All roads lead to convergence? Tracing the many paths from laggard to leader

This paper studies unconditional convergence across sectors using a newly constructed 12-sector cross-country panel of value added and employment covering the past five decades. Guided by standard neoclassical growth theory, it estimates sector-level convergence rates and examines the contributions of within-sector productivity growth and structural change. It incorporates a shift-share term into the specification to attribute the share of productivity growth by sector and their within- and between-sector components. The paper also uses unsupervised machine learning to understand the general patterns of growth and convergence over the last three decades. This paper offers four key findings. First, using a new dataset, it affirms the literature’s recent harmony of findings that economy-wide unconditional convergence is present and has been for several decades. Second, many sectors display unconditional convergence, especially highly-tradeable services sectors like business (including ICT), finance and real estate. Third, within-sector productivity improvements in the agricultural sector explain a third of convergence over the period 1990 to 2018, while labour reallocation into manufacturing explains around 10 percent. The remaining 57 percent is attributable to a diverse range of sources. Fourth, this paper tests both theory-based and machine-learning-derived country groupings, and finds evidence of unconditional convergence in every group, albeit at different speeds.