Abstract The paper is concerned with tests for Granger causality and for cointegration between (log) money and (log) output in Germany. Augmented Dickey—Fuller tests show that only second differences of both variables are stationary and only first differences (and not levels) are cointegrated. Furthermore it is shown that there is most times feedback between both variables (in terms of second differences) and therefore neutrality is rejected. Finally the effects of anticipated and non-anticipated money on output are analysed. Within this framework neutrality of anticipated money is also rejected.