The work presented in this dissertation can be grouped around three major themes. The first theme relates to risk, the second theme relates to asset pricing, whereas the third theme relates to serial correlation of asset returns. The three chapters of this dissertation investigate these themes Chapter Two analyses the behavior over time of market risk, aggregate idiosyncratic risk and correlations in portfolio of Taiwan listing stocks and studied pattern of aggregate correlation between the 3 most important Taiwan stock index and Taiwan value-weighted index. We find (1) Idiosyncratic risk is trended upwards; (2) The conditional stock returns correlation process is asymmetric. The implication of our finding is (1) It takes more stocks to achieve a given level of diversification; (2) Diversification strategies perform poorly in bear markets. Chapter Three investigates the role of the asset co-skewness and conditioning information in asset pricing. First, I estimate long-run predictive regressions of asset returns to test whether aggregate idiosyncratic risk is a price factor of industrial returns. Then I use data on Taiwan 19 industry portfolios to fit various assets pricing models. I find (1) the cross-sectional ctional correlation between 2 i £] (the gamma coefficient from the 3M-CAPM equation) and 3 i ϕ (the interaction coefficient from the CCAPM equation) is positive and fairly large. (2) The firm-level volatility is a good proxy for cay as conditioning information variable. (3) The gamma coefficient can pick up the extent of beta co-vary with the market wide excess-return over the business cycle. (4)among 19 industrial returns, the 2 industrial returns can be explained by 3M-CAPM; the 7 industrial returns can be explained by CCAPM; the 5 industrial returns can be explained by 3M-CAPM+CCAPM, Others can¡¦t be explained by either of three models. Chapter Four examines the impact of positive feedback trading behavior of the investors on the short-term dynamics of return for four Taiwan index futures contracts by utilizing the framework of the model developed by Sentana & Wadhwani(1992). Use of the Asymmetric Nonlinear Smooth Transition GARCH Model demonstrates that positive feedback trading of investors is the main determinant of short-term dynamics of return for Taiwan index futures contracts. Moreover, it shows that positive trading is more intense during market declines than it is during market advances due to extensive use of spot-loss trading for investors. Finally it is shown that the sophisticated professional investors intend to take positive feedback trades wave so that they lead to increase positive feedback trading in Taiwan index futures since the government opened the enterprises for managed futures.

Cette thèse est organisée en trois chapitres. Dans le premier chapitre, qui est co-écrit avec Ilze Kalnina, nous proposons un test statistique pour évaluer l’adéquation de la volatilité idiosyncratique comme mesure du risque idioyncratique. Nous proposons un test statistique qui est basé sur l’idée qu’un bon proxy du risque idiosyncratique devrait être non correélé à travers les actifs financiers. Nous démontrons que l’estimation de la volatililité est sujet à des erreurs qui rendent le test non standard. Nous proposons un modèle à facteurs qui permet de réduire sinon éliminer les corrélations dans la volatilité idiosyncratique, avec comme ultime but d’ aboutir à un facteur qui satisfait mieux aux critères souhaités du risque idiosyncratique. Dans le deuxième chapitre de ma thèse, qui est co-écrit avec Christian Dorion et Pierre Chaigneau, nous proposons une méthodologie pour étudier l’importance des risques d’ordres supérieurs dans la valorisation des actifs financiers. A la suite de Kraus and Litzenberger (1976) et Harvey and Siddique (2000a), beaucoup d’études ont analysé l’aversion aux risques de skewness et kurtosis de façon inconditionnelle. Dans ce chapitre, nous proposons une méthodogie qui permet de faire une analyse conditionnelle assez précise de l’aversion au risques d’ordres superieurs. Notre étude complémente la littérature dans la mesure ou nous étudions aussi la valuation des risques d’ordre plus élevé que la kurtosis à savoir l’hyperskewness et l’hyperkurtosis qui sont théoriquement valorisés dans certaines fonction d’utilité comme le CRRA. Dans le dernier chapitre de ma thése, j’étudie la structure à terme de la prime de risque pour le risque de co-skewness, un risque qui mesure l’asymmétrie systématique dans les actions individuelles. Nous y proposons une méthode assez générale qui permet de faire une analyze mutli-horizon contrairement à la plupart des études existantes. / This thesis is organized in three chapters. In the first chapter (which is co-authored with Ilze Kalnina), we propose a statistical test to assess the adequacy of the most popular measure of idiosyncratic risk, which is the idiosyncratic volatility. Our test statistic exploits the idea that a “good" measure of the idiosyncratic risk should be uncorrelated in the cross-section. Using in-fill asymptotics, we study the theoretical properties of the test and find that it has a non-standard behaviour due to various biases induced by the latency of the idiosyncratic volatility. Moreover, we propose a regression model that can be used to reduce if not eliminate the cross-sectional dependences in assets idiosyncratic volatilities. The second chapter of my thesis is the fruit of a colaboration with Christian Dorion and Pierre Chaigneau. In this chapter, we study the relevance of higher-order risk aversion in asset pricing. The evidence in Kraus and Litzenberger (1976) and Harvey and Siddique (2000a) has spurred the literature on the estimation of the risk premiums attached to skewness and kurtosis risk in addition to the standard variance risk. However, most of these studies focus on the estimation of unconditional premiums or average premiums. In this chapter, we propose a methodology that allows to accurately estimate the time-varying higher-order risk aversions using options prices. Our study complements the literature as we also study the higher-order risks beyond the kurtosis such as hyperskewness and hyperkurtosis risks which are valued by a CRRA investor. . In my third chapter, I study the term-structure of price of co-skewness risk. Co-Skewness risk captures the portion of the stock returns asymmetry that arises as a result of market returns asymmetry. I propose a general methodology that allows to study the multi-horizon pricing of this risk in contrast to many existing studies.