did.

The same dynamic was evident in Senate race polls. In Marist’s and Quinnipiac’s early-summer surveys, Braley led Republican Joni Ernst by 13 percentage points among college-educated voters, but he trailed among non-college voters by an average of 3 points. (Not all polls release racial breakdowns, but about 95 percent of Iowa voters are white, so college-educated voters are a very good proxy for white college-educated voters — same with non-college educated voters.) Braley led overall by an average of 2 percentage points; he was doing better than Obama across the board, although he was being dragged down by a lack of success among non-college whites.

Obama still has a problem with non-college whites in Iowa, but Braley’s standing with them has shifted: He’s now doing equally good (or bad) among Iowans of different educational levels. In an average of the past two Marist surveys, Braley has trailed among college graduates by 5 percentage points. He has done slightly better among non-college voters, losing them by an average of just 1.5 points.

So, Braley is trailing because he lost ground among college-educated voters over the summer. The voting patterns in the Senate race no longer match the patterns in Obama’s approval polling. If they matched, Braley would be toast.

Of course, Braley isn’t doing quite well enough among either college or non-college graduates to win. He’s a modest underdog, with a 36 percent chance of winning, according to our Senate forecast. But if Braley pulls off the victory, it will probably because he was able to separate himself from Obama among working-class whites.