一类面板模型中部分结构变点的检测和估计

Detection and estimation of partial structural change in a class of panel models

给出了解决面板回归模型中存在未知部分结构变点时的检测、估计和推断问题的新方法,得到了结构变点的相合估计及渐近分布。基于最小二乘方法构造全局最小残差平方和,并用于检测结构变点。该方法适用于纯结构变点模型和部分结构变点模型,并且即使面板数据只有很少的个体,估计的相合性也可保证。Monte Carlo模拟结果验证了理论的正确性。基于中国股票市场上参与股改的404家工业企业的实证研究表明:应首先检测结构变点是否存在,然后才能分析解释数据;忽略检测结构变点会产生有偏估计,甚至出现伪回归。

A new method for detecting and estimating partial structural change at an unknown common time in random effect panel regression model was considered. And the rate of convergence and asymptotic distribution of structural change estimator was derived. The least squares estimator was constructed as global minimizers of the sum of squared residuals in order to identify structural changes. The structural changes can be well estimated even in the short panel data. Monte Carlo simulations were carried out to verify the theoretical result numerically. Empirical study with 404 China enterprises which participate in share reform showed that structural change exist or not should be detected firstly, and then data analysis and explanation is meaningful. Bias estimator or spurious regression would appear if neglect to detect structural changes.