In this post, you'll learn everything you need to know about fair lending regression analysis. In it, you'll learn who regression analysis is a good fit for, when and why to consider it, and much more. In keeping with the back-to-school theme, you'll also have a chance to get a refreshed, updated Fair Lending Regression Analysis Primer.
If you're in compliance, you've probably heard about regression analysis. That said, unless you're a statistician like me, just the thought of running regression might make you nervous. But regression analysis doesn't have to be scary.
In this post, you'll learn the answers to 9 important questions about fair lending regression analysis and whether it can help you.
At the end of this post, you'll also have a chance to download the Regression Analysis Primer, which provides even more in-depth details about regression analysis, and when it can help! Let's get to those 9 questions...
Regression analysis is a statistical model used to better understand the relationship between different variables. It helps us look at different factors or variables to identify causes and account for discrepancies.
For fair lending, it analyzes different factors or variables (like DTI, LTV, and FICO) in aggregate to explain the disparities, and determine what is important to focus on and dig into further. It allows you to really analyze and understand other factors that may help explain decisioning and pricing disparities.
Regression is an ideal tool for analyzing loan data to determine if prohibited basis factors had any bearing on either the credit decision or the price of the loan.
The primary objective of regression analysis is to explain disparities in the data with more in-depth analysis of different factors.
Regression is intended to eliminate the need for line-by-line comparison. It looks at everything in aggregate and hopefully can use the variables to explain the disparity.
For fair lending, the goal is to identify similarly-situated applicants or borrowers, and determine if they received a similar decision or price. Regression will determine if age, race, gender, ethnicity or another prohibited basis factored into the decision to make the loan, or its pricing.
Regression analysis can:
Regression analysis is a great fit for institutions who:
Read also: 9 Fair Lending Compliance Training Essentials [Free Checklist]
In fair lending compliance, you're looking at lots of credit characteristics, and regression can help identify which ones are relevant to explaining disparities.
Remember, disparities don't always mean discrimination, but analyzing your data is the only way to know for sure.
Technically, you could analyze any set of loan data using regression. However, regression analysis is typically used for HMDA analysis, indirect auto lending analysis, credit card, and consumer loan analysis.
The data needed depends on the customer's goals, and the type of lending being analyzed. The key data points needed to run regression analysis for both underwriting and loan pricing may include: Credit Score, Loan-to-Value (LTV), Debt-to-Income (DTI), Interest Rate, APR, Loan Officer ID, Branch ID and Loan Term. Depending on your institution and your risk profile or the types of regression analysis you're conducting, different data fields may be needed.
HMDA/mortgage regression analysis for underwriting and pricing will need (at a minimum):
For auto lending regression analysis, we're either conducting decisioning analysis or interest rate analysis. Indirect auto lending regression analysis will need:
For decisioning, we will analyze any decisioning factors that your institution considers important.
To analyze dealer mark-up, we would conduct a different type of analysis called comparative analysis. To do this, we would need the data on the rate sheet. That typically is the credit score, loan term, and vehicle age.
Credit card and consumer lending regression analysis will need:
Yes! This is part of what makes regression such a powerful tool - it can really accommodate a lot of different variables.
For example, some other factors that may influence the credit decision or pricing, and may need to be considered are:
As mentioned above, regression is not an effective tool for small data sets. The larger the file, the more meaningful the predictive regression model. We recommend at least 1,000 records.
In addition, regression will not work when there is little or no variation in the dependent or independent variables. For example, if the data contains:
Regression is a deep dive! Before starting, consider the quality or integrity of your data, and determine what type of analysis is best for you. At least a cursory review of data integrity is useful for expediting the process. For example, make sure that everyone has a state, action, and that there are no negative credit scores.
Before conducting regression, start with a more basic fair lending analysis. That way, you'll know where to focus your efforts. If you're only lending to white, non-Hispanic borrowers, regression won't help you much. Or, if you have a redlining issue, regression is not an appropriate tool for managing risk.
Consider regression when your origination, denial, or price disparity rates are high, indicating potential fair lending risk.
Regression analysis is a powerful fair lending compliance tool that Ncontracts has provided to financial institutions for years. We work with many of the top mortgage companies, financial institutions, and law firms to conduct fair lending regression analysis and reduce compliance risk.
If you're interested in learning more about how we can help you with your fair lending regression analysis, please click here to set up a time to talk with an expert. Ncontracts' fair lending regression analysis is different, because it is supported by guided analysis reviews, a team of experts, and even help during exams, if needed. This custom analysis delivers insights that your team can use to actively manage and reduce risk.
In the meantime, check out this Regression Analysis Primer: