Paper: arXiv 2410.22382

Authors: Chris Lam

Abstract

Alternative data provides valuable insights for lenders to evaluate a borrower’s creditworthiness, which could help expand credit access to underserved groups and lower costs for borrowers. But some forms of alternative data have historically been excluded from credit underwriting because it could act as an illegal proxy for a protected class like race or gender, causing redlining. We propose a method for applying causal inference to a supervised machine learning model to debias alternative data so that it might be used for credit underwriting. We demonstrate how our algorithm can be used against a public credit dataset to improve model accuracy across different racial groups, while providing theoretically robust nondiscrimination guarantees.

Complexity vs Empirical Score

  • Math Complexity: 5.5/10
  • Empirical Rigor: 5.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper provides a thoughtful theoretical bridge between legal doctrines of discrimination and Judea Pearl’s Causal Bayesian Networks framework for alternative credit data. It provides an empirical evaluation on a public mortgage dataset to demonstrate bias reduction while balancing model accuracy.

Research Flowchart

  flowchart TD
    A[Goal: Debias Alternative Data for Credit Underwriting] --> B[Input: Public Credit Dataset & Alternative Data]
    B --> C[Method: Apply Causal Inference to Supervised ML Models]
    C --> D[Process: Identify & Remove Illegal Proxy Variables]
    D --> E[Outcome: Robust Nondiscrimination Guarantees]
    D --> F[Outcome: Improved Model Accuracy Across Racial Groups]

Cited References (42)

  1. Using Alternative Data to Evaluate Creditworthiness (2017) — link
  2. Using Alternative Data in Credit Risk Modelling (2023) — link
  3. What’s in my FICO® Scores? (2024) — link
  4. Invisible Primes: Fintech Lending with Alternative Data (2022) — link
  5. The Use Of Cash-Flow Data In Underwriting Credit: Market Context and Policy Analysis (2020)
  6. Give Credit Where Credit is Due: Increasing Access to Affordable Mainstream Credit Using Alternative Data (2006)
  7. On the rise of FinTechs : Credit scoring using digital footprints (2020)
  8. Buying and Payment Habits: Using Grocery Data to Predict Credit Card Payments (2021) — link
and 34 more
  1. When words sweat: Identifying signals for loan default in the text of loan applications (2019)
  2. Big data’s disparate impact (2016)
  3. Proxy Discrimination in the Age of Artificial Intelligence and Big Data (2019)
  4. What is Proxy Discrimination? (2022) — link
  5. A Causal Bayesian Networks Viewpoint on Fairness (2019) — link
  6. Causal Conceptions of Fairness and their Consequences (2022) — link
  7. Causal Fairness Analysis (2022) — link
  8. Fairness and Machine Learning: Limitations and Opportunities (2023)
  9. Causal Inference in Statistics (2016)
  10. The book of why (2020)
  11. Addressing Traditional Credit Scores as a Barrier to Accessing Affordable Credit (2023)
  12. Give Me Some Credit!: Using Alternative Data to Expand Credit Access (2023)
  13. Using publicly available information to proxy for unidentified race and ethnicity (2014)
  14. Keeping Fintech Fair: Thinking About Fair Lending and UDAP Risks (2017)
  15. Interagency Fair Lending Examination Procedures (2009)
  16. The seven tools of causal inference, with reflections on machine learning (2019) — link
  17. 5 Cs of Credit (2023) — link
  18. 5 Cs of Credit: What They Are, How They’re Used, and Which Is Most Important (2023) — link
  19. Counterfactual Fairness (2018) — link
  20. The Sensitivity of Counterfactual Fairness to Unmeasured Confounding (2019) — link
  21. Fair Lending Monitorship of Upstart Network’s Lending Model (2024)
  22. On Testing for Discrimination Using Causal Models (2022) — link
  23. Data Point: Credit Invisibles (2015)
  24. Fairness Through Awareness (2011) — link
  25. How big data is unfair: Understanding unintended sources of unfairness in data driven decision making (2014) — link
  26. Discrimination-aware data mining (2008) — link
  27. Avoiding Discrimination through Causal Reasoning (2018) — link
  28. Unlawful Proxy Discrimination: A Framework for Challenging Inherently Discriminatory Algorithms (2024) — link
  29. Discrimination law (2022)
  30. Path-Specific Counterfactual Fairness (2018) — link
  31. Fair Inference On Outcomes (2018) — link
  32. Algorithmic Fairness (2018)
  33. Direct and Indirect Effects (2013) — link
  34. National Survey of Mortgage Originations POC (2024) — link

0 of 42 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.