Do large recessions reduce output permanently?

The slow recovery following the 2008/2009 recession has led to renewed interest in the question whether deep recessions lower real GDP permanently or whether we can expect a rebound to earlier trend levels. Using a recent quantile autoregression unit root test we check whether shocks to real GDP have permanent or temporary effects. In contrast to earlier studies this approach takes into account that the transmission of a shock might depend on the sign and the size of the shock. Large recessionary shocks might have a different effect than smaller recessionary or expansionary shocks. We do not only test the unit root hypothesis at the conditional mean of GDP, but also in the tails of the distribution where the lower tail corresponds to large recessions. The test has more power than conventional unit root tests. We find that positive and negative shocks including large recessionary shocks have permanent effects on output. Therefore, a rebound of GDP to its pre-crisis trend level is unlikely. Current output gap estimates based on deterministic trends are likely to be too negative and inflation forecasts based on these are likely to be too low. -- Unit root tests ; quantile autoregression ; GDP ; recessions ; asymmetries