摘要
假设资产系统风险(市场贝塔)随经济状态变动,建立具有动态参数的条件CAPM并应用广义矩方法进行横截面定价检验。研究发现相对于CAPM、消费CAPM、投资增长、三因素模型等经典资产定价模型,条件CAPM具有明确的经济含义和较好的解释能力。在5个备选状态变量中,贝塔随每一个变动都对横截面收益有显著的解释能力,相对而言,上海银行间同业拆借利率(L)、广义货币供应量增长率(ΔlnM)等资金因素在中国证券市场有更重要的影响。本文发现小公司的系统风险一直高于大公司。小公司对经济状态更加敏感,因此它在经济不景气时的风险相对大公司更高。投资者应关注资产风险的动态变化。
CAPM (Capital Asset Pricing Model) has been widely adopted in the finance theory. However, some studies have shown that CAPM theory is ineffective at explaining the "size effect" and the "book-to-market effect" from cross-sectional returns. Among all models to improve CAPM, Fama and French ' s three-factor model has higher explanative power for cross-sectional average returns. The CAPM was derived in a hypothetical economy in which investors live for only one period, Therefore, studies based on CAPM theory often assume that market beta (systematic risk loading) remains constant over time. In the real world investors live for many periods. Therefore, a firm's systematic risk is likely to vary over the business cycle and the beta value to measure the systematic risk changes over time. In consideration of the reality, this paper assumes that the market portfolio is conditionally mean-variance efficient, the expected return on an asset is linear with its conditional beta in every period, and conditional beta varies with state variables. This paper has three research objectives. There are several classical pricing models in the international literatures and practices, including CAPM, consumption CAPM, three-factor model and investment-based model. The first objective is to test how these models perform in Chinese stock market. This study assumes that conditional betas vary with Shanghai Interbank Offered Rate (L), generalized money supply growth rate (AlnM), consumption growth rate (AlnC), fixed assets investment growth rate (Alnl) and consumer price index (CPI). The second objective is to explore the ability of conditional CAPM to explain the cross section of average stock returns. In the conditional CAPM setting, conditional beta is a function of state variables. Thus, the time-series of conditional betas, which reflect the varying pattern of asset systematic risks during business cycle, can be obtained from the function. The third objective is to dynamically analyze the pricing anomalies such as "size effect" through conditional betas over time. Theoretical models imply that the expected returns should be linear to asset betas with respect to pricing factors. There is an enormous body of empirical research that examines these linear asset pricing relations. A conceptually simple and more general solution is the generalized method of moments (GMM) approach, which is robust for both conditional heteroskedasticity and serial correlation in returns as well as in factors. In order to make asymptotically valid inferences, this paper adopts the GMM approach. This paper has accomplished three objectives. Firstly, although the three-factor model has the highest explanation power among the four classical models, it fails to pass the model specification test. Secondly, the conditional CAPM performs better than those classical models. Conditional betas varying with each of five candidate state variables have significant impact on average returns. These candidate state variables have different effects. When L or AlnM is state variable, the conditional CAPM can pass the model specification test and has higher explanation power. In addition, the conditional CAPM can dynamically analyze the pricing anomalies. According to the dynamical analysis, systematic risks of small firms are always higher than big firms' and are much higher when business cycle is in trough. Small firms are more sensitive to economic states. In summary, the results show that the expected return of an asset is not only determined by the average level of risk but also by the varying pattern. Therefore, it is important for investors to pay attention to time-varying risks.
出处
《管理工程学报》
CSSCI
北大核心
2012年第4期137-145,共9页
Journal of Industrial Engineering and Engineering Management
基金
国家自然科学基金重点资助项目(70532003)