Total factor productivity and the convergence hypothesis

We study the convergence, or lack thereof, of total factor productivity and real gross domestic product (GDP) per worker for a pooled (cross-section, time-series) sample of developed and developing countries, adding breadth and depth to the convergence debate. We first estimate total factor productivity from a parsimonious specification of the aggregate production function involving output per worker, capital per worker, and the labor force, both with and without the stock of human capital. Then we test for absolute and conditional convergence of total factor productivity and real GDP per worker, using cross-section and cross-section, time-series data. Fixed-effect estimates across countries convert the cross-section test of absolute convergence into a pooled test of conditional convergence, since it controls for country-specific effects. Our tests consider both beta- and alpha-convergence. Our findings support both absolute and conditional beta-convergence of total factor productivity, but only conditional convergence of real GDP per worker. Further, sigma-convergence tests must measure absolute convergence, since conditional convergence assumes that an equilibrium dispersion of total factor productivity or real GDP per worker exists. We find mixed evidence for absolute sigma-convergence. (C) 2002 Elsevier Science Inc. All rights reserved.