As all good social scientists know, to analyze a problem properly one needs hard data, the harder the better. The United States government thinks so too, and is a prolific producer and consumer of statistics. If one wishes to do research on the military expenditures of the Latin American countries, for example, one has available not one but at least four sets of statistics complied by or for the federal government.

At this point, however, difficulties emerge: these sets of statistics not only are not identical, but often differ so drastically that they lead to diametrically opposed conclusions.