Predictive Modeling of Household Credit Risk and Fear of Denial: A High-Dimensional Analysis Using PCA and XGBoost on the 2022 Survey of Consumer Finances (SCF)

This study employed machine learning methods to predict household credit access concerns using comprehensive financial and demographic data from the 2022 Survey of Consumer Finances (SCF), analyzing 4,595 households to examine which characteristics predict whether families have been turned down for credit or feared credit denial in the past five years, a critical measure of financial vulnerability affecting approximately 84% of surveyed households. Two classification models were developed and compared: an XGBoost gradient boosting model and a logistic regression model, both using 263 principal components derived from the original feature space. The XGBoost model (Model A) achieved exceptional predictive performance (AUC = 0.9885, accuracy = 96.65%, precision = 97.52%, recall = 82.01%), substantially outperforming the logistic regression model (Model B: AUC = 0.7955, accuracy = 80.34%, precision = 44.93%, recall = 78.37%), demonstrating that credit access concerns follow highly systematic patterns. Feature importance analysis revealed that asset-based financial indicators dominated predictions, with Equity to Income, Homeownership, Credit application history, Emergency Savings, and Leverage Ratios emerging as the top five predictors, while behavioral and historical factors particularly payment delinquencies and prior credit experiences exhibited substantial importance, supporting path-dependent theories of financial exclusion. Race ranked 14th among predictors, suggesting that observed disparities operate substantially through differential economic circumstances, though structural barriers persist. The findings have immediate implications for financial institutions, which can deploy similar predictive models to identify at-risk customers for targeted financial counseling, develop alternative credit scoring approaches accounting for asset ownership and emergency savings, and design early warning systems flagging households with declining liquid assets or increasing leverage ratios for preventive assistance. Financial advisors can use these insights to prioritize asset-building strategies over simple income increases, emphasize emergency savings establishment as critical for credit access maintenance, counsel clients on credit history management following past delinquencies, and recognize that payment delinquencies create enduring barriers requiring proactive rehabilitation.