We Just Averaged over Two Trillion Cross-Country Growth Regressions
IMF Working Papers, July 1, 1999
Source details
- Canonical URL
- We Just Averaged over Two Trillion Cross-Country Growth Regressions
Other formats
Bibliographic details
- Authors: Mark F. J. Steel, Eduardo Ley
- Published: July 1, 1999
- Series: IMF Working Papers
- DOI: https://doi.org/10.5089/9781451852493.001
Summary
- Authors: Mark F. J. Steel, Eduardo Ley
- Date: July 1, 1999
- Core investigation: Model uncertainty in cross-country growth regressions using Bayesian model averaging (BMA).
- Main summary points:
- The posterior probability is distributed among many models, suggesting the superiority of BMA over any single model.
- Out-of-sample predictive results support the superiority of BMA.
- In contrast with Levine and Renelt (1992), results broadly support the more “optimistic” conclusion of Sala-i-Martin (1997b) that some variables are important regressors for explaining cross-country growth patterns.
- The variables identified as most useful for growth regression differ substantially from Sala-i-Martin’s results.
Methodology and Analytical Approach
- Method used: Bayesian model averaging (BMA).
- Evaluation: Posterior model probability distribution and out-of-sample predictive performance used to assess model uncertainty and the performance of BMA relative to single-model approaches.
Key Findings and Interpretations
- Findings:
- Posterior probability distributed among many models.
- BMA outperforms any single model according to posterior distribution and out-of-sample predictive results.
- Evidence aligns more with Sala-i-Martin (1997b) than with Levine and Renelt (1992) on the existence of important regressors for cross-country growth.
- The specific variables found most useful differ substantially from those in Sala-i-Martin (1997b).
Content in this bundle
- We Just Averaged over Two Trillion Cross-Country Growth Regressions - WP/99/101