The permanent income hypotheses is one of a class of assumptions that economists make about how people smooth their spending over time. These assumptions predict that people prefer to spread their consumption smoothly over their life based on their expected lifetime earnings. Variations of the permanent income hypothesis are frequently used to simplify the mathematics behind economic models so that they are easily solvable.

But this begs a question: Are the permanent income hypotheses and its variants reasonable assumptions? Because this hypothesis was developed and popularized in an era before data were sufficient to make them readily testable, it is important to periodically revisit these long held beliefs as new information becomes available. In a new National Bureau of Economic Research working paper, Michael Gelman and Matthew D. Shapiro at the University of Michigan, Shachar Kariv and Steven Tadelis at the University of California, Berkeley, and Dan Silverman of Arizona State University does just that—looking at whether and how government workers smoothed their consumption during the 2013 government shutdown.

The authors examine the consequences of the 16-day federal government shutdown in October 2013, when Senate Democrats and House Republicans could not pass a spending bill in large part due to disagreement over defunding the Affordable Care Act. According to an Office of Management and Budget report, the shutdown resulted in a combined total of 6.6 million lost work days amounting to $2.5 billion in compensation and lowered that quarter’s Gross Domestic Product by billions of dollars.

Using detailed data on spending and earnings from the Mint program (a free website/app used to track individual finances), the authors find that consumption dropped sharply among government workers. This spending drop occurred even though the shutdown would have little if any impact on their lifetime earnings because they knew they would be paid later for the time they did not work. The reaction of the workers, then, is at odds with commonly used versions of the permanent income hypothesis.

The authors did find that some of the drop in consumption was because people delayed paying certain bills, such as their mortgage bill or some credit card debt. While this allowed some of the workers to partially smooth their spending over time, there still was a real drop in consumption by most of the government workers impacted by the shutdown. Thus, the majority of these workers did not smooth out their earnings over time based on their expected long-term earnings but rather immediately reacted to the situation at hand.

This finding is important because it provides real insight into how people respond to income shocks, which has implications for how we think about credit constraints, unemployment, and other economic issues. Because people respond to short-term shocks in their income by dramatically cutting back on their spending, programs such as unemployment insurance that replace income immediately for displaced workers could be much more effective at smoothing consumption than temporary tax cuts that are not seen until April 15. Currently, strong interpretations of the permanent income hypothesis predict the unemployment policies and tax cuts deliver equivalent outcomes.

This new work by Gelman, Kariv, Shapiro, Silverman, and Tadelis is important because it pushes back against a set of flawed but long-held beliefs about how the economy works that can lead to bad policy. By bringing data to the debate, these authors are helping to advance the science of economics.