Changfeng Tao陶长风

Database数据库

Data I have built for my research. Public versions will be posted here once they are ready.这里放我在研究中构建的数据，公开版本整理好后会发布在本页。

Exchange Comment Letter Database沪深交易所问询函数据库

In preparation整理中

A letter-level dataset of the comment letters issued by the Shanghai and Shenzhen Stock Exchanges and the corresponding firm replies, built for my doctoral dissertation. Every letter can be traced back to its source, its original text, and how it was matched.为博士论文自建的函件级数据集，覆盖沪深交易所发出的问询函与公司回函。每封函件都可以追溯到来源、原文与匹配方式。

  • 21,061letters封函件
  • 3,722firms家公司
  • 133,601questions个问题
  • 181letter-level measures个函件级指标
  • 94.8%of letters with replies函件有回复
Unit of observation观测单位
Letter, with question-level and firm-year layers函件，另有问题级和公司-年两层
Sample period样本期
January 2015 to December 20242015 年 1 月至 2024 年 12 月
Coverage覆盖范围
Inquiry letters, letters of concern, and work letters issued by the Shanghai and Shenzhen Stock Exchanges to firms listed on the Main Board, the STAR Market, and ChiNext沪深交易所发出的问询函、关注函与工作函，覆盖沪深主板、科创板、创业板
Sources数据来源
The CNRDS database, the exchanges' websites, and listed-company announcements on CNINFOCNRDS 数据库、交易所官网、巨潮资讯网上市公司公告
Text-based measures文本指标
Inquiry intensity, inquiry topics, textual features, reply characteristics, and reply stance问询强度、问询主题、文本特征、回函特征与回函立场
Related research相关研究
The Regulatory Effects of AI-Related Comment Letters人工智能主题问询函的监管效应研究
Public version公开版本
A de-identified public version will be posted on this page.脱敏后的公开版本将发布在本页。

Technical Challenges技术瓶颈

Building the database meant solving several problems: letters missing from the CNRDS database, long texts truncated by Excel, documents not issued by the exchanges, single letters spread across several announcements, and unstructured text.

解决了 CNRDS 数据库漏收部分函件、长文本被 Excel 截断、主体混杂、一函多文、非结构化等问题。

Flow chart in Chinese: three data sources (CNRDS, the exchanges' websites, and CNINFO announcements) feed a cleaning and merging step that addresses four technical challenges and produces a letter-level master table, question-level and letter-level measures, and a firm-year panel.
Figure 1. How the database is built (in Chinese).图 1　问询函数据底座的构建流程

Outputs产出

Stacked bar chart in Chinese of letters per year from 2015 to 2024 by source (CNRDS, exchange websites, CNINFO reconstruction), with 21,061 letters, 3,722 firms, and 181 letter-level measures built on 133,601 questions.
Figure 2. Letters by year and source, and key outputs, January 2015 to December 2024 (in Chinese).图 2　各年函件数（按来源）与主要产出，2015 年 1 月至 2024 年 12 月