The Relative Efficiency of Public and Private Firms in a Competitive Environment: The Case of Canadian Railroads Author(s): Douglas W. Caves and Laurits R. Christensen Source: Journal of Political Economy, Vol. 88, No. 5 (Oct., 1980), pp. 958-976 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/1833143 . Accessed: 23/05/2014 02:47 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of Political Economy. http://www.jstor.org This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions The Relative Efficiencyof Publicand Private Firmsin a Competitive Environment: The Case of CanadianRailroads Douglas W. Caves and LauritsR. Christensen Universityof Wisconsin-Madison The efficiency of public and private firms is usually compared in industries which have heavy regulation and limited competition. In this paper we present a case study in which the effects of property rights can be isolated from the effects of regulation on noncompetitive markets. We compare the postwar productivity performance of the Canadian National and Canadian Pacific Railroads. Contrary to the predictions of the property rights literature, we find no evidence of inferior performance by the government-owned railroad. We conclude that any tendency toward inefficiency resulting from public ownership has been overcome by the benefits of competition. I. Introduction A major theme in the recent literature on the economics of property rights is the notion that public ownership is inherently less efficient than private ownership. Alchian is a leading proponent of this view, and his 1965 paper is a frequently cited discussion of the issues The authors acknowledge research support from the U.S. National Science Foundation and the Canadian Transport Commission (CTC). Members of the CTC staff have provided considerable assistance throughout the study. The authors are grateful for the cooperation and encouragement received from K. W. Studnicki-Gizbert, John Heads, and George Hariton, and the research assistance of John Gibberd and Vivian Wey, all of the CTC. However, views expressed in this paper do not necessarily coincide with the views of the CTC or the National Science Foundation. The authors also wish to thank Joseph Swanson, Sam Peltzman, Frank Gollop, and an anonymous referee for helpful comments and Philip Schoech, Michael Tretheway, and Beth Van Zummeren for research assistance. 1 See Furubotn and Pejovich (1972) for a survey of the property rights literature. Journal of Political Economy, 1980, vol. 88, no. 5] ? 1980 by The University of Chicago. 0022-3808/8018805-0001$01.50 958 This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions THE CASE OF CANADIAN RAILROADS 959 involved. The essential argument is based on the fact that public ownership is diffused among all members of society, and no member has the right to sell his share. Given these aspects of public ownership, there is little economic incentive for any owner to monitor the behavior of the firm's management. In contrast, it is argued, the ownership of private firms is concentrated among fewer individuals, each having the right to sell his shares; and thus the owners have incentives to scrutinize management to ensure efficiency in the production of goods or services. Parallel to the theory which predicts efficiency differentials on the basis of type of ownership is a theory which predicts efficiency differentials on the basis of market structure. This theory, which is most widely discussed in the industrial organization literature, predicts superior efficiency in markets characterized by effective competition among firms. The essential element of the theory is that in competitive markets productive efficiency is a prerequisite for survival-at least for privately owned firms. The two classifications of ownership, public and private, and the two classifications of market structure, competitive and noncompetitive, provide four categories into which firms might fall. Three of the four categories have been extensively studied. The fourth category, government-owned firms operating in a competitive environment, has received much less attention. The principal reason is undoubtedly that few enterprises fall into this category. Nonetheless, the study of such firms has the potential to yield important policy insights. An important question of public policy is how the four categories of firms rank in terms of relative productive efficiency. There is little doubt that expert opinion would rank competitive private firms as the most efficient and noncompetitive public firms as the least efficient.2 Ranking the remaining two categories would be more difficult. There is no clear consensus as to whether public firms facing competition behave more like their private counterparts or more like their noncompetitive government counterparts. Both views can be found in the literature. Peltzman (1971, p. 147) suggests: "The differences between government monopolies and government firms with private competitors might be greater than the differences between government firms and private firms in competition with one another." A similar opinion is expressed by Spann (1977, p. 75): "One would expect competition to exert some market pressure on government enterprises to hold down costs . . . and to eliminate some of the 2 This discussion abstracts from the complicating factor of scale economies. It is possible that for a particular industry efficiencies resulting from a competitive market structure would be more than offset by inefficiencies resulting from the sacrifice of scale economies. This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions 960 JOURNAL OF POLITICAL ECONOMY opportunities for discretionary behavior on the part of bureaucracies." A contrary view is provided by De Alessi (1974, p. 7): "The managers of political firms . . . are less constrained by market considerations . . . and find it easier to obtain subsidy and to mask bad management under the guise of fulfilling other 'social' goals. Government firms . . . can survive for long periods . . . [with] grossly inefficient management."9 There have been numerous comparative studies of the behavior of public and private firms.3 The studies are generally of service industries which are heavily regulated. In some cases there exists a degree of direct or indirect competition between private and public firms. But we are not aware of any study of public and private firms coexisting in a competitive environment. Thus the findings of previous studies reflect a mixture of effects of property rights, regulation, and limited competition. Our objective in this paper is to perform a case study in which the effects of differing property rights can be isolated from the effects of regulation and noncompetitive markets. To achieve this objective we have chosen to study the Canadian National (CN) and Canadian Pacific (CP) Railroads. They are very large railroads of roughly equal size. The CP is a privately owned railroad which spanned the North American continent in 1885. The CN is a crown corporation, wholly owned by the Canadian government. The CN became a nationwide competitor of the CP in 1923, when the government took over and consolidated the operations of several failing railroads. The CN and CP have received the bulk of railway revenues in Canada for over half a century; currently they account for approximately 90 percent of gross railway revenues. Since World War II the CN and CP have faced stiff competition from other modes of transportation. During the postwar period restrictions on the railroads' ability to compete for traffic have been removed. Thus the Canadian railroad industry provides an attractive case in which to study the efficiency effects of property rights in a competitive environment. The best single measure of productive efficiency is total factor productivity (TFP)-real output per unit of real resources expended.4 In this paper we estimate both the rates of growth of TFP and the relative levels of TFP for the CN and CP during the period 1956-75. We find that both railroads have achieved high rates of growth in TFP. Contrary to what is predicted in the property rights literature, we find no evidence of inferior efficiency performance by 3Examples, in addition to the papers already cited, include Davies (1971, 1977), Clarkson (1972), Ahlbrandt (1973), Frech (1976), Lindsay (1976), Pashigian (1976), and Savas (1977). 4 For a general discussion of productivity measurement, see Fabricant (1974). This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions THE CASE OF CANADIAN RAILROADS 961 the government-owned railroad. In fact, our evidence indicates that the CN has achieved larger gains in productivity than the CP since 1956. In the late 1950s and early 1960s the CN had a level of productivity approximately 90 percent as high as the CP, but this gap has been closed. We conclude that in the case of Canadian railroads the beneficial effects of competition have been sufficient to overcome any tendency toward inefficiency resulting from public ownership. II. The Canadian Railroad Industry5 The CP became a transcontinental railroad in 1885, largely as the result of massive aid from the Canadian government. The major impetus was a desire to tie British Columbia to the other provinces and to facilitate development of all of Canada's western provinces. During the last half of the nineteenth century numerous other railroads were established, also with generous amounts of government construction, financing, subsidies, and land grants. By World War I it was clear that the Canadian rail network was severely overexpanded. After the war the government took over three large privately owned rail systems, which were near bankruptcy, and amalgamated them with the existing Canadian Government Railways. Thus the Canadian National Railways had an inauspicious beginning as a government-owned firm with massive amounts of debt and extensive overlapping facilities. Prior to World War II Canada's two dominant railroads enjoyed substantial market power and, accordingly, were heavily regulated by the Canadian government. By the 1950s, however, their market power had declined considerably. The railroads found themselves facing stiff competition for freight traffic from highway and waterway carriers, and for passenger traffic from private passenger cars, buses, and airplanes. Under the burden of pervasive economic regulation the railroad industry was unable to respond effectively to the rising competition from the other transport modes, which were essentially unregulated. During the 1950s the railroads gained limited relief from regulation through provisions allowing them to negotiate "agreed" charges with individual shippers. The movement toward deregulation gained substantial momentum in the early 1960s with the report of the MacPherson Commission, which urged that prices be determined by the competitive market rather than by government regulation. By 1967 most rate regulations had been swept away. The chief impedi- 5 For further discussion of the Canadian railroad industry, see Purdy (1972) and Heaver and Nelsen (1977). This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions 962 JOURNAL OF POLITICAL ECONOMY ment to fully competitive rates was the retention of statutory rates for hauling grain and flour to be exported. Other than statutory grain rates, the railroads are free to set rates between the limits of variable cost and two and one-half times variable cost.6 Although Canadian railroads have great leeway in the setting of rates, they do not have complete freedom to adjust their physical plant or the services they offer. There are restrictions on the abandonment of track and on the discontinuation of passenger service. The restrictions on discontinuation of passenger service have not been severe, and it appears that the importance of this mode of transportation will continue its steady decline in Canada. The abandonment of uneconomic trackage is a more difficult issue. Both the CN and CP have large amounts of lightly used track. The CN appears to have a more severe problem in that it inherited excess trackage from its predecessors. The magnitude of the difference in the situation of the CN and CP is suggested by the fact that tonmiles per mile of track for the CP exceeds the CN figure by approximately 10 percent.7 Up to 1967 there was a gradual abandonment of branch lines by both railroads. However, by 1967 it became clear that many branch lines in the prairies were uneconomical because of statutory grain rates, which were far below variable costs. As part of the 1967 National Transportation Act which ended rate regulation, thousands of miles of prairie branch lines were protected against abandonment. In the three western provinces of Alberta, Manitoba, and Saskatchewan 9,591 miles of CN track and 8,512 of CP track were protected by the 1967 legislation.8 The related problems of statutory grain rates and restrictions on abandonment of prairie branch lines cannot be dismissed lightly. However, the bulk of railroad operations are free from regulatory interference. Furthermore, for present purposes, it is important to note that the grain-hauling problem affects both railroads. If there is any differential impact, it appears likely that the CN is more heavily burdened due to the greater size of its rail network in the prairies, much of which has never been profitable. Other than the possible distorting effects of the enforced uneconomical haulage of grain, we are not aware of any factors which would undermine the validity of a comparison of economic efficiency based on measures of TFP. Both railroads operate coast to coast, serving all major industrial areas. The railroads appear to face similar levels of competition from other transport modes. It appears that neither railroad has an inherent oSee Heaver and Nelsen (1977) for further discussion. 7E.g., in 1970 CP net ton-miles per track mile were 1.74 million vs. 1.59 million for the CN. 8 Purdy 1972, p. 264. This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions THE CASE OF CANADIAN RAILROADS 963 advantage in terms of low-cost traffic, but we investigate this possibility below. Finally, we address the possibility that one of the railroads has a cost advantage due to differential scale economies. The CN is somewhat larger in size than the CP; thus the existence of scale economies would provide the CN with the possibility of lower-cost operation. It seems clear, however, that both railroads are so large that any possible scale economies have been fully exploited by each of them.9 Therefore, our comparison ought not be affected by the difference in size between the CN and CP. III. Methodology for Measurement of Total Factor Productivity Christensen and Jorgenson (1970) proposed the following index of total factor productivity: In (TFPk/TFPI)= 2( RI Ril )ln (YiklY(l) (1) _ St )In (XikIXil), where k and 1 are adjacent time periods, the Y's are output indexes, the X's are input indexes, the R's are output revenue shares, the S's are input cost shares, and the i subscripts denote the individual outputs or inputs. Diewert (1976) has shown that (1) is the exact index procedure which corresponds to a homogeneous translog production or transformation function. Caves and Christensen (1980) have further shown that no restrictions of separability or neutral technological change are implicit in (1). Caves, Christensen, and Swanson (CCS) (1980) have noted that it is not justifiable to use (1) to measure TFP in the railroad industry. The problem is that the revenue shares in (1) are used as estimates of the elasticities of total cost with respect to the individual outputs. This procedure is satisfactory only if the price of each output is equal to its marginal cost of production. It is widely accepted that prices for railroad services do not reflect marginal costs of production. Thus CCS proposed that railroad TFP measurements make use of estimated output cost elasticities in place of revenue shares: 9Griliches (1972) has argued that large U.S. railroads do not have available any unexploited scale economies. Both the CN and the CP are larger than all but a few U.S. railroads. This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions 964 JOURNAL OF POLITICAL ECONOMY In(TFPk/TFP,)= X/2(I[ D ) + In(Yiklyi) (2) _-E ( S~+ik2 )In(Xlk/Xil)We follow CCS in using (2) to compute TFP for the CN and CP. As pointed out by Jorgenson and Nishimizu (1978); formulas such as (1) and (2) can be used to make both time-series and cross-sectional comparisons of TFP. In the case of cross-sectional comparisons, the indexes k and I are interpreted as different firms rather than different time periods. We follow Jorgenson and Nishimizu in choosing a base year (1963) to carry out a comparison of the levels of CN and CP productivity. The growth rates of CN and CP productivity are used to extend the level comparison to earlier and later years. IV. Productivity Estimates Our primary productivity estimates distinguish two indexes of railroad output and five indexes of railroad inputs. The two output indexes are freight ton-miles and passenger-miles. The five input indexes are labor, structures (including right-of-way), equipment (including rolling stock), fuel, and materials. A detailed description of the sources and methods used to develop our data base is contained in Caves and Christensen (1978). We have relied heavily on the annual reports of the CN and CP filed with the Canadian Transport Commission (CTC). The CTC provided us with access to these reports and made available supplementary data which were essential for completion of the study. The annual reports follow the Uniform System of Accounts which was instituted in 1956. Accounting procedures and reporting practices before 1956 were significantly different from those instituted in 1956. Thus, our study is limited to the period from 1956 to 1975, the most recent year for which annual reports were available when our research was being carried out. The major task in the data development involved the estimation of capital input for structures and equipment. The procedures which we have used are very similar to those suggested by Christensen and Jorgenson (1969). Cost elasticities with respect to output levels are not directly observable. They must be estimated in order to implement our approach to productivity measurement. The most attractive approach to obtaining cost elasticities is the estimation of a multiproduct cost function using cross-section data. There are not enough Canadian railroads to provide data for such estimation; however, CCS (1980) have used cross-section data from the U.S. railroad industry to estimate the This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions THE CASE OF CANADIAN RAILROADS 965 structure of rail costs. We have taken the approach of using their estimated equations to infer Canadian cost elasticities. The CCS estimates were developed from cross-section data for 1955, 1963, and 1974.10 We use their estimated coefficients along with data on CN and CP output levels and input prices to estimate the cost elasticities for ton-miles and passenger-miles for the CN and CP in 1956, 1963, and 1974.11 The estimated cost elasticities and their standard errors are presented in table 1.12 The cost elasticities are then interpolated between the cross-section years in order to provide annual weights for the productivity indexes. We now have all the series required to obtain productivity estimates from (2). We summarize this information in table 1 by presenting figures for 1956, 1963, and 1974.13 The first three rows of table 1 contain the inputs and outputs of the CN relative to the CP. The next six rows contain the average annual percentage rates of growth of CN and CP inputs and outputs between the cross-section years and over the full sample period. The final six rows contain the cost shares and estimated cost elasticities for the CN and CP for the cross-section years. Between 1956 and 1963 there was no growth in ton-miles for either the CN or CP, and passenger-miles declined substantially. During this period both railroads were able to make large cuts in their fuel usage. This reflects the rapid replacement of steam locomotives by more fuel-efficient diesel locomotives. Diesel locomotives also required substantially less labor for operation and maintenance, which partially accounts for the decline in total labor input. The major difference 10 The numbers of firms included in the samples for these years were 58, 56, and 40, respectively. They employed the generalized translog multiproduct cost function, proposed by Caves, Christensen, and Tretheway (1980), to obtain the estimated cost elasticities. This cost function has the same form as the translog multiproduct cost function (Burgess 1974; Brown, Caves, and Christensen 1979) except for output levels, where the Box-Cox metric is substituted for the natural log metric. This generalization permits the inclusion of firms with zero output levels for some products. In railroad applications, it permits the inclusion in the sample of firms with no passenger output. 11Although CCS did not compile U.S. cross-section data for 1956, their analysis showed that the formula for the cost elasticity as a function of output and input prices was independent of time. Therefore to estimate 1956 cost elasticities for the CN and CP we have inserted 1956 output levels and input prices for the CN and CP into the estimated 1955 cost-elasticity equation. 12 The estimated multiproduct cost functions for all 3 years display significant scale economies at low output levels but not at high output levels. Constant returns to scale imply that the freight and passenger cost elasticities must sum to unity. In the region of freight and passenger output levels produced by the CN and the CP, the hypothesis of constant returns to scale cannot be rejected. 13 The figures for all years are presented in the data appendix of Caves and Christensen (1979), which is available on request from the Social Systems Research Institute, 1180 Observatory Drive, Madison, Wisconsin 53706. This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions 0S - -14 1- < n 00- f C o a O~~0000 t~~~~~~0~~~Ic0O0 C 0)~~ . CO~~~~~~~~~~~~~~~~~~~~~~1 t - . H~~~~~~~~~~~~~~~~~~CNX1-3c--- >~~~~~~~~~~~~~~C C C> O C14GNOcn t ooin 0 co 't 00 0 0 v 4: _ _ _ 3 or n o O t < W1 O O 0. 00) 00 00 0q 0)n c u OO) CO H t oo 0 0 E-4~~~~~~~~~~~~~- 0Q) H - 0 ~~~x QL00\1 04 000CZ0~ O 0 cA' C 0 1 '0n q ",\[ oo00 tf 00 c>0 4-OT -C ) C enst_ 0 i 0 .....r C _ Cll . .. .. . )I_ _ <' 'tse:( ( e S C ) O C >()C 14 Ul - -~~~~~~4Mr CO- " X~ 0 0 0~~~~~z ~- - 4 0- 0 CO U U U U U U U U U U V U 4 - 0t ) c" cn"00 ?- en 1~c0~0 Q~ OC O'00 i < ~~~~S -- - CO 040000~~~~~~~~~-- - 00 0 ~~~~~~~~.0 -00 ~ ~ ~ o 0l00-in~ 0000t ~ ~ x Ic I a. Z0XFt- n E 0 0 00~~~0 4O00z 966 This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions THE CASE OF CANADIAN RAILROADS 967 between the railroads during the 1956-63 period was that the CN stock of structures increased while the CP stock decreased. After 1963 both railroads experienced large increases in freight traffic. Both railroads also showed increases in inputs of all factors except labor, which continued to decline. The major difference between the railroads in this period was in the growth of passenger service-the CP experienced a large decrease in passenger-miles while the CN experienced a modest increase. However, neither of the figures reflects the large gyrations in passenger traffic after 1963. In 1964 both railroads successfully stimulated additional passenger traffic with discount pricing schemes. For the next few years the CN made an all-out attempt to make a success of their passenger service. The 1967 World's Fair (Expo) in Montreal, which generated large amounts of traffic, marked the end of this era. In 1967 CN passenger-miles were more than twice the 1963 level. Thereafter the CN apparently joined the CP in the belief that there was no possibility of obtaining adequate revenues from passenger service. By 1973 its passenger-miles had declined almost to the 1963 level.14 Since the ratio of CN to CP passenger output changed so dramatically from 1956 to 1974, the weights which we assign to passenger output in equation (2) take on great importance. In Section V we investigate the sensitivity of our results to the sampling error underlying our estimated cost elasticities. In the first three columns of table 2 we present our estimates of the growth of CN and CP productivity and their relative levels. Our estimates indicate that between 1956 and 1965 CN productivity was between 80 and 90 percent as high as that of the CP. Both CN and CP productivity was stagnant from 1956 to 1962 but then surged upward in 1963 and 1964. From 1964 to 1968 CN productivity increased rapidly while CP productivity again stagnated. In this 4-year period CN productivity rose, relative to CP productivity, from 0.85 to 1.12. From 1968 through 1975 the CP's productivity growth exceeded that of the CN. At the end of the period we find CN productivity slightly below that of the CP. V. Robustness of the Output Indexes and Cost Elasticities The output indexes and cost elasticities used in computing our productivity estimates are both open to criticism. The problem with the cost elasticities is that they are based on econometric estimates and 14 See Purdy (1972) for further discussion of the approaches taken by the CN and CP to the provision of passenger service. This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions A S U ;000 0000 00 0000 ~~~~~~~~Q) > ~H HO VZa clM cl.NN e Z H_ v n , 5 v A U oo oDooooI ooinoo oo 0 >~>0 q 0c*nG L > o X U CO~~~~000 00 GO00 GO00 - co00 00O t z~~~~ o o o o n C .0 t-0 0 CD G H ~~~~~~~~~~i n-i c c~ 968 This content downloaded from 147.251.185.161 on Fri, 23 May 2014 02:47:32 AM All use subject to JSTOR Terms and Conditions 0 -o - o 0 0 0 0 - - 0 -.Z -_ - _ - - - V. . l . . . . . . . . . .. . in n cn Gn C1 GmC: Ic 't q O: in in - "t cq X U) C n _ oe -00 &d& r-4 rGMj c - - I c cq 0 Qc qCM t- 0) 't O 0%roo 0 C) Cz o - 0 C) 0 0 O C -o ~~~0c~1:- Qtp oo0 ci0- 000 uic~C > e X U