We analyse the short-term impact of social distancing measures on the US labour market, using a panel threshold model with high frequency (weekly) data on unemployment across US states. We find that changes in the restrictiveness of mandated social distancing, as measured by the Oxford Stringency Index, exert a strong immediate impact on initial unemployment. The unemployment rate is not immediate affected but follows within a very short time (two to four weeks). We also document a substantial asymmetry between tightening and easing: the impact of tightening restrictions is twice as large as that of easing them. The state of the endemic, proxied either by cases or fatalities, constitutes a marginal factor.