摘要
为解决面板数据分析中的异质效应和门限效应,文章提出了一个新的经济计量模型:面板数据门限分位数回归模型(PTQR)。PTQR模型不仅能够测度不同分位点处的门限效应,而且能够揭示解释变量对响应变量整个分布的非线性影响。通过数值模拟,将其与面板数据门限均值回归模型进行比较,发现PTQR模型极大地提高了估计和预测的准度与精度。将PTQR模型应用于政府干预与县域金融发展关系研究,结果表明:政府干预对县域金融发展具有促进作用,且县域金融发展水平越高,促进作用越强;政府干预对县域金融发展的影响呈现非线性特征,即较少的政府干预具有较大的正向边际影响,而较多的政府干预会导致边际效应递减,但县域金融发展水平越高,边际效应递减的幅度越低。
In order to solve the heterogeneity effect and threshold effect in panel data analysis,this paper proposes a new econometric model:panel data threshold quantile regression model(PTQR),which can not only measure the threshold effect at different quantiles,but also reveal the nonlinear influence of explanatory variables to whole distribution of response variables.Through numerical simulation,the paper makes a comparison with the panel data threshold mean regression model,finding that the PTQR model greatly improves the accuracy and precision of the estimation and prediction.Finally,the paper uses the PTQR model to analyze the relationship between government intervention and county-level financial development.The results show that government intervention has a promoting effect on county-level financial development,and the higher the level of county-level financial development,the stronger the promoting effect,that the influence of government intervention on the financial development of county shows nonlinear characteristics,that is,less government intervention has a greater positive marginal impact,and more government intervention will lead to a diminishing marginal effect,but the higher the level of county financial development,the lower the extent of diminishing marginal effect.
作者
蔡超
冯林
薛伟
Cai Chao;Feng Lin;Xue Wei(School of Statistics,Shandong Technology and Business University,Yantai Shandong 264005,China;School of Finance,Shandong University of Finance and Economics,Jinan 250000,China)
出处
《统计与决策》
CSSCI
北大核心
2021年第7期46-51,共6页
Statistics & Decision
基金
国家自然科学基金资助项目(71303139)
山东省社会科学规划项目(18DTJJ01)
关键词
面板数据门限分位数回归
县域金融
政府干预
threshold quantile regression of panel data
county-level finance
government intervention
