It was during Bill Clinton’s successful 1992 presidential campaign that the phrase “It’s the economy, stupid” entered the popular political lexicon.

The saying was coined by Clinton advisor James Carville, who argued that the ’92 campaign should simply focus on the problem of rising unemployment, as Americans’ basic economic situations would drive their perceptions of the state of the country and the quality of leadership provided by President George H.W. Bush, Clinton’s opponent.

Since that time it’s been taken as an article of faith that one of the most important drivers of voters’ perception is economic conditions on the ground. But this election cycle may bring this belief into question.

According to Peter Atwater, socionomist and president of the research and analysis firm Financial Insyghts, this is because the correlation between actual economic conditions and people’s perception of the economy is breaking down, and perhaps wasn’t all that strong to begin with. “The transmission of our economic reality into our economic perception is not straightforward,” Atwater wrote in a recent blog post. He points out that there is a strong correlation overall between the economic confidence index compiled by polling firm Gallup and the unemployment rate. But when you dig deeper into the data, the relationship looks much weaker.

Atwater decided to analyze the relationship between confidence and the unemployment rate on a year-by-year basis. He found that the “relationship between ECI and Unemployment appears to change year to year—sometimes dramatically.” The following chart illustrates this fact, with each line representing another year of the recent economic recovery. As you can see, sometimes economic confidence goes up when unemployment goes down (as represented by the lines that slope downward from left to right), but sometimes just the opposite happens (as represented by the upward-slopers).

screen-shot-2016-09-14-at-12-45-13-pm

Atwater talked to an analyst at Gallup who tried to chalk these strange relationships to unusual events, like the debt ceiling crisis of 2011 and the government shutdown of 2013. Another theory, posited for the negative relationship between confidence and unemployment last year, was slowing growth, because Americans may take economic facts other than the jobless rate into consideration when deciding how they feel. But what about the fact that average wages rose faster last year than any year since 1967, when the Census Department began measuring the statistic? Surely rising pay should increase economic confidence, but according to Gallup it didn’t.

This research should worry Hillary Clinton, if you believe that voters’ perceptions of the economy drive their decision whether or not to support the incumbent party. And Atwater gives some reasons why Clinton supporters might latch on to this disconnect as reason to worry that positive economic data will not help her. He writes:

I would also caution against viewing specific events as contributors to our level of confidence . . . While [it] may seem intuitive that the Debt Ceiling Crisis and the Government Shutdown negatively impacted confidence in 2011 and 2013, respectively – it is far more likely that both events were in fact symptoms, if not outright consequences of falling confidence. When confidence is weak, zero sum thinking naturally takes hold and reaching a political compromise is especially difficult . . .

Many events that would intuitively suggest even weaker confidence levels to come frequently mark THE low in confidence. Rather than causing confidence to fall further, these events instead mark the bottom. For example, as [the following] chart shows, the 2008 banking crisis didn’t cause American confidence to fall further; it reflected that confidence had fallen already to an extreme. The collapse of Lehman Brothers – not to mention the “spontaneous” formation of the Tea Party – was a symptom of exceptionally weak social mood.

screen-shot-2016-09-14-at-12-57-19-pm

Atwater, then, is pretty sure that it’s not events that are causing changes in confidence, but changes in confidence that are driving events. But what actually causes waves of optimism and pessimism in people is much harder to understand. This thought, Atwater argues, can be unsettling, because people like to know why things are they way they are. But Atwater’s research has led him to believe that these changes, while mysterious, are real and important. They cannot be argued away by pointing to reality, because human beings aren’t necessarily convinced by objective facts. And that could spell bad news for the Clinton campaign come November.