A Note on the Measurement of Growth Rates in Cross-National Research

A common approach to the explanation of growth rates in crossnational research employs a first-difference (or panel) model. However, when the basic variables employed in such a model are badly skewed, the estimated disturbances are likely to be heteroscedastic. When this happens, the parameter estimates become inefficient, and conventional tests of their statistical significance are biased. This research note briefly reviews the problem, suggests a remedy, and draws on two recent empirical studies of the economic growth of nations to illustrate the substantive importance of the issues involved.