Productivity growth and some of its determinants in the deregulated U.S. railroad industry.

• Author: Bitzan, John D.

This study analyzes the effects of an important postderegulation innovation on rail freight productivity: the elimination of cabooses and related crew members. It also analyzes the overall growth of productivity in rail freight between 1983 and 1997 (using a translog rail cost function estimated over a sample of Class I railroads between 1983 and 1997). The results indicate that elimination of cabooses and associated crew members from freight trains reduced costs by 5-8% on the typical Class I railroad in 1997, equivalent to an annual cost saving of $2 billion to $3.3 billion for all Class I railroads. Moreover, if Class I railroads had no other technological advances since 1983, their 1997 costs (with 1997 factor prices) would have been 36-43% higher than they in fact were. Finally, the results show that overall productivity growth in rail freight did not decelerate between 1983 and 1997; if anything, it accelerated slightly.

1. Introduction

   Many researchers agree that productivity growth among U.S. freight railroads accelerated after the Staggers Rail Act of 1980. (1) Yet beyond such a basic point, important questions remain. How long did the acceleration of productivity growth last? Had it dissipated by the 1990s? Where did the productivity growth come from? Can any of it be tied to specific innovations directly or indirectly related to deregulation, or could the growth be a synchronistic coincidence, not related to regulatory reform?

   To the end of better understanding such issues, the following essay does two things. First, it presents new evidence on postderegulation rail productivity change, extending to the late 1990s. Second, it isolates the contribution to productivity and lower costs of a crucially important technological innovation implemented in the years shortly after deregulation, (2) namely, the ability to run trains with crews of two members, rather than four or five, and the elimination of the caboose, traditionally needed both for safety at the end of a freight train and as quarters for the two or three crew members now technologically obsolete.

   There is a broader motivation for this work as well. Productivity growth is often measured as a residual in a growth-accounting exercise or a time trend in a cost function. Yet over the years, economists have often referred to that residual or time trend as a "measure of our ignorance"; something is causing productivity to grow, but it is not understood (for an early such critique, see Jorgenson and Griliches 1967).

   For this reason, there is a long, rich tradition in empirical economics of trying to isolate the effects of specific innovations on productivity rather than simply measuring it as a residual or time trend. Many studies, the seminal example of which is perhaps Griliches (1957), have tried to isolate the effects of specific innovations on productivity change. The present paper thus endeavors to contribute to this larger literature in economics as well as to a better understanding of railroad costs and productivity since the Staggers Act.

   Studies of the effects of crew size innovations on rail productivity are few and the results tentative enough to merit further study. Keaton (1991) simulates possible freight service improvements that might occur from elimination of cabooses and associated crew members but does not attempt to estimate the cost savings. The one previous study of rail cost savings from reduced operating crew expenses for freight trains over the 1980s and 1990s is the work of Martland (1999). Using aggregated accounting methods, he finds that changes in line-haul labor expenses from reductions in crew sizes, plus changes in the size of crew districts, yielded a net savings of $3.3 billion per year as of 1995 compared with 1983. However, Martland's estimates, useful as they are, are based on accounting guesses rather than microeconomic analysis and/or econometrics, and we believe that a clearer picture of both productivity change and changes in this specific technology can be achieved though estimation of a cost function derived from economic theory. (3)

   To analyze these issues of productivity change and the structure of production in rail freight, the model we use is a translog cost function. Before specifying the cost model, it is necessary to explain more clearly what the change in technology was and how it worked. In other words, why were larger crews and cabooses once thought necessary, and what changes in technology caused these larger crews and cabooses to become obsolete? These issues in technology and institutions are discussed in section 2. The model is summarized in section 3, and data and estimation issues are considered in section 4. Section 5 presents results of estimation, and section 6 presents analysis of the implications of the results for rail productivity change. Section 7 presents conclusions and implications for further research and for policy toward the railroad industry.

2. Important Changes in Rail Technology

   As mentioned previously, in order to understand the causes of recent productivity growth in the post-Staggers rail industry and to inform the specification of our econometric model, it is important to review briefly what the most important changes in the underlying technology were over this period.

   Railroads during the steam era (which ended in the 1950s) needed crews of four or five (or even more) people to operate a freight train, and until well into the 1980s, unions and labor laws forced railroads to continue these staffing policies, though they were not technologically required. Various technological changes eliminated the need both for cabooses and for more than two crew members to operate the freight train. First, the need for a fireman disappeared with the end of steam locomotives on freight trains. (4) This both reduced the overall crew size and freed space in the cab of the locomotive for another crew member.

   Furthermore, the safety and control functions required of cabooses and the crew members that occupied them were no longer required by the 1970s because they were automated via electronic controls. (5)

   Martland (1999) has pointed out the upshot of this: For most freight trains in most places, there are only two crew members (an engineer and a conductor), and both ride in the cab of the locomotive. In addition to these savings, work rules for line-haul freight trains were also changed: Traditionally, crews worked only the lengths of divisions, which were 100 miles or less per day for freight trains. In recent years and with fast freight trains, that often meant crews got paid a full day's pay for two hours of work. Work rules were also changed in the 1980s to try to compensate crew members by the hour worked rather than based on a 100-mile day, and the railroads were allowed to save some of the money gained. We now turn to the tasks of analyzing the effects of these changes on railroad costs and productivity.

3. The Model

   The model developed here is a translog cost function, with all the desirable properties of that form, including flexibility, ease of interpretation, and the ease of inclusion of both multiple outputs and technological variables. However, the present model differs from any previous uses of such functions to analyze rail productivity change in several ways.

   First, the present model allows measurement of the contribution of a specific (and very important) technological innovation: the elimination of cabooses and the freight train crew members that go with them. This innovation was allowed for by new technology that eliminated the need both for cabooses tending the safety of the train at the end of the train and for the crew members that tended to that safety. Finally, concurrent with that change, other changes occurred, including elimination of previous confinements of crew members to rigid operating divisions. All these changes are summarized by one variable, the ratio of caboose miles to freight train miles, or, approximately, the fraction of freight train miles operated with cabooses. The way in which this variable is incorporated into our model is discussed in the following.

   Second, unlike some previous studies of rail productivity change over the past two decades, the present paper incorporates multiple outputs, including through train miles, way (local) train miles, and unit train miles. Certainly, useful studies have been done with only one output variable, but one can theoretically justify use of multiple outputs because it allows for more precise control for effects of economies of scale and scope relative to the effects of technological progress. (6) Although this paper does not focus on economies of scale and scope, (7) we nevertheless believe that controlling for them as precisely as possible will allow a more accurate measure of productivity change than would otherwise be possible. The reason for this is that the present method allows for more possibilities of costs varying with each type of output than a single-output model. (8)

   Third, the present analysis includes data into the late 1990s. As previously stated, there is an ambiguity as to whether the acceleration of productivity growth that occurred after the Staggers Act slowed as of the late 1980s or continued into the 1990s. On one side of this issue is Wilson (1997), who finds that rail productivity growth slowed by the late 1980s. On the other hand, analysis by the Surface Transportation Board (1997), combined with other work cited by Oum, Waters, and Yu (1999), indicates that productivity growth accelerated between 1991 and 1995 compared with previous recent years. (9)

   Additionally, the two contributions of Bereskin (1996, 2001) can also be taken together to imply an acceleration of rail productivity after 1993, though he does not explicitly address this issue in his papers. Specifically, Bereskin (1996), analyzing rail productivity change using a translog cost function, finds that, from...