Research Says Broker Originated Loans Underperforming

Researchers at Columbia University have released a new study indicating that mortgage loans originated by large banks during the housing boom were 50% less likely to go sour than those originated though their brokerage channel. 

The research covered the period from 2004 to 2008 and used data from over 720,000 loans of all types that were originated within the bank. Data analysed included information on loan terms, borrower demographic and financial information, property information, and monthly performance data through January, 2009.  The loans were heavily concentrated in the Midwest.

Data was divided into four subsamples by the loan origination channel (bank or broker) and loan documentation level (full documentation or various levels of reduced documentation including no documentation.  This resulted in samples labelled Bank/Full-Doc, Bank/Low-Doc and a matching pair for broker originated loans.

The study’s authors, Wei Jiang, Ashlyn Aiko Nelson, and Edward Vytlacil, maintain that their sample does not over-represent subprime loans.  On a national basis sub-prime loans represent 18 to 21 percent of all loans.  In the sample the subprime representation was stable at across years at 14-15 percent.

The bank studied was not identified but was among the top ten mortgage banks in 2006, specialized in low-and-no-doc loans, and suffered one of the largest losses in the industry during the crisis.  Loans originated by the bank since the beginning of 2004 reached a delinquency rate of 28 percent by the beginning of this year.

As the housing boom grew so did the banks involvement in mortgage lending.  In 2004, 20,000 loans were originated through all channels.  At the peak in 2006 the total was 154,000.

The bank’s rapid expansion in mortgage lending was driven by broker originations and by low-doc loans.  At the beginning of 2004 brokers originated 75 percent of the banks loans; by the end of 2006 it was 94 percent.  During the same period the percentage of low-doc loans increased from 39 percent to 75 percent.

The authors uncovered two types of what it calls agency problems in mortgage lending which they say constitute the fundamental causes of high delinquency rates.  The first is a misalignment of incentives for brokers who do not bear the long-term consequences of low quality loans.  The loans in the broker subsamples have delinquency probabilities that were 10-14 percentage points – or more than 50% - higher than the bank sub-samples  A binary decomposition of the data found that 75% of the Bank-Broker gap could be accounted for by borrower characteristics and the rest by (unobserved) heterogeneity.  This means that the higher delinquency rates among brokered loans could be explained by broker penetration of poorer quality borrower pools as measured by credit scores, LTV, and debt ratios than those borrower pools penetrated by the bank.

Comparing documentation subsamples for both Brokers and Bank, the research not surprisingly found the Low-Doc samples had worse performance than the Full-Doc samples by a factor of 5-8 percent.  The heterogeneity mentioned earlier explains nearly all of this difference.

The Low-Doc analysis did not find that those loans necessarily compromised lending standards but that there was less careful verification of some of the reported information such as owner occupancy status and less rigorous screening along hard to quantify measures such as major expenditures not included in the debt ratio.  Thus the second agency problem is between the lender (Bank or Broker) and the borrower who can hide or even misrepresent unfavourable information because of lax verification procedures.

Other research indicated that the quality of low-doc loans is significantly lower than full-doc loans thus lowering the ability of information to predict loan outcomes.  This was particularly evident in Broker/Low-Doc loans where there appear to be both looser standards and less diligent verification.  The authors “conservatively” estimated that the median of income exaggeration along low-doc borrowers was about 20 percent.

The final finding was a significant difference among delinquency rates by ethnicity.  Hispanics had a 4 to 11 point higher rate relative to white borrowers and Blacks a 3 to 4 percent difference.  The authors concluded that this had no roots in discrimination or in different standards but were explained by differences between white and minority borrowers, possibility because the latter lacks previous home buying experience or much background dealing with mainstream financial institutions.

Mortgage News Daily's Managing Editor Adam Quinones adds perspective..

"When interpreting the finding of this research one must remember that this report looks back on the mortgage industry during the heart of the housing boom. At that time many a roaming salesman had entered the industry looking to make a quick dollar. Since then the lending environment has drastically evolved. The findings of this report do not reflect the amount of diligence required to compete in today's real estate market.  To put it simply, only the strong have survived."