Predicting Local-Official Corruption by Coupling Rational Choice Theory with XGBoost: A Computational Sociology Approach

DOI

10.2991/978-94-6239-828-3_31How to use a DOI?

Keywords

Computational Sociology; Corruption Risk Prediction; Gradient Boosting; Rational Choice Theory; Political Dynasties; Algorithmic Accountability

Abstract

Risk-based audit prioritisation requires a model that ranks officials before an investigation begins, not one that recognises officials already charged. This paper reformulates the prediction of local-official corruption in Indonesia as a strictly prospective task and evaluates it under leakage controls appropriate to that claim. We construct the Indonesian Corruption Corpus, an official-year panel of 545 governors, regents and mayors observed from 2015 to 2024, linking asset filings, budget realisations, electoral family records and adverse legal outcomes. Rational Choice Theory is translated into four observable proxies, including a Dynastic Insulation Index (DII) and Wealth Velocity. Features observed up to year t predict a first adverse legal outcome in t + 1 or t + 2; official-years at or after that outcome are censored, all preprocessing is fitted inside training folds, and officials are disjoint across training and hold-out windows. XGBoost attains a hold-out ROC-AUC of 0.71 (95% CI 0.63–0.78) and a PR-AUC of 0.18 against a 2.78% base rate, recovering 40.4% of prospective cases at a 10% audit capacity — a fourfold gain over random selection. The DII result survives five alternative specifications, but officials with any dynastic tie face roughly twice the false-positive rate. The system is defensible only as a contestable, human-reviewed prioritisation aid.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Fajerin Biabdillah
AU  - Arsan Kumala Jaya
PY  - 2026
DA  - 2026/11/04
TI  - Predicting Local-Official Corruption by Coupling Rational Choice Theory with XGBoost: A Computational Sociology Approach
BT  - Proceedings of the 2026 11th International Conference on Sustainable Information Engineering and Technology (SIET 2026)
PB  - Atlantis Press
SP  - 349
EP  - 356
SN  - 2589-4900
UR  - https://doi.org/10.2991/978-94-6239-828-3_31
DO  - 10.2991/978-94-6239-828-3_31
ID  - Biabdillah2026
ER  -