In his 1976 Presidential Address to the American Finance Association, Merton Miller provided a compelling argument that there currently exists no viable theory of the optimal capital structure of an individual firm. This argument follows from the critique he presented of existing models of capital structure and from the theory he outlined of the optimal aggregate capital structure of the economy as a whole. That theory depends on the existence of different marginal tax rates for individuals and a tax-free security. Professor Miller pointed out that he was motivated to develop his hypothesis by the apparent inadequacy of a (if not the most) popular explanation for capital structure at both the micro and the aggregate level: the tradeoff between the tax advantages of debt and the cost to the firm's security holders of the bankruptcy process. He observed that neither the tax advantage of debt nor the costs of bankruptcy may be quite what they seem at first glance. When the corporate income tax and the differential taxation of regular income and capital gains are taken into account, then the tax advantage of debt is reduced. Moreover, the limited empirical evidence from actual bankruptcies suggests that the real costs to security holders of bankruptcy may be really rather low. And the recent discussion by Haugen and Senbet [6] suggests that most of the costs attributed to bankruptcy are really costs of liquidation of the firm's assets and not relevant to the capital structure decision.

In this study, alternative real and simulated market indexes are examined as proxies for the “common factor” required by the Sharpe portfolio selection model [13]. The ex post performance of efficient and well-diversified portfolios generated by the model based on the different indexes is compared. The results indicate no significant difference in performance between real and simulated indexes, although the degree of diversification is much lower for portfolios based on indexes which relate well to the universe of securities. It is also shown that portfolios which are selected according to the Sharpe model (regardless of the index) outperform strategies which call for investing in the market portfolio.

The existence of seasonality in security rates of return has implications for both the study of market efficiency and tests involving return models. The existence of seasonal asset returns may be an indicator of market inefficiencies. In an efficient market, investor arbitrage should remove any excess seasonal return an asset receives over a comparable asset of equal risk. The presence of seasonal returns, however, does not necessitate market inefficiency. For example, an expected seasonal return may exist in an efficient market simply because of anticipated seasonal patterns embedded in its underlying determinants. Tax regulations, government monetary policy, seasonal information lags, or risk adjustments have all been advanced as determinants of seasonal movements in return. No matter what the basis for return seasonality or the extent of market efficiency, if seasonality in asset returns exists, then these returns do not follow a strict stationary process within the year. Statistical models analyzing asset returns may use this information to improve model specification. For instance, Kinney and Rozeff [16] have shown that large efficiency gains in estimating portfolio betas can be achieved using time stratified estimates which explicitly incorporate seasonality in 4 stock returns.

The relationship between the rate of inflation and the interest rate has been a topic of research for quite some time. A breakthrough in the analysis occurred years ago with Irving Fisher's hypothesis that the nominal interest rate fully reflects the available information concerning the possible future values of the rate of inflation. Others have extended Fisher's original insight to explain further the interaction between the rate of interest and inflation. For example, Mundell [26] uses the Pigou real-balance effect to hypothesize that the real rate of interest is inversely related to the rate of inflation.

Since the beta systematic risk coefficient and the standard deviation are both important statistics in the received capital market theory [22] and the received option theory [1], considerabe effort has been expended on obtaining empirical estimates of these statistics [30]. The ordinary least squares (OLS) technique is typically utilized to estimate beta as the regression coefficient of a simple linear regression. However, the OLS betas for common stocks were found to be disconcertingly unstable over time [5, 6, 13, 15, 25]. But, whether the OLS beta or an adjusted beta were used, the regression statistics could still only explain less than half of the variability of most New York Stock Exchange (NYSE) stocks' returns (more specifically, R2 < .5).

Of the many conglomerate studies to date, some have dealt with the risk-return performance of conglomerates in the context of the capital asset pricing model [2,7,10,14], others have considered the motives for the formation of conglomerates [4,5,6,13], and still others have examined the operating characteristics of conglomerates [9,12,15]. Within the last group, Weston and Mansinghka [15, p. 928] argued that the primary motivation for conglomerate formation is defensive diversification, “…defined as diversification to avoid adverse effects on profitability from developments taking place in the firm's traditional product market areas.” Another motivation is provided by Levy and Sarnat [4] and Lewellen [5] who demonstrated that the only economic gain from a purely conglomerate merger may be the increased debt capacity resulting from the combination of entities having imperfectly correlated earnings streams.

In a recent article Fabozzi and Francis [3] presented the results of empirical tests designed to determine if the regression coefficients of the single-index market model were significantly different in bull and bear markets. Using three alternative definitions of bull and bear markets, Fabozzi and Francis (FF) concluded the coefficients of the single-index market model were not significantly different in the two types of markets.

There is strong theoretical support for the notion that prices in a perfect capital market will vary randomly ([22], [16]). However, the existence of some nonrandomness in stock prices is well documented, see ([10], [11], [13], [14], [20]). Of special importance for this study is the research by Young [24] who finds predominantly negative autocorrelation for a sample of securities using a monthly differencing interval. Autocorrelation coefficients are often used as a measure of nonrandom price behavior; negative autocorrelation is an indication of price reversals.

The properties and characteristics of financial ratios have received considerable attention in recent years with interest primarily focused on determining the predictive ability of financial ratios and related financial data. Principal areas of investigation have included the prediction of corporate bond ratings [13, 20, 23, 34], and the anticipation of financial impairment [1, 2, 3, 5, 6, 7, 18, 19, 29, 32, 33, 35]. Related studies have examined the characteristics of merged firms [25, 28], the differencesin financial ratio averages among industries [9, 10], whether firms seek to adjust their financial ratios toward industry averages [15], the relationship between accounting-determined and market-determined risk measures [4, 8, 24], and the influence of financial ratios on analysts' judgments about impending bankruptcy [14, 17]. The general conclusion to emerge from these various research efforts is that a number of financial ratios have predictive and descriptive utility when properly employed.

This paper is concerned with costs of capital in mathematical programming formulations of the problem of capital budgeting under capital rationing. It shows that there is a serious error in the method outlined by previous authors for converting the shadow prices from the solution of the dual into measures of the firm's marginal internal opportunity rates. In addition to demonstrating that the traditional approach leads to erroneous and nonsensical results, this paper presents a correct procedure for determining these rates.

The theoretical desirability of stochastic dominance (SD) as a decision rule is well established [1, 3, 4, 7, and 11]. However, implementation of SD as a decision rule has been hindered seriously by the lack of an optimal search algorithm [8]. An optimal search algorithm is desirable since it takes the distribution of returns for a group of assets and determines the optimal proportion of each asset which should be combined to provide efficient combinations. For example, for a given expected value (variance) the mean-variance (EV) algorithm builds the portfolio with the smallest (largest) variance (expected value). The EV algorithm determines which assets should be combined and the proportion of the total investment that should be invested in each asset. An analogous algorithm does not exist for SD.

Since the formulation of the portfolio selection problem by Markowitz [12] as a parametric quadratic programming problem, considerable effort has been devoted to obtaining operational portfolio management models. Research has involved: (1) characterizing the return generating process in terms of index models; (2) specifying special-purpose algorithms such as the critical-line method of Markowitz [13] or the solution procedure of Jucker and de Faro [11]; (3) using linear programming formulations to approximate solutions to the nonlinear programming problems such as Sharpe [20, 22] and Stone [25]; and (4) converting portfolio selection models into portfolio management models designed to revise an existing protfolio subject to transaction costs using heuristics such as Smith [24] or revision formulations such as Pogue [16, 17] and Stone and Reback [27].

In an important article in 1971, Fisher and Weil [4] demonstrated that it is possible to immunize a portfolio of default-free coupon bonds against unexpected interest rate changes so that at the end of the planning period the investor will realize at least the return expected at purchase. Immunization may be achieved by constructing a portfolio whose average duration is equal to the length of the investor's planning period. The computation of duration that produces immunization is dependent on the nature of the assumed stochastic interest rate shocks. Fisher and Weil derive the duration that will produce immunization for additive shifts in the yield curve under instantaneous compounding, e.g., g(t) + λ where g(t) is the instantaneous interest rate at time t and λ is a random shift parameter.

In the June 1978 issue of this Journal, Herbst [2, p. 363] cautioned the reader to beware of a unique, real internal rate of return (IRR) because such a return is “an incorrect measure of the return on investment” for a mixed project. While it is true, as discussed by Teichroew, Robichek and Montalbano (TRM) [4], and now by Herbst, that a mixed project may have a unique IRR, several additional observations developed by TRM should be added to Herbst's discussion. These comments will clarify the implications of Herbst's paper for a decision maker who has a mixed project with a unique real IRR.

In a recent article [1], Beedles suggests that the valuation process for cash outflows (or negative benefits using his terminology) is, in some sense, different from the valuation process for cash inflows. This result, however, is not consistent with the assumption of perfect capital markets. Any cash outflow from one firm represents a cash inflow to some other firm(s) or investor(s). Consequently, any difference in the valuation processes for cash outflows and cash inflows will create profitable arbitrage possibilities.