It is assumed that the problems of market research have to do with determining trends in demand for products with a view of anticipating future demand and that the problems of advertising are (1) the dissemination of information concerning available products, and (2) the creation of demand for specific products.
If the assumptions are correct, the problems of market research and of advertising are special cases of general problems of mass behavior. Influence on mass behavior can be exerted most effectively if the dynamics of such behavior is properly understood.
There are strong indications that theoretical-mathematical methods can significantly advance our understanding of such matters.
The applications of statistical analysis to mass behavior are, of course, not new. Insurance practice rests solidly on such applications.
The soundness of insurance practice shows that not only "involuntary" aspects of mass behavior are subject to strict probabilistic laws (deaths, disease, fires, accidents, etc.) but even "voluntary" ones such as thefts, embezzlements, etc. Nor have statistical methods been confined to the problems of insurance. These methods are constantly being extended to the determination of public opinion, prediction of election results, etc.
Much publicity is given to spectacular "failures" of such predictions with a concomitant disparagement in some sectors of the press of attempts to apply quantitative methods to human behavior. It must be pointed out, however, that elections are all-or-none events; a small error in predicting a distribution of votes can result in a "total" error in predicting the outcome of an election, even though the actual quantitative error is so small as to leave no doubt concerning the efficacy of the method.
All these applications of quantitative methods have, however, been seriously limited, because they are almost exclusively confined to the statistical analysis of purely empirical data. What is desired in understanding mass behavior is a derivation of laws of such behavior, which would allow the prediction of behavior on the basis of a few observations, not merely by treating the observations as a "sample" and extrapolating the results, but by treating them as parameters in an equation, in which social dynamics is taken into account.
To give an example, suppose we wish to determine the spread of a piece of information through a population. The empirical-statistical method requires taking a series of sample readings which can be extrapolated to describe the whole population. The knowledge of the dynamics of information spread on the other hand would require perhaps only one or two such readings at the initial stage from which the entire course of the spread could be predicted. …