Output convergence revisited: new time series results on industrialized countries

Abstract

Cross-country output convergence is re-examined using a flexible concept of unit roots. While the presence of a constant unit root in output-differences implies nonconvergence, the presence of a stochastic unit root on the contrary implies convergence. Using the output-differences between the USA and the other 14 OECD countries, we find output divergence only for the USA/UK and USA/Sweden country-pairs.

Acknowledgements

Yau acknowledges the research support from the National Science Council of the Republic of China (NSC86-2415-H-030-005-T).

Notes

¹ Exceptions include Cunado et al . (2003), who allow fractional root in the output differences; Linden (2002), who uses a nonlinear model. Nahar and Inder (2002) also demonstrate that standard unit root test is inappropriate and propose a new methodology. They all find convergence of outputs between the USA and some OECD countries.

² Before conducting the STUR test, we use Kalman filter to estimate (1) with an AR(1) of α _t . The mean of the absolute values of the estimated ranges from 0.056 to 0.119. Therefore, the approximation in (2) is appropriate. The mean of the estimated α _t ranges from 0.990 to 1.000.