Abstract
In this article we use Bayesian classification and finite mixture models to extract information from Levine's (Citation2002) cross-country database and reconsider the relationship between financial structure and long-run economic growth. Our methods, based on statistical similarities and multi-dimensional structures, allow for parameter heterogeneity across the countries in Levine's database and yield substantially different findings than Levine's regarding the relationship between financial structure and economic performance.
1This article builds on material from Karl Pinno's PhD dissertation at the University of Calgary.
Acknowledgements
We would like to thank John Boyce, Herbert Emery, Daniel Gordon, Zuzana Janko, Michael Robinson and Jean–Francois Wen for useful comments. Serletis also gratefully acknowledges support from the Social Sciences and Humanities Research Council of Canada.
Notes
1This article builds on material from Karl Pinno's PhD dissertation at the University of Calgary.
2 The section draws heavily from documentation that accompanies the AutoClass program.
3 Each of the nine paired sorts was modelled as a covariant normal.