The standard approach to the estimation of unemployment persistence assumes that quantile parameter heterogeneity does not matter. Using panel quantile autoregression techniques on state-level data for the United States (1980-2010), we suggest that it does.