ency of such changes is often considered to be towards undermining those powers of private initiative and of voluntary coöperation which are our Saxon birthright. On the contrary, the high-handed procedure of many corporations holding public franchises, the excessive profits they often actually secure and more often are firmly believed to secure, the power exercised by such corporations in municipal or even in state affairs, when their interests are at stake, are arguments entirely aside from the present phase of the subject, and yet of great, perhaps of decisive, weight in forming the final conclusion.
It must, I think, be admitted that it is not a function of government, whether state or municipal, to undertake enterprises for profit. Some other end more clearly within the range of governmental action must be found in order to warrant the proposed extension of powers, but, that found, the profit or lack of profit may have a practical, argumentative force. For example, the recent legislation in South Carolina on the liquor traffic cannot be justified merely by showing the state management to be profitable, but its profits, if permanent and large, will have a tendency to convince the taxpayer of the wisdom of the change.
After merely noting that such considerations are here irrelevant, I pass to the central question: How may the economic productivity of municipal and private enterprises be measured and compared?
In the first place the method must be statistical. While in the natural sciences various methods of measurement are employed, in the social sciences all methods of measurement are statistical. Social phenomena are so variable in time and place that we never meet the same concrete condition twice, as we do in the natural sciences. Hence no unvarying units are possible. On the contrary, the variations are usually wide and often inexplicable. But if any trustworthy generalizations are to be obtained, they must come through the elimination of these variations. This is secured by the so-called law of large numbers, the basis of statistics. It rests on an arbitrary division of
