Detecting anomalous matches in professional sports: a novel approach using advanced anomaly detection techniques
Dr Dulani Jayasuriya, Business and Economy, Business School; Mohamed Ayaz, PhD candidate, Department of Civil and Environmental Engineering; Michael Williams, PricewaterhouseCoopers, Auckland, New Zealand
Abstract
This study investigates the increasing use of digital footprints (DFs) by lenders in the U.S. residential mortgage market. From 2013 to 2018, the proportion of lenders utilizing borrowers’ DFs rose significantly from 6% to 34%. The research demonstrates that DFs can considerably reduce lenders’ overall risk and mitigate discriminatory lending practices, providing a substantial informational advantage over traditional lending methods.
Introduction
With the rapid growth of the Internet, the utilization of collected data, including digital footprints (DFs), has become a critical issue in academic and political debates. This study is set against the backdrop of increasing internet usage, where an estimated 89% of American adults are active online. The research aims to explore the role of DFs in the financial sector, particularly in the U.S. residential mortgage market, where FinTech lenders have been at the forefront of incorporating technology into the loan approval process.
Contribution to literature
The research contributes to various areas, including FinTech lending, shadow banking, financial intermediaries, information asymmetry, consumer behavior prediction, residential mortgages, and discriminatory lending. It is the first study to focus specifically on the impact of DF users on riskiness, discriminatory lending, and borrower characteristics in the U.S. residential mortgage market.
Number of DF users in US residential mortgage market.
Methodology and hypotheses
The study is structured around several main research questions, focusing on the informational advantages of DF usage and its implications on borrower riskiness, market segmentation, and discriminatory lending practices.
Hypothesis 1A: DF use is negatively associated with default rates. The study employs logistic regression to test whether DF users have superior knowledge in selecting borrowers, potentially reducing default risks.
where Defaultijst is equal to one if the payments for mortgage i by lender j in state s originated in time t ever become delinquent for 90 days or longer over our observation period (Fuster et al., 2019).
Hypothesis 1B: DF use is negatively associated with prepayments. This hypothesis is tested using logistic regression to determine if DF users select borrowers who minimize overall risk, including prepayment risk.
where Prepaymentijst is an indicator equal to one if the mortgage i by lender j in state s originated in time t had a zero balance (i.e., was closed) before its official maturity date. DFjt indicates whether the lender is classified as a DF user.
Hypothesis 1C: DF use is negatively associated with interest rates. The study uses OLS regression to assess whether DF usage impacts the interest rates charged to borrowers.
where Rateijst is the interest rate (in percentage terms; i.e., the rate multiplied by 100) indicated by the origination documents for mortgage i by lender j in state s originated in time t. Since our dependent variable is continuous, we frame our estimation as an OLS regression. DFjt indicates DF users.
Hypothesis 2: There is significant market segmentation among DF users’ borrower base. This hypothesis is tested using logistic regression to examine borrower characteristics associated with DF users.
To this end, we estimate the following logistic regression:
Our dependent variable DFjt indicates whether the lender is a DF user. BCijst is specified to include all variables that can aid in explaining borrower characteristics. This includes traditional risk measures: FICO, LTV and DTI; indicators for whether the borrower was insured, and whether he/she was buying their first home; borrowers’ choice of origination (retail, correspondent or broker); loan purpose (purchase or refinance); the borrower’s occupancy status (primary, investment, secondary); and the log of the loan amount.
The study also focuses on discriminatory lending, exploring if DF users exhibit discriminatory traits in the mortgage market
Conclusion
The study concludes with a detailed analysis of empirical results, offering insights into how DF usage by lenders impacts various aspects of the mortgage market, including risk assessment, market segmentation, and potential discrimination. The findings suggest that DFs can be both advantageous in reducing risks and controversial due to data privacy concerns. The implications are significant for professional bodies, banks, governments, and regulators, highlighting the dual nature of digital data usage in the modern financial landscape.
For more detailed analysis, interested readers can refer to the original publication link https://onlinelibrary.wiley.com/doi/10.1111/acfi.12946.
References:
Fuster, A., Plosser, M.C., Schnabl, P. & Vickery, J.I. (2019) The role of technology in mortgage lending. The Review of Financial Studies, 32(5), 1854–1899. Available from: https://doi.org/10.1093/rfs/hhz018
A news article also published in the University regarding this research work. website. https://www.auckland.ac.nz/en/news/2022/08/08/getting-personal–how-digital-footprints-are-changing-the-mortga.html
