standard products in every group but one. The average for the seven groups showed a rise of 16.1 percent for trade-marked articles as against only 8.5 percent for standard products.

Perhaps the most striking feature of this relationship is the consistency with which it expresses itself. This consistency manifestly precludes any possibility that purely chance factors are operating to cause this difference in behavior.

It is important to emphasize that, even for trademarked products, a marked correlation exists between frequency of change and amplitude of movement. Table I shows that the decline in price experienced by trade-marked products during the depression increased sharply and regularly with frequency of change. In each case, however, the extent of the decline was of approximately the same order as that suffered by standard commodities whose apparent flexibility, judged only by frequency of price change, was considerably greater. Trade-marking thus appears to carry with it

the power to postpone but not to avoid the making of price adjustments to meet changing competitive conditions. The rigidity displayed by the prices of trademarked articles and that of the prices for standard commodities thus seems to be, to a degree, different in kind.

If the price of a standard product changes but rarely, it is probable that the total amplitude of its fluctuations is also narrow. If, on the other hand, comparatively few price changes are recorded for trade-marked commodities, it is still entirely possible that it will respond fully, though with some delay, to changing general economic conditions. This behavior suggests that the ability of producers of standard products to control the frequency of price change is associated with the ability to restrict actual price fluctuations within narrow limits. Trade-marking, on the other hand, appears to permit stability of price quotations in the absence of the ability to limit the amplitude of price fluctuations.

The statistical information available in the public utility field leaves much to be desired. While it has been possible to collect selling price series for most of the major public utilities, these series for the most part relate only to the country as a whole or to a sample thereof, and thus possess the frailties that characterize any over-all average. In some cases it has been impossible to obtain even a single price series for the entire period of years. In other cases, notably the telephone industry, it has been possible to secure an index for only a single state or area; and for one important industry, motor transport, no worthwhile data are available. Satisfactory series for water transportation are also lacking. Moreover, it has been impossible to obtain adequate data for certain divisions of the utility markets; thus the electric, gas and telephone data relate only to the residential sales of those services. In the case of certain of the measures employed, the accuracy of the data themselves is only approximate as a measure of average price behavior. This is especially true of railroad freight rates.

There is a further inherent deficiency in virtually all of the price statistics presented. The unit of service chosen for such industries as freight transportation or residential telephone service, is not a constant during a period of time. Quality factors, such as speed, convenience, safety and regularity of service are inevitably modified, so that precisely the same thing is not being purchased over a period of years. While it has been impossible to make allowance for changes of this sort in the price series which are presented, it should be recognized that an influence is operative which in many cases amounts to a reduction in price in that a better quality of service is obtained. Changes in quality, however, are usually gradual and of significance only over a period of years rather than during a short-run cyclical period of price fluctuation. Moreover, the same criticism may be advanced toward a large number of price data included in such measurements as the wholesale price and cost of living series of the Bureau of Labor Statistics, with the result that when different price series are compared there is a tendency for errors to be mutually compensatory.

Table I gives in summary the results for each series for 1929, 1932, and 1936.

Table II gives an index of freight rates for 1900 to 1936. To our knowledge there has never been published an accurate index of the freight rate level. It has been customary to employ the revenue per freight 1 Appendix 4 was prepared by John D. Sumner, assisted by R. G. Lorenz.

ton-mile as a crude approximation of movements in the level of freight rates, but this index is subject to error in several respects. The changes in average revenue per ton-mile may result not only from adjustments in freight rates, but also from changes in the average distance for which commodities are transported,2 and shifts in the composition of traffic between high and low revenue commodities. Either an increase in the averge length of haul, or a greater proportion of low revenue traffic will tend to reduce the average revenue per ton-mile. A fourth and probably less important variable in the situation is the changing distribution of originated traffic between different sections of the country. Inasmuch as the rate level is not uniform through the United States, an increase in the relative proportion of low rate eastern territory traffic, for example, would operate to lower the revenue per ton-mile.

An adjustment has therefore been prepared which involves a partial correction for two of these factors, the change in length of haul and the shifting commodity composition of traffic. Adjustment for changes in length of haul was made as follows: (1) The average distance of haul for all tonnage carried by the railroad network (i. e., by the railroads treated as one system) was tabulated for the period 1900-1936; (2) A composite progression rate schedule, showing the general effect of distance on the amount charged, was obtained from the Interstate Commerce Commission. This progression rate schedule is not precisely accurate for any rate area, commodity group, or year, but is thought to be roughly representative of the effect of distance upon the level of charges.*

From the composite rate schedule a hypothetical total revenue for each length of hau! was then computed. Inasmuch as freight rates progress in blocks or steps each comprising a number of miles, it was necessary to interpolate to find a charge for the average haul for each distance in (1) above. It was assumed that for each rate-distance block the rate was centered at the midpoint of the block. (3) This hypothetical revenue, divided by the average length of haul in each year, provided a revenue per ton-mile for each year. (4) This annual revenue per ton-mile was then indexed,

The freight rate is typically based upon a tapering principle, by which the total charge increases much more slowly than does the distance of carriage.

The Interstate Commerce Commission should not be held responsible for the methods employed in the computation of this index, although members of the staff were generous with advice and suggestions.

The representativeness of the schedule is probably less for pre-war than for postwar years.

using 1926 as a base. The index thus obtained constituted a correction factor showing roughly the extent to which, assuming no changes in the level of freight rates as such, or in the composition of traffic, actual revenue per ton mile would have changed, due solely to changes in the average length of haul.

Adjustments for changes in the commodity composition of traffic hauled were computed in the following manner: The tonnage originating on Class I railroads was tabulated, broken down according to the six-fold classification of the Interstate Commerce Commission. This classification includes carload shipments of agricultural, animal, mineral, forest and manufacturing and miscellaneous products and less-than-carload freight. The percentage of each commodity group to the total, was then computed and a ton-mile revenue, constant for all years, was assumed for each of the commodity groups. These revenue data, suggested by the Bureau of Statistics of the Interstate Commerce Commission, are believed to be fairly representative of differences between the commodity groups. The assumed tonmile revenues were as follows: Products of agriculture, 1.142 cents; products of animals, 1.958 cents; products of mines, .802 cent; products of forests, .853 cent; manufacturing and miscellaneous, 1.369 cents; and lessthan-carload, 4.155 cents. Using the percentage composition of traffic for each year for weighting purposes, a theoretical average revenue per ton-mile was computed to take account of changes in the relative importance of each major commodity group. This revenue per ton-mile was then indexed, using 1926 as a base. This index provides a correction factor showing the extent to which, assuming no change in the level of freight rates or length of haul, average revenue per ton-mile might be expected to vary, due to changes in the composition of traffic.

The two adjusted indexes were then multiplied together to obtain a combined correction factor. The actual revenue per ton-mile, tabulated from the annual Statistics of Railways in the United States, was then divided by the correction factor for each year and the result indexed on a 1926 basis.

Several objections can be brought against each of these two adjustments.

Against the adjustment for length of haul, the following objections may be made:

(a) When there is a shift in the relative importance of the major commodity groups, there is usually an accompanying change in the average length of haul. If one makes adjustment for these shifts in commodity composition of traffic, this adjustment must be made on the basis of the average revenue per ton-mile, which in turn partially reflects the average haul of that commodity. A partial duplication results from making another adjustment for changes in the average length

of haul. The net result of this error may tend either unduly to increase or decrease the index.

(b) The average length of haul during the period was increased in part as a result of changes in the method of reporting traffic. Greater use of through billing and increases in the dimensions of individual railroads have contributed to an apparent increase in haul which did not actually take place. That is, a shipment going from point A to point B may previously have been counted as two originated shipments on account of transfer from one railroad to another, but later is counted as one shipment. Informed sources consider this fictitious increase to be of minor importance.

(c) The rate schedule is not a straight-line progression, but rises more sharply on the short hauls, and approaches the horizontal on the long hauls. The average haul is around 300 miles. At this point, the rise in revenue per marginal ton-mile is less sharp than at shorter hauls. In recent years the lengthened haul is due largely to loss of short-haul traffic to the trucks. Obviously, this means a greater shrinkage in average revenue per ton-mile than can be compensated for by corrections for the lengthened haul at the 300-mile point. This influence is somewhat counterbalanced by a rising length of haul due to the increase of very long hauls, where the rise in marginal revenue per tonmile is even smaller than at 300 miles.

(d) The assumed progression scale of freight charges, while believed to be substantially representative of the true progression, is not for any year, nor for all rate areas, identical with it.

The adjustment for the changing commodity composition of traffic is subject to the following major defects:

(a) The adjustment made is incomplete in that it was impossible to correct for shifts in the relative importance of high and low revenue freight within each of the six major commodity groups. A relative increase in the tonnage of manufactured and miscellaneous products, for example, tends to increase actual revenue per ton-mile for all freight traffic because this group of commodities moves at a higher than average revenue per ton-mile. This apparent rise in freight rates is eliminated by the correction factor described above, but only to the extent that the relative increase of traffic in manufactured products did not occur in products within that group which were carried at a higher (or lower) revenue per ton-mile than the assumed average of the manufactured products as a whole. Thus, if the increase in major group tonnage is solely due to an increase of high revenue commodities within the group, the correction is incomplete because it does not sufficiently deflate the actual revenue per ton-mile of all freight traffic. The rise in the latter resulted partly from an increase in the relative importance of a

high revenue group-manufactured products; this rise the correction factor eliminates. In part, however, the rise was caused by an increase of manufactured products transported at a revenue per ton-mile higher than the assumed average of their group; this increase is not eliminated by the correction factor, and, in view of the inadequacy of available data, cannot be.

It is difficult to judge the quantitative importance of this factor. While no correction can be made for the changes within each major commodity category, these shifts are undoubtedly not all in the same direction. In consequence the errors offset one another to some extent. It is improbable, however, that changes in the composition of each of the major categories of railroad traffic, due both to the evolution of the national economy and to the development of rival forms of transportation, are sufficiently random in character for the effects to cancel out entirely

(b) Originating tonnage of various commodities is not necessarily in the same proportion as the ton-miles of those commodities. The effect of this is to assign greater weight in the index to the commodity groups with the shorter hauls (and hence higher revenues per ton-mile) than would be the case if adequate data were available for the ton-mile significance of each group.

(c) The commodity statistics were taken from the data for Class I railroads only. If Class II and III railroads were included, the result would undoubtedly be somewhat different. Class II and III railroads probably originate and carry a larger proportion of bulky, low revenue traffic. But they carry only a small proportion of the total traffic in any case, and cannot affect the results substantially.

The quantitative significance of these various defects in the adjusted index cannot be accurately determined from existing data. To a certain extent the direction of error seems clear. There seems little doubt, for example, that the length of haul has increased in part by virtue of the loss of short-haul traffic to motor trucks, and that the imperfect adjustment in this respect tends to overstate declines in rate levels. The opposite effect may result from inadequate allowances for other imperfections described above. All that is claimed for this adjusted index is that it shows the direction of error in the unadjusted revenue per tonmile index of freight rates, and gives some indication of the probable extent of that error. Moreover, its behavior seems to show more accurately the effect of the major freight rate level cases decided by the Interstate Commerce Commission during the 1930's 5 than

The 15 Percent Case, 1931, 178 I. C. C. 539, 179 I. C. C 215, and 191 I. C. C. 361, permitted emergency increases in freight charges beginning Jan. 4, 1932. Again, in Emergency Freight Charges, 1935, 208 I. C. C. 4 and 215 I. C. C. 439 (1936), the Interstate Commerce Commission permitted emergency increases in freight rates until December 31, 1936. The effects of later adjustments, of course, are not relevant to this record which ends with 1936.

does the unadjusted index of average revenue per tonmile.

The use of an average freight rate index necessarily obscures the highly divergent behavior of thousands of particular class and commodity rates which together constitute the rate structure. Among agricultural commodities, for example, a striking contrast exists between the generally stable behavior of livestock rates, and rates on cotton which declined over 40 percent from 1926 to 1935. As stated by the Interstate Commerce Commission in 1933, "The lowering of rates, however, has not been uniform, many rates not having been reduced at all since 1922, and others having been reduced as much as 50 percent or more." 7

Table III shows revenue per passenger-mile and number of passenger-miles for class I railroads, 1911-36. Revenue per passenger mile is believed to be a fairly accurate index of passenger rates. Changes in rates are not perfectly measured by this index. Changes in the distribution of traffic between regular and special-rate travel, between low and high rate areas, and related changes, as well as changes in rates, affect the average revenue per passenger-mile.

Table IV presents an index of fares from 1913, and of fares and volume of traffic from 1917 to 1936, for street railways.

8

Table V shows indexes of the residential price of manufactured gas, 1913-36, and the volume of residential consumption, 1929-36. The series is not entirely homogeneous. For the period 1923-36, inclusive, it represents data published by the Bureau of Labor Statistics, which are based on the cost of 30.6 and 10.6 therms, respectively, of gas in 25 cities. This thermal basis of prices insures accuracy in the sense that virtually no change is involved in the quality of the service purchased. The amount, 30.6 therms, is deemed characteristic of the use of gas for cooking and water-heating purposes, while 10.6 therms is typical of range use only. To this index has been spliced an earlier index of the Bureau of Labor Statistics of the cost of 3,000 cubic feet of manufactured gas for household use in selected cities. The behavior of the spliced series in overlapping years is so similar that there seems to be little inaccuracy involved in the combination of the two.

Table VI presents data on local telephone rates and use. Unfortunately, there is no available index of telephone rate behavior for the entire United States; an index of local telephone rates for the State of Wisconsin has therefore been used. Prepared from data supplied by the Wisconsin Public Service Commission, it represents the price in some 55 Wisconsin cities, weighted

U. S. Department of Agriculture, Agricultural Statistics, 1936, p. 407.

7 General Rate Level Investigation, 1933, 195 I. C. C. 5 (1933), p. 67.

8 Changes in the Retail Price of Gas, 1923-36, Bulletin 628, Washington, 1936. • Bureau of Labor Statistics, bulletins on Retail Prices, annua).

according to population and class of residential telephone service. The index, while representing rate behavior in only one State, is believed to be fairly representative, particularly for the period since the War. The rate behavior shown in this index checks with the following comment in the congressional report on communications companies:

From information available, the trend of rate changes of all telephone companies during the 11-year period from January 1, 1922, to December 31, 1932, was upward, in the case of local exchange rates. Most of the increases, however, occurred during the first 5 years of this period, or during the years 1922 to 1926, inclusive. An upward movement in toll rates also occurred during the years 1922 to 1926, inclusive, followed by some reductions. It is significant that, since 1929, rates, for the most part, have been stationary

* *

Approximately 1,034 rate changes were made by the 22 large regional telephone companies of the Bell System during the 11-year period from January 1, 1922, to December 31, 1932, 797 of the changes affecting local rates and 237 of the changes affecting toll rates. Approximately 501 of the changes affecting local rates were in the nature of increases and approximately 296 were reductions. Of the 237 changes affecting toll rates approximately 83 were in the nature of increases and 154 were reductions. Thirty-seven of the rate reductions were involuntary.

The effect of these rate changes was an upward revision of rates, as hereinbefore pointed out with reference to most telephone companies, increases having been accomplished principally during the years 1922 to 1926, inclusive.10

A further check was obtained by examining the annual reports of the American Telephone & Telegraph Co. for the years 1930 to 1934, inclusive. If one takes account of all the rate reductions referred to in these reports and includes as reductions all cases referred to as pending before courts or commissions, the resultant reduction amounts to less than 2 percent of the total operating revenues of the Bell System in 1934.

A sample index of pipe line rates on crude petroleum is presented in table VII. The index is based on the rates charged by the Sinclair Pipe Line Co. and its predecessor companies between points in Oklahoma and Kansas in the mid-continent field to Whiting (Chicago refining district), Wood River (St. Louis district), and Cleveland. The rates to these points have been combined in a simple arithmetic average. While restricted in its composition, the index is believed to be fairly representative of pipe line rate behavior from the important mid-continent field to northeastern points. A similar, but somewhat more inclusive, tabulation was employed by the Federal Trade Commission in 1927.11 There seems to be a high degree of uniformity in the movement of pipe line rates on crude petroleum. The index presented, however, is not representative of the movement of crude oil from the Gulf Coast to eastern refining points by waterway shipment, nor is it necessarily representative of

10 73d Cong., 2d sess., H. R. 1273, part III, No. 1, Washington, 1934, pp. 932 to 933.

pipe line rates in the California and other producing

areas.

The data on the price of electricity presented in tables VIII and IX represent the average cost of 25, 100, and 250 kilowatt-hours in selected cities; the Federal Power Commission deems these amounts representative of the use of electricity for lighting and small appliances, plus refrigeration and for these two plus cooking, respectively.12 The data for 1924-1936 are taken from the Commission's study of rate trends in 132 cities of 50,000 or more population.13 To these price series have been spliced data drawn from a study by Mr. W. G. Vincent 14 of rate trends in the 51 cities included in an index formerly compiled by the Bureau of Labor Statistics. 15 The methods employed in Mr. Vincent's study are said to be identical with those used by the Federal Power Commission.16 While the number of cities is smaller, this study includes approximately 80 percent of the population covered in the report of the Commission. The latter investigation shows a somewhat sharper downward trend in the overlapping years 1924-36 than does the Vincent study.

The index of the price of 25 kilowatt-hours is applicable to the largest number of residential users, although due to the promotional character of electric rate structures, it is a less sensitive index than those for 100 and 250 kilowatt-hours. Each of these average price indexes conceal many variations in rate behavior." During the 1930's the reductions in residential rates appear to have been greater than those for either commercial or industrial power.18

11 Petroleum Industry, Prices, Profits, and Competition, 70th Cong.. 1st sess., S. Doc. 61, Washington, 1928, pp. 36, et. seq.

12 Trends in Residential Rates, Washington, 1937, p. 21.

13 Ibid., p. 21. These cities contain a population of 37,533,000-88 percent of the population of cities of 50,000 or more inhabitants, and 46 percent of the urban population in communities of 2, 500 or more inhabitants.

14 "Rate Reductions," Edison Electric Institute Bulletin, June 1936, p. 217. Mr. Vincent is a vice president of the Pacific Gas and Electric Company.

15 The Bureau now publishes in its retail price bulletins the price of specified typical amounts of electricity in selected cities. The older index was based upon the price of varying most popular consumption amounts of electricity. Inasmuch as elec tricity rates vary with consumption volume, this method is inadequate as a measure of rate changes only.

16 The average is weighted according to the population of the cities in 1930; their total population in that year approximated 30,000,000. Three cities were omitted, 1906-16. "In order to determine the trend of domestic rates over a long period I secured for each of the 51 cities listed * the amount charged in that city for the same monthly consumption (domestic) used by the Federal Power Commission in its rate survey and under the same specifications, as of Jan. 1 of each year W. G. Vincent, op. cit. The data were secured directly from the companies, p. 224. 17 A careful measurement of price reductions for specified quantities of electricity from 1930 to 1933 is presented by L. G. Cannon and D. F. Estes, "The Trend of Electric Utility Rates: 1930-33," Journal of Land and Public Utility Economics, November 1934, p. 359. The material presented in this article shows reductions of 4.6, 6.4, and 8.8 percent in the cost of 25, 50, and 100 kilowatt-hours, respectively, of residential consumption. Reductions in the cost of 100 kilowatt-hours ranges from no change to as high as 30 percent (Delaware) in the various States.

18 In the commercial or small power market reductions were 5.3, 6.3, and 4.6 percent for 145, 1,440, and 4,500 kilowatt-hours, respectively. In the case of industrial or large light and power, the decreases were 2.1, 3.2, and 4.2 percent for quantities of 7,200, 48,600, and 432,000 kilowatt-hours, respectively. Again there was a great deal of divergence in rate behavior between the several States. Ibid.