Requiem for the unit root in per capita real GDP? Additional evidence from historical data

We present new evidence on the uncertain nature of nonstationarity – that is, trend stationarity vs. difference stationarity – of aggregate per capita real output, by submitting to a composite testing procedure a 20-country sample over an historically relevant time span. We find that the degree of uncertainty associated with the presence of a unit root appears to be well exceeding that shown by other studies conducted so far on cross-country historical data. For almost all the countries in our sample, inference appears to be strongly dependent on the type of test one makes use of, so that conclusions reached at early stages of a composite confirmatory testing procedure have to be frequently discarded at subsequent stages. Our reading of these findings points towards rejecting the assumption of temporal homogeneity of per capita GDP time series over long time spans, a prerequisite implicitly assumed in all studies looking for invariant statistical properties like stochastic or deterministic nonstationarity.First version received: October 2002/Final version received: September 2003