​Segmenting Political Connection and Earnings Management: A Clustering and Regression Approach in IDX Firms

Abstract

This study aimed to segment 140 non-financial companies listed on the Indonesia Stock Exchange (IDX) for fiscal year 2024 based on political connections and earnings management, and to test the consistency of this segmentation through Ordinary Least Squares (OLS) regression. Political connection was measured through five indicators, while earnings management was measured through discretionary accruals and two real-activities indicators, with normal-model coefficients estimated on a cross-section of 852 IDX-listed firms, Exploratory Factor Analysis (EFA) and k-means clustering were then applied to the resulting factor scores, followed by ANOVA/Kruskal-Wallis tests and OLS regression. Exploratory factor analysis yielded a robust two-factor structure (Kaiser-Meyer-Olkin = 0.798, cumulative variance 72.6%), and k-means clustering produced a statistically optimal two-cluster solution (Silhouette = 0.616) that separated the sample purely by political connection, with no significant earnings-management difference (p>0.49), a three-cluster solution revealed one bootstrap-stable cluster with a significant income-increasing pattern (p<0.001). The findings were more consistent with the political shelter hypothesis than with political capture at the cluster profile and ANOVA/Kruskal-Wallis levels, although a robustness check using clusters formed from the political-connection factor alone found no significant partial association, although the single-year cross-sectional design limited interpretation to association rather than causation. This study contributed methodologically by cross-validating the factor structure and complementing cluster segmentation with regression evidence.

Keywords

Earnings Management Exploratory Factor Analysis K-Means Clustering OLS Regression Political Connection