一类协整模型的分位数回归及应用

Quantile Regression and Application for a Class of Cointegration Models

协整理论在非平稳时间序列分析中已经得到了快速发展和广泛应用.然而,这些理论大多数都是建立在非稳健的普通最小二乘框架下.本文考虑带有平稳协变量和线性时间趋势项协整模型的稳健估计和应用.采用分位数回归方法,给出了模型的估计步骤并且得到了估计的渐近分布.同时,推导出一个完全修正的分位数估计,用来消除序列相关和长期内生性的影响.从稳健性和精确性两个方面,使用Monte Carlo模拟对模型估计的有限样本性质进行了检验.进一步,模型和回归方法被应用于两个经济实证研究中,所得结论与经济理论相一致.

Cointegration theories have obtained great development and wide applications in nonstationary time series analysis.However,most of these theories are established in the setting of non-robust ordinary least squares framework.This paper considers the robust estimation and application for a class of cointegration models with stationary covariates and a linear time trend.By using quantile regression technology,the estimation procedures and asymptotic distributions of those models are obtained.In addition,a fully modified quantile estimator is adopted to eliminate the influences of serial correlation and long-run endogeneity.Monte Carlo simulations are performed to assess the finite sample properties of the suggested estimators from two aspects of robustness and accuracy.Furthermore,the cointegration models and the quantile technology are applied to two empirical studies.The empirical conclusions are consistent with economic theories.