New evidence on the income convergence among OECD countries

This paper illustrates the importance of accounting for cross-correlations among panel data sets in examining stationarity. Strazicich et al. (2004) utilized the LM unit root tests with two trend shifts to examine the income convergence among twenty-one OECD countries. We revisit their analysis using a more recent dataset and two novel unit root tests that account for cross-correlations in the errors. Our study reveals that incorporating common factors yields different results from models neglecting them. Our findings show no significant evidence of stochastic income convergence among the OECD countries from 1870 to 2022. Recent studies show that failing to control for cross-correlation can lead to bias in panel estimation. A similar issue applies to unit root tests commonly used to examine convergence and integration effects.