Deterministic, stochastic and segmented trends in aggregate output: A cross-country analysis.

This paper examines whether output per capita in 126 countries is better described as trend or difference stationary using formal statistical tests. Appropriate finite-sample critical values are constructed to evaluate the test results. Depending upon whether one uses solely a test with a trend stationary null, or solely one with a difference stationary null, one obtains very different conclusions regarding the long run dynamics of GDP. This outcome suggests that it is useful to consider the tests complementary, rather than competing. Based on results from both tests, we find that when a definite characterization of GDP can be made, it is very likely to indicate a difference stationary process. However, the likelihood of making definite conclusions does vary positively with both the income level and the quality of data.