Problems of Definitional Dependency: The Case of Administrative Intensity

In the process of defining their variables, researchers often produce dependencies between them by redundantly including component variables. The result is that a null hypothesis of "no relationship" cannot be tested by comparing the observed correlation with zero. The potentially misleading consequences of this practice are discussed in a detailed analysis of the size-administrative intensity literature. Suggestions are made for dealing with the problem. When researchers translate their ideas into formalized theories with testable hypotheses, the required conceptual development often results in the creation of variables which are defined as functions of component variables. We shall refer to these as "complex variables." Percent literate is an example. Variables of this kind are often created when variations in one component variable are seen as theoretically significant only with reference to variations in some other variable. This usually calls for standardization. Suppose, for instance, the researcher defines evolution in terms of energy consumption. Obviously comparisons may be highly misleading if population size is not taken into account. So the researcher may define the level of evolution as "per capita energy con-