This paper uses a rich set of microeconomic labour market data—the 1988­90 Labour Market Activity Survey published by Statistics Canada—to test whether there is negative duration dependence in unemployment spells. It updates and extends similar work carried out by Jones (1995) who used the 1986­87 Labour Market Activity Survey. Applying hazard model estimation, the analysis finds some evidence of negative duration dependence at the microeconomic level, which is consistent with the de-skilling hypothesis of hysteresis. These microeconomic estimates of negative duration dependence are used to compute macroeconomic estimates of hysteresis in unemployment. The results suggest that hysteresis effects from de-skilling are very small at the macro level, contributing less than 0.1 percentage points to the aggregate unemployment rate. The small estimated size of this hysteresis effect may explain why evidence of hysteresis has been so difficult to find at the macroeconomic level. The paper also shows that Unemployment Insurance (UI) benefits reduce the probability of exiting from unemployment and that unemployment duration does not seem to be prolonged by reservation-wage effects.