I. INTRODUCTION AND PURPOSEEMPIRICAL TESTS of the now famous but still controversial market model, specified in [17,18], have been generated quite extensively in the United States. This--simple single-index model has recently been applied and empirically tested in foreign capital markets in Europe [15,22], Japan [11] and probably is being tested in numerous other environments. Indeed, Solnik has extended the concept to a world capital market structure [23, 24].The purpose of this paper is to examine, on a comprehensive basis, the market model for the French Stock market (the Paris Bourse) and to compare on an exact basis our results with equivalent U.S. studies. In so doing, we will examine data for 316 common stocks traded continuously on the Bourse for the period January 1, 1964 to November 30, 1971 (approximately eight years). In addition, we examine briefly the price index behavior of two "minor" data bases consisting of 35 industry groupings and 9 large sector groups. The market index utilized is the Campagnie des Agents de Change (C.A.C.) market value weighted index consisting of 430 stocks.'In our analysis, we utilize rate of return data for weekly, monthly, and in some cases quarterly intervals. The rate of return for French stockholders is of the form,
Rit-Pt -Pt, + Dt (1 + 0.5) Pt-' where Rit= Rate of return (before taxes) for security i in period t Pt = Price at end of period t * The authors acknowledge the assistance of Chambre Syndicale Agents de Change (Paris) who generously provided data and financial assistance and the Centre d'Enseignement Superieur des Affaires (Jouy-en-Josas) for the computer support necessary to construct the data base and test the model. Finally, to John McDonald, Didier Pene, Bruno Solnik and Randy Westerfield for valuable comments on an earlier version of this paper, "Comparative Market Model Analysis: France and the United States," Salomon Brothers Center For In this paper we merely refer to the industry and sector results for comparative purposes and the reader is referred to [2, 3] for more extensive reporting. These industry groupings comprise the same 430 stocks as the market index. The smaller individual stock sample, however, does not change composition over the entire-period-a factor conductive for intertemporal portfolio comparisons but a potential cause of statistical bias for averaging purposes (discussed in section 2). 1495 1496 The Journal of Finance Pt-, = Price at end of period t -1 Dt = Cash dividend2 paid during period tSection II examines the stationarity and stability of the Beta risk parameter from an individual security standpoint both in terms of short term and longer term forecasting potential. Section III extends the analysis to portfolios of securities. The market model's explanatory power is explored in Section IV. Whenever appropriate, we compare the reliability of results over various time horizons and for different rates of return period intervals. Results from U.S. stock market studies will form our comparative analysis.
II. EMPIRICAL RESULTS-I...