Determinants of Economic Growth Redux: A Measurement Error Model Averaging Approach

Gernot Doppelhofer, NHH

Determinants of Economic Growth Redux: A Measurement Error Model Averaging Approach (joint with Ole-Petter Moe Hansen and Melvyn Weeks)

This paper estimates income per capita using recent vintages of the Penn World Tables (PWT) and investigates the robustness of determinants of long-run growth rates in a cross section of countries. We propose a novel Measurement Error Model Averaging (MEMA) approach that accounts for measurement error in international income data, model uncertainty over the determinants of long-term growth and allows for outliers in the form of heteroscedastic model errors. Generally, we find that newer vintages of the PWT are more precisely measured than older vintages, and that income in richer countries tend to be more accurate then poorer countries. We find that eighteen variables are robustly related to economic growth using different prior assumptions about the measurement of incomes across PWT vintages and countries. The results are robust to alternative prior assumptions about outliers and the inclusion of growth determinants, as well as inclusion of PWT vintages.