We estimate the elasticity of vacancy duration with respect to posted wages, using data from the near-universe of online job adverts in the United Kingdom. Our research design leverages firm-level wage policies that are plausibly exogenous to hiring difficulties on specific job vacancies, and controls for job and marketlevel fixed-effects. Wage policies are defined based on external information on pay settlements, or on sharp, internally-defined, firm-level changes. In our preferred specifications, we estimate duration elasticities in the range −3 to −5, which are substantially larger than the few existing estimates.

Using micro data for the UK and Germany, we provide novel evidence on the cyclical properties of reservation wages and estimate that wages and reservation wages are characterised by moderate and very similar degrees of cyclicality. Several job search models that quantitatively match the cyclicality of wages tend to overpredict the cyclicality in reservation wages. We show that this puzzle can be addressed when reservation wages display backward-looking reference dependence. Model calibrations that allow for reference dependence match the empirically observed cyclicality of wages and reservation wages for plausible value of all other model parameters.