Replication code for:
"The Dark Side of Connectivity: How Board Interlock Networks Are Associated with Accounting Risk Through Director Mobility"
Jihwan Woo and Nari Kim
Journal of Business Research (2026, submitted)
We track how director mobility creates and dissolves board interlock edges over 20 years (2004–2023), and test whether restatement risk co-moves through these temporal connections. Using 561,306 director events and 154,920 edge lifecycles from WRDS, we find that firms receiving directors from restatement companies face significantly higher subsequent restatement risk — an effect that survives multiple identification strategies.
Board interlock networks at three time points. Red = restatement firms; blue = clean firms. The network grows denser over time, and restatement firms are distributed throughout.
Top: Restatement rates for exposed (red) vs. clean (blue) groups. Bottom: Difference in percentage points. The contagion effect peaks in 2010–2014 and declines thereafter.
Effect size decreases monotonically as controls become more stringent, while statistical significance is maintained — characteristic of a robust finding.
| Hypothesis | Finding | Effect Size | p-value |
|---|---|---|---|
| H1 | Temporal contagion exists | +7.0 pp (exposed vs clean) | < 0.001 |
| H2 | Dose-response relationship | +0.25 pp per exposed edge | < 0.001 |
| H3 | Contagion declines over time | +12.9 pp (2010–14) → +2.1 pp (2020–23) | < 0.001 |
| H4 | Contaminated edges dissolve faster | −0.95 years | < 0.001 |
| H5 | Cross-community edges transmit more | +3.7 pp (cross vs within) | < 0.001 |
| Test | Result |
|---|---|
| Strict fraud (SEC investigation / fraud) | Significant (p < 0.001) |
| Multiple contagion windows (0, 1, 2, 3 yr) | All significant |
| Non-executive directors | Strongest effect (+9.9 pp) |
| Same-auditor control | Effect persists for different-auditor pairs (+4.2 pp) |
| Industry × Year FE | Key variable remains significant (t = 3.32) |
| Logit specification | Marginal effect confirms LPM |
All data are sourced from Wharton Research Data Services (WRDS) and require an institutional subscription. Data are not included in this repository due to licensing restrictions.
| Dataset | WRDS Table | Records |
|---|---|---|
| Directors & Officers | audit.feed17_director_and_officer_changes |
561,306 |
| Restatements | audit.feed39_financial_restatements |
28,662 events |
| Auditor-Company | audit.feed01_audit_opinions |
71,480 |
| Compustat Financials | comp.funda |
48,110 firm-years |
pip install wrds pandas numpy networkx scipy matplotlib# 1. Download data from WRDS (requires credentials)
python code/00_download_wrds_data.py
# 2. Build yearly interlock networks and edge lifecycles
python code/01_build_edge_lifecycles.py
# 3. Temporal contagion analysis (H1, H3, H4)
python code/02_temporal_contagion.py
# 4. Firm FE regression and matched counterfactual (H2)
python code/03_causal_analysis.py
# 5. Robustness tests
python code/04_robustness.pyAll scripts include checkpoint functionality — if interrupted, re-running resumes from the last completed step.
Results are saved as JSON files in results/:
| File | Contents |
|---|---|
temporal_network_results.json |
H1, H3, H4 results |
causal_analysis_results.json |
H2 (Firm FE + matching) |
robustness_results.json |
Strict fraud, windows, Industry×Year FE |
community_detection_results.json |
H5 (community structure) |
reviewer_response_results.json |
Director roles, same-auditor, restatement types |
dynamic_network_results.json |
Edge birth analysis |
Code: MIT License
Data: Not included. Available from WRDS with institutional subscription.
@article{woo2026darkside,
title={The Dark Side of Connectivity: How Board Interlock Networks Are Associated
with Accounting Risk Through Director Mobility},
author={Woo, Jihwan and Kim, Nari},
journal={Journal of Business Research},
year={2026},
doi={10.5281/zenodo.18872817},
note={Submitted}
}