Download Full Text (756 KB)


Most contemporary macroeconomic models account for unemployment by making the simplifying assumptions that 1) there is an equilibrium level of unemployment and that 2) when the economy is not at that level it will tend towards equilibrium. Implicit in these models is also the assumption that the actual behavior of unemployment does not affect the equilibrium level. This paper joins a growing number of economists pointing out that such assumptions are false: the equilibrium does depend on past behavior, a trait called hysteresis. This paper considers the hysteresis hypothesis by using an iterated version of OLS to construct a series for equilibrium unemployment. Regression analysis shows strong evidence that actual unemployment does affect its equilibrium level. This paper also focuses on one of the specific channels through which hysteresis supposedly works, called ranking. The ranking hypothesis asserts that while unemployed, workers’ skills degrade as they lose their connection to the labor market, so employers rank their potential new hires based on duration of unemployment. If the ranking hypothesis is true, then when average duration of unemployment rises, workers are less hirable and the equilibrium level of unemployment should also rise. Using statistical filtering to analyze the timing of changes in unemployment duration and the equilibrium level, this paper finds evidence for the ranking hypothesis and the hysteresis hypothesis. However, incorrect timing of events provides strong evidence against the connection between the ranking and hysteresis hypotheses, despite both hypotheses likely being true. Since unemployment is not self-correcting, policy must step in.

Publication Date



Unemployment; Hysteresis (Economics)


Labor Economics | Macroeconomics

Primo Type

Conference Proceeding

Jobs Don’t Grow on Trees