Consumer automobile loans dataset

A dataset containing the performance records of 58,118 consumer automobile loans with original terms of 72 to 73 months. The data was obtained from four publicly traded asset-backed securities (ABS) bonds. This dataset is subject to left-truncation (loans must survive long enough to be included in the ABS trust) and right-censoring (the trust may be unwound before all loans terminate), and it features competing risks of default and prepayment.

Usage

aloans

Format

aloans

A data frame with 58,118 rows and 7 columns:

risk_cat

Credit risk band based on the loan's Annual Percentage Rate (APR). Categories include "super_prime" (0-5%), "prime" (5-10%), "near_prime" (10-15%), "subprime" (15-20%), and "deep_subprime" (20%+).

Z

Time-to-event, representing the loan termination age in months.

Y

Left-truncation time, representing the loan age in months when the ABS trust began making payments to investors.

C

Right-censoring indicator (1 = right-censored, 0 = exact termination observed).

D

Default indicator for competing risks (1 = default, 0 = prepayment).

bond

Categorical variable specifying the source ABS bond: CARMX (CarMax Auto Owner Trust 2017-2), AART (Ally Auto Receivables Trust 2017-3), SDART (Santander Drive Auto Receivables Trust 2017-2), or DRIVE (Drive Auto Receivables Trust 2017-1).

Source

Data was originally compiled from the SEC's Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system. The replication data repository is available at https://github.com/jackson-lautier/credit-risk-convergence/.

References

Lautier, J. P., Pozdnyakov, V., & Yan, J. (2024). On the convergence of credit risk in current consumer automobile loans. Journal of the Royal Statistical Society Series A: Statistics in Society, qnae137. doi:10.1093/jrsssa/qnae137 .