In retrospect, it’s amazing that the 2008 presidential bid of Sen. John McCain (R-Ariz.) was as strong as it was. Incumbent president George W. Bush was deeply unpopular, largely a continuation of frustration over the Iraq War. The economy was crumbling. Between January and November 2008, the unemployment rate rose from 5 percent to 6.8 percent; within a year, it hit double digits. Barack Obama’s historic candidacy would have been important regardless of when he ran, but he couldn’t have asked for a better environment in which to do so.

A new study from researchers at Georgetown University and Columbia Business School, though, suggests a factor to that victory that was much more important than the spiking unemployment rate: the ability of Americans to get mortgage credit.

My first reaction upon reading the analysis from Alexis Antoniades and Charles W. Calomiris was that the link between mortgage availability in 2008 and the election results was at least intuitively clear. The economic collapse was in large part a function of banks extending mortgage credit to unqualified individuals and then selling those bad mortgages as a separate, bundled product that got a pass from credit-rating agencies. The people holding those mortgages defaulted, kicking off a cascade of losses for financial firms. There’s the link.

But it’s simpler than that. Antoniades and Calomiris looked at the change in mortgage credit supply — in broad terms, the ability of people to get mortgages — in each county between 2004 and 2008 and found a direct link between how much that supply contracted in a county and its presidential vote. That held even, controlling for other factors such as demographics and homeownership numbers. What’s more, they found a similar link in the 2012 election, after another four-year period in which the mortgage credit supply had contracted in most counties.

However! There was no effect in the other two elections the researchers looked at, those in 2000 and 2004. During the four years before those races, the mortgage credit supply had expanded.

“The way we made sense of it,” Calomiris said when we spoke by phone last week, “was to say, if you get your mortgage approved, you don’t credit the other people. You just look at that as natural. You’re a good guy, you do well, you deserve to have your mortgage approved. But if you get your mortgage rejected? You’re more — for cognitive dissonance reasons, let’s say — more likely to try to be blaming others for that.”

So Bush got blame for the mortgage credit crunch in 2008 — but didn’t get political benefit from it in 2004.

How big was this effect? Big. Calomiris compared it to the unemployment rate, which, as noted, is often cited as an important factor in presidential voting.

“With respect to the effect of incumbent parties in presidential elections, if the unemployment rate gets better, that is lower, the incumbent party does better. If the unemployment rate gets higher, the incumbent party does worse,” he said. “We found that the contraction in the supply of mortgage credit … had an effect that was five times as large as the unemployment rate.”

They even put that into concrete terms related to the 2008 election.

“In the counterfactual world that our model imagines,” Calomiris said, “if you hadn’t had any change in the supply of credit from 2004 levels to 2008 levels … Barack Obama still would have won, but he would have had a much narrower margin of victory. He would have lost half of the votes in the swing states that could have reversed the election.”

McCain would have won North Carolina and might have won Indiana. (The black bars on the graphs above show margins of error.) Still not enough to win Ohio or Florida — but in Florida, McCain could have closed the gap with Obama by nearly 100,000 votes.

The obvious question is why this would be such an important factor in presidential voting. Why is this five times as important as the unemployment rate, in Antoniades’s and Calomiris’s estimation?

“In unemployment,” Calomiris explained, “you may be unemployed, but, if so, you’re always in the minority of people.” As noted above, the unemployment rate in November 2008 was under 7 percent — high for an unemployment rate, but still directly affecting less than 7 percent of those actively working or looking for work. “You may know people some who are unemployed. The unemployment rate is a much more, for many people, an abstract economic statistic,” he continued. “But getting your mortgage rejected is not an abstract economic statistic. That’s something a lot of people experience.”

About two-thirds of Americans have mortgages. The overlap among owning a home, age and income — and the fact that not moving often means you don’t have to remember to update your voter registration — means that homeowners are more likely to vote.

Data from the Current Population Survey shows the extent of that effect: Three-quarters of voters in 2016 lived in owner-occupied homes, places where one of the residents was the person who was paying off a mortgage or who had already paid it off. And people living in such homes were much more likely to turn out.

This doesn’t explain the effect by itself. Homeownership — including paying a mortgage — means being much more likely to vote. Interestingly, a 2012 analysis from researchers at the University of Wisconsin and Marquette University found that unemployed people were less likely to vote, though overall turnout increased as unemployment rates rose.

“An increase of one percentage point in turnout at the county level increases turnout by about .13 percentage points, a modest effect,” researchers Barry Burden and Amber Wichowsky write. “As in the previous models, a one-point change in state unemployment raises turnout by more than half a point.” The effect is smaller than that for a downward shift in the mortgage credit supply, where, Calomiris said, a one-point contraction led to a six-point drop in support for the incumbent party relative to four years prior.

Again, a world in which mortgage credit is plentiful doesn’t necessary equal a world in which the parties of incumbent presidents always win. Nor did the contractions leading into 2008 lead to Obama’s victory, nor, obviously, did the contraction leading into 2012 lead to his defeat. This analysis, though, suggests that mortgage credit supply might be a more important factor to consider moving forward than the rate of unemployment.

Scott Clement and Emily Guskin contributed to this report. The description of the study from Burden and Wichowsky has been corrected.