Factor analysis, a statistical technique of growing popularity in political science, has recently found application in the June, 1966 issue of this Review in an attempt to identify U.N. voting blocs. The hand is that of Professor Bruce M. Russett, the author of a number of outstanding contributions toward the systematic and quantitative study of international relations. The following paragraphs are designed to be cautionary in nature, not destructive, and endeavor to suggest: 1) that factor analysis has questionable value as a tool for the study of complex phenomena; 2) that even if acceptable as a tool, it has peculiar faults when applied to U.N. bloc voting; and 3) that even if acceptable as a tool for the study of bloc voting in the U.N., an application somewhat different from the one chosen by Professor Russett might have been more profitable. In the process it will be argued that the procedures of the sort profferred by Professor Arend Lijphart for determining U.N. voting blocs are preferable.

1. If appealing in its seeming promise to make order of disorder and sense of nonsense, factor analysis proves to be exasperating in practice. For while it is true, that, as Professor Russett has noted elsewhere, “the method is completely objective” in the sense that the decision as to which variables are to “load” on which factors is made entirely by the impartial machinery, the criteria which the machinery applies in arriving at this decision are manipulated by the operator—and depending on these criteria, the outcome can vary widely. Thus factor analysis can generate not one, but a number—in fact an infinite number—of solutions to the same problem. In choosing among these solutions the one most likely to represent truth (if that is what one is interested in), the analyst is commonly urged to select the solution which makes the most “sense”—in other words to embrace the objective solution which is most subjectively appealing. Therefore, one can get out of factor analysis almost whatever one wants to get.