About November 1st a very severe industrial depression began, reducing the amount of the traffic in some instances as much as forty per cent in the winter of 1920-21. Traffic for the year 1921 was about twenty-five per cent less than that for 1920-this decrease being by far the greatest ever recorded in one year. The result was that earnings during this period were entirely subnormal.

(2) THE RELATION BETWEEN TRAFFIC DENSITY AND COST OF TRANSPORTATION

The preceding method of analyzing the law of increasing returns did not fully illustrate the workings of the principle from the earliest stage of railroad development. To do so it would be necessary to study the average rates during the period 1830 to 1860, during which the system of steam transportation had its beginnings. While the early roads were largely experimental, nevertheless, the figures would be of interest to indicate the operation of the principle when transportation was in its infancy. But a study of this period seems impracticable owing to the incompleteness of the records. Also, it is doubtful whether the workings of the principle could be clearly discerned even if the figures were available, because earnings and rates were affected to such an extent by public support given the roads (land grants, etc.), by the rebate system, the numerous changes in organization, etc.

To avoid these arbitrary factors which enter into an historical study, and to obtain a more complete analysis of the principle of increasing returns, it may be worth while to compare the conditions and results as of some recent year for various railroads in different conditions of size and growth. By analyzing the characteristics of typical railroads of different types as revealed in the group averages, we can ascertain the workings of the principle of increasing returns as based

1 The 1907 panic and depression resulted in a decrease of about eight per cent and the outbreak of the war in 1914 caused about the same decline.

upon modern conditions and without respect to historical changes in operating methods. The principle will thus be limited in application, being based on magnitude of operations only and without respect to historic changes in the development of transportation, technique, and the character of the traffic, and the results may thus be more easily interpreted.

This method of statistical analysis will also enable us to study the relations between rates and traffic density for individual commodities (which relations could not be compared over a long period of years because the data are available only for a limited period).

The accompanying graphs are based on a group classification of one hundred and ten large railroads according to the average ton-mile receipts for five commodities and for total traffic in the year 1914.1 While, as already indicated, there is a wide variation in individual receipts per ton-mile for the same commodity, even for roads in the same territory or on a competitive basis, the graphs indicate that when a number of roads which receive about the same ton-mile receipts are grouped together, such average receipts bear an approximate relation to the average density of traffic for the given commodity, as well as to the average length of haul. Only one out of the five commodity graphs (cotton) fails to show a fairly consistent relationship, probably due to the fact that the heaviest density for this traffic is on Southern roads. Even here, however, the factors of length of haul and density for all freight show the anticipated trend.

The graph for "All Commodities" seems to be more consistent than those for the individual commodities, as might be anticipated, but in interpreting this graph it must not be forgotten that there are varying proportions of different commodities represented at different points in the curves.

1 See detailed explanation in note on page 77, which applies to the method of preparation of these graphs. The year 1914 is the latest for which results are obtainable.