On Predicting Mortgage Defaults
March 2019

Happy daylight-savings week everybody! Hopefully you’ve managed to catch up on that lost hour of sleep and are all ready for some light reading about credit risk.

With many of you making nice progress on your CECL implementation, we have noticed the dialogue gradually shifting away from theory and more towards practice. The rubber is starting to hit the road for a good number of institutions, which is great! Consequently, the questions we’re now being asked often begin with “how come” rather than “what if.” One of the most common questions we encounter is, “what caused the increase/decrease in my loss rate from last month to this month?”. The answer is often simple, but can be difficult to track down.

For Probability of Default (PD) methods, it can be helpful to fully understand which factors are at play in determining expected loss. I’d like to help out by describing the loan-level and macroeconomic factors that we have found to be most influential in the First Mortgage PD model.

  • Delinquency: No shocker here.
  • Loan Age: Believe it or not, the age of the mortgage (in months) is one of the most highly correlated factors considered—old mortgages don’t charge off.
  • Original LTV: Not surprisingly, the LTV at origination is a strong predictor of default.
  • Credit Score: This is arguably the strongest predictor.
  • Original Balance: The size of the loan matters! Interestingly, the relationship doesn’t appear to be linear. The riskiest original balance is in the 300-400K range. Risk decreases as you move above or below this range
  • Interest Rate: As expected, the interest rate (relative to what would be expected based on credit score, national averages, etc.) is a strong predictor of prepayment. This is the primary reason it is included in our models. However, this variable is also strongly correlated with default! Borrowers that are paying significantly more than they should be for a mortgage are riskier.
  • Debt to Income: Even after all the above are considered, simple economics shine through with this factor. People who push the boundaries of their income with large amounts of debt are much more likely to default when hard times come.
  • Unemployment: When the economy thrives and borrowers are employed, they make their payments.
  • House Price Changes: We have found that the average monthly change in house prices since origination ranks among the top predictors of default.

As you observe the dynamics of your PD calculations for First Mortgages, check out this list and use Visible Equity’s Loan Analytics platform to dive into your data and find patterns among these important factors.

Related Blog Posts
Further Your Education