Citi: The mortgage underwriter’s tale
He explains how between 2005 and 2007 Citi’s underwriting standards simply fell apart, as the securitization professionals in the investment bank would override or ignore Citi’s own underwriters:
In the third quarter of 2006 the Wall Street Chief Risk Officer started changing many of the underwriting decisions from “turn down” to “approve.” This was done either personally or by direction to the underwriters. This artificially increased the approval rate on the sample. This higher approval rate was then used as justification to purchase these pools.
In the sample on one $300+ million Merrill Lynch subprime pool the underwriters turned down 716 mortgages as not meeting Citi policy guidelines. The Wall Street Chief Risk Officer personally changed 260 of these “turn downs” to “approved.” The pool was purchased…
Still another $320 million Merrill Lynch pool was purchased with an approval rate of 72%. Citi policy required a minimum approval rate of 90%.
Bowen also explains the sneaky way in which a pool of mortgages which was 60% bad was reported to the Third Party Origination Committee (“TPO”), which had overall responsibility for managing the selling mortgage company relationships, as being 95% good:
I spent time with the QA management and underwriters to better understand the QA processes. I learned that there were actually two categories of “agree” decision, with only the total of the two agree decisions being reported to TPO committee.
There was the “agree” decision, meaning the Citi underwriter agrees with the selling mortgage company underwriter that the file meets Citi policy criteria.
And there was an “agree contingent” decision, meaning that the Citi underwriter agrees with the original underwriting decision. But the decision is contingent upon receiving documents that are missing from the file, and those documents confirm the conditions underwritten.
An an example, the selling mortgage company underwriter may have approved a mortgage file showing a 45% debt to income ratio, which was within Citi policy criteria for the product. However, the required proof of income documentation confirming the borrower income used in the underwriting decision might be missing from the file. In this instance the Citi underwriter would assign an “agree contingent” decision to the file. The agree decision would be contingent upon receiving the income documentation proving the income utilized in the originating underwriter decision.
The total of the “agree” and “agree contingent” decisions would be reflected as the overall “agree” rate when reported to TPO Committee. This overall agree rate was the only agree rate reported to TPO through June 2006. And it was believed by the underwriters I interviewed that over half of the files had “agree contingent” decisions, meaning over half of the files were missing policy-required documents.
The QA process was very manual and lacked any automated reporting. The manager relied upon manual tally sheets, manually added, to produce the aggregate reporting given to TPO committee.
After significant effort, it was determined that the 5% disagree, 95% agree originally reported to June TPO was incorrect. It should have been 5% disagree, 55% agree contingent, and 40% agree. In other words, 5% were not underwritten to Citi policy and another 55% were missing policy- required documents.
Bowen sent an email to Bob Rubin and other senior Citi executives in November 2007. It had the subject line “URGENT — READ IMMEDIATELY — FINANCIAL ISSUES”, and it explained that Citi had yet to recognize enormous mortgage-related losses. I’m looking forward to the commissioners asking Bowen for some tips on what questions they should put to Rubin tomorrow. That testimony is going to be very interesting indeed.