Social Rates of Return on American Railroads in the Nineteenth Century

Abstract

W T rHAT was the social rate of return on investment in American railroads during the nineteenth century? An impressive list of economists have attempted to answer that question: Paul David, Albert Fishlow, Robert Fogel, Lloyd Mercer, and Marc Nerlove.2 All of their results are unfortunately incorrect. This article attempts to explain the nature of their errors and the reasons why social rates of return may not be a useful tool of analysis for the economic historian. The question which measures of social rates of return would attempt to answer is eminently reasonable: Would the American economy of the nineteenth century have been better off with a little more, or a little less, investment in the railroad sector? The answer involves three distinct steps: (i) the average social rate of return for railroad investment must be estimated, (2) some technique must be devised for transforming that average rate into a marginal rate of return on the last dollars spent on railroad investment, and finally (3) a norm must be developed against which that marginal rate can be compared. Essentially the procedure is the conventional one of comparing marginal relationships to determine whether investment returns were higher or lower in the railroad sector than elsewhere. One popular variation of this procedure is somewhat less reasonable. The second step noted above can be omitted provided that the question is restructured to ask not about marginal relationships but rather: Would the economy have been better off without the entire railroad system ? Such a large hypothetical reallocation of investment dollars might well have seriously altered the returns on investment alternatives and thereby the willingness of the community to save and invest. The task of determining what those dollars could have earned if diverted to uses other than the railroad therefore becomes extremely complex. Despite these difficulties, most of the writers to be considered have not pursued their calculations beyond estimating an average social rate of return for the railroad sector. For this reason, Part i will be confined to elaborating the formulae required for calculating average social rates of return for a single year and for a number of years. These formulae will then be used in Part ii to examine the