The Pitfalls of Per Capita

The use of per capita measures in aggregate data analysis has been giveni little theoretical justification. This paper argues that unless there is such a justification, the employment of such transformations can lead to greatly distorted results. It is shown that such transformations can lead to either spuriously high or low correlations among sets of variables. Furthermore, there is no precise formula for determining the extent of any such biases. Even when the distortions are not great, they may lead the researcher to erroneous conclusions about the strength of the relationships among his variables. It is argued that per capita transformations are not always undesirable, but that greater attention needs to be paid to the theoretical justifications for employing such indices.