I am conducting a Fair Lending self-evaluation. It's a very simple version where I am basically looking at the average rate for specific multiple loan products for different groups. My question is--how much of a difference between rates for the control group vs the prohibited basis group raises a red flag for you? For example, with used cars, Males had a 0.1% lower rate than females. Do I need to investigate this further or is it small enough to shrug off? Is there any guidance on what the threshold should be, and is it based on the sample size? Do I need to do more complex statistics?

I am not aware of any formal guidance, but when the CFPB was doing all the indirect auto work, they informally suggested disparities of 0.10% or greater were problematic. I am aware of at least one DOJ referral for a persistent disparity of 0.03%. My institution uses 0.05% or 5 bps for determining the need for file reviews.

Regardless of which disparity you use, I recommend pairing it with a test of statistical significance and file reviews to investigate the disparities, especially if you aren't using regression analysis.

Agree with InFairness. Are you talking about the rate, or APR?
We ran into one situation where the overall APR's to females was higher than the rates to males. It was determined this occurred for 2 reasons.
1 - Females had a tendency to be conservative with purchases and finance a lower dollar amount.
2 - Females opted (and we verified they made the decision) for shorter term loans.

Since there was an application fee, the shorter term and lower dollar amount had a greater impact on the APR.