Standard econometric tests for whether money causes output will be meaningless if monetary policy is chosen optimally to smooth fluctuations in output. If U.S. monetary policy were chosen to smooth U.S. output, we show that U.S. money will not Granger-cause U.S. output. Indeed, as shown by Rowe and Yetman [2002. Identifying policy-makers objectives: an application to the bank of Canada. Canadian Journal of Economics 35 (2), 239-256], if there is a (say) 6 quarter lag in the effect of money on output, then U.S. output will be unforecastable from any information set available to the Fed lagged 6 quarters. But if other countries, for example Hong Kong, have currencies that are fixed to the U.S. dollar, Hong Kong monetary policy will then be chosen in Washington D.C., with no concern for smoothing Hong Kong output. Econometric causality tests of U.S. money on Hong Kong output will then show evidence of causality. We test this empirically. Our empirical analysis also provides a measure of the degree to which macroeconomic stabilisation is sacrificed by adopting a fixed exchange rate rather than an independent monetary policy. © 2007 Elsevier Ltd. All rights reserved.