Heterogeneity of automobile consumer finance propensity in China: new evidence from explainable machine learning

Abstract

Purpose The purpose of this paper is to study the differentiated sources of consumer credit propensity through real transaction record data. Design/methodology/approach This study involves innovative use of automobile finance down payment ratio (DPR) to quantify consumer credit tendency. Based on the real transaction data of an auto finance company in China, random forest (RF) and gradient boosting regression (GBR) were constructed. Feature importance (FI) and partial dependence plot (PDP) were used to test the main factors leading to the difference of consumers’ preference in credit choice. Findings The fitting of RF and GBR is better than ordinary least squares (OLS). FI and PDP show that young people generally have low assets, while customers with weak loan repayment ability have great debt pressure, and their willingness to choose credit is relatively low to ensure the quality of life. Customers with mortgage loans will be more inclined to accept credit products, and high commodity prices will raise the current financial burden, thus enhancing the choice tendency. Besides, stress and hedonic mechanism are obtained by FI, and the hedonic mechanism does not play a completely positive role in the western region and the middle-aged group. Originality/value This paper quantifies consumers’ credit tendency through the real transaction records, which makes up for the subjective bias caused by questionnaire survey data. With the help of machine learning and interpretability tools, we re-identify the source of heterogeneity of automotive consumers’ credit propensity.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 15 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.