1980s Endogenous Growth Theory Discovered in the 1960s

I may be a geek, but this is my idea of fascinating. A number of core ideas of endogenous growth theory that made up a massive (and ongoing) area of macroeconomic research in the 1980s and onwards were discussed but ignored in a barely-read article in the 1960s, Frankel (1963) (gated link here). This was unearthed by Edmund Cannon in 2000 (gated link here):

An important strand of the growth literature of the last decade or so is the assumption that factors which can be accumulated indefinitely may have marginal returns which do not fall to zero for society as a whole. The importance of this assumption lies in the consequence that a society’s growth rate will depend upon its propensity to save…

…The model in Frankel’s paper is of the “AK” variety, where the aggregate production function is linear in capital due to externalities at the firm level. As such it anticipates many of the features of the models found in Paul M. Romer (1986), Robert E. Lucas, Jr. (1988), and Sergio Rebelo (1991)…

…Frankel’s paper has lain unnoticed for the last 26 years…

Cannon goes on to discuss possible reasons why this paper was ignored at the time, but seems to conclude that this is essentially a mystery:

Why the paper was ignored at the time remains a bit of a puzzle and perhaps serves as a demonstration of the role of chance in the research and growth processes.

Ironically, the fact that an excellent pioneering paper on endogenous growth theory went unnoticed suggests R&D yield is substantially driven by chance, suggesting economics cannot do a good job of explaining technological growth, which in turn is a big argument in favour of Solow-style growth models in which technological growth is exogenous.

PS. for anyone currently formally studying growth or otherwise curious about Frankel's early discovery, I strongly recommend reading Edmund Cannon's short and very readable paper.


  1. Gabriel M. Says:

    Will do!

    Great guest blogging, by the way!

  2. Anonymous Says:

    The fact that R&D yield may be substantially driven by chance is not a good reason for arguing that economics will not be able to model the process of growth. Economics routinely models processes which have random elements.
    For example, if R&D yield is a random process, the moments of the distribution (e.g.mean, variance) could be modelled as endogenous functions of the capital stock, number of skilled workers, number of universities etc.