Dispelling ESG Investing Risk Misconceptions

This replication package provides the synthetic data, code, and instructions required to reproduce the main empirical results of the above-mentioned research paper. The package is designed to facilitate transparency and enable independent verification of all statistical findings reported in the study.
The empirical analysis is organized into three main blocks:
1 – Computation of Risk Metrics (R)
Risk measures are computed from daily fund return data obtained from data sources. The daily returns are first processed to construct fund-level risk metrics. These measures are then aggregated and transformed to a semester frequency.
All data processing and risk metric construction for this step are implemented in R. The corresponding scripts and intermediate datasets are provided in the R replication package.
2 – Data Merging and Empirical Analysis (Stata)
The semester-level dataset constructed in Step 1 is merged with fund-level information from Morningstar funds.
The main empirical analysis is conducted in Stata using the merged dataset. All Stata do-files required to replicate the empirical results are included in the Stata replication package.
3 – Random Forest Model Analysis(R)
A Random Forest model is estimated in R using the merged and cleaned dataset.
All scripts necessary to replicate the machine learning analysis are included in the R replication package.

本复现包提供了复现上述研究论文主要实证结果所需的合成数据、代码与操作指南。本包旨在提升研究透明度，支持独立验证该研究报告的全部统计发现。

本次实证分析分为三大核心模块：
1. 风险指标（Risk Metrics）计算（基于R语言）
风险指标的计算基于从数据源获取的每日基金收益率数据：首先对每日收益率进行预处理以构建基金层面的风险指标，随后将这些指标聚合并转换为半年度频率。本步骤的全部数据处理与风险指标构建工作均通过R语言实现，相关脚本与中间数据集已包含于R语言复现包中。
2. 数据合并与实证分析（基于Stata）
将步骤1构建的半年度层级数据集与来自晨星（Morningstar）基金的基金层面信息进行合并。本次核心实证分析采用合并后的数据集，通过Stata软件完成。所有用于复现实证结果的Stata do文件均已包含于Stata复现包中。
3. 随机森林（Random Forest）模型分析（基于R语言）
基于已合并与清洗后的数据集，通过R语言构建随机森林模型并完成参数估计。所有用于复现该机器学习分析的脚本均已包含于R语言复现包中。