Thursday, October 11, 2012

The latest on growth accounting

This the early days of the Solow growth model, we are accustomed to the decomposition of GDP growth into the contributions of capital, labor and the "Solow residual," sometimes interpreted as technology or total factor productivity. Roughly, each contributed a third, and this distribution has changed relatively little once measurements or models have been refined, for example in adding other forms of capital, such a human capital or public capital.

This is revisited by Robert Tamura, Gerald Dwyer, John Devereux and Scott Baier (with an impressive data appendix) who reassemble data from 168 countries and are especially carefully in constructing new estimates for human capital. Instead of adding years of schooling and possibly years of experience, they use a human capital production function that depends also on the parents' human capital and the world's frontier. In the end, the residual accounts for much less than previously, at most a third in some extreme specification, a tenth in the other extreme. The rest is split quite evenly between physical and human capital. What this means, is that technology, institutions and whatever else you want to throw in the Solow residual accounts for much less than we previously thought in per capita output growth and in differences of output per capita across countries.


Vilfredo said...

The question in undoubtedly important, but can we really rely on results from data that is very likely to be of poor quality?

Kansan said...

EL, you have yourself pointed out the dangers of such panel data: Why growth regressions are so scary