Cross-country regressions explaining output growth often obtain a negative
effect from inflation. However, that result is not robust, due to the selec
tion of countries in sample, temporal aggregation, and omission of conseque
ntial variables in levels. This paper demonstrates some implications of the
se mis-specifications, both analytically and empirically. In particular, fo
r most G-7 countries, annual time series of inflation and the log-level of
output are cointegrated, thus rejecting the existence of a long-run relatio
n between output growth and inflation. Typically, output and inflation are
positively related in these cointegrating relationships: a price markup mod
el helps to interpret this surprising feature. Copyright (C) 2001 John Wile
y & Sons, Ltd.