摘要
价值不断提升的政府网站内容数据不仅可以描绘政策注意力,也为中央政策向地方层级扩散的测量与评估提供了新的机遇.在我国多层级政府组织治理模式下,地方政府对中央政策的贯彻落地是政策生效的前提条件.对纵向政策扩散的有效测量和评估将有助于理解政策扩散机制,提升政策落地效果.本文基于全国省、市级政府门户网站每日内容更新数据,通过概率主题建模方法建构主题概率矩阵,刻画政府对不同主题的注意力分配差异,并基于概率主题建模结果构建函数测量地方政府对中央政策的扩散速度与扩散程度.本文讨论了测度建构的原理和细节,并引入机器学习方法进行鲁棒性检验,通过多政策主题扩散的混合回归分析了影响短周期政策层级扩散的因素.研究以测度建构为突破口打通文本数据挖掘到有价值公共管理知识的“中间层”,对政策信息学在政策扩散及评估监测中的应用前景进行了初步探索.
The increasing significance of government website content data not only depicts policy attention,but also provides new opportunities for measuring and evaluating the diffusion of central governmental policies to the local government.For China's multi-level governmental organization and governance mode,the implementation of central government policies by local governments is a prerequisite for the policies to take effect.Effective measurement and evaluation of vertical policy diffusion will help understand the policy diffusion mechanism and enhance the effectiveness of policy implementation.Based on content data updated daily on government portals at provincial and municipal levels,this article constructs a topic probability matrix of government website content by probabilistic topic modeling method,characterizes the difference in the distribution of government attention to different topics,and constructs functions based on the results of probabilistic topic modeling to measure the diffusion speed and diffusion degree of local governments to central policies.This article discusses the rationale and details of the measurement,introduces a machine learning approach for robustness testing,and analyzes the factors affecting the hierarchical diffusion of short-term policy through a mixed regression of the diffusion of multiple policy themes.The study bridges the“middle layer”from textual data mining to valuable public management knowledge,using measurement constructs as a breakthrough,and provides a preliminary exploration of the application of policy informatics in policy diffusion,evaluation and monitoring.
作者
张楠
黄梅银
罗亚
马宝君
ZHANG Nan;HUANG Mei-yin;LUO Ya;MA Bao-jun(School of Public Policy and Management,Tsinghua University,Beijing 100084,China;Laboratory of Computational Social Sciences and State Governance,Tsinghua University,Beijing 100084,China;Nankai Business School,Nankai University,Tianjin 300071,China;Shanghai Key Laboratory of Brain-Machine Intelligence for Information Behavior,Shanghai 201620,China;School of Business and Management,Shanghai International Studies University,Shanghai 200083,China)
出处
《管理科学学报》
CSSCI
CSCD
北大核心
2023年第5期154-173,共20页
Journal of Management Sciences in China
基金
国家自然科学基金资助项目(91646103,72293571,71974111,72172092)。
关键词
政策信息学
概率主题建模
注意力分配
政策扩散
policy informatics
probabilistic topic modeling
attention distribution
policy diffusion
