Many recent papers use Bayesian Model Averaging to address model uncertainty, but (Ciccone and Jarociński, 2010) have questioned the approach on theoretical and empirical grounds. They argue that a standard ‘agnostic’ approach is too sensitive to small changes in the dependent variable, such as those associated with different vintages of the Penn World Table (PWT). This paper revisits their theoretical arguments and empirical illustration, drawing on more recent vintages of the PWT, and introducing an approach that limits the degree of agnosticism.
Figure 4: Distributions of Posterior Inclusion Probabilities across data sets.
James Rockey and Jonathan Temple, Growth Econometrics for Agnostics and True Believers European Economic Review 81, January 2016, 86–-102