传统均值角度下研究的动态面板数据模型会受经典假设条件的约束,将动态面板数据与分位回归数模型相结合,不仅可以解决约束问题,而且能更加全面地描述响应变量条件分布的全貌.文章引入自适应惩罚项,并应用工具变量构造了自适应惩罚的动态面板分位回归方法,证明了该方法得到的估计量具有大样本性质.同时蒙特卡洛模拟结果表明自适应惩罚的方法相较于传统的方法更加有效.文章最后对中国大中城市商品房销售价格与各地人均国民生产总值的关系进行案例分析,发现两者之间存在正反馈机制.

The study of dynamic panel data is mainly based on the conditional mean regression methods, but constrained by classical assumptions. The quantile regression model for dynamic panel data considered, not only solves the problem with constraint, but also fully describes the conditional distribution of the response variables. This paper presents a method to study dynamic panel data with instrument variables based on adaptive penalty quantile regression, and proves that the proposed estimator has large sample properties. Monte Carlo simulation study shows that the proposed method is better than traditional methods. In the end, the paper analyzes the case study on the relationship between the selling price of commercial houses and per capita GDP in large and medium cities of China, and founds that there is a positive feedback mechanism between them.