Knowledge And Decisions

One of the elements in the public success of the Malthusian doctrine which has proved equally serviceable in other politically “relevant” doctrines has been the display of cognitively irrelevant statistics. The second edition of Malthus’ Essay on Population was several times larger than the first, due to the addition of masses of data. These data were never used to test the Malthusian theory but to illustrate or apply it. In Malthus’ own words, the data are intended to “elucidate the manner” in which his theory operates, to “examine the effects of one great cause” — the population principle — but not to test the principle itself. Any population size or growth rate would be consistent with the principle: “The natural tendency to increase is everywhere so great that it will generally be easy to account for the height at which the population is found in any country.”¹⁰⁸ No matter what the data show, he would be “right.”

This decorative display of numbers which in no way test the central premise continues in modern, more sophisticated, statistical studies. A noted study of the economic effects of racial discrimination begins by simply defining “discrimination” as all intergroup differences in economic prospects.¹⁰⁹ It then proceeds to elaborate mathematically and statistically in the light of that premise, but never testing the premise itself. All intergroup differences in cultural orientation toward education, work, risk, management, etc., are simply banished from consideration by definition. Discrimination in this context becomes simply a word denoting statistical results, though of course the very reason we are interested in discrimination, in its usual sense, is because it refers to intentional behavior whose moral, political, and social implications concern us. That social and political concern is implicitly appropriated for statistical results that depend on numerous other factors as well.

Such arbitrary attribution of causation by definition is a special case of a more general problem that plagues statistical analysis. Whenever outcome A is due to factors B and C, by holding B constant, one can determine the residual effect of C on A. The problem is that A may also be affected by factors D, E, or F, etc., and if they are not specified in the analysis, then all of their effect is wrongly attributed to C. Moreover, even the attempt to hold B constant may fail in practice. Theoretical variables may be continuously divisible, but actual statistics may be available only in discrete categories. In comparing two groups who differ on a particular variable (male and female differences in height, for example), attempts to hold that variable constant by comparing individuals with the same value of the variable (the same height) may mean in practice comparing individuals who fall in the same discrete intervals (between five and six feet, for example). But groups whose distributions differ across specified intervals can also differ within those respective intervals. The average height of males and females who fall in the interval from five feet to six feet is probably different (males in that interval being taller than females in the same interval), despite the attempt to hold them constant. Therefore some of the effect of the variable supposedly held constant will appear statistically as the effect of some residual variable(s). This residual method of analysis has great potential for misstating causation, through inadequate specification of the variables involved, either inadvertently or deliberately. Whether one’s preferred residual explanation is discrimination, genetics, schooling, etc., deficiencies in the specification of alternative variables are rewarded with more apparent effect from the preferred residual variable. The ultimate extreme of this is to implicitly hold all other variables constant by arbitrarily defining one variable as the variable and using this definition as if it were a fact about the real world, by using the same word normally used to describe that fact — “discrimination” in this case. The political benefits of this cognitive deficiency may be illustrated not only by the reliance of national political figures and institutions on the advice of the economist using this technique, but also his academic success in promoting a conclusion consonant with academics’ social and political vision, however cognitively questionable. It is a technique — and a result — common in other fields, as will be noted again.