CORNELL UNIVERSITY LIBRARY Digitized by Microsoft® BIIBEID OF RAILWAY ECONOMICS ' E$tablistied by Railways of the United States for the Scientific Study of Transportation Prpblems FRANK HAIQH DIXON . CHIEF. STATiaTlCIAN JULIUS H. PARMELEE STATISTICIAN ,-,-< Arguments For aitdlsiiiiisl Limiialion of Length of Freight Trains ,,^^ .Cornell University Library HE1801.B95 A6 ^'^'iJUmimmiitS/', ^'"' aoainst limitation of olin 3 1924 030 121 044 ConsecutiTe No. 92 Miscellaneous Series No. 23 WASHINGTON, D. C. 1916 Digitized by Microsoft® BULLETINS OP THE BUREAU OF RAILWAY ECONOMICS MONTHLY BTJLLETIITS The monthly summary of Revenues and Bxpens.es of Steam Roads in the United States has been issued regularly since July, 1910. Following is the list of bulletins for the last twelve months : Consecutive Number. Monthly Report Series. Month of — 77. 58. March, 1915. 78. 59. April, 1915. 79. 60. May, 1915. 80. , 61. June, I915.--In turns : E^lscal 1915, 82. 62. July, 1915. 84. 63. August, 1915. 85. 64. September, 1915. 86. 65. October, 1915. 87. 66. . November, 1915. 89. 67. December, 1915, 90. 68. January, i915, 91. 69. February, 1915. •Insert with Bulletin — Operating Ra- The following issues are Out of Print : Consecutive Monthly Report Number. Series. Month t>t— 7. 5. October, 1910. »• 7. December, 1910i. 13. 10. March, 1911. 19. 13. June, 1911. 22, 15. August, 1911. 51. 87. June, 1913. (Continued to page 3 of coyer.) Digitized by Microsoft® Arguments For and Against Limitation of Lengtli of Freiglit Trains Bureau of Railway Economics Washington, D. C. Hay. 1916 Digitized by Microsoft® ^ ^ '^ "^ i; r. / ^' 3 ^ y Digitized by Microsoft® TABLE OF CONTEXTS. Page. Introduction 5 Origin and extent of movement to secure train-limit legislation 5 Long trains one feature of progress in railway transpor- tation 6 Number of cars per freight train 7 Capacity of freight cars 8 Train loads 8 Size and power of locomotives 9 Strength and efficiency of track, bridges, and other structures 10 Summary of Arguments For and Against Limiting the Length of Trains 11 Arguments for limiting train length 11 Arguments against limiting train length 12 Relative Safety of Long and Short Trains as Operating Units. . 14 Transmission of hand signals 14 Failures of draft rigging or brake gear, parting of trains, etc 16 Buckling of trains 17 Detection of broken car rigging en route 18 Possible promptness of control of long trains 19 Conclusion 20 Statistics Regarding Comparative Number of Accidents to Long and Short Trains 21 Analysis of the causes of accidents, with respect to the effect of long trains 28 The human factor in accidents 34 Recapitulation 37 The Effect of Increased Train Density Upon the Risk of Accident 39 Why increased density tends to increase accidents 39 Statistical evidence of the relation of accidents and casual- ties to train density 41 Digitized by^l^icrosoft® 4 Relation to train density of collisions and derailments. ... 44 Relation to train density of casualties to trainmen and passengers 45 Relation to train density of casualties to trainmen on the road 46 Relation to train density of casualties to passengers 48 Relation to train density of other casualties 50 Relation to train density of the risk of accidents and casualties 5^ Summary of evidence on the effect of train density upon accidents and casualties 5^ The greater relative importance of train density in the safety of railway operations 52 Decrease in casualties in relation to service performed by railways 53 Summary of evidence on relation of train density to accidents 55 Economic Considerations Involved in Train-limit Legislation. 56 Much of recent railway investment would be rendered valueless eg Many economies in operating expenses would be lost .... 57 Additional investment would be made necessary 58 Financial effects of train-limit legislation 58 Effect on rates paid by the public eg Appendix g^ Statistical proof of the relation to train density of acci- dents and casualties gj Digitized by Microsoft® INTRODUCTION. Origin and Extent of Movement to Secure Train-Limit Legislation. Following in some states the defeat, and in others the repeal, of so-called "full-crew" laws, there has been a widespread effort on the part of the legislative representatives of the railway labor brother- hoods to secure substitute legislation. This has taken the form of bills, introduced in numerous states, to limit the length of freight trains. The grounds on which they urge these proposals are almost solely considerations of safety. They claim that the number of casualties on railways to all classes of persons, but especially to trainmen and passengers, is increasing and that this is due in sub- stantial measure to the increasing number of cars handled in a single train. In the attempts to secure so-called "full-crew" legislation, efforts were directed toward having the law prescribe the minimum num- ber of men it should be lawful to employ on trains containing speci- fied numbers of cars. The bills that have been advocated more re- cently have sought to fix either the maximum length of a train or the maximum number of cars in a train. In some cases the limit has been set at trains a half mile long; in som° '^'^ 3" <.iHB. in Illinois in 1915 lUc successtul opposition of the railways to the passage of a bill limiting trains to 50 cars was met by an amend- ment placing the. maximum at 75 cars. The bill as thus amended also failed. During the winter and spring of 1914-1915 bills to limit the length of trains were introduced in the legislatures of 20 states. All of them failed to pass. These states were : California, New Jersey, Colorado. New York, Georgia, North Carolina, Illinois, North Dakota, Indiana, Ohio, Iowa, Pennsylvania, Kansas, South Carolina, Michigan, South Dakota, Minnesota, Utah, Nevada, Digitized by Microsofmconsin. (S) In Arizona legislation limiting freight trains to 70 cars is in force. This is the only state in which such legislation has been passed. long Trains One Feature of Progress in Raihvay Transportation. The proposal to limit the length of freight, trains, whether for reasons of safety, as claimed, or otherwise, fundamentally concerns the whole scheme of railway operating methods. An intelligent consideration of the merits of the proposal must, therefore, take account of the important role that long freight trains have come to play in railway progress. The increase in the length of freight trains has been an important factor in the steadily growing use of large train units; and large train units have constituted the cardinal feature of the increased efficiency of railway transportation in the United States. Through- out every industry, efforts to reduce the cost of a unit of product have given wide extension to large scale methods of production. The application of these methods to transportation, as rapidly as volume of business permitted, has been made through larger and larger steamships and longer and longer trains. This use of long freight trains not only has reduced the necessity ^f '•'"•■"o manv freight rates to meet the marked increase in wages and prices generally, v^u \^^^ j^o^mUt^j- <;nme-_— toJ^w^^i'^n^ m lates. This benefit from large train units has inured to practically the entire population, for substantially all people are today depend- ent, directly or indirectly, on railway transportation for a great part, if not all, of their supply of commodities. Railway employees, considered separately as one group, have been benefited to the ex- tent that the economies from large trains have enabled the carriers to meet successive advances in wages, notwithstanding the public's restraint upon increases in rates. Such a two-fold pressure upon the railways, namely, public demand for a lowering of rates and employees' demand tor increases in wages, could not have been met without progressive economies in operation resulting mainly from the handling of freight in larger units. Further, the increase in train load, caused in large part Ijv the use of longer trains, has resulted in savings that have added to thp fund from which permanent betterments and improvements to the Digitized by Microsoft® properties have been made, for the benefit not only of the pubUc, but also of the employees working thereon. The place of large train units, as the central factor in increased efficiency in railway operation, can best be understood by a brief review of those facts of railway development that are associated with the movement. Xuiuber of Cars per Freight Train. The average number of cars per freight train on all the railways of the United States increased from 26.8 cars in 1904 to 34.3 cars in 1914, an increase of 28 per cent. On the railways of the Eastern District the increase was from 28.6 to 36.4 cars ; on the railways of the Southern District, from 23.8 to 30.7 cars; and on the railways of the Western District, from 26.1 to 33.8 cars. The facts are shown in Table I. TABLE I. Freight Cars Per Train.* ,,.„ n«;*«„q ctot»c Eastern Southern Western Yea,, United States. District. District. District. igo4 26.8 28.6 23.8 26.1 1905 27.6 28.8 24.4 27.5 1906 27.9 29.3 24.4 27.8 1907 27.2 28.8 23.1 27.0 1908 28.7 30.9 24.3 28.1 1909 30.2 32.4 26.5 29.3 1910 29.9 32.0 26.8 28.9 1911 30.8 32.8 27.8 30.1 1912 31-8 33-9 28.3 31.1 1913 .32.7 34-7 29.1 32.0 1914 34-3 36.4 30.7 33-8 Increase 1914 over 1904. 7.5 7.8 6.9 7.7 *Computed from Statistics of Railways in the United States, Interstate Commerce Commission, 1904 to 1914. Averages exclude mixed train mileage. Of course, these averages include many roads with comparatively light traffic. The larger railways show averages much greater than these. And in those sections of the country where the traffic is heavy, especially where the volume of a single class of freight war- rants the handling of solid trains of minerals like coal, and of agri- cultural products, freight trains not infrequently include from 50 to 75 cars, and even mJ?^itized by Microsoft® s On branch lines the service requires that trains be run frequently regardless of the smallness of the number of cars which it is neces- sary to handle. The average number of cars operated as a train is affected by the fact that such an average covers trains run in both directions and the number of cars run in one direction one day may be much greater than the number to be moved in the opposite direc- tion. Livestock trains and those carrying perishable freight fre- quently are and must be moved in small units. It is only on main arteries where the great volume of business moves that long trains are common. Capacity of Freight Cars. The average capacity of freight cars on the railways of the United States increased from 30 tons in 1904 to 39 tons in 19 14. The average load carried by freight cars increased from 17.7 tons in 1904 to 21. 1 tons in 1914. Here again this average falls far short of indicating the capacity and loading attained in the sections of greatest density of freight business, and especially in the trans- portation of minerals. In 1914, 30 per cent of all freight cars had a capacity of 50 tons or over. Table II shows the increase in aver- age capacity and loading from year to year. TABLE II. Av'ERAGE Capacity of Freight Cars and Tons Carried Per Loaded Car.* Year. Average ca- Average load Year. Average ca- Average load pacity (tons). per car (tons). pacity (tons). per car (tons). 1904... 30 17-7 19IO. . . 36 19.8 1905... 31 18. 1 I9II. . . 37 19-7 1906. . . 32 18. g 1912 . . 37 20.2 1907... 34 19-7 191.? 38 21. 1 1908. . . 35 19.6 1914- •■ 39 21. 1 1909... 35 19 -3 *Statistics of Railways in the United States, Interstate Commerce Commis- sion, 1904-1914; Bulletin 81, Bureau of Railway Economics. Train Loads. The effect of heavier car loading and increased number of cars per train is shown in the train load. This has been increased at an even greater rate than either of the component factors. In 1904 the average load carried per freight train on the railways of the Digitized by Microsoft® United States was 307.8 tons. By 1914 this increased to 451.8 tons, or nearly 50 per cent. In the Eastern District, including the region of heaviest traffic, the average train load in 1914 was 537.3 tons. The progress of this growth is shown in Table III. TABLE III. Average Tons Per Train.* Year. Tons per train. Year. Tons per train . 1894 179-8 1905 322.3 1S9S 189.7 1906 344.4 1896 198.8 1907 357.4 1897 204.6 1908 351.8 1898 226.5 1909 362.6 1899 243. 5 1910 380.4 1900 270.9 1911 383.1 1901 281.3 1912 406.8 1902 296.5 1913 445.4 1903 310.5 1914 4SI-8 1904 307.8 Increase 1894 to 1014, 151.3%. Increase 1904 to 1914, 46.8%. On some lines the average train load is over 1,000 tons, and train loads of minerals from 3,500 tons to 5,000 tons are not uncommon. Sice and Power of Locomotives. To haul such heavy trains the size and power of locomotives have necessarily been increased. The average weight, exclusive of ten- der, increased from 62 tons in 1904 to 83 tons in 1914, and the aver- age tractive power increased from 22,804 pounds in 1904 to 30,420 pounds in 1914. The steady increase in tractive power is shown in Table IV. TABLE IV. Average Tractive PovifER of Locomotives, 1904-1914.* Year. Pounds. Year. Poutids. 1904 32,804 I910 27,282 190=; 23,666 191 1 27,949 igo6 24,741 1912 28,634 1907 25,823 I913 29,702 1908 26,.384 1914 30,420 1909 26,634 ♦Statistics of Railways in the United States, Interstate Commerce Com- mission, 1904-1914. Digitized by Microsoft® lO In 1914 there were in operation on the railways of the United States 775 locomotives of the Mallet type with an average weight, exclusive of tender, of 197 tons and average tractive power of 79,021 pounds. A number of these exceeded 100,000 pounds in their tractive power. Track, Bridges and Other Structures. The steady introduction of larger, heavier and stronger equip- ment and the constant increase in weight of freight trains have required corresponding changes of every other part of the railway plant. Tracks have been relaid with heavier rails. Bridges have been replaced by stronger ones capable of bearing the heavier trains. Curves have been reduced or eliminated. Grades have been cut down. Passing tracks have been rebuilt to accommodate the longer trains. Water supply and coaling stations have been enlarged to adapt their capacity to the greater water and coal con- sumption of locomotives, and have been relocated to suit the greater distances fixed between stops in the interest of more efficient opera- tion. Finally, terminal facilities, round-houses and shops have been improved and enlarged to accommodate the heavier equipment, or in some instances have been relocated to reduce the length of run. In order that at the outset a comprehensive view of the problem of limiting train length may be had, the arguments for and against the proposal will be stated in order and without discussion. Digitized by Microsoft® II SUMMARY OF ARGUMENTS FOR AND AGAINST LIM- ITING THE LENGTH OF TRAINS. Arguments for Limiting Train Length. The principal arguments presented for train-limit legislation may be summarized as follows : 1. It is difficult to transmit signals by hand or lantern from the rear of a long train to the engine. Hence, there is frequent failure of enginemen to receive or to understand such signals, and an in- crease of accidents by reason of such failure. 2. Because of the greater aggregate weight of the cars in a long train, there is more strain on the draft gear of the cars near the locomotive. Therefore, there is more risk of accident from the pulling out of drawbars or the breaking of knuckles or knuckle- pins. Also, the greater momentum of a long train puts a harder strain on the brakes of the forward cars when the train is stopped and hence entails greater possibility of failure of brake gear. 3. In coming to a stop, especially when stopping suddenly as in an emergency, or when, by the bursting of an air hose or otherwise, brakes are automatically set, the liability of the train "buckling" is greater in u. long train than in a short one, and the risk of collision from passing trains is increased. The reason assigned for this claim is the fact that in applying air brakes from the locomotive the brakes on the cars nearest the engine set first and then, in turn, those next behind, and so on until the rear cars are reached. Now, the interval of time between the first check upon the speed of the forward cars and the final slackening of the rear cars is greater on long than on short trains. Hence it is urged that there is greater opportunity and risk of a "bunching of slack," in the course of which the unchecked momentum of the rearward part of the train can exert a pressure upon the already retarded forward cars suffi- ciently powerful to force the intervening cars to one side. Because of the greater amount of "slack" in the draft gear of long trains, more severe shocks result from the sudden stopping of the train. This is claimed to be particularly the case when the brake action on the forward cars is more efifective than on the cars in the rear part of the train. In addition to the risk of buckling. Digitized by Microsoft® 12 it is claimed that such severe shocks are liable to cause injury to men upon the rear of the train, throwing them from the cars or knocking them down in the caboose. In short trains there is less slack and therefore, it is maintained, less liability of injury resulting from emergency stops. 4. Liability to accident from defective equipment is greater in long than in short trains, it is contended, because : (a) The strain upon draft gear, brake rigging and the body of the car is greater in long than in short trains. Hence, the liability of parts of cars becoming defective is intensified by the use of long trains. Not only is the number of such defective parts in a train likely to bear a certain relation to the number of cars in the train, but (b) There is greater likelihood that such defects will escape detection in the inspection by trainmen during stops on the road, since the cars in a long train are likely to be inspected more hur- riedly than in a short train. 5. More time is required for stopping and starting long, heavy trains, when it is necessary to do so to avoid accident ; hence, there is greater risk of accident from situations that might be harmless in the case of short trains. Arguments .Igainst Limiting Trnin Length. All of the arguments advanced for train-limit legislation are directed against the alleged danger of the long train as a unit of operation. They take no account of other factors affecting safety of operation as a whole that would be influenced by restrictions on train length. For this reason the case against train-limit legisla- tion distinguishes two groups of arguments on the question of safety : 1. The claim that, whether for any or all of the five reasons given, the operation of a long train necessarily or ordinarily in- volves more risk than that of a short train is not, it is contended, sustained either by an analysis of the various elements affecting safety, or by the available statistical evidence. 2. Reducing and limiting the length of trains necessarily means that more trains must be operated to handle existing business and that there must be even more than a proportionate increase in their Digitized by Microsoft® 13 present number in order to care for additional business in the future. Crowding the railway lines with more trains, a necessary conse- quence of restrictions in the number of cars operated as a single train, will give more chances for collisions, derailments, for misun- derstanding of orders because of the greater number of orders, and for any and all such causes of accidents as are likely to arise in connection with the more frequent stopping, starting, and passing of trains. In the following pages the points thus briefly stated will be con- sidered more at length. Digitized by Microsoft® 14 RELATIVE SAFETY OF LONG AND SHORT TRAINS AS OPERATING UNITS. The claim that long trains are less safe to operate than short trains is based on five arguments, which will now be discussed in order. Transit I isston of Hand Signals. It is claimed that it is difficult to transmit intelligible signals from the rear of a long train to the engineman on the locomotive. This claim would seem, at first thought, to have much force, but its force is greatly impaired, if not destroyed, by the changes in methods of operation and train control which have accompanied increases in the length of trains. Even if the claim is true, it has little bearing, it is insisted, on the question of limiting train length. In the class of service usually performed by long trains, the necessity of exchanging signals between the rear and the engine is much less .frequent than on the shorter trains that do junction or local work. Long trains are ordinarily operated in through service. They move as a unit between division points with few or no stops for local service at stations. To avoid interruptions in their move- ment, they ordinarily have the right of way over shorter trains per- forming station service. In certain instances they have the right of way over the less important passenger trains. Except to take on supplies of coal or water, or for unforeseen contingencies, they usually make no stops between division points. And these few occasions for stopping are previously known by the engineman without need of communication from other trainmen at the rear Ordinarily, therefore, there is little occasion for the transmission of signals by hand or lantern from one end of the train to the other. It is the custom on most railroads to require an exchange of signals between the front and rear ends of trains when passing stations or through interlocking plants. Such rules are carried out success- fully even where loo cars are handled as a single train. Section foremen, track-walkers, agents, operators, towermen and other em- ployees stationed at such points are in position to give warning to enginemen and trainmen of any unusual circumstances likely to cause accident to the train. Digitized by Microsoft® IS Further, practically all cars of all trains are equipped with air- brake apparatus controlled from the engine. In case of an emer- gency stop, the engineman can call out the rear brakeman for flag duty by the locomotive whistle. He is recalled by the same means. In the case of need for the transmission of signals in the reverse direction, which on the road is relatively infrequent in this class of service, the engineman can be signaled, if need be, with the air signal from the caboose, or the train can be stopped by the applica- tion of the brakes from any car, though this is a somewhat haz- ardous operation. So, also, when a train is maneuvering, fully made up with air hose connected, there is always a chance to avoid the consequences of failure of a signal from the rear to the locomotive b}- air signal from the caboose. In terminal yards the length of the train, as finally made up, does not enter into consider- ation, for it then consists of several parts, each of which is ordi- narily under the control of a member of the yard switching force who is relatively not far distant from the engine. Finally, actual tests indicate that hand signals can be successfully transmitted from the rear of long trains to the locomotive, even under adverse weather conditions. In a hearing in February, 191 5, before a committee of the Kansas legislature, C. W. Kouns, General Manager of the Eastern lines of the Atchison, Topeka & Santa Fe, described tests made with a train of 79 cars, as follows : "We had a series of tests made on January 26, which was a foggy, damp, dark day. The train [crew] consisted of an engineer, fireman, two brakemen and a conductor. The first test was made with one brakeman on top of the car next to the engine and the other on the caboose, 3,137 feet apart. Every signal made was cor- rectly answered and the engine was correctly moved on it on all of those tests at that distance. In the second test, we placed the men on the ground alongside the train because it would probably be more difficult to signal on the ground by hand than it would on top of the cars. These men were 3,188 feet apart; every signal was correctly transmitted to the engineer, who made his movements, without a flaw, upon each signal transmitted. Every man partici- pating was more than forty years old." It may be contended that such a test as this, planned beforehand and made with the employees in a specially attentive state of mind, is not entirely representative of all situations. But, as already ex- Digitized by Microsoft® i6 plained, the occasions for hand signaling usually arise, in the nature of things, in connection either with scheduled stops or special stops on some signal or situation ahead perceived by the engineer. In either case the train crew's attention is specifically drawn to the signaling to be done. The situation is practically identical with that in the special test. That test indicates that, whatever the diffi- culties presented by long trains, they can be overcome with due care and attention. Hence, the only additional risk from length of train in hand signaling is that springing from the decline in alertness that arises when operations become habitual. But the remedy for this would seem to be a recognition by the employees of their own claim that operations with long trains are somewhat different from ordi- nary operations and that they require therefore such a degree of care as would be taken in a special test. Considering, then, the relative infrequency of occasions for hand signaling in the service in which long trains are used, and consider- ing that on far the greater part of such occasions as do arise any special risk due to length of train can be met by recognizing that there is special risk and by giving the work special care and atten- tion, railway managers insist that limiting the length of trains is hardly a rational solution for the difficulties of hand signaling. Failures of Draft Rigging or Brake Gear, Parting of Trains, etc. It is claimed that increasing the number of cars in a train in- creases the liability of accident to trainmen on account of the greater danger of failure of draft gear or brake rigging, or of part- ing of trains. This assumes that the increase in the strength of the cars themselves, as well as improvements in draft gear and braking apparatus, have not kept pace with their increased number per train. It can be answered that probably no features of the equipment of cars have been subjected in recent years to more careful study and improvement, to render them adequate to the augmented load, than draft gear and brake rigging. Both have received the benefit of the combined study of manufacturers and railway officers, expert engineers and employees. Both the construction of the freight car in detail and the sufficiency of its attachments are subjects of con- stant investigation by the Master Car Builders' Association. Changes in standards and recommended practice follow closely the appearance of a necessity for increased strength. Digitized by Microsoft® 17 If it were admitted that failures of draft gear or brake apparatus are more frequent than formerly, this, it is said, would indicate merely that the improvement in the design of cars for strength of construction and equipment has not yet fully met the demands of heavier trains. Obviously, the remedy for such a situation lies in bringing the strength of equipment, including that of attachments, up to the standards that progressive experience is showing to be necessary in the operation of long trains, rather than in drastic pro- posals that would result in throwing away all the economic advan- tages secured from long trains. This progress, as already said, the railways are rapidly making. Meanwhile it is common practice, so far as it is necessary to use them in long 'trains, to provide for the location of the weaker cars at points in the train where they will be subject to the least strain. But ordinarily they are employed in service in which shorter trains fulfill the requirements of traffic. Buckling of Trains. It would be correct to say that, as an abstract proposition, and other things being equal, long trains are subject to greater risk of buckling than short trains. For it is true, as a general statement, that the longer the train the longer will be the interval before the application of the brakes on the rear cars will cause a slackening of their speed uniform with that of the forward cars. It is also true that, if frequent stops were made by long trains under emer- gency conditions and if there were a considerable number of empty cars on the front end of the train, there might occur such a "bunch- ing of slack" as would produce a severe shock from the impact of the cars in the rear. But, it is urged, speaking concretely and with regard to actual conditions, long trains are not made up and oper- ated "other things equal" with short trains. In long and heavy through trains there is not the same necessity for making up in "station order" as in the case of short trains performing local serv- ice. Instructions are therefore commonly given for the distribution of empty cars through the long train to prevent the possibility of the slack "bunching." Moreover, air-brake manufacturers have anticipated non-observance of such instructions and furnish a "light and load" brake which maintains the same relative braking power upon the empty car as upon the loaded car. Again, it is pointed out that emergency stops are less frequent Digitized by Microsoft® i8 with long and heav\- trains than with short trains, because of the nature of their customary service and because of the rights allowed them on the road ; that air brake hose does not regularly burst and set brakes suddenly; that enginemen are invariably instructed, in applying brakes, to manipulate the controlling valve with regard to the length and weight of the load behind the tender. If the brake is properly manipulated, there is no reason why in ordinary service a long train should not be brought to a stop with little or no shock and without undue strain upon car sills or draft gear. It thus ap- pears that, because the service in which long trains are ordinarily engaged is of such a character that stops are infrequent and the speed is lower, the occasions and conditions in which sudden stops might produce buckling are few. In ordinary stops the situ- ation is entirely within control of the engineman, whose skill and carefulness are the determining factors. So far as strains upon brake apparatus are concerned, it is as- serted, the argument is in favor of the long train. Short trains ordinarily move at higher speed between stops than long trains and in stopping subject brake equipment to more severe strains. A method of administration could hardly be defended which should prepare for the occasional situations in which emergency stops might be disastrously made, or for occasional carelessness on the part of the engineer in ordinary stops, by restricting, in all cases, such an important factor of railway operating efficiency as length of trains. Detection of Broken Car Rigging En Route. The argument for limiting train length on the ground that break- ages in draft gear or brake rigging, or other defects, occurring on the road are more liable to escape detection in long trains than in short, has this weakness. The risk claimed from this source may be regarded as due either to more frequent breakages en route under the strain of long trains, or to greater difficulty of detecting such as do occur, in view of the length of the train to be inspected in the time available during stops. With respect to the first view, it is claimed that the proper remedy, — already being applied, — is to con- tinue the steady improvement in the strength of equipment and the steady substitution in long trains of newer and stronger cars for older and weaker ones. In this view, the risk, whatever it may be Digitized by Microsoft® 19 is temporary. With respect to the second point of the argument, rigid terminal inspection of trains before departure leaves for con- sideration only breakages that develop on the road. As to these, it is to be noted that they occur under the sudden strains of such oper- ations as stopping, starting, and backing ; and that, in the service in which long trains are used, these operations are relatively infre- quent. Hence the conclusion is reached that whatever the addi- tional risk of a permanent character from the difficulties of inspect- ing long trains may be, it cannot be of an appreciable amount, and thus would be far from justifying the policy of a limitation of train length. Possible Promptness of Control of Long Trains. The final argument for limiting train length is that long trains can be stopped or started only with comparative slowness. Hence, in any of these recurrent situations in which a collision or some other accident is impending and can be avoided by promptly stop- ping a moving train, or starting one at rest, the chance of avoiding such an accident is reduced by the greater time necessary with a long train to perform the needful movement. This claim is true to the extent that brakes must be applied to long trains more grad- ually than is permissible with short ones, in order to avoid danger of buckling, as already explained, and to the extent that a heavy train cannot be gotten under way so promptly as a light one. But, as between a train of nearly fifty cars and one of many more than fifty cars, there should ordinarily be no difference in the require- ment of careful manipulation of brakes. A sudden or severe ap- plication is not permissible in either case except in emergency. Then the emergency must be met regardless of the length of train. Over against the additional risk from this source, railway men strongly urge that there should be considered the great increase in the num- ber of threatening situations that would follow the multiplication of trains on the road, if their length is limited, in order to move the same volume of business. This point is developed more fully later on. But it needs no explanation to see that if the opportuni- ties for accidents are increased, the additional accidents actually happening on that account may well be far in excess of the number that would be avoided through the greater mobility of short trains. Digitized by Microsoft® 20 Conclusion. The foregoing general analysis of the arguments against long trains shows the following weakness to be common to all except the last. The hazard in train operation lies mainly in maneuvering — in starting, stopping, backing and switching — and in speed. But the service in which long trains are used is one in which these oper- ations are relatively infrequent, and in which the trains run at a more uniform speed, which is slower than that commonly attained by shorter trains in local and way service. Therefore, it is consid- ered obvious that the maximum possible effect that the length of the trains in through service can have on that large proportion of casualties which arise from train operations is less than might at first seem possible in view of the arguments advanced. This conclusion will appear, also, from an examination of such data as are available for making a comparison of the number of casualties with long and short trains. Entirely pertinent statistics showing the comparative measure of risk in the operation of long and short trains are not available. There are, however, some sta- tistics that have an important bearing on the question. Digitized by Microsoft® 21 STATISTICS REGARDING COMPARATIVE NUMBER OF ACCIDENTS TO LONG AND SHORT TRAINS. The advocates of train-limit legislation support their principal claim that long trains have caused an increase in accidents by refer- ence to the accident statistics of the Interstate Commerce Commis- sion. But no definite and relevant comparisons have been produced by them to show that such casualties have shown a tendency to in- crease with the increase in length of trains. The extent to which such statistics have been employed is limited usually to general statements that casualties due to railway operations are increasing from year to year. When more specific and definite comparisons have been made, they have failed to distinguish accidents that might, on the basis of their general reasoning, be considered as possibly related to long trains and accidents that cannot, on any reasoning whatever, be even remotely affected by long trains. It ought to go without saying that, to prove anything concerning the danger as- serted to lie in long trains, the statistics offered in proof must be limited to accidents of such a nature as can be seen to have some relation to the length of the train, and must cover sufficient time and sufficiently varied conditions to be representative of actual tenden- cies and not abnormally aft'ected by temporary or local influences. An example of neglect in all these respects is found in accident statistics oft'ered by representatives of the Brotherhood of Railway Trainmen, who appeared at a hearing before a committee of the Kansas legislature in February, 191 5, in support of a bill limiting trains to not more than fifty cars. They stated that, according to the quarterly Accident Bulletin of the Interstate Commerce Com- mission for July, August, and September, 1901, 366 employees were killed and 1,679 employees were injured during the quarter; and that, according to Accident Bulletin No. 51, for the months of Jan- uary, February, and March, 19 14, 752 employees were killed and 37,315 were injured during that quarter. The first thing to note about these citations is that they are in- accurate statements of what the Interstate Commerce Commission bulletins show. The 366 killed in 1901 were trainmen only. The number killed, among all employees, was 615. The number stated as the injured does not appear in any of the tables in the Bulletin Digitized by Microsoft® 22 cited. Of trainmen alone, there were 4,619 injured, and of all em- ployees there were 8,361 injured. In 1914, the figures quoted m- clude all those casualties in railway shops, or otherwise than in purely railway occupations, known as "Industrial accidents," which were not reported by the Interstate Commerce Commission in 1901, but had become a part of the returns by 1914. These amounted to 79 killed and 24,679 injured. To get statistics that are really com- parable with those for 1901, all such industrial accidents — which can be in no way whatever affected by the length of trains — ^must be omitted. That leaves 673 killed and 12,636 injured in the first quarter of 1914, to be compared with 615 killed and 8,361 injured in the third quarter of 1901. Even after correcting these inaccuracies, and taking only the statistics just given as more nearly representing the same classes of employees in the two years, the statistics are still without significance for the claim that the growth of long trains has caused an increase in casualties. In the first place, three months, or even a full year, for that matter, is far too short a time to reveal a surely representative picture of average condi- tions. And in this case, the two periods were most unhappily chosen. The three months of July, August, and September, 1901, were the very first three months of the existence of the department of the Commission that studies accident reports and prepares the accident Bulletins. It is conceded by everybody that the statis- tics gathered during, not only the first three months, but the first year or so, were distinctly incomplete because the organization for gathering, sifting, and compiling them was not fully worked out for some time. It will be clear to everyone, then, that no significance for the present question can attach to a comparison between the first quarter's reports and the latest quarter's reports in the Acci- dent Bulletins of the Interstate Commerce Commission. But the comparison cited is even more fundamentally defective in that it neglects to distinguish between accidents fairly subject to inquiry as to whether affected by the length of trains, and accidents that can have only a very remote relation to length of trains, if any at all. There are many trainmen injured who have nothing to do with long trains, and many of the casualties to employees who are connected with trains are in no way due to the length of the trains. Yard trainmen, for example, have nothing to do with trains on the Digitized by Microsoft® 23 road, but handle long trains in yards only when such trains are in separate sections. Therefore, it hardly needs to be argued that such comparisons of casualties prove nothing as to the increase in danger claimed to have followed the growth in number of long trains. The increase in total casualties to trainmen may just as well have been due to many causes in no way related to length of train as to causes that may pos- sibly be so related; and the increase may just as well have occurred while there was an actual decrease in the number of casualties to those employees whose risk, it is claimed, is affected by the length of the trains. Even though allowance be made for all these classes of employees, and the statistics limited to casualties to trainmen on the road (who, of all employees, can most reasonably be assumed to be affected by train length), still the comparison would prove nothing, because so many of the casualties even to trainmen are due to causes in no way connected with length of train, and because no separation is made between casualties to trainmen on long trains and those to trainmen on short trains. Statistical verification of the claim that long trains are more hazardous to operate than short trains would require, at the very least, that the data for casualties in connection with long trains be separated from those in connection with short trains, both classes of trains being operated under substantially the same condi- tions. Otherwise no comparison is possible between the risk at- tendant upon each class of trains in itself. But this need is not met by any statistics regularly reported by the Interstate Commerce Commission, because the distinction between long and short trains has never been observed in its compilations. In the absence of such comprehensive data, various railways have classified the casualties on their lines according as they occurred on long or short trains, and have presented the results at hearings before legislative committees that have considered bills for limiting train length. Several of these statements are significant, though the data are so arranged that they are not readily comparable with each other. At a hearing before a committee of the General As- sembly of Virginia on a bill proposing to limit all freight trains to 50 cars, N. D. Maher, vice-president of the Norfolk & Western, presented a statement for that road covering a period of two years showing the number of trains of over and under 50 cars and the Digitized by Microsoft® 24 number of deaths and injuries resulting from their operation. The statistics were as follows: Trains and Train Mileage. Number of trains of more than 50 cars (freight) i6.7l2 Number of trains of 50 cars or less (mcludmg passenger) 31. 112 Total mileage of freight trains of more than 50 cars 4.173,027 miles. Total mileage of trains of 50 cars or less (mcludmg passenger trains) .. 7,770.293 miles. Total passenger train mileage 5.137.293 miies. Total mileage of freight trains of SO cars or less 2,033,000 miles. Casualties to Employees. Trains of more than 50 cars kjljed i. injured 179 Trains of 50 cars or less killed 20, injured 521 Casualties to Other Than Employees. Trains of more than 50 cars Wiled 8, injured 16 Trains of 50 cars or less killed 76, injured 193 Of the 20 deaths and 521 injuries to employees caused by trains of so cars or less, 5 deaths and 114 injuries resulted from the move- ment of passenger trains, leaving 15 deaths and 407 injuries due to the movement of freight trains of 50 cars or less. In the cases of persons other than employees, 24 deaths and 42 injuries caused by trains of 50 cars or less resulted from the movement of passenger trains, leaving 52 deaths and 151 injuries caused by freight trains of 50 cars or less. There were thus in the two years 9 persons killed and 195 injured by freight trains of more than 50 cars, and 67 persons killed and 558 injured by freight trains of less than 50 cars. The distribution of casualties between the two classes of trains, compared with the distribution of number of trains between the two classes, including all trains, is then as follows : Percentage Distribution of Trains and Casualties. (Norfolk & Western.) Class of trains Number of trains. Per cent of total. Number killed. Per cent of total. Number injured. Per cent of total. Trains of more than 50 cars Trains of 50 cars or less 16,712 31,112 35 6S Di gitiTdci hy Mi 96 rosoft<§ 8.6 91.4 195 714 21.5 78.5 25 It appears from this comparison that, while there were more than one-half as many trains of over 50 cars as of trains of 50 cars or less, including passenger trains, the number of fatalities in connec- tion with the long trains was less than one-tenth of the number in ■connection with the short trains, and ofjnjuries slightly more than one-fourth. This discloses in a general way that the casualties on long trains were relatively fewer than on short trains. By how much fewer is more forcefully and definitely apparent when it is stated that, for an equal number of long and short trains, the number of persons killed in connection with long trains on the Norfolk & West- ern was only about one-sixth of the number killed in connection with short trains, while the number injured was only one-half as many. The data in this case are complete enough to permit casualties to passengers and employees to be considered separately and with rela- tion to the mileage of each class of trains. The results can also be distinguished as between the passenger and freight service. Elimi- nating from the calculation the numbers of casualties in connection with the 'movement of passenger trains and relating the remaining numbers of casualties, connected with freight service, to the mileage of the two classes of freight trains, the result is as follows : Percentage Distribution of Freight Trains and Casualties. (•Norfolk & Western.) Mileage. Per cent of total. Eraploj-ees. Class of trains. Killed, Per cent of total. Injured. Pel cent of total. Freight trains of more than 50 cars Freight trains of 50 cars or less 4,173.027 2,633,000 61.3 38.7 IS 6.2 93-8 179 407 30. S 69- 5 class of trains. Passengers. Killed. Per cent of total. Injured. of total. Freight trains of more than 50 cars . Freight trains of 5° cars or less. . 52 133 86.7 16 151 Digitized by Microsoft® 9.6 90.4 26 In this comparison it appears that, though the mileage of freight trains of more than 50 cars was 61 per cent of the entire mileage of freight trains, the number of fatalities to employees in connec- tion with these trains was only 6.2 per cent of the total number, and fatalities to passengers only 13.3 per cent of the total. Put in more definite terms, the data show that, for an equal number of miles run by long and short trains, the number of employees killed in connection with long freight trains was less than 5 per cent of the number killed in connection with short trains, while the number injured was only 28 per cent. During a hearing before a committee of the Virginia Assembly, George P. Johnson, then general manager of the Chesapeake & Ohio, presented a statement relating to that road and covering a service of 4,896,326 train-miles for the year ending June 30, 1913, showing the relative numbers of casualties resulting from the oper- ation of trains of more than 50 cars and of trains of 50 cars or less, as follows : Percentage Distribution of Trains and Casualties. (Chesapeake & Ohio.) class of trains. Number of trains. Per cent of total. Num- ber killed. Per cent of total. Number injured. Per cent of total. Trains of more than 50 cars Trains of 50 cars or less 12,073 13.099 48 52 4 31 II. 4 88.6 18 163 9-9 90.1 This statement shows that 12,073 trains, or 48 per cent of the whole number operated, consisted of more than 50 cars, and 13,099, or 52 per cent, consisted of 50 cars or less, including passenger trains. While the number of long and short trains was thus nearly equal, the number of fatalities in connection with the long trains was only one-eighth of the number in connection with short trains, and the number of injuries less than one-ninth as great. A summary has been made by the railways operating in Illinois showing for the year 1913 the number of employees killed or seriously injured on all trains run in that state by 2-? roads TTn^c^ Digitized by Microsoft® ^ ^ ^^^^^ 27 are subdivided to show the relation of numbers of casualties occur- ring in connection with the operation of trains of over 50 cars and trains of 50 cars or less. The data in detail are shown in the accompanying table : Percentage Distribution of Trains and Casualties. (Illinois Railways, 1913.) Class of trains. Number of trains. Per cent of total. Num- ber killed. Per cent of total. Number injured. Per cent of total. Trains of more than 50 cars Trains of 50 cars or 236,702 413,510 36 64 23 43 34.8 65.2 46 149 23.6 76.4 Here again, though the number of trains of more than 50 cars was 57 per cent as great as the number of trains of 50 cars or less, the number of fatalities in connection with the long trains was only 53 per cent as great as in connection with the short trains, and the number of injuries less than one-third as great. Or, put differ- ently, for the same number of trains of each class, the number killed in connection with long trains was only about nine-tenths of the number killed in connection with short trains, while the number in- jured was only about one-half. It may be argued that statistics of this character are not conclu- sive because of the difference in the character of the services per- formed by the two classes of trains. Reference is made to the fact that short trains include all local and way service trains in connec- tion with which the service performed by trainmen includes sources of risk not found in the same degree in the through service per- formed by long trains. But if, as this argument concludes, long trains are used in a class of service in which, owing to its character, the risk of accident is relatively less, it follows that this particular argument gives relatively little or no ground for legislation to reduce the length of the trains used in this service. Furthermore, the fact that there are elements in the character of the service ordinarily performed by short trains that render their Digitized by Microsoft® 28 operation more liable to produce casualties, it is explained by the opponents of train-limit legislation, would argue for a possible increase in casualties if short trains were substituted for long ones. The long train ordinarily moves at a regular rate of speed, which is slower than the speeds often attained by short trains. It ordinarily makes fewer stops, and therefore incurs less risk of accidents in starting and stopping. If train length were restricted as proposed, the shortening of trains would necessarily increase those risks of accident which arise from the operation of relatively short trains at higher rates of speed. Analysis of the Causes of Accidents, with Respect to the Effect of Long Trains. That long trains do not tend to increase the risk of accidents in any material degree, if at all, is indicated, not only by the general considerations previously stated, but also by an analysis of the causes of accidents to trainmen, as assigned in the statistics of the Interstate Commerce Commission. The causes of accidents to trainmen, as reported by the Commission, fall into fifteen principal classes. Casualties classified according to cause, with sub-classes in the case of such main classes as could, by any possibility, be affected by the length of trains, are presented in Table V following, which applies to the fiscal year ended June 30, 1915. In the table, those immediate causes are italicized which may be reasonably con- sidered as having been themselves due to length of train in more than rare instances. Casualties to trainmen, only, are given because they are more likely to be influenced by the length of trains than are accidents to any other class of persons. In the case of collisions and derailments, the Commission does not classify the casualties in detaU according to causes. Digitized by Microsoft® 2y TABLE V. Causes of Casualties to Trainmen. Year Ended June 30, 1915. Number of casualties Cause. Number of to trainmen. 1 1 Killed. Injured. I. Collisions 3,538 435 60 1,199 a. Rear' (p. 30)^ b. Butting' (p. 30) 282 c. Trains separating (p. 30) .... 302 d. Miscellaneous (p. 30) 2,518 2. Derailments 6,849 114 1,174 a. Defects of roadway (p. 32).. 1,507 b. Defects of equipment (p. 32). 3,416 (i) Broken or burst wheel. 335 (2) Broken flange 346 (3) Loose wheel 100 (4) Miscellaneous wheel defects 86 (S) Broken or defective axle or journal 367 (6) Broken or defective brake rigging 390 (7) Broken or defective draft ^ear 280 (8) Broken or defective side bearings 141 (9) Broken arch bar 222 (10) Rigid trucks 177 (11) Failure of pozver brake apparatus, hose, etc.. 353 (12) Failure of couplers 2ig (13) Miscellaneous causes.. 400 c. Negligence of trainmen, sig- nalmen, etc. (p. 30) 297 d. Unforeseen obstruction of track, etc. (p. 30) 244 e. Malicious obstruction of track, etc. (p. 30) 70 f. Miscellaneous causes (p. 30). 1,31s 3. Accidents to trains, cars or en- gines, except I, 2 and 4 (p. 24) 4 203 4. Bursting of. or defects in, locomo- tive boilers, etc. (p. 24) 13 429 5. Accidents to roadway or bridges not causing derailment, such as fires, floods, landslides, etc. (P- 24) 'Some rear-end collisions may conceivably have occurred to long trains after being forced to stop by reason of breakage due to their great length. Some rear and butting collisions may conceivably have happened because of greater delay in stopping or starting trains of great length. But it is not believed, as a practical matter, that such accidents could have been numerous enough to deserve recognition in this table. 'Page numbers refer to J.C. C, Accident Bulletin, No. 56 (1915). Digitized by Microsoft® 30 Number of accidents. Number of casualties to trainmen. Killed. 6. Coupling or uncoupling cars (ex- clusive of accidents with air or steam hose) (p. 33) a. Adjusting coupler with foot.. b. Adjusting coupler, cars acci- dentally started c. Careless manipulation of un- coupling lever d. Cars not equipped with auto- matic coupler e. Coupler broken, using link and pin or chain f. Coupling damaged cars g. Coupling with chain or other emergency appliance on curve too sharp for auto- matic coupling h. Coupling with chain or other emergency appliance be- cause of uneven track i. Coupling or uncoupling safety chain /. Fingers or hand caught be- tween uncoupling lever and body of car k. Uncoupling without using lever (unnecessary) /. Uncoupling without using lever (lever out of order) . m. Foot caught in frog, switch, or guard rail n. Opening or closing knuckle when cars were near to- gether, miscalculated speed. 0. Opening knuckle when cars were near together, engine accidentally started p. Opening knuckle, part of de- fective coupler fell on foot. q. Opening knuckle, lost foot- ing r. Riding on car to uncouple, slipped off J. Struck by object at side of track t. Caught by unexpected move- ment of car, due to slack running in u. Caught by unexpected move- ment of car, due to misun- derstanding in giving hand signals V. Uncoupling moving cars and lost footina Injured. 1,949 176 60 39 35 43 16 28 38 4 107 4 22 IS 160 9 93 54 4 68 4 45 I 41 155 12^ Digitized by IVIicrosoft® 31 Number ol casualties Number of to traiumen. accidents. , -^ . Killed. lujured. Zi.'. Parts hard to move, causing delay i 45 X. Went between cars unneces- sarily and contrary to rule. 8 82 y. Hand caught between project- ing load and end of next car .:. No witness (fatal accident) . . aa. Other causes hh. Unexplained While doinpf other work about trains (not in shops or engine houses) or while attending switches (p. 35) a. Shaking grates fc. Firing engine (raking fire, shoveling coal into fire, etc.) c. Coaling engine d. Taking water at water cranes e. The working or action of re- verse levers i. Scalded by water from squirt hose : g. Throwing switches h. Poling cars 1'. Coupling or uncoupling air or steam hose (includes the turning of angle cocks) .... y. Using hand brakes k. Loading or unloading freight, baggage, etc I. Cinders in eye m. Stepping on or stumbling over objects, stepping in holes, slipping, etc., on or at side of track n. Unexpected or abnormal movement of trains, cars, or engine o. Struck by objects on or at side of track (not fixed structures) p. Struck while riding in or on trains, cars, or engines, by trains, cars or engines on adjoining track 4 309 q. Struck while riding in or on trains, cars, or engines, by projections from traiiis,_ cars or engines on adjoining track 2 102 r. Miscellaneous 12 2,731 Digitized by Microsoft® 9 4 112 S8 76 17,770 704 I 1,831 238 I 440 542 219 838 128 12 I 412 1,098 I 1,463 1,060 18 3,384 33 2,082 I 189 Number of Cause. accidents. 8. Coming in contact, while riding on cars, with overhead bridges, tunnels, or any signal appa- ratus, or any fixed structure above or at the side of the track (p. 24) 9-10. Falling from, or getting on or off, cars or engines (p. 34) . . a. Fell from roof of box car by reason of — (i) Defect in car (2) Ice or snow (3) Parting of train. (4) Derailment, collision, or shock due to abnor- mal movements of cars other than those under s preced- ing (5) JVhile setting brakes... b. Fell from — (6)-(7) Freight car other than box car. ... (8) Engine or tender (9) Passenger car (10) Engines, tenders, or cars (all kinds) at rest (11) Miscellaneous causes.. (12) Not clearly explained. (13) Slipped getting on moving trains or cars. (14) Jumping off moving trains (is) Jumping from engines or cars anticipating collision, derailment, • or other accident . ... (16) Fell from engines or cars by reason of de- fective handholds and sill steps (17) Getting on or off mov- ing engine (18) Caught in frog, guard rail, or switch II. Other accidents on or around trains (p. 24) (Only one of twelve subclasses of causes appears to have any possibility of connection with train-lengths, viz : "Unex- pected or abnormal move- ment of trains, cars, or en- ,aDvmcroso. IS, and thus affect the 51 casualties from train accidents that happen to other employees than trainmen on the road. Also, it is quite probable that increased train density would swell the number of casualties sustained in connection with moving trains, but not in coUisions, derailments, or other acci- dents to trains themselves. The more trains there are running, the more occasions there are for casualties to employees in connection with the numerous duties of operating trains, or working about yards, stations or warehouses where trains pass or move about. The more, too, are the occasions for passengers and others to be struck by passing trains in stations, and especially at highway crossings. Such casualties as these result from many other causes and occasions besides train density, so that train density alone is not responsible for a sufficiently large proportion of all such casualties to make the total from all these causes vary in very close agreement with it. However, although it is thus not possible to find, in the data that are available, any statistical support of the conclusion deduced from other than statistical considerations, the conclusion has firm ground in those considerations to sustain it. Relation to Train Density of the Risk of Accidents and Casualties. In all of the foregoing statistics train density has been compared only with the total number of the accidents and casualties consid- ered. But the claim of railway men is, not merely that the number of accidents and casualties will increase with train density, but also that the risk will increase. That claim is supported by the consid- eration that occasions or opportunities for accidents and casualties increase more than in proportion to the increase in number of trains. This argument cannot be given statistical verification, although it would seem a perfectly sound deduction from the facts. The diffi- cultiesin the way of statistical verification lie partly in the lack of a satisfactory measure of risk and partly in the concurrent effect of other influences on risk besides that of train density. Yet it cannot be doubted that there is an added risk run by each trainman and each passenger when the number of trains is increased, and vice versa. That is to say, train-limit legislation, by making train density higher than it would otherwise be, will not only result in accidents to more trains and casualties to more persons than otherwise, but will increase the chances of each train and of each trainman and passenger meeting with disaster. Digitized by Microsoft® 52 Summary of Evidence on the Effect of Train Density Upon Acci- dents and Casualties. A consideration of the causes and occasions of accidents and casualties, as related to train density, and an examination of the statistics thereon, thus leave no doubt, it would seem, that the num- ber of collisions and derailments and the number of casualties there- from to trainmen on the road and to passengers are largely and directly affected by the density of trains upon the railways. The same consideration of the causes and occasions of accidents and casualties leads directly to a similar conclusion with respect to many of the other train accidents, besides collisions and derailments, and of the casualties therefrom, and to many of the casualties to all employees and passengers, and even other persons, that occur in accidents connected with moving trains, but not train accidents, so called. These all appear to be either directly or indirectly influ- enced by the density of trains. This conclusion stands, regardless of the fact that the number and concurrent effect on such accidents and casualties of other influences than train density are so great as to prevent the statistics from revealing the relation of such accidents and casualties to train density. Considering now that it has been shown that a limitation upon the length of freight trains would unavoidably cause an increase in the number of trains, there seems to be adequate basis for the con- tention that, if the length of freight trains be limited by law, train density will be greater than it otherwise need be. To the extent that train density is greater, it will cause the number of accidents and casualties to be greater than they need be. And not only the number, but also the risk of accident and casualties would be increased. The Greater Relative Importance of Train Density in the Safety of Raihvay Operations. Since that result would follow a limitation of train length railway men are convinced that the claim of employees concerning the dan- ger of operatmg long trains becomes of minor importance For even ,f it be conceded that long trains are more hazardous units of Digitized by Microsoft® 53 railway operation than short trains, it appears that the safety of railway operation as a whole is chiefly determined by the much more important factors involved in train density. They contend that, whatever the shortening of trains might conceivably accomplish in reducing the risk of casualty in connection with the operation of the individual train, it would be much more than oflfset by the increase in risk of collision and other accidents, and of casualties therein, that would result from the increase in the number of trains operated. It appears from the statistical comparisons presented in the fore- going that there is a close relation between train density and num- ber of casualties to trainmen and to passengers. It is admitted by railway officials that other influences have been exerted in the direc- tion of securing a greater degree of safety in railway operation ; but since fluctuations in train density in either direction are ordinarily accompanied by changes in number of casualties, the influence of train density appears to predominate. Train density, under any given conditions, and especially so far as freight trains are con- cerned, will ordinarily depend upon the number of cars which it is permissible and practicable to operate as a single train. The fewer cars per train there are, the greater the train density and the more numerous in proportion are the accidents. Decrease in Casualties in Relation to Service Performed by Railways. There is another aspect of the argument on safety that should be noted if the matter is to be considered in its entirety. The rail- way business is not unlike many others, on which the public is de- pendent, in that loss of life and personal injury are unavoidable incidents of the operations by which its service is rendered. Since, therefore, casualties are to be expected, the only question is one of steadily reducing them in proportion to the service performed. Now that is exactly what the railways have been doing, and one of the principal means of doing it has been the use of long train units. A summarv comparison of the increases in railway traffic as a whole over a period of years, and in the numbers of casualties to employees and passengers, is presented in Table XV The period covered is from 1905 to 1914, inclusive. Since compari- sons between individual years may be objectionable on the ground Digitized by Microsoft® 54 of the unrepresentative character of one or both of the years selected, the present study presents annual averages of the data for two periods of five years each^igos to 1909 and 1910 to 1914, inclusive in each case. The comparison of the data for 1905-1909 with that for 1910-1914 indicates the general tendency of number of casualties in relation to volume of service performed by the rail- ways. The entire period has been especially characterized by m- creases in the length and capacity of freight trains. TABLE XV. Summary of Increases in Raiiway Traffic in Relation to Numbers of Casualties. Comparison of Averages for Five-year Period 1905 to 1909 with Averages for Period 1910 to 1914, Inclusive* Increase per cent Average per year Average per year 1905-1909 to 1905-1909. lglo-1914. 1910-191+. Freight ton miles 215,225,000,000 272,641,000,000 26.7 Freight train miles 585,300,000 625,633,000 6.9 Passengers one mile ... . 26,976,000,000 33,739,000,000 25- 1 Passenger train miles.. 492,030,000 581,890,000 18.3 Total freight and pas- senger train miles. .. . 1,077,000,000 1,208,000,000 12.2 Number of employees.. 1,502,945 1,725,842 14-8 Mileage of roads 226,533 249,388 10. i Number of employees per TOO miles of line.. 663 692 4-4 Number of employees : Killed 3,447 3,i86 7.6 Dec. Injured 54,357 55,047 i -3 Number of passengers : Killed 469 353 24.7 Dec. Injured Il,9i7 15,047 26.3 *Computed from Statistics of Railways in the United States, Interstate Commerce Commission, and from Accident Bulletins for the years included. "Industrial Accidents" are excluded. The comparisons show that the increases in casualties have been much smaller in proportion than the increases in the volume of railway service rendered, except in the number of passengers in- jured. In fact, in the last five-year period covered there were abso- lute decreases in fatalities to both passengers and employees, not- withstanding an increase of about 25 per cent in the average amount of passenger and freight traffic handled and of over 14 per cent in the average number of employes. In view of the sharp reduction Digitized by Microsoft® 55 in number of passengers killed, part of the increase in number in- jured may reasonably be regarded as due to improved methods of reporting the details of unimportant casualties. The reduction in fatalities leads also to the conclusion that the severity of passenger casualties as a whole has been considerably reduced. Suiumary of Evidence on Relation of Train Density to Accidents. From this examination of the experience of railways over a period of years, there would seem to be no question that the number of collisions and derailments, and the number of casualties to passen- gers and trainmen in train accidents, all bear a well-marked relation to total train density. The relation which has been demonstrated to exist between train density and casualties resulting from railway operation appears clearly to be one of cause and efifect. It appears also that notwithstanding progressive increases in vol- ume of business and in number of employees, there has been no cor- responding increase in number of casualties incident to railway operation. On the contrary, in recent years there is shown a tend- ency to decrease in the number of casualties in comparison with the increase in volume of railway business. Because of the increase in length of many trains, the increase in volume of business has not been accompanied by a corresponding increase in accidents and casualties. Digitized by Microsoft® S6 ECONOMIC CONSIDERATIONS INVOLVED IN TRAIN- LIMIT LEGISLATION. If it could be shown that the operation of long trains entailed increases in casualties, purely economic considerations would be forced to give way to the more important consideration of safety. But it appears from the foregoing discussion of the arguments on safety that the net effect of increase in the length of trains is not to increase the hazards of operation, but so far as is determinable has the opposite tendency. This being the case, it becomes proper and pertinent to point out briefly some economic considerations involved. Much of Recent Railway Investment Would Be Rendered Valueless. The tendency of transportation on both water and land always has been toward the use of larger units of operation, the main pur- pose being to reduce the cost per ton of rendering the service. The development of the large unit in railway operation has involved large investments of capital in forms designed solely to permit the use of larger train units. The productiveness of many of these investments would be nullified by legislation designed to arrest this form of economic advance. Investment in Roadivay and Track. — To enable them to haul longer and heavier trains railway companies have made large in- vestments for the reduction of grades and the elimination of curva- ture in their main track and for the construction of long passing and yard tracks. The investment for these purposes would be rendered practically valueless by the enactment of legislation that would not permit the operation of long trains, the very purpose for which the investment was made. Investment in Car Construction and Equipment. Increased strength of car construction, among other things the substitution of steel for wood as a structural material, has been an essential element in increasing the capacity of the unit of transportation Improvements in brake systems, couplings, etc., have had the pur- pose and effect of making the train more nearly a unit both physically and as to control, than formerly. These features of progress in car construction and equipment have permitted an in- Digitized by Microsoft® 57 crease in the number of cars hauled in a train without correspond- ingly increased liability to casualties resulting from failure of the car body, or of connections between cars, or of connections between cars and locomotive. Some of the principal advantages derived from the use of steel under-frames and center sills, of improved draft gear, of brake apparatus highly specialized with reference to its required service, and other features of strength in individual parts of equipment, would be greatly reduced and the increased in- vestment in them rendered largely unproductive, if the number of cars handled in single trains should be limited as proposed. Investment in Larger Locomotives. — The tractive power of loco- motives has been greatly increased in order that a longer and heavier train may be hauled by a single locomotive. Such increases in tractive power have been obtained only at an increased cost per loco- motive. The additional tractive power of these large locomotives and the additional investment made to secure it would become valueless under the proposed legislation. Investment in Engine Terminals. — That portion of a railway's in- vestment in roundhouses and in all such facilities as turn-tables, which has been made to adapt them to the accommodation of large locomotives, would be rendered unproductive. Many Bconoinies in Operating Expenses Would Be Lost. Limitation of train length would prevent many reductions in operating expenses now being effected by means of large train- loads. In 1890 the average number of tons handled per train in the United States was 175. i. The average revenue of the railways per ton per mile was .941 cents. In some sections of the country it was as high as 1.651 cents. At the average rate per ton per mile and at the average tonnage per train in 1890, the railways received $1.65 per freight train per mile. In 1914 the average number of tons of freight hauled in a train was 451.8. The average revenue per ton per mile was .733 cents. The earnings per freight train per mile at this lower average freight rate were $33 1. or twice as great as they were at the higher average rate charged in i8go. These fig- ures indicate clearly how the increase in the length and capacity of trains has enabled the railways to make greatly reduced rates to the shipper, pay greatly increased wages, and at the same time steadily Digitized by Microsoft® 58 improve their service. Without the increase in the length and capacity of trains, either rates could not have been reduced, or wages could not have been so increased, or the railways would have been compelled to economize at the expense of service or suffer financial ruin. Additional Investment Would Be Made Necessary. Legislation limiting the length of freight trains would not only render valueless a great part of recent investments, but would re- quire additional investment on the part of the railways to accom- modate their operations to changed conditions. It would necessitate an increase in the number of locomotives equal to the increased number of trains required to handle a given traffic. The increased number of locomotives would make necessary re- modeling of locomotive terminal facilities in order to provide addi- tional stalls in roundhouses and more track room in yards propor- tionate to the increase in number of locomotives. Additional tracks, sidings, switches and signal apparatus would be required in a ratio indicated approximately by the increase in number of trains that would result from limitation of train-length. If traffic were equal in both directions and at all times of the day, and if every train met every other train, meeting points, alone, of trains would increase as the square of the number of trains. Addi- tional passing points, with trains in the same direction, but of differ- ent speed, would cause the total of both meeting and passing points to increase in even a greater ratio. And meeting and passing facili- ties would have to be provided accordingly. Since the actual con- ditions are not so uniform as that, the increase in the number of meeting and passing points might be more or less than this, but there would be a material increase. Financial Effects of Train-limit Legislation. When a bill to limit the length of freight trains to 50 cars was pending in the legislature of Illinois in 191 5, data were compiled bv 23 roads, having 8,781 of the 12,611 miles of railway in that state showing the estimated effect that would be produced by such leg-is- Digitized by Microsoft® 59 lation on their investment and operating expenses. These data, summarized, were as follows : Investment made valueless : a. For grade reduction and heavy rail $30,631,943 £°'' siding and terminal track extensions 6,727,453 c. For heavy motive power 8,075,160 Total $45,434,556 Increase in operating expenses per year : a. Increase in payroll, enginemen and trainmen $2,266,059 b. Increase in payroll, other employees 976,141 c. Increase in cost of fue' 1,025,480 Total $4,267,680 New capital required : a. For locomotives $4,007,080 b. For passing and terminal tracks, etc 4,138,850 c. For locomotive terminal facilities 3,955,000 Total $12,100,930 d. For second, third, and fourth track requirements in ten years $43,495,360 Reduced to a mileage basis, it appears from this statement that the railways of Illinois placed the amount of investment that would be rendered valueless by a limitation of the length of freight trains to 50 cars, at $5,198 per mile ; increase in operating expenses per year, $488 per mile; immediate investment of new capital, $1,384 per mile; amount of new capital the investment of which would be made necessary in 10 years, $4,976 per mile. Whether similar legislation would produce proportionate effects on all the more than 250,000 miles of railway in the United States can only be con- jectured. Effect on Rates Paid by the Public. Railways have onlv one source from which to derive the means for paying their fixed charges and operating expenses. This is the revenue received by them from the public for the transportation of passengers and freight. Since legislation limiting the length of trains would increase their fixed charges and operating expenses, it is clear that it would, in the long run, make the passenger and freight rates which they would have to charge the public, and would be legally entitled to charge it, higher than would otherwise be necessary. Digitized by Microsoft® Digitized by Microsoft® Appendix. Statistical Proof of the Relation to Train Density of Accidents and Casualties. The statistical proof that train density has an important effect on :he number of accidents and casualties can be briefly indicated. The comparison with train density of the number of accidents and jf casualties, as specified in the text, has been made with full recog- lition of the fact that other factors, as well as train density, affect :he number of accidents and casualties. But, since it is not possible, :o eliminate these other factors entirely in order to observe the rela- ion between train density alone and accidents and casualties, it has )een necessary to compare the general trend of accidents and casual- ies from all causes and influences with the general trend of the sin- gle influence, train density, and make allowance in the conclusions tor the probable effect of the other causes. Most important of these other influences are the human factor md the numerous safety appliances or other improvements in condi- ions affecting safety, such as second and other additional tracks. \s to the human factor, it is reasonable to assume that the "safety- irst" movement has made employees somewhat more careful, and lence that accidents and casualties have become somewhat fewer :han they would have been otherwise. As to safety appliances, these have doubtless more and more pre- sented accidents as the use of such appliances has increased and :heir improvement has progressed. Moreover, the steady increase n second, third, and other tracks tends to reduce the chances of iccident from congestion of the line with trains. The effect of such cumulative changes in these other factors vould be to reduce the total number of accidents and cas- ualties below what it otherwise would have been by reason of the nfluence of train density. The actual statistics, therefore, would lot reflect the effect of train density as closely as if there had been 10 such changes in these other conditions affecting safety. Hence :he dependence of accidents and casualties on train density may be issumed to be even greater than that indicated by the statistics in ■he form in which they are available. Digitized by Microsoft® 62 The most definite and precise method of detecting whatever rela- tion, — or, more accurately, whatever correlation, — there may be be- tween two series of numbers that appear to vary in agreement with each other, is a mathematical computation which gives a result that measures the degree or extent to which the two series of numbers are correlated. This result is known as the coefficient of correla- tion. If the two series of numbers are completely correlated, the coefficient of correlation will be i. If they are not completely de- pendent, one upon the other, the coefficient of correlation will be a decimal less than i, its amount being smaller the less the correlation between the two series of numbers. If the two series of numbers vary without any relation whatever to each other, the coefficient of correlation will be o. Thus the coefficient of correlation is virtually a measure, in percentage, of the dependence of one series of num- - bers on the other. In the present case, one of the two series of numbers to be tested is that of the train density, year by year, considered as a condition affecting the number of certain classes of accidents and casualties. The other series of numbers is that of the accidents and casualties, year by year, that are thought to depend in part upon the degree of train density. Now the numbers in this latter series are, as already noted, affected by influences other than, but concurrent with, train density. These other influences may, and doubtless usually do, change their upward and downward tendencies quite independently of the changes in train density. Therefore, the statistics that are available of accidents and casualties, being subject to these other con- current influences also, show a dependence on train density only to the extent that it is one of the predominant influences. That is to say, the coefficient of correlation for train density and any class of accidents or casualties tends to be lower than it would be if all these other influences should remain unchanged from year to year, and the effect of train density alone on the class of accidents or casualties were thus separately tested. Therefore, whatever the coefficient of correlation between the available statistics on accidents or casualties and those on train den- sity, it may be assumed to be less than the true measure of the inde- pendent efl^ect of changes in train density. It is commonly held among statisticians that a coefficient of cor- relation of about .50 indicates a correlation between the statistics Digitized by Microsoft® other reasoning, that one series of statistics depends upon the other. And the higher the coefficient is above .50 the greater the evidential value of the statistics. On the other hand, a coefficient appreciably less than .50, and especially as lov^^ as or lower than .35, is not considered as indicating a sufficient degree of correlation between the two series or groups of statistics to support an argument that either series depends on the other. And if the coefficient is as low as .25, and especially if as low as .10, the statistics cannot be considered as affording any support whatever to a postulated relationship between them. How- ever, this does not mean that a low coefficient of correlation neces- sarily disproves the proposition. That would depend on how nearly each series of statistics represented one factor only in the problem. If either series represented more than one factor, it might well be that the relation could be shown, from other than statistical argu- ments, to be true for that one factor, and yet the relative weight of that factor be so small that the relation would not be reflected in the statistics. Turning now to the available statistics on train density, accidents, and casualties, given in the text, the coefficient of correlation found in each case is shown in the following table. To save needless de- tail of demonstration, only the total casualties, including fatalities and injuries, are represented in this table, although the number of killed and of injured is separately shown in the text. Coefficients of Correlation with Train Density of the Specified Classes of Accidents and Casualties. „ f Coefficient »Jui° Page. Item correlated with train density. of '3'''^- Correlation. VIII. .. Collisions V^III. . . Derailments [X. . . Total casualties to trainmen on the road occurring in collisions X. . . Total casualties to trainmen on the road occurring in derailments SCI. . . Total casualties to trainmen on the road occurring in all train accidents ■ ■ Sir. . . Total casualties to passengers occurring in collis- ions .- • SCIII. • • Total casualties to passengers occurring in derail- ments •.•••■. : • Y^YV, . . Total casualties to passengers occurring in all train accidents Digitized by Microsoft® 61 62 49 66 6r 90 67 79 Digitized by Microsoft® BULLETINS OF THE BUREAU OF RAILWAY ECONOMICS MISCELLANEOUS PUBLICATIONS (Continued from page 2 of cover.) ConsecutlTe Miscellaneous Nnmljer. Series Number. ^ 4. 1. A Comparative Statement of Physical Valuation and Capitalization. 21. 7. The Cost of Transportation on the' Erie Canal and by Rail. 34 12. ■ A Comparative Study of Railway Wages and the Coat qf Living in the United States, the United Kingdom, and the Principal Countries of Contiriental Europe. 39. 13. Comparison of Capital Values-rrAgrlculture, Manu- factures, and the Railways. 45. 14. Railways and, Agriculture, 1900-1910. 62. 16. List of Publications Pertainiag to Groveminent Own- ership. 73. 18. The Arguments For and Against Train-Crew Legis- lation. . 81. 20./ Statisttcs of Railways, 1904-1914, United States. 83. 21. Comparative Railway Statistics, United States and Poreigil Countries, 1912. 88. 22. Summary of Railway Returns for the Fiscal Tear Ending June 30, 1915. 92. 23. Arguments For and Against Limitation of the Length of Freight Tratas. The following Miscellaneous Publications are Out of Print and nO copies can be supplied: Consecutive Number. Miscellaneous Series Number. Consecutive Number. Miscellaneous Series Number. 6. 2. 24. 9. 11. 3. 28. 10. 'l5. 4. 31. 11. 17. 5. 63. 15. 18. 6. 66. 17. 28. 8. 75. 19. Digitized by Microsoft® Digitized by Microsoft® Digitized by Microsoft®