A note on the power of money-output causality testsy

The role of money and its effects on national output have generated a voluminous amount of literature (Blanchard, 1990; Lucas, 1996; Sargent, 1996). Since Friedman and Schwartz (1963) rekindled the research on the effects of money on aggregate output, numerous studies have aimed to characterize and establish the interactions between money and output. Theoretical models are constructed to show that money can affect output via different channels, including unanticipated monetary shocks, real and nominal rigidities, and menu costs. However, the accumulated empirical results, which are derived from different speci®cations, different sample periods, and data from different countries, still do not give a de®nite answer to the question of whether money affects output. Also, there are diverse opinions on how money affects output. Most empirical studies adopt a vector autoregression (VAR) framework or a modi®ed one to analyze the causal relationship between money and output following the two seminal studies by Sims (1972, 1980). A basic regression equation has output as the dependent variable and lagged output and money as regressors. The basic speci®cation is often augmented with additional macroeconomic variables as controlling variables. The statistical signi®cance of the lagged money variable is used to evaluate the effects of money. Depending on the choices of sample periods, controlling variables, and data OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 63, 2 (2001) 0305-9049