Econometric Analysis of Event Studies

本研究的目的是检查事件研究中出现的一些计量经济学问题，这些问题通常用于调查特定事件对公司股价或其发行的债券价格的影响。首先，分析了在兴趣模型中包含多种生成变量的理论影响。使用生成变量对估计量的一致性和估计量的标准误差的渐近影响表明，除其他外，取决于第一阶段和第二阶段回归中的观测数，以及所使用的生成回归量的类型。对这些问题的解决方案进行了详细的描述。其次，使用蒙特卡罗研究分析了使用一些生成变量的小样本影响。研究发现，一些工具变量估计器比普通的基于最小二乘的估计器具有更小的偏差，但与生成变量系数相关的检验的大小和功率似乎不受生成变量存在的影响。相反，当生成的解释变量应包含在结构方程中时，与非生成的解释变量的系数相关的试验规模就会大大扭曲。第三，使用事件研究和非事件研究数据说明了生成变量问题的各种解决方案。这些解决方案包括系统估计，使用White或new - west校正调整标准误差，以及工具变量估计。最后，对债券价格在事件研究中对日本数据的使用进行了调查。对于调查与新债券发行有关的任何错误定价的事件研究，发现场外报价更为合适。

The purpose of this study is to examine some of the econometric problems that arise in event studies which are typically used to investigate what impact a particular event has on the firm's share price or the price of bonds it has issued. First, the theoretical impact of including in the model of interest a variety of generated variables was analyzed. The asymptotic impact of using a generated variable on the consistency of estimators and on the standard errors of estimators is shown to depend among other things on the number of observations in the first and second stage regressions, and the type of generated regressor used. Solutions to these problems are carefully described. Second, the small sample impact of using some generated variables was analyzed using a Monte Carlo study. It was found that some instrumental variable estimators had smaller biases than ordinary least squares based estimators, but the sizes and powers of tests associated with the coefficient of the generated variable do not seem to be affected by the presence of the generated variable. In contrast, the sizes of tests associated with the coefficient of the non generated explanatory variable are considerably distorted when the generated variable should be included in the structural equation. Third, a variety of solutions to the generated variable problem are illustrated using event study and non-event study data. These solutions include system estimation, adjusting standard errors using White or Newey-West corrections, and instrumental variable estimation. Finally, the use of bond prices in event studies for Japanese data was investigated. For event studies investigating any misspricing associated with a new bond issue, over the counter quotations are found to be more appropriate.