The past two decades have seen a proliferation of mathematically sophisticated portfolio selection models. Of these, the mean variance (MV), expected utility, and growth optimal (GO) models have received the bulk of attention.

Edward Miller [5], expanding on the work of Williams [8], Smith [6], and Lintner [4], has proposed a direct relationship between a stock's “risk” and its “divergence of opinion.” Under conditions of uncertainty, potential investors in a stock arrive at different assessments of expected return. Thisvariation in expectations is characterized as the stock's divergence of opinion. Miller argues persuasively that at a point in time a stock's price does not reflect the expectations of all potential investors, but rather the expectations of only the most optimistic minority who are trading the issue. As long as this minority can absorb the entire supply of stock, an increase (decrease) in divergence of opinion-leaving the average expectation unchanged-will increase (decrease) the market clearing price.

A recent issue of the Journal of Financial Economics (June/September 1978) is devoted to reporting anomalous evidence regarding market efficiency. This study may qualify under the same heading. Here we examine the market response to information about firms whose future is assessed as being extremely problematic.

Much of modern portfolio theory rests on conclusions drawn from the original form of the Capital Asset Pricing Model. Fundamental to the conclusions of this model is the assumption of perfect competition among investors; i.e., all investors possess approximately the same small amount of wealth such that equilibrium price cannot be influenced significantly by the demand of any of the investors. Today's security market, however, is characterized by individuals and large institutional investors such as insurance companies and investment funds. Although institutions represent a very small fraction of all investors in the market, institutional investors in 1977 held 34.3 percent of all outstanding stock. By the very magnitude of the dollar transactions effected by these large investors, prices can and are affected dramatically. Because today's security market is composed of investors exhibiting extreme differences in wealth, the United States securities market probably is not perfectly competitive as assumed in portfolio theory. Consequently, investment theory must be extended to reflect vast wealth differences among investors. To achieve this end, modifications are made to the original Capital Asset Pricing Model. Equilibrium conditions are examined and conclusions are drawn as to how portfolio theory must be altered to include price affecting ability by a segment of the investors in the market.

The importance of accounting information on security price determination is of interest to both security analysts and accountants. Beaver [3], Downes and Dyckman [6], Gonedes [12], Beaver and Manegold [4], and others have investigated the possible relationships between accounting information and market information. Rosenberg [26] has shown the existence of extra-market components of covariance in security returns while Simkowitz and Logue (S–L) [28] have derived the interdependent structure of security returns. However, none of this research has explicitly investigated how the empirical results can be affected by alternative accounting profitability measures within an industry simultaneously.

Over the past years the beta coefficient has been widely used as a measure of systematic risk in investment and portfolio analysis. The validity of using the beta coefficient as the proper measure of systematic risk is dependent upon the assumption that the beta coefficient is stationary over time. Unfortunately, this assumption has been challenged by a number of empirical studies which have found the beta coefficient to be unstable over time. Examples of such empirical investigations are those documented by Blume [4], Levy [12], Levitz [11], Baesel [2], Altman, Jacquillat, and Levasseur [1], and Roenfelt, Griepentrong, and Pflaum [16]. Most recently, Fabozzi and Francis [9] reported that some security beta coefficients tend to be random over time. Their findings also support the regression tendency of the beta coefficients towards the mean over time, as found by Blume [4]. Thus, because the beta coefficient is changing over time, the use of the ordinary least-squares (OLS) method in investment and portfolio analysis will yield an inefficient estimate of systematic risk. Furthermore, the OLS estimates of security and portfolio residual risks will be influenced by the variability of beta coefficient. Therefore, the purpose of this paper is to investigate the relationship between the variability of the beta coefficient and portfolio residual risk, and hence to provide a real picture of the process of portfolio diversification under the condition of beta nonstationarity. It is shown that the use of the OLS method to estimate security and portfolio residual risks will produce an incorrect conclusion that larger residual risks tend to be associated with higher variability in the beta coefficient.

This study focuses on the risk-return characteristics of investments in the common stocks of U.S.–based multinational corporations (MNCs) and U.S. national corporations (NATLs). Findings follow from a comparison of the risk-adjusted performance of MNCs and NATLs using the framework of the capital asset pricing model (CAPM). Results of this comparison challenge assertions of earlier writers that marginal benefits or advantages accrue from investments in MNCs as compared to NATLs.

The search for a distribution which accurately describes the behavior of stock price returns has generated a considerable amount of controversy. While it is well known that the traditionally used assumption of lognormality deviates in systematic ways from the empirically observed—the latter has fatter tails and a larger concentration of mass near zero—none of the alternatives that have been proposed over the years (Stable Paretian—Mandelbrot [5], Poisson mixture of; lognormal distributions—Press [10], scaled T distribution—Praetz [9], lognormal with nonstationary variance—Rosenberg [11], subordinate stochastic process—Clark [2]) has gained general acceptance.

This page replaces the original in Richard Roll's article, “Orthogonal Portfolios,” in the December 1980 issue. Tear out and insert.

Since the orthogonal frontier extends indefinitely with nonzero slope, every nonefficient portfolio has an orthogonal portfolio with the same mean return.