I examine three composite analyst forecast of earnings per share as proxies for expected earnings. The most current forecast weakly dominates the mean and median forecasts in accuracy. This is evidence that forecast dates are more relevant for determining accuracy than individual error. Consistent with previous research, I find analysts more accurate than time-series models. However prior knowledge of forecast errors from a quarterly autoregressive model predicts excess stock returns better than prior knowledge of analysts' errors. This is inconsistent with previous research, and is anomalous given analysts' greater accuracy.

[The role of ordinary options in facilitating the completion of securities markets is examined in the context of a model of contingent claims sufficiently general to accommodate the continuous distributions of asset pricing theory and option pricing theory. In this context, it is shown that call options written on a single security approximately span all contingent claims written on this security and that call options written on portfolios of call options on individual primitive securities approximately span all contingent claims that can be written on these primitive securities. In the case of simple options, explicit formulas are given for the approximating options and portfolios of options. These results are applied to the pricing of contingent claims by arbitrage and to irrelevance propositions in corporate finance.]

The role of ordinary options in facilitating the completion of securities markets is examined in the context of a model of contingent claims sufficiently general to accommodate the continuous distributions of asset pricing theory and option pricing theory. In this context, it is shown that call options written on a single security approximately span all contingent claims written on this security and that call options written on portfolios of call options on individual primitive securities approximately span all contingent claims that can be written on these primitive securities. In the case of simple options, explicit formulas are given for the approximating options and portfolios of options. These results are applied to the pricing of contingent claims by arbitrage and to irrelevance propositions in corporate finance. The role of complete contingent-claims markets in the optimal allocation of risk bearing is well known [Arrow (1964) and Debreu (1959)] and is the cornerstone of the economic theory of financial markets [Mossin (1977)]. As a consequence, it becomes important from a practical as well as a scholarly perspective to determine how complex the securities markets must be in order to achieve the allocational efficiencies of complete markets. The literature on this question has grown to be sizable. Much of this literature has been reviewed in John (1981, 1984) and Amershi (1985). A seminal contribution concerning the complexity of complete securities markets was made by Ross (1976) in analyzing the role of conventional options in com-

This study is an ex post performance test comparing the accuracy of an American model to a European model for valuing listed options. Specifically, the Geske and Johnson American put valuation model is compared with the Black and Scholes European put model. On average, both models undervalue, relative to market prices, put options. However, the Geske and Johnson model values are significantly closer to market prices than are the Black and Scholes values.

"We argue that the postwar baby boom [in the United States] caused substantial fluctuations in both the economic rewards to education and educational attainment over the last 3 decades. If substitutability between young and old workers diminishes with education, the present value of lifetime earnings for a boom cohort is depressed more for highly educated workers, reducing incentives for educational attainment. The opposite is true for pre- and postboom cohorts. The diminishing substitutability hypothesis explains the declines in both the returns to college and college completion rates in the 1970s and predicts a substantial increase in educational attainment for postboomers."

Abstract. This empirical study is a response to the FASB's call for further research into the properties of alternative accounting measurement methods. There exists no empirical evidence on the time‐series properties and predictability of general price level (GPL)‐adjusted annual earnings data. In order to assess the effect of the monetary gains or losses on the stochastic properties of GPL‐adjusted data, earnings were calculated both with and without inclusion of the monetary gains or losses. The results of the time‐series analysis of the earnings series across the alternative accounting methods indicated that (a) just under one‐half of the historical cost (HC) series followed a random‐walk‐type process with most of the remainder being autoregressive, and (b) the GPL series showed substantially fewer following a random walk process with a corresponding increase in the number of stationary series (modeled as autoregressive or white noise processes). The predictive ability results were consistent with the time series findings. That is, application of a random‐walk model to the HC series indicated that these series were fairly well‐represented by the random walk, but a similar application to the GPL series confirmed that these series were not as well‐represented by a random walk. Résumé. L'étude empirique qui fait l'objet du présent article a été effectuée en réponse à l'invitation du FASB à poursuivre les recherches sur les attributs des méthodes de mesure comptable de rechange. Il n'existe aucune démonstration empirique des attributs des séries chronologiques et de la valeur prédictive des données relatives aux bénéfices annuels indexés sur le niveau général des prix (N.G.P.). En vue d'évaluer l'incidence des gains ou des pertes monétaires sur les propriétés stochastiques des données indexées sur le N.G.P., les bénéfices on été calculés à la fois avec et sans la prise en compte des gains ou des pertes monétaires. Les résultats de l'analyse chronologique des séries de données relatives aux bénéfices à travers les différentes méthodes comptables ont donné lieu aux constatations suivantes: a) à peine moins de la moitié des séries de données au coût d'origine (C.O.) ont connu des variations de type aléatoire, la plupart des autres séries étant autorégressives, et b) beaucoup moins de séries de données indexées sur le N.G.P. ont affiché des variations aléatoires avec une augmentation correspondante du nombre de séries stationnaires (modélisées comme étant autorégressives). Les résultats relatifs à la valeur prédictive étaient compatibles avec les résultats de l'analyse chronologique. En d'autres termes, l'application d'un modèle aléatoire aux séries de données au C.O. a permis de constater que ces séries sont assez bien représentées par la méthode aléatoire, mais l'application du même modèle aux séries de données indexées sur le N.G.P. confirme que ces séries ne sont pas aussi bien représentées par la méthode aléatoire.

On introduit le concept de processus aleatoire vectoriel presque integre. Ce type de processus aide a travailler vers une theorie asymptotique generale de la regression pour des series temporelles multiples