Standard unit-root tests of the hysteresis hypothesis specify a unit root under the null against the stationary alternative of the natural-rate hypothesis, making the two theories of unemployment mutually exclusive over the sample period. In this paper, we allow switches between hysteresis and natural-rate theory using the Kejriwal, Perron, and Zhou test. The null hypothesis of the test is that the unemployment rate is I(1) throughout the sample, and the alternative hypothesis is that the unemployment rate changes persistence [i.e., switches between I(0) and I(1) regimes]. We apply the test to the unemployment rate of 20 metropolitan statistical areas (MSAs) and the USA. We use monthly observations over the period 1990:1–2016:12 and apply the test to seasonally unadjusted and seasonally adjusted data. Important differences exist between these tests. We find that with seasonally adjusted data, the Great Recession associates with a change in persistence from I(0) to I(1) in eight MSAs and the USA and to a change from I(1) to I(0) in six MSAs. Conversely, with seasonally unadjusted data, the Great Recession only associates with a change in persistence from I(0) to I(1) in four MSAs and to a change from I(1) to I(0) in three MSAs. This differential resilience to the shocks of the Great Recession provides a new aspect of the heterogeneity of the US labor markets.