摘要
财政分权体制下的地方政府支出竞争行为已成为我国学界关注的热点。当前中国正处在由追求简单经济增长向追求社会民生转变的阶段,提供教育公共服务则是保障社会民生的重要方面。本文研究了我国地市级地方政府可能存在的教育支出相互竞争行为。在理论部分本文提出一个简单的、中国式的"自上而下"标尺竞争模型来解释地市级政府之间教育支出的相互影响机制:上级政府通过比较教育支出来评价地市级政府的相对教育政绩甚至决定地方官员是否连任,从而引发了地市级政府之间的标尺竞争。在文章实证部分使用2007年至2010年312个地级市教育支出等数据构建空间面板数据模型,并在固定效应框架下运用GS2SLS和MLE方法实证估计了地市级地方政府教育支出反应方程。实证结果符合理论模型的预期:地市级政府在教育支出上确实存在标尺竞争,经济距离相近地区之间出于标尺竞争的模仿激励要高于地理相邻地区。此外本文还发现,以财政自主度衡量的财政分权显著地减少了地方政府的公共教育供给。
This paper focuses on the local city government's ( local city is an administration level just below provincial level) behavior on the education input. We mainly explore whether or not the interaction exists on education expenditure among local governments within the same province. In theoretical part of this paper, we give a simple theoretical mode/of Chinese- Style yardstick competition "from the top" to explain why there may exist the interactions among local governments over the education expenditure. Based on this theoretical model, this paper further uses the education expenditure data of 312 local cities in China from 2007 to 2010 to construct a spatial panel data model. And we apply GS2SLS and MLE methods under the fixed-effects framework to estimate the parameter of the reaction equation for education expenditure. The empirical results are consistent with our theoretical prediction; that is, there actually exists a yardstick competition among local city level governments over the education expenditure. The interesting thing is that interaction among the neighbors defined by similar economic background is more significant than that among the geographical neighbors. In addition, we find that, for the local city level governments, the more of fiscal decentralization ( measured by fiscal autonomy index) , the less of incentive for more input in public education.
出处
《经济研究》
CSSCI
北大核心
2013年第11期127-139,160,共14页
Economic Research Journal
基金
国家自然科学基金(70871073,71171073)
上海财经大学“211工程”四期重点学科建设项目的资助
关键词
地市级政府
教育支出
标尺竞争
空间面板数据模型
Local Municipal Governments
Education Expenditure
Yardstick Competition
Spatial Panel Data Model