|
Abstract:
|
We consider modeling determinants of long-run macroeconomic growth in a Bayesian model averaging context. Historically, there have been a variety of factors that have led to growth and the need to incorporate model uncertainty has been recognized by Sala-i-Matin (1997), Fernandez et al. (2001) and Eicher et al. (2007). We build on these results by considering growth from the viewpoint of graphical models, which introduce greater parsimony in model choice than regression models. Model search is significantly harder in this context and we review the use of the stochastic search algorithm MOSS, as developed by Dobra and Massam (2008) and Lenkoski et al. (2008). We then extend the analysis considerably by showing that the technique of instrumental variables can naturally be incorporated into graphical model search, thereby confronting endogeneity and model uncertainty simultaneously.
|
