This paper uses the Backus and Kehoe (1992) data to test the long-run
implications of the neoclassical stochastic growth model for Canada, u
sing Johansen's (1988) maximum likelihood approach for estimating and
testing long-run steady-state relations in multivariate vector autoreg
ressive models. Although two long-run equilibrium relationships are de
tected for the 1929-1983 subsample, the hypotheses that the log ratios
of consumption to output, and investment to output, are stationary ar
e both rejected.