A new discussion paper by Kevin S. Nell and A. P. Thirlwall, KDPE 1703, March 2017.
‘New’ (endogenous) growth theory seeks to explain growth rate differences between countries outside the confines of orthodox neoclassical growth theory, but also to rehabilitate the neoclassical model with diminishing returns to capital by introducing other variables into the equations to explain why there has not been unconditional convergence of per capita incomes across the world as predicted by the basic neoclassical (Solow) growth model.
It is argued that because output growth is by definition equal to a country’s ratio of investment to GDP times the productivity of investment, if the investment ratio is included in a new growth theory equation, all that new growth theory is doing is testing for why the productivity of investment differs between countries. But the productivity of investment is never treated as the dependent variable. It is easy to do so, however, by dividing the whole equation by the investment ratio. This has the added advantage of being able to test directly the hypothesis that the productivity of investment falls as investment rises and as countries get richer (the neoclassical assumption of diminishing returns to capital), without relying on the sign on the initial per capita income variable, a negative sign on which could be the result of catch-up or faster structural change in poor countries than rich and not diminishing returns. The model is tested using the general-to-specific model selection algorithm, Autometrics, taking a sample of 84 countries over the period 1980-2011 and 19 different independent variables that have been highlighted in the growth literature. Nine of the independent variables turn out to be significant, the most important of which turn out to be export growth; education; latitude, and political rights. There is no evidence of diminishing returns to capital i.e. that the productivity of investment in rich countries is lower than in poor countries.
You can download the complete paper here.