A refinement in the specification of empirical macroeconomic models as an extension to the EBA procedure

Ross Levine and David Renelt's (LR) paper [1992] investigate the "robustness" of the relationship between growth, investment, and variables of interest using Learner's [1985] Extreme Bounds Analysis (EBA). LR claim that few economic variables have a robust relationship with either long-run cross-country growth rates or investment's share of GDP. As something resembling an extension of McAleer et al. [1985] and Hoover and Perez [2004], we use a battery of statistical tests to search for models that may be less subject to specification error. We conclude that there are alternative models based on LR that have better statistical properties as well as give rather different results. Copyright © 2006 The Berkeley Electronic Press. All rights reserved.