The productivity slowdown of the early 1970s continues to puzzle economists. A frequent explanation of this puzzle is that mismeasurement of output has worsened enough to help account for the apparent shortfall of output growth. Griliches (1994) highlighted one channel through which this worsening measurement could occur. He raised the possibility that—because output growth in the service sector likely is undermeasured—the rising share of services has led to greater undermeasurement of overall economic growth. This paper demonstrates that this argument is of little quantitative significance. Even under assumptions most favorable to the hypothesis, the rising share of services has had only a small impact on measurement error. These results—along with evidence from Baily and Gordon (1988)—make mismeasurement of output an improbable explanation for the productivity slowdown in aggregate data.

This paper reexamines duration dependence in U.S. business cycles using parametric hazard models. Positive duration dependence would indicate that expansions or contractions are more likely to end as they become older. This paper provides statistically significant evidence of positive duration dependence for expansions before World War II and contractions after World War II. The evidence is stronger than in earlier research utilizing nonparametric techniques, because certain nonparametric techniques have low statistical power against the type of duration dependence found in this paper. Evidence is also presented suggesting that expansions became longer, on average, after World War II, while contractions became shorter.

The paper first reconciles a variety of specification tests for partial adjustment money demand models and points out a fundamental identification problem which makes it impossible to distinguish between the real and nominal partial adjustment models if inflation has an independent effect on the long-run demand for money. The paper also finds that empirical estimates of simple partial adjustment models have some undesirable properties and then considers the shortand long-run effects of inflation in a more general distributed lag model.