Unemployment persistence and quantile parameter heterogeneity

We argue that a random-coefficients representation of the classical Barro’s model of unemployment dynamics can be used as a theoretical basis for a panel quantile autoregressive model of the unemployment rate. Estimating the latter with State-level data for the United States (1980–2010), we find that (i) unemployment persistence increases along quantiles of the conditional unemployment distribution; (ii) disregarding State-fixed effects implies an overestimation of unemployment persistence along unemployment quantiles; (iii) a macroeconomic shock changes not only the location but also the dispersion of the distribution of the State unemployment rates; (iv) a federal policy equally applied in each State can reduce unemployment inequality among States; (v) “hysteresis” and “natural rate” hypotheses can co-exist along quantiles of the unemployment distribution, with the former being not rejected at upper quantiles. In sum, while the standard approach to the estimation of unemployment persistence implicitly assumes that quantile parameter heterogeneity does not matter, we suggest that it does.