We analyse Polish active labour market policy (ALMP) training programmes from a macroeconomic (regional) point of view. The effects of training programmes on the outflows from unemployment and the effects of all ALMP programmes on the outflows from employment (to identify displacement effects) are estimated. The variety of specifications presented is revealing. In contrast to other studies on Poland, we show that it can make a difference to the estimates whether current ALMP expenditure is excluded from the set of regressors to reduce the endogeneity problem, whether lagged dependent variables are included to take account of the dynamics, and whether fixed or random effects models are estimated. The empirical evidence gives some tentative support to the view that public training programmes can be used to reduce unemployment.