【作者】 闫彬彬；

【导师】 孙焱林；

【作者基本信息】 华中科技大学 ， 数量经济学， 2013， 博士

【摘要】 向量自回归模型自Sims(1980)引入计量经济学以来,因其在结构分析方面独有的优势,得到广泛应用。目前,VAR模型及SVAR模型已经成为宏观计量分析的主要工具。然而,现实经济是复杂的,即便经济变量遵循确定的规律,用模型表示它们也将是极为困难的。在实证分析中,使用SVAR模型经常产生一些违背经济常识的结论,比较典型的如“价格之谜”、“利率之谜”、“汇率之谜”等。造成此现象的原因可能在于传统的SVAR模型设定形式可能过于苛刻,一方面,其参数是固定不变的,无法反映经济结构的变化,另一方面,约束条件通常以等式形式出现,约束条件过强。为解决上述问题,本文引入了符号约束的时变参数VAR模型。在识别机制方面,模型利用广为接受的经济常识建立符号约束条件,在模型参数设定方面,模型引入了系数和方差协方差矩阵的动态演进机制,以反映经济现实中的结构变迁特征。符号约束的时变参数VAR模型的优势在于,通过宽松的模型设定,采取让经济数据说话的形式,能最大限度避免模型误设,在结构冲击识别上,许多经济变量在经济中存在同期交互效应,此时等式约束将无法识别各个冲击,而借助符号约束则可以实现不同结构冲击的识别。本文首先对SVAR模型进行梳理,然后在此基础上对符号识别下VAR模型的建立、估计和运用进行了详细的分析。接下来,本文将VAR模型扩展至TVP-VAR模型,进而将符号识别机制引入该模型,并对其进行详细的阐述。在对TVP-VAR模型参数的动态变化特征进行一般性的设定时,模型引入了马尔科夫转移向量来判断模型参数是否发生变化、何时发生变化,以获取模型结构变化特征的直接证据。TVP-VAR模型的估计是一个繁重的工作,本文详细说明了估计方法。信贷市场是联接我国金融体系与实体经济的重要桥梁。现有的信贷理论和经验分析均表明,我国信贷市场具有明显的波动性特征,信贷冲击的传导具有显著的不对称性。然而实证领域仍缺乏工具来刻画信贷市场的这些特征。在完成了符号识别的TVP-VAR建模的理论工作之后,本研究利用符号约束,对我国信贷的供给冲击和需求冲击进行识别,并对不同冲击的性质、动态变化特征以及传导机制进行分析。研究结果表明,我国信贷冲击存在明显的时变特征,在政策取向上,政策的时效性应是重要的考量因素；不同时期我国经济的波动具有显著差别,经济过热和经济趋冷都会使波动性显著变化；不同冲击的传导效应不同,政府应根据不同时机有针对性的选择政策措施,以使政策效力达到最大。本文的贡献在于以下几个方面：第一,国内关于符号约束机制和TVP-VAR模型的研究才刚刚起步,本文将两者结合起来,从理论上进行详尽阐述,丰富了进行实证研究的工具；第二,本文将时变系数、随机波动以及马尔科夫状态转移融入符号识别的VAR模型中,有效的提高了模型完整性和兼容性,在方法论上具有一定价值；第三,本文对信贷传导的非线性和非对称性理论进行了总结,并结合我国经济现实对我国信贷传导的非线性特征进行理论分析；第四,本文关于我国信贷供求冲击的研究结论对宏观政策分析具有重要的参考意义。更多还原

【Abstract】 Since the introduction of the VAR model by Sims(1980), VAR has been widely used because of its unique advantages in structural analysis and has become the main tool of macroeconomic analysis. However, even if the economic variables is determined to follow the simple law, it would be extremely complex to model them. Under the complex reality, using the SVAR model in the empirical analysis generated something contrary to the conclusions of the economic common sense, such as "the mystery of the price","interest conundrum","exchange rate puzzle" and so on. The reason for this phenomenon may be that the form of the SVAR model is too harsh. On one hand, the fixed parameters can not reflect the changes in the structure of the economy, on the other hand, constraints condition usually set by equation which is too strong for the identify mechanism.To solve these problems, this dissertation studies the sign restriction on time-varying parameters VAR model. As for the identification mechanism, the model uses the widely accepted economic common sense to create sign constraints in impulse response function. As for the model parameters, the model introduces the dynamic evolution mechanism of the coefficient and the variance-covariance matrix to reflect structural changes in the economic reality. The advantage of sign restriction on time-varying parameters VAR Model is its ability to avoide misspecification maximum by put less restriction on the model and let the data determine which the model is better. In addition, the equality constraints are limited when economic variables interact at the same period, while the sign restriction can recognize the influence of different shocks separately.The dissertation retrospects the SVAR model firstly and then anatomy how to establish, estimate and apply the sign restriction on VAR model. Next, explains the TVP-VAR model in detail, and introduces the sign constriction into the model. In order to obtain direct evidence of the model structure changes, the model uses Markov transfer vector to determine whether and when the model parameters change before the setting of the dynamic characteristics of the TVP-VAR model parameters. The estimation of TVP-VAR model is a heavy work, the dissertation explains the estimation method step by step.Credit market is an important bridge to join financial system and the real economy. Existing credit theoretical and empirical analyzes show that China’s credit market has the distinct character of volatility and the conduction of credit shocks has significant asymmetry. In the empirical field, however, there is still a lack of tools to characterize these features of the credit market. After the theory analysis of sign restriction on the TVP-VAR model, the dissertation identify the loan supply shocks and demand shocks, and analyses the nature, dynamic characteristics and conduction mechanism of the different shocks. The dissertation found the credit shocks have the distinct character of time varying, For the policy makers, the timeliness of policy should be an important consideration; the economic fluctuations varied vastly during different periods, economic overheating and increasingly cold will be significant change volatility; the government should be more focus on the regulatory role of the market, and to exercise demand management policies prudent to enable the effectiveness of policies.Contributions of the dissertation can be stated as follow:First, domestic research on the mechanism of sign restriction and TVP-VAR model is just underway, the dissertation combines the two methods, elaborates the theory in detail and riches empirical tools; Secondly, the dissertation integrates the time-dependent coefficients, random fluctuations and Markov state transition into the sign restriction on VAR model, improve the integrity and compatibility of the model which has a certain value in methodology; Thirdly, this thesis summarizes the nonlinearity and asymmetry of the credit market theorically; Finally, research conclusion on the impact of the loan supply and demand in China has an important reference significance for macro-policy analysis.更多还原