Datasets:

Changahou
/

CSBA

license: cc-by-4.0
language:
  - en
  - zh
tags:
  - accounting
  - corporate-governance
  - china
  - supervisory-board
  - audit-committee
  - a-share
  - cninfo
pretty_name: CSBA - China supervisory board abolition
size_categories:
  - 1K<n<10K
configs:
  - config_name: transitions
    data_files: 04_abolition_transitions.csv
  - config_name: announcements
    data_files: 03_abolition_announcements.csv
  - config_name: approval_dates
    data_files: 05_approval_dates.csv
  - config_name: sample_flags
    data_files: 06_sample_flags.csv

CSBA — Board reform and the abolition of the supervisory board

Replication data for the paper "Board reform and the abolition of the supervisory board: evidence from China's 2025 governance wave" by Xiangdong Liu (Jinan University, ORCID 0000-0002-9438-9285), Jiahao Chen (Jinan University, ORCID 0000-0003-4673-1311), Zhenyan Liu (University of Manchester, ORCID 0009-0007-5672-2060) and Jinfeng Pan (Guangdong University of Finance, ORCID 0009-0008-1950-7465).

Code and documentation: https://github.com/ChenJiahaoJNU/CSBA

What this is

China's revised Company Law, effective 1 July 2024, lets a listed firm abolish its supervisory board and vest the statutory oversight functions in the audit committee of the board. No existing database records the reform, so the transition file was reconstructed from public CNINFO announcements. Within a single year 2,482 A-share firms announced an abolition.

Files

file	rows	what it is
`01_cninfo_raw_hits.csv`	2,674	raw CNINFO search results before classification
`02_classified_candidates.csv`	2,431	hits classified as abolition-related
`03_abolition_announcements.csv`	2,522	the firm's own abolition announcements
`04_abolition_transitions.csv`	2,482	firm level, first and last announcement, cohort
`05_approval_dates.csv`	2,482	board proposal date and shareholder approval date
`06_sample_flags.csv`	4,546	the risk set, public columns only

`04_abolition_transitions.csv`

One row per firm that announced an abolition.

column	meaning
`secCode`	6-digit stock code
`secName`	firm name as printed in the filing
`first_ann`, `last_ann`	first and last abolition announcement dates
`n_ann`	number of abolition announcements by that firm
`has_charter`	whether the announcement also amends the articles
`cohort_month`, `cohort_quarter`, `cohort_year`	derived from `first_ann`

`05_approval_dates.csv`

The sweep behind the paper's measurement-error section.

column	meaning
`code`	6-digit stock code
`ann`	first abolition announcement, which is the board's proposal
`meet`	general meeting date, from the firm's own resolution announcement
`n_res`	resolution announcements found in the 180-day window
`status`	`ok`, `no_date`, `no_orgid`, `query_fail`, `worker_error`

2,377 of 2,482 resolved. Median gap 102 days, interquartile range 21 to 147. Only 35.7 per cent of firms vote in the same calendar quarter as their proposal, which is why the paper makes no timing claim.

`06_sample_flags.csv`

The 4,546-firm risk set with the licensed covariates stripped out. 2,072 firms abolished, or 45.6 per cent, which is the figure the paper reports.

Licence and what is deliberately absent

Everything here was reconstructed from public CNINFO disclosures. A-share filings are public documents, and this data is released under CC BY 4.0.

The paper's explanatory variables come from a commercial vendor, whose licence permits use for replication but not redistribution. They are therefore not included. That covers return on assets, leverage, total assets, institutional ownership, information asymmetry, risk-disclosure intensity, the CNRDS ESG rating, MD&A text similarity, analyst coverage, stock-price crash risk and audit opinion. Obtain them under your own licence; the pipeline in the code repository reads them from a vendor/ folder.

You can reproduce the transition file, the sample construction counts, the cohort distribution and the approval-date analysis from this data alone.

Known limitations, stated by the authors

The treatment date is the board's proposal, not the shareholder vote. Across all 2,482 firms the two differ by a median of 102 days.
The approval-date extraction is automated and unvalidated by hand. Nine of 2,377 resolved firms (0.4 per cent) yield an impossible negative gap, so the file is sound distributionally and not firm-by-firm. Do not quote an individual firm's gap without checking that firm's filing.
Keyword recall is imperfect. A single-phrase query would have missed 193 genuine announcements; this release uses a nine-keyword union.
The information-asymmetry result does not reproduce with alternative proxies (crash risk, analyst coverage, |ROA − CFO|). The paper reports this as a failure.

Citation

If you use this data, please cite the paper, and this dataset:

Liu, X., Chen, J., Liu, Z., and Pan, J. Board reform and the abolition of the
supervisory board: evidence from China's 2025 governance wave. China Journal of
Accounting Research, forthcoming.

@misc{csba2026,
  title  = {Board reform and the abolition of the supervisory board: evidence from
            China's 2025 governance wave},
  author = {Liu, Xiangdong and Chen, Jiahao and Liu, Zhenyan and Pan, Jinfeng},
  year   = {2026},
  url    = {https://huggingface.co/datasets/Changahou/CSBA}
}