The recent literature on convergence started from the observation by Romer that the growth rate of an economy appears to exhibit no correlation with the initial value of its per capita income. Barro marshalled an impressive battery of regressions showing that a negative correlation between initial income and growth rate could be observed when this correlation was taken conditionally upon a set of variables, the most significant of which was the level of school enrolment. This was initially interpreted as an indication that poor countries could catch up with rich countries, if only they were initially educated enough. In Discussion Paper No. 1163, Programme Director Daniel Cohen investigates this issue further, drawing on Barro's findings, but proceeding beyond one-dimensional tests of the `convergence hypothesis' to analyse a two-dimensional set of differential equations in which physical and human capital evolve simultaneously.

The paper offers new tests of the `convergence hypothesis'. It first analyses the pattern of growth of measured inputs (human and physical capital conventionally measured by an inventory method) and shows that these tests sustain the hypothesis. On the other hand, when the pattern of growth of revealed inputs (physical capital and Solow residual) is analysed, the convergence theory is rejected. In order to understand what lies at the heart of this discrepancy, the paper shows that poor countries fail to catch up with rich countries not so much because they fail to raise their school enrolments (or the UN-conditional convergence of the stock of measured inputs would not hold), but because the law of motion of human capital embodies a `stock of knowledge' which they fail to raise adequately.

Tests of the `Convergence Hypothesis': Some Further Results
Daniel Cohen

Discussion Paper No. 1163, April 1995 (IM)