The Determinants of Long-Run Economic Growth: A Conceptually and Computationally Simple Approach

Durlauf, Johnson, and Temple (2005) forcefully argue that the empirical analysis of economic growth is one of the areas of economics in which progress seems to be hardest to achieve and where only few definite results are established. Large sets of potentially relevant candidate variables have been used in empirical analysis to capture what Brock and Durlauf (2001) refer to as theory open endedness of economic growth. A large variety of different approaches has been and is used to identify variables relevant for economic growth. Many of the contributions employ model averaging estimators to tackle the uncertainty about the relevant variables. Sala-i-Martin (1997a) runs two million regressions and uses a modification of the extreme bounds test of Leamer (1985), used in the growth context earlier also by Levine and Renelt (1992), to single out what he calls ‘significant’ variables. Fernandez, Ley, and Steel (2001) and Sala-i-Martin, Doppelhofer, and Miller (2004) use Bayesian model averaging (BMA) techniques to identify important growth determinants. The former perform Bayesian averaging of Bayesian estimates, introduced by Leamer (1978), whereas the latter perform Bayesian averaging of classical estimates, proposed by Raftery (1995).