THE REGRESSION PORTFOLIO: A STATISTICAL INVESTIVATION OF A RELATIVE DECLINE MODEL

This paper reports the results of empirically testing a share selection system based upon the assumption that the stock market over-values recent performance. This system, which is called the Regression Portfolio, had modest success in yielding above-average income but was less successful in its capital performance. The results provide some interesting insights into the behaviour of the London Stock Exchange and the performance of quoted companies, during the period 1967–72, in addition to suggesting further work on share selection models. Dans ce papier, on révèle les résultats de l'analyse à laquelle on a soumis un système sélectif d'actions fondé sur l'hypothèse que le marché des valeurs surestime les derniers rendements. Si ce systéme, appelé‘portefeuille régressif’, a relativement permis d'offrir des bénéfices supérieurs à la norme moyenne, il n'a pas su par contre justifier son entière efficacité en ce qui concerne la valeur en capital. Excluant le fait qu'une recherche plus approfondie s'impose sur les moèbles sélectifs d'une action, les résultats fournissent des renseignements intéressants quant au comportement de la Bourse de Londres et au rendement des sociétés cotées en Bourse, au cours de la période 1967–72. Dieser Beitrag berichtet über die empirischen Teste eines Aktienauswahlsystems, welches davon ausgeht, dass der Börsenmarktneue Aktienrentabilität überbewertet. Dieses System, das Regressionsportefeuille heisst, hatte mässigen Erfolg bei der Auswahl von Dividenden, war aber weniger erfolgreich bezüglich der Aktienkurszuwachsrate. Die Ergebnisse beiten interessante Einblicke in das Verhalten der Londoner Börse, sowie in die Leistung von notierten Aktiengesellschaften in den Jahren 1967–72. Darüberhinaus zeigen sie neue Richtungen für weitere Porschungsarbeit auf diesem Gebiet.     

A STATISTICAL MODEL OF CHANGES IN ASSET PRICES EMPLOYING INTRADAY DATA: A RECURSIVE APPROACH

This paper develops a statistical model of changes in asset prices employing intraday data. The procedure proposed in this paper is an alternative to the Hausman, Lo, and MacKinlay (1992) ordered probit model. Similar to the ordered probit model, our model also contains the linear regression model as a special case. However, compared to the ordered probit model, our specification is parsimonious. Parsimony does come at a cost, but for certain applications where, for example, a benchmark return is needed in intraday studies, there is value in terms of the computational effort required and the method's robustness to various empirical microstructure phenomena. The parsimony is achieved when statistical implications arising from intraday structural changes, which may due to such factors as concentrated trading patterns, are incorporated into a statistical model.