【作者】 项云帆；

【导师】 王少平；

【作者基本信息】 华中科技大学 ， 数量经济学， 2010， 博士

【摘要】 金融变量具有非对称性,时变性的特征,无论是新兴金融市场、还是成熟金融市场中,交易的金融工具以及不同金融市场之间具有互补性和替代性,因此,宏观经济变量对金融市场收益的影响具有不对称性。本文首先运用MS-VECM模型分析中国宏观经济变量对股票市场收益的影响,并运用Bootstrap仿真检验对估计参数进行检验。实证发现可将状态分为：我国金融市场发展早期的状态1、我国金融市场受政策干扰而变化频繁的状态3、以及经济健康发展且稳定的状态2。结果表明宏观经济变量对我国股票市场收益在不同状态下影响不同,具有不稳定性,但在状态内具有一定持续性。Fisher假说在我国股票市场发展较成熟时成立；汇率上升对股票市场收益具有正的影响；货币供给量M1对股市收益影响则随着状态不同而变化；在早期对我国股票市场,相关的替代性资产或者金融市场(如无风险利率、标准普尔指数收益)对股市收益具有反向影响,我国金融市场发展到一定阶段时具有正向影响。然后,以现值模型为基础,用每股盈余取代每股股利,运用状态空间模型和卡尔曼滤波方法对我国上证指数、沪深300指数和中小板指数进行分析不可观察的投机泡沫变量,检验我国股市是否因为存在泡沫而导致股指偏离其市场基本价值,并对泡沫的大小、形成速度等特征和存在期间进行分析,实证结果发现我国股指较难用市场基本价值解释,股指较长时间存在正泡沫,泡沫累积过程较长,但泡沫破灭速度较快,且导致较短时间内存在负泡沫。中小板指数显然比上证指数和沪深300指数投机泡沫成分大,上证指数和沪深300指数泡沫特征类似。估计贴现因子系数表明泡沫累积难以持续,终会破灭。随后,考虑到CAPM的动态特征以及时变性,用贝叶斯动态线性模型分析CAPM参数的时变性,本文用面板贝叶斯动态线性模型分析CAPM,面板贝叶斯动态线性测度方程方差、状态方程方差矩阵均为未知,因此,本研究先以Gibbs抽样技术以及马尔可夫链蒙特卡罗方法(MCMC)运用动态线性模型滤波方法估计出测度方程、状态方程的方差矩阵,然后运用此方差矩阵,利用所有信息集,用平滑滤波估计出时变β。本文发现商业银行的CAPM中风险溢价因子β波动幅度显然比非商业银行金融机构低。在牛市时,CAPM中风险溢价因子β较长时间为正,而在熊市,CAPM中风险溢价因子β较长时间为负。从投资角度来看,长时期内β均值为零。最后,本文用面板分位数回归方法分析不同分位点时CAPM中β的特征,面板分位数回归同样对于传统的检验并不有效,因此本文运用Bootstrap仿真方法来检验参数估计的有效性。并和面板固定效应估计的值进行比较,发现固定效应面板CAPM估计的β值高于0.6分位点β值,说明我国投资者对于不同风险偏好不同。在不同的分位点下,β值为正值,并Bootstrap仿真检验β值统计上显著。而且随着分位点从低到高,β值逐渐增加,但在0.3-0.6分位点间较为平稳。均值面板回归估计的β值显著地大于中值处的估计的β值,说明股票市场收益分布并不是对称的。单个股票超额收益与市场组合报酬显著地正相关。本文的主要贡献在于：首先,本文探讨了股票市场收益受通货膨胀、外围股市以及其他宏观经济因素的影响,证实我国股票市场收益受通货膨胀的影响较大,并且在不同状态下,通货膨胀和股票市场收益不同。其次,对我国股票市场泡沫进行了测度。第三,运用面板贝叶斯动态线性模型分析CAPM时变参数的非对称性特征。最后,探讨CAPM中不同分位数的β的特征。更多还原

【Abstract】 Financial variables has characteristic of asymmetric and time varying In financial market which complementary and substitution each other. Here we examines the influence of Macroeconomic variables on stock market returns in emerging stock market, in P. R. China with cointegrated vector autoregressive Markov-switching model(MS-VECM). Also bootstrap simulation is used to test the signification of parameter estimated. The results is three states which is determinant:(1) the developing early stage (state 1); (2) macroeconomic is healthy and stable development (state 2); (3) the financial market is influenced by macroeconomic policy. Fisher hypothesis only exists in mature stage. We found (a) exchange rate has positive effect about return of stock index; the comsumer price index, (b) the money supply have different affect on stock return in different regime. (c) free risk rate and Standard & Poor 500 index have negative effect in early stage and positive effect in mature stage on stock returnSecondly an present-value model is used as the basic model, to analysis the unobserved variable-speculative bubbles of Shanghai stocks index and Husheng300 index and mid and small capital stocks index. state space model is used to and the parameter is estimated by kalman filter. Per-share earnings is proxy of dividends per share in this paper. The bubbles is tested whether existed or not, and its character such as size and formulating speed, the period is also analyses. The result is that the stock index can’t be explained by market foundation value. The stock index is positive bubbles in a long period and cumulative a long time. But its deflated abruptly also. Then negative bubbles existed markedly short-lived. Mid and small capital stocks index is more speculative bubbles than others. The discount factor estimated show that the bubbles can’t sustain for ever and will collapse.Then dynamic and time varying of capital asset pricing model (CAPM) is considered. CAPM is analysis using Bayesian dynamic linear model, then extended to panel data Bayesian dynamic linear model. However, the variance matrix of both measure equation and state equation are unknown. Markov chain monte carlo and gibbs sampling technology is used to estimate two variance arrays. The smooth filter is used to estimate all time varying parameters of panel data in CAPM. The results is beta which is risk premium factor, of commercial bank have lower volatility than that of non-commercial bank. Risk premium factor beta In up market, in CAPM is positive in long time and negative in downside market and zero mean value.Finally, extreme value or value at risk is researched. In the chapter, CAPM is analysis by panel data quantile regression. The beta in different quantile which estimated by panel quantile regression is compared with which estimated using fixed effect panel data. The beta value which estimated by fixed effect panel data is higher than that of 0.6 quantile. The traditional test such as t test can not be used to test panel data quantile regression, so bootstrap simulation also be used to test significant of parameter. The beta value which estimated by panel data is higher than that of 0.6 quantile. Because of financial variable which is asymmetric of, the beta value is getting bigger accompany with growing quantile. The distribution of financial variable is not symmetry.The contribution of this dissertation is:(1) the stock return is influenced by which maroeconomic variables and how to be influenced; (2)the speculative bubble is measured with dynamic linear model in China stock market; (3)the time varying and asymmetric parameter of CAPM is analysis using panel data Bayesian dynamic linear model;(4) the difference quantile beta is analysis using panel data quantile regression.更多还原