Testing Growth Convergence with Time Series Data— a non-parametric approach

A non-parametric time series testing is suggested to analyse the convergence of international output per-capita gaps. Non-parametric tests are based on signs and ranks of time series properties of output differences. The methods are applied to logs of USA percapita income differences for 16 OECD countries from 1900-97. In contrast to the results obtained by Bernard & Durlauf (1995) for the period 1900-87, convergence of output gaps was evident for the majority of countries. However, the trends in 1970-97 and 1987-97 are noticeably more complicated than the homogeneous convergence found in the pre-1970 period. The results indicate that widening USA gaps are now more likely to emerge than steady-state or narrowing gaps.