Investor Horizons and the Timing of Leverage Adjustment

Published: 3 November 2026| Version 2 | DOI: 10.17632/fjkn2z4y63.2
Contributors:
,
,
,

Description

This replication package accompanies “Investor Horizons and the Timing of Leverage Adjustment” (Economic Modelling, manuscript ECMODE-D-26-00470R1). The study examines whether the composition of long-horizon mutual-fund holdings is associated with the timing of firms’ leverage adjustment. Using Chinese A-share firms observed from 2015 to 2023, the analysis combines firm financial and market data, institutional shareholding disclosures, fund portfolio reports, and reported fund purchases and sales. The paper measures long-horizon fund ownership from disclosed fund positions and four preceding half-years of portfolio turnover. It studies both the annual response of firm leverage to a pre-existing leverage gap and the timing of fund position changes within the year. The preferred firm-level sample contains 15,908 firm-year observations for 4,183 firms. The archive contains the complete Python and Stata analysis code, LaTeX source files, figures, tables, aggregate results, audit files, variable documentation, software requirements, and a consolidated README. It also contains synthetic replacement files for the restricted input data. These files preserve the documented field structures, English variable labels, identifier formats, relational roles, missing-value conventions, and approximate source moments. They are synthetic observations and were not used to produce the published estimates. The firm-account, market, institutional-holding, and fund-report inputs are Guotai An China Stock Market & Accounting Research (CSMAR) data. The licensed raw files, fund-stock position records, and row-level analysis panels are not redistributed because they are subject to the provider’s data-use terms. Users with authorised access to the relevant CSMAR modules can reconstruct the published estimates by following the source map and processing sequence in README.md. Users without licensed access can validate the public data schemas, inspect the documented construction, run the verification scripts, and compile the manuscript and Online Appendix.

Files

Steps to reproduce

1. Download and unzip the replication package, then read the single `README.md`, which documents the files, variables, source modules, restrictions, software, run order, and table/figure scripts. 2. Public verification requires Python 3.11+: python3 revision/code/build_synthetic_replication_data.py python3 revision/code/validate_synthetic_replication.py bash revision/code/run_public_check.sh This validates the synthetic schemas and compiles the main manuscript and Online Appendix. Synthetic files verify data structures and code paths only; they cannot reproduce the published estimates. 3. Exact reconstruction requires authorised Guotai An (CSMAR) access to: - listed-company financial and market files; - Institutional Investor Shareholding Detail, retaining institution-category code P9401; - Fund Portfolio Stock Purchases and Sales, `FUND_PTF_BSSTOCKTOTAL`; - public-fund master, net asset value, manager-history, listed-fund-share, and portfolio-history files for identifier matching. Record the data vintage, table names, field dictionary, download date, reporting thresholds, and licence terms. Place authorised files under `data/licensed`, or set `LICENSED_DATA_DIR` to an authorised local directory. 4. From the package root, run: python3 revision/code/rebuild_raw_fund_horizon.py python3 revision/code/build_rebuilt_horizon_analysis_panel.py python3 revision/code/run_rebuilt_horizon_flexible_speed_audit.py stata-mp -b do revision/code/run_rebuilt_horizon_composition_wild_bootstrap.do python3 revision/code/run_rebuilt_horizon_local_projections.py stata-mp -b do revision/code/run_rebuilt_horizon_local_projection_webb.do python3 revision/code/fund_exit_gap_loo_gate.py python3 revision/code/run_reported_trading_horizon_exit_timing.py python3 revision/code/rebuilt_horizon_debt_action_decomposition.py python3 revision/code/build_manuscript_exhibits.py python3 revision/code/build_online_appendix_exhibits.py 5. Compile the documents: cd revision/manuscript latexmk -pdf -interaction=nonstopmode -halt-on-error main.tex latexmk -pdf -interaction=nonstopmode -halt-on-error online_appendix.tex 6. Compare generated outputs with `revision/results/` and `revision/manuscript/`. Stata 18+, `reghdfe`, and `boottest` are required for Stata steps. Licensed CSMAR files and row-level derivatives must not be redistributed. Synthetic files are for public verification only.

Institutions

Categories

Economics, Finance, Corporate Finance

Funders

Licence