I. The problem, 547. — II. Cost of capital: whose opportunity cost? 548. — III. A few formal relationships, 549. — IV. A diagrammatic approach to cost of capital, 552. — V. Transactions costs and taxes in practice, 555. — VI. A few comments on the opportunity loci, 559. — VII. Characteristics of an optimal financial structure, 561. — VIII. The real marginal cost of debt and equity, 564. — IX. A measure of the cost of capital, 567. — X. Capital as a weighted average, 569. — XI. Conclusion, 571. — Appendix A, 571. — Appendix B, 575.

IN recent years, the term structure of interest rates has become an increasingly important area for economic research. As a continuing balance-of-payments deficit has forced United States authorities to search for new policy weapons, theorists and statisticians alike have taken another look at the possibility of twisting the rate structure. Far from converging toward a single set of answers, however, this research has intensified pre-existing controversy. While there is substantial agreement among recent empirical investigators that expectations do play a prominent role in determining the rate structure,1 considerable disagreement rernains regarding the proper emphasis and interpretation to give to investor expectations. Particular areas of dispute with relevance for monetary policy include the following. Do investors actually form expectations and do they appear to rely on them in appraising what maturity ranges of the yield curve are attractive for investment, or are preferences for particular maturities largely explained by institutional considerations? 2 Will rates tend to equal the average of short-term rates expected by the market over the period to maturity of the long-term debt, as has been argued by the pure expectationists, or must risk aversion be introduced to explain the prevailing pattern of market yields? 3 Do the expectations of market participants tend to be identical or does there tend to be a wide dispersion of forecasts? A dispersion of expectations would imply that exogenous changes in the maturity composition of the debt can influence the rate structure even in the absence of risk aversion and practical impediments to lender and borrower mobility among maturity areas.4 Obviously, since these issues are empirical, so wide a range of hypotheses cannot be tolerated in the long run. In blocking the development of a reliable consensus, the major obstacle has been the lack of independent evidence on investors' expectations. The data that would resolve the controversy have remained buried in the unobservable unconscious of market participants. In order to test for the importance of expectations, it has been necessary first to generate artificially a representative series of expected rates.5 This paper describes an attempt to measure market expectations directly. The data presented were gathered by means of a mail survey conducted on April 1, 1965. A copy of the questionnaire is included in the appendix. The sample consisted of 119 banks, 16 life insurance companies (LICO's) and 65 nonfinancial corporations (NFC's). To insure that the more important participants in the government securities market would be sampled, we concentrated on large firms. The bank sub-sample includes primarily the largest banks in a sample of 500 banks constructed by Baxter and Shapiro.6 Similarly, the 65 NFC's surveyed * We are greatly indebted to the National Science Foundation for financial support, to Professor D. G. Luckett for helpful criticism, and to Professor N. D. Baxter for valuable advice in the preparation and distribution of our questionnaire. We also wish to thank Professors Baxter and H. T. Shapiro for allowing us to use their sample of investing institutions, and Mr. D. B. Rubin for programming assistance. Computations were performed on computer facilities supported by NSF grant GP 579. 'See, for example, Meiselman [12]; de Leeuw [3]; Kessel [7]; and Modigliani and Sutch [14]. 2 Though he believes that expectations may be important in the very short run, Culbertson [2] is widely recognized as the leading exponent of this view. 'See Meiselman [12, Chapter 2]; Hicks [5, pp. 138-139 and 144-147]; Wood [16, pp. 165-166]; and Kessel [7]. 'See Luckett [8]; and Malkiel [10]. The only exception of which we are aware is an unpublished survey conducted in 1943 for the National Bureau of Economic Research by Donald Woodward and described in Wallich [15, p. 493]; and Meiselman [12, p. 11] . This survey asked 200 experts to estimate a single average long-term rate over the next two decades. 6 This sample includes almost all of the nation's 50 largest banks. See Baxter and Shapiro [1, pp. 483-496] for a description of the design of their sample. All banks in the Baxter-Shapiro sample with 1962 deposits of over $200