¹ Moody's Public Utilities Manual (New York: Moody's Investors Service, 1967), p. v.Google Scholar

² Childs, John F., Long-Term Financing (Englewood Cliffs, N. J.: Prentice-Hall, 1961), pp. 10 and 11. Also see his discussion on pp. 166–168.Google Scholar

³ Sidney Homer, Solomon Brothers & Hutzler, letter dated December 12, 1967.

⁴ Harold, Gilbert, Bond Ratings as Investment Guide (New York: Ronald Press, 1938), pp. 9–13.Google Scholar

⁵ Cohen, Jerome B. and Zinbarg, Edward D., Investment Analysis and Portfolio Management (Homewood, Ill.: Irwin, 1967), p. 359Google Scholar. Their treatment of ratings on pages 356–362 is an excellent introduction to this subject.

⁶ Annual Report of the Advisory Investment Board of the Public Employees' Retirement System of Iowa, March 4, 1964.

⁷ Donaldson, Gordon, Corporate Debt Capacity (Boston: Graduate School of Business, Harvard University, 1961), pp. 98–101.Google Scholar

⁸ Gilbert Harold, op. cit.

⁹ Hickman, W. Braddock, Corporate Bonds: Quality and Investment Performance (New York: National Bureau of Economic Research, Inc., 1958) Occasional Paper 59.Google Scholar

¹⁰ Atkinson, Thomas A. and Simpson, Elizabeth T., Trends in Corporate Bond Quality (New York: National Bureau of Economic Research, 1967Google Scholar) Studies in Corporate Bond Financing. For a critical review of this volume, see the review by Sloan, Peter E., Journal of Finance, March 1968, pp. 218–220Google Scholar. Mr. Atkinson is presently the Director of Economic Research for the American Bankers Association.

¹¹ Fisher, Lawrence, “Determinants of Risk Premiums on Corporate Bonds,” Journal of Political Economy, June 1959, pp. 217–237Google Scholar.

¹² The use of an additive rather than a multiplicative form of the probability function presumes that the effect of each explanatory variable is independent of the values of all other explanatory variables. For example, the effect of leverage on a firm's bond rating is assumed to be independent of its profitability (although the bond rating need not be independent of profitability). This assumption, though probably not strictly valid, seems plausible since the literature on the subject typically treats leverage, profitability, earnings instability, etc. as independent factors to be considered in bond rating.

¹³ Walter, James E., The Investment Process (Boston: Harvard University, 1962), p. 433Google Scholar.

¹⁴ Moody's Investors Services uses ten bonds in preparing each of its averages with the exceptions of Aaa Industrials (for which only eight bond issues are used) and Aaa Railroads. The latter series was discontinued as of December 18, 1967, because there was an insufficient number of issues outstanding. Also, some of the railroad bonds used had only small amounts outstanding. (Moody's Bond Survey, December 25, 1967, Vol. 59, No. 52, page 28.) During the period 1961 through 1966, only five or six different Aaa railroad bond issues were used, depending upon what bonds were outstanding and active. Less than ten Aa railroads were used during most of the years considered in this study.

¹⁵ See, for example, Childs op. cit., pp. 10–12.

¹⁶ The authors are very well aware of the limitations of accounting data that result from changes in the income tax regulations, variations in accounting practices such as the treatment of extraordinary gains and losses and interperiod income tax allocations, and different procedures required by federal and state regulatory bodies such as public utility commissions. Nevertheless, adjusting accounting data to a consistent basis — even if that were possible — would not be appropriate for the present study. Such adjustments are not made systematically by either the rating agencies or by those who rely on such ratings (as far as the authors can learn). Furthermore, the various state codes that define legal investments for insurance companies and other fiduciaries do not consider these accounting limitations.

Differences in organizational and financial practices also affect the inter-firm and inter-industry comparability of financial data. For example, railroads lease much of their rolling stock on the basis of equipment trust obligations. Some industrial corporations also utilize long-term leases to different degrees in order to obtain the use of fixed assets. Since long-term leases and equipment trust obligations are not included among the long-term liabilities in corporate balance sheets, apparent exactness of the leverage ratio is compromised. Finally, no attempt is made to prepare leverage ratios on the basis of market values of securities because practicing security analysts do not look at such ratios as far as academicians can learn and because of the problems in evaluating direct placements.

¹⁷ Earnings coverage, X₅, does not appear in Tables 5 and 7 because X₅ is essentially a composite measure of leverage (X1) and profitability (X₂). Therefore, one would not expect X₅, on the one hand, and X1 and X₂, on the other hand, to make independent contributions to the explanation of bond ratings. Our results are consistent with this expectation. The correlations between X₅ and X1 and X₂ are high in comparison to the correlations among other explanatory variables. And, the addition of X₅ to equations that included X1 and X₂ brought no improvement or only marginal improvement in the fit of the equation (as measured by R2 or the standard error or estimate).

However, X₅ was substituted for X1 and X₂ in a set of alternative regressions. The results of this substitution are discussed briefly at the end of this section.

¹⁸ The multiple correlation coefficients (R's) are all statistically significant at the .01 level.

¹⁹ Warner, op. cit., pp. 60–61 employs a similar procedure to gauge the fit of estimated probability functions to the original sample observations.

²⁰ Secondary samples for utility and rail firms were not drawn for two reasons. First, the procedure employed to predict ratings can be illustrated adequately with the sample of industrial bonds. Second, additional secondary samples would have required much additional data collection and computation.

²¹ As was the case with the industrial samples of firms, explained variation was generally lower for Y₂, Y₃, and Y₄ than for Y₁. Relatedly, the coefficients of individual explanatory variables were less frequently significant in the Y₂, Y₃, and Y₄ equations than in the Y₁ equations.