You are here

Why HMDA Data Needs to Change

Adam Rust's picture

Posted August 26, 2010

The Dodd-Frank bill will require lenders to disclose more data about their lending, but the fundamental problems with HMDA remain largely unresolved. Dodd-Frank says that it will collect, and then disseminate, the following new categories within an updated HMDA by no later than 2012:

  • age of borrower
  • borrower credit score
  • total points and fees payable at origination
  • the spread between the loan's interest rate and the corresponding treasury note of similar maturity
  • value of the collateral pledged against the loan
  • non-amortizing loan features
  • length before loan reset (months)

Those are some good ideas. I think that there is going to be a substantial discussion about

credit score. In this day and age, releasing discrete credit scores could be a real problem. I can just imagine zillow providing your credit score next to information about the sales price of your home. That isn't going to happen, because releasing the actual score would be illegal. What did Congress mean, then? It's something that will be decided by a phalanx of bank lobbyists, a small coterie of consumer advocates, and the Federal Reserve.

Plenty of things still are not addressed, however. I spoke with one lender of a Big Four Bank. Mr. Big Four was pretty upset with the existing HMDA design. The gist of his critique was that it was designed in a way that made his bank look bad. Fair enough that he would have that opinion. It creates false positives and false negatives.  For instance, if you have an adjustable-rate loan, HMDA will report the interest rate at the level of its pricing in the first  month. That means that ARMs generally appear to have low pricing. Moreover, HMDA doesn't flag loans as "fixed" or as "ARM," or even as interest-only.  All of those exotic loan products that were invented after the late 90s are not accounted for in HMDA, because it was designed in 1991 (save for one adjustment to pricing data in 2004).

I had to agree with Big Four. For years, I've done HMDA research and I've had to constantly explain why HMDA data puts out findings that say that Countrywide and IndyMac were the biggest prime rate lenders in the country. There ended up being three kinds of banks:

prime rate lenders: they would have high declination rates, and big disparities in the availability of capital across different demographic groups.When they did make loans to populations with riskier profiles, they would make a lot of high-cost loans.

subprime lenders: they would come across as very friendly to just about all people, with high approval rates among all kinds of people. In some instances, they would have higher rates of "high-cost" loans, but only if they made a lot of fixed-rate loans. If they were running a 2-28 shop, then they'd be golden in HMDA.

lenders with both prime and sub-prime channels. These lenders would have the worst of both worlds. It would appear to be a double standard, both in terms of approving loans and also in terms of making high cost loans.

HMDA created real problems. HMDA could work when you were speaking with or about one bank, because the "unknowns" were (alert: Rumsfeldism approaching) largely "known unknowns."

It was less effective in comparing banks. It was incredibly difficult when trying to address the lending issues that dominated the last ten years. HMDA had no sensitivity to the quality of lending, or to the inputs that were utilizing in underwriting in an era of risk-based pricing. Congress' intention to have information on credit score would address the latter, but I doubt if it will be implemented at the level of the borrower. Short of that, HMDA could be back where it started.

Mr. Big Four wants to go ahead and put the underwriting data out there. His logic is that it will differentiate his bank from the sub-prime lenders. That would save what he considers to be the hindrance of un-deserved lawsuits. He is hoping that the Federal Reserve will see it his way, and create a set of variables that get at underwriting. Those would include:

  • credit score
  • loan-to-value
  • debt-to-income
  • existence of a bankruptcy in the past seven years
  • late mortgage payments in the last three years
  • foreclosure in the last five years

There are some other indicators that make a difference in underwriting, but that are already included. Owner-occupancy is revealed in HMDA. Big Four knows that his bank turns down a lot of investor loans.

If HMDA can satisfy the issue of underwriting, it will be a far more useful database. I would like to see more about loan terms and property characteristics - if for example some bank is doing stated-income loans, then the public should have some concern - but underwriting is in and of itself a huge advancement.