Using vector autoregressions on U.S. time series for 1957-1979 and 1983-2004, we find government spending shocks to have stronger effects on output, consumption, and wages in the earlier sample. We try to account for this observation within a DSGE model featuring price rigidities and limited...