1170 The Methods of Growth Econometrics
Acknowledgments
This chapter updates and extends our earlier survey, Durlaufet al.(2005). Durlauf
thanks the University of Wisconsin and National Science Foundation for financial
support. Johnson thanks the Department of Economics, University of Wiscon-
sin for its hospitality in Fall 2003, during which some of the preparation for this
chapter was completed. Temple thanks the Leverhulme Trust for financial support
under the Philip Leverhulme Prize Fellowship scheme. Finally, we thank Stephen
Bond and William Brock for useful discussions.
Notes
- See Temple (2000b) and Brock and Durlauf (2001a) for a conceptual discussion of this
issue. - This independence assumption is sometimes defended either on theoretical grounds or
because the parameter estimates are consistent with those predicted by the augmented
Solow model. - Given the role it plays in the analysis of convergence, the initial income variable logyi,0
is usually distinguished from the other Solow variables. - Other early studies include Baumol (1986), DeLong (1988), Grier and Tullock (1989),
Kormendi and Meguire (1985) and Marris (1982). - Note that while some failures of exchangeability call into question the interpretation
of the regression, this is not always the case. For example, a heteroskedastistic error,
while violating exchangeability, does not undermine the interpretation of the point
estimates of the parameters. See Draperet al.(1993) for further discussion of the role of
exchangeability in empirical work. - See the discussion in Brocket al.(2003) of the Ellsberg paradox.
- For further discussion of extreme bounds analysis, see Temple (2000b) and the references
therein. - In this discussion, we will assume that one of the models in the model spaceMis the
correct specification of the growth process. When none of the model specifications is the
correct one, this naturally affects the interpretation of the model averaging procedure. - Fernandez, Ley and Steel (2001b) provide a general analysis of proper model specific
priors for model averaging exercises. - Sachs and Warner (1995) use this variable as an index of overall openness of an economy.
- The posterior inclusion probabilities of single variables ignores their interdependence
and so may be criticized for reasons similar to those we have raised with respect to model
space priors that assume, for inclusion in a given model, conditional independence
across variables. Doppelhofer and Weeks (2007) propose ways to measure the jointness
of variable inclusion. Lettingkandldenote the events “variablerkappears in the true
model” and “variablerlappears in the true model,” and usingkandlto denote the
complements of these events, the authors propose the jointness statistic
Jk,l=log⎛
⎝μ(
k
∣∣
l,D,M
)μ(
k∣∣
l,D,M)·μ(
k∣∣
∣l,D,M)μ(
k∣∣
∣l,D,M)⎞
⎠to measure the degree of dependence between two candidate variables. The authors find
that positive dependence is a common feature of the candidate growth determinants
studied by Sala-i-Martinet al.(2004).