¹ All the data used in this article are in real terms using 1860 as the base year. That is, the data have been inflated to take account of the postwar deflation. The index used was taken from Ethel D. Hoover “Retail Prices after 1850,” in National Bureau of Economic Research, Trends in the American Economy in the Nineteenth Century (Princeton, 1960), 142, 162.Google Scholar A definition of total assets in nineteenth century railroad accounting would be worth an article in itself. In broad terms total assets included all tangible assets such as property, equipment, materials on hand, and a host of other items. It also included construction accounts, securities of other companies owned, accounts due, cash and bills on hand, and special items. Not surprisingly, the total assets of weaker roads often included dubious and spurious items.

² These figures are calculated from Poor, Henry V., Manual of the Railroads of the United States for 1894 (New York, 1894), xvGoogle Scholar and passim.

³ For those who might not be familiar with some terms used in this article, the following brief note might be of some use.

In linear regression analysis, which is in the form of Y = a + bX, b (regression coefficient) is the slope which indicates the rate of change in Y per unit of X. The intercept (a) is a magnitude of Y when X is zero. In both a and b, they are not actual or “observed” values but values calculated by the linear regression equation fitted by the standard least squares method. Coefficient of determination (r²) indicates the “goodness of fit” and its value ranges from — 1 to 1. The term “t-test” refers to the test of the “reliability” of the regression coefficient (b) and is obtained by dividing the value of b by the standard error of b.

In terms of Table 1, the equations were calculated in the form of A = a + bt, where A is total assets and t is time in year unit. The term “trend lines” in the heading of the table is a standard expression for those regression equations in which X is t. The expressions used for intercept and slope (initial amount of total assets and annual rate of increase of total assets) are, of course, calculated and not actual values. The Notes to Table 1 state that both the results of t-tests and r ^2's are significant at the .01 level. This simply means that the “confidence coefficient” was 99 per cent, i.e. there is only a one out of hundred chance of obtaining such high values for t-test and r²had the original data been entirely random. Later in this article, the. 01 level is replaced by the .05 level. This means that those values of t-test and r² have a lower value of “confidence coefficient,” i.e. 95 per cent. All through the article, the standard expression “at given n (sample size)” is eliminated.

⁴ The negative intercepts seen for seven roads indicate that the total assets in those roads rose sharply from the calculated value of the initial year. As the original data confirm, this is due to an accelerated growth in later years which has an effect of tilting the slope upward and tending to understate the value of earlier years. But given the observed high r², the fit of these lines was generally good.

⁵ The three roads in which rate of increase for equity exceeded that for debt deserve special explanation. For two of the roads, 13 and 17, the data are misleading. Both actually had a considerably higher ratio of increase for debt than for equity, but financial reorganizations caused a massive rescaling of the ratios in favor of equity. In the case of road 13, two such reorganizations took place during the period. Road 1 is a somewhat unique case paralleled by only a few southern roads. Its debt was so small that it could generate enough internal capital to supply its growth needs during the territorial era. In 1881, on the verge of the great expansion era, it was leased to another system on highly favorable terms.

⁶ Note also that all t-test results are significant at the .05 level. It should be added here, for those readers who might object to our particular regression coefficients, that they are used because of their intuitive appeal. Theoretically, it would be desirable to evaluate each b after allowing for a specified confidence interval; i.e., after evaluating each b by, say, a two-tail test at t.025 level. We use the ratios we do, however, because our sole purpose is to depict general trends and not to test econometric propositions. It goes almost without saying that a critical test for a trend line is the t-test. Our discussions in the text are done in terms of r ^2's again because of their intutitive appeal at no cost to the substance of our discussions.

⁷ Normally, operating expenses referred to those current expenses necessary to keep the road a going concern, such as supplies, salaries, materials etc. But accounting procedures varied widely among individual roads and at different points in the time period. For example, many roads sometimes included capital account expenditures in the current expense account. Other lines began including tax payments in operating expenses later in the period. Numerous exceptions to the general statement can be found, with each requiring its own explanation.

⁸ Here, as with operating expenses, the term “fixed charges” has a general explanation not always applicable to individual cases. Basically, it constituted all fixed obligations against the road, such as interest and tax payments. These might also include such things as lease payments, payments on endorsed bonds of other companies, and any regular annual obligation. The company's profit was thus figured by deducting all fixed charges from net earnings.

⁹ The literature on regulation and the rate problem is voluminous. For the southern roads in particular see Joubert, William H., Southern Freight Rates in Transition (Gainesville, 1949)Google Scholar and Potter, David M., “Historical Development of Eastern-Southern Freight Rate Relationships,” Law and Contemporary Problems, XII (Summer, 1947), 416–48.CrossRefGoogle Scholar

¹⁰ A survey of eighteen leased roads connected with our sample roads revealed only three cases in which the leased road earned enough to meet its guaranteed dividend after paying all expenses and fixed charges. Even the Georgia Railroad met its guaranteed 10 per cent dividend only once in twelve years.

¹¹ Study of the data makes it clear that the Charlotte, Columbia & Augusta dividend was not the product of unexpectedly high earnings or reduced expenses. Other sample roads where doubtful dividends were declared are the Central of Georgia, the East Tennessee, Virginia & Georgia, and the Louisville & Nashville.

¹² For sketchy and somewhat inaccurate accounts of the Central of Georgia's plight, see Daggett, Stuart, Railroad Reorganization (Cambridge, 1908), 169–88CrossRefGoogle Scholar, and Campbell, E. G., The Reorganization of the American Railroad System (New York, 1938), 96–106.Google Scholar

¹³ Recall that the data are all in real terms.

¹⁴ This concept will be developed in more detail in another article to be published in a forthcoming issues of the Business History Review.

¹⁵ This broad pattern of growth is traced in Stover, John F., The Railroads of the South, 1865–1900 (Chapel Hill, 1955)Google Scholar, which emphasizes the thesis that southerners were losing control of their railroads to “northern men and money.” While true to some extent, the couching of the struggle for control in such vague sectional terms seems to miss the point of what is happening and makes use of Stover's thesis limited and crude.

¹⁶ For explication of these terms, see Johnson, Arthur M. and Supple, Barry E., Boston Capitalists and Western Railroads (Cambridge, Mass., 1967), 8–10, 181–91, 333–46.CrossRefGoogle Scholar

¹⁷ An excellent insight into the nature and diversity of these linkage investments, including the Lenoir City Project, can be gained by studying the Charles M. McGhee Papers (Lawson-McGhee Library, Knoxville, Ohio). McGhee was involved in numerous southern roads, especially the East Tennessee, Virginia & Georgia and the Memphis & Charleston, over the entire period.

¹⁸ There are numerous examples of this practice. The most notorious case, perhaps, was the Georgia Company swindle. See Campbell, Reorganization, 97–104, and Daggett, Railroad Reorganization, 162–66, 177–78.

¹⁹ The variables we have considered include those discussed in this article and several others, including net worth, tonnage, mileage, operating expenses, fixed charges, cost per ton-mile, earnings per ton-mile, and operating ratios. Though we intend to evaluate the data for each railroad, a few results calculated for the aggregate data (twenty roads), as shown below, suggest our line of thought.

M is mileage, T is tonnage, and Ct is cost per ton-mile. The numbers in parentheses are standard errors.

Where all notations (K, N, F, n) are the same as in the text; where δK is K_t — K_t−1; and where t is the weighted mean rate of dividend.

As seen in equation (1), T is much less meaningful than Ct (with the expected negative sign) in explaining N, while M is the most important of the three variables. Though there are various difficulties involved in making Ct comparable within the sample roads, the differences in explanatory value for the Ct of each road is an important item for further investigation. Equation (2) yields the expected result. As seen in our Table 4, dividends are unimportant in explaining δK and N/K + F yields a “wrong” sign. For many roads, this form of equation yields statistically insignificant results. But this is another item that needs to be evaluated carefully vis-à-vis the entrepreneurial functions of various lines.

The authors are happy to correspond with anyone interested in a more detailed description of our research plans and in the results of our calculations.