Replication data and code for: Analyst Individualism and the Readability of Investment Research: Evidence from China

Description

Replication materials for "Analyst Individualism and the Readability of Investment Research: Evidence from China". The public archive contains four derived analysis-ready datasets, Stata estimation code, variable dictionaries, table-to-data mappings and reference results. Analyst names, report titles and bodies, emails and URLs are excluded. Analyst and source row/event identifiers are replaced with package-local labels; no identity crosswalk is distributed. Original named research inputs are retained separately by the authors and are not part of this public archive. The main full-model readability sample contains 60,326 observations from 2007–2022. Text and market analyses use their documented subsamples. The later market design contains 5,120 events, including 612 isolated events, from 2009–2022. Table 11 uses an earlier event design; Tables 12–13 use the later two-component readability design. See the README and table mapping for these distinctions. A Stata rerun of 44 models matched 679 available reference coefficient rows, standard errors and p values within 0.000002. The package also documents the check following identifier replacement. Optional selection-weighted models are supplied separately and were not part of the 44-model rerun. Construction scripts are provided as references and require upstream files and path configuration. The principal data sources are CNRDS and CSMAR; name-character frequency references are derived from ChineseNames. The depositor has confirmed permission to redistribute the research data and code.

Steps to reproduce

1. Extract Replication_Data_and_Code.zip. 2. Install Stata 18 and reghdfe with its dependencies. 3. Set Stata’s working directory to the extracted Replication_Data_and_Code folder. 4. Run: do code/run_models.do 5. Compare verification/coefficients.csv with the reference results and documentation. 6. The optional code/run_ipw_sensitivity.do estimates selection-weighted text models separately. See README.txt for sample periods, identifier handling and source-data access.

Institutions

• University of Macau

  Macau, Macao

Categories

Licence