一种协整时间序列的动态因果效应估计与推断方法

An Approach of Causal Effect Evaluation and Inference for Co-integration Time Series

为了进一步规避经典面板数据模型因果推断方法中对照样本异质性的问题,以及中断时间序列方法可能存在的"伪回归"问题和"二阶矩平稳"等条件的限制,本文借助时间序列的协整理论,并结合误差修正模型和结构突变检验方法提出了一种基于协整时间序列的实验设计方法,扩展了政策因果效应估计与推断方法。该方法将传统的趋势平稳过程中断时间序列扩展到随机趋势过程,不仅可以推断和估计处理效应,而且解决了政策效应的暂时性和持续性识别问题;另外,协整时间序列的因果效应推断方法更适用于不存在有效对照样本的政策因果效应分析。此外,本文评估了上海市和重庆市的房地产税试点政策,研究发现,长期来看试点政策对两市的商品房价格增长具有抑制作用,但政策效应表现出显著的差异性,上海市的试点政策对抑制房价上涨的力度大、起效快速,但重庆市的房产税政策存在较长期的时滞。

In order to avoid the bias of the heterogeneity of control samples in the causality inference method based on the classic panel data model and the possible"pseudo-regression"of interruption time series(ITS)method,and to relax the restriction on the result variable"second-moment stability",the time series can be used to estimate and infer the policy causality.Using the co-integration theory of time series,this paper proposes a causal effect inference method for co-integration time series based on the error correction model and tests of structural breakpoint.This method expands ITS from the traditional trend-stationary process to a stochastic trend process,can not only infer and evaluate the policy treatment effect,but also solve the identification problem of temporary and continuous policy effect,and the method is more suitable for the analysis of policy causal effect in the absence of effective control samples.In addition,the empirical analysis of Shanghai and Chongqing pilots shows that the property tax pilot policy has a long-term inhibitive effect on the price growth of commercial housing,and the policy effect of the two cities shows a significant difference—the effect on Shanghai is strong and quick,while the effect of Chongqing’s policy has a longer time lag.