The flaws in the pipeline through which Citigroup moved mortgages into mortgage-backed securities may create far more opportunities for investors to demand their money back than the bank is expecting.
In our article Tuesday, John Carney and I wrote about loan acquisition channels at Citigroup. Based on description of the flaws of those channels, we believe that Citi is far more exposed to mortgage put-backs than is commonly thought.
A mortgage put-back occurs when an investor in a mortgage-backed security successfully demands that a bank repurchase the underlying mortgages that do not adequately satisfy the representations and warranties the bank made about the mortgages when it originally sold the security. A bank can be forced to repurchase the mortgage at par under certain circumstances.
Citi says our view has "absolutely zero basis in fact." So let's take a deeper look at just one of their loan channels to examine Citi's potential exposure.
In April of 2010, former Citigroup executive Richard Bowen testified before the Financial Crisis Inquiry Commission (FCIC). Bowen's testimony under oath before the FCIC reads like a roadmap to Citi's correspondent loan acquisition channels, and has been invaluable in our understanding of Citi's loan acquisition process.
Bowen was Senior Vice President and Chief Underwriter for Correspondent and Acquisitions for Citifinancial Mortgage, and then later was promoted to Business Chief Underwriter for Correspondent Lending in the Consumer Lending Group. In this latter capacity, he was responsible for 220 underwriters, and was charged with responsibility for underwriting over $90 billion a year in mortgages. (Correspondent Lending is one of four channels Citi used to add mortgages and mortgage backed securities to its overall portfolio).
According to Bowen's testimony the Correspondent Lending group had two sub-channels: Underwritten Flow and Delegated Flow. In both sub-channels mortgages are submitted individually to Citigroup. (Loans are not purchased in bulk through either of these sub-channels.)
The principal difference between these two sub-channels within Correspondent Flow is that Citi directly underwrote the mortgages in the Underwritten Flow sub-channel, but did not directly underwrite the Delegated Flow channel loans.
It may be helpful to review the concept of 'underwriting'. In loan parlance, underwriting is the act of determining eligibility for a loan. When we talk about a loan being underwritten, it does not imply that the loan has been approved: a loan can be underwritten and then rejected. (This differs from the typical usage of the word 'underwritten' in the securities world, where completed deals are often referred to as underwritings.)
As discussed above, loans in the Underwritten Flow sub-channel had all their underwriting done directly by Citi. In this sub-channel, Citigroup would verify that individual loan files met their credit criteria. Citigroup underwriters would also validate that loan documents required by Citigroup were present , according to Bowen's testimony .
In delegated flow, however, the underwriting function was delegated to the "correspondent mortgage company"—essentially outsourcing the underwriting process to the seller of the mortgage. According to the testimony, this sub-channel contained only prime loans.
The loan volume within Citi's Delegated Flow sub-channel was quite large: $50 billion a year in mortgage loans were purchased. The Delegated Flow sub-channel made use of 1,600 mortgage companies as submitters for loans. (An organization called the Third Party Origination Committee, or TPO committee, was charged with the responsibility of managing the relationships with the selling mortgage companies.)
Under this arrangement, Citigroup maintained the right to force the selling mortgage company to repurchase mortgages Citi deemed defective. (A defective mortgage is a mortgage that either does not comply with Citi's underwriting criteria, or does not contain all required loan policy documents.)
Citigroup would perform sampling of loans within Delegated Flow to determine whether loans underwritten by the correspondent mortgage companies were in compliance with Citigroup's standards. Sample loans were the underwritten again, this time directly by Citi employees, to determine whether the correspondent mortgage company had correctly performed the process the first time, according to Bowen's testimony.
Citi would then assign an 'Agree' or 'Disagree' determination to the sample loans they had underwritten. According to Bowen's testimony, Citigroup had a policy in place during the year 2006 whereby a minimum of 95 percent of loans purchased through the Delegated Flow channel received an 'Agree' decision.
In addition to assigning a mandatory 'Agree' rate on 95 percent of the loans, Bowen stated in his testimony that there were irregularities in the sampling process for Delegated Flow at Citi.
Bowen's testimony states that the 'Agree' determination was further subdivided into two additional classifications: The first was a straight 'Agree' determination— meaning the mortgage was compliant with Citi's underwriting criteria, and all necessary paperwork was present in the loan file; the second was an 'Agree Contingent' determination, where the loan appeared to be compliant with Citigroup's underwriting standards— however, a final determination was contingent upon receiving documents that were missing from the loan file.
According to Bowen, loans bearing the 'Agree' status—as well as loans bearing the 'Agree Contingent' status—were rolled up together when reporting the overall 'Agree' rate to the TPO committee that supervised the sampling and purchase of loans within the Delegated Flow sub-channel.
Bowen goes on to say in his testimony that during an internal review of Delegated Flow's sampling procedure, it was determined that while the 95 percent 'Agree' target was reached, that figure did not represent the true percentage of compliant loans. In reality, only 40 percent of all loans were truly compliant and consistent with the 'Agree' determination, he said.
The other 55 percent should have received the 'Contingent Agree' determination.
While it's easy to become mired in the complicated details of the loan acquisition process, the salient point is this: 55 percent of the loans in this sub-channel were missing required policy documents—and therefore by definition defective.
Of the $50 billion a year in loans acquired through the Delegated Flow channel, 80% were sold to investors. That means that a total of $40 billion a year in loans were sold to investors through this sub-channel alone.
Bowen states in his testimony that the largest purchaser of these loans were Fannie Mae and Freddie Mac.
According to Bowen's testimony, Citigroup made representations and warranties to Fannie and Freddie about these loans. Under those representations and warranties, Citigroup may be required to repurchase mortgages from Fannie and Freddie in the event that the mortgages turn out to be defective.
Bowen states that, under the terms of the representations and warranties made by Citi, any mortgage that does not conform to Citigroup's underwriting criteria—or has missing loan documentation—can be considered defective once it is identified.
What were the defect rates based on the sampling? According to the testimony, the overall defect rate of loan purchased through the Delegated Flow sub-channel was 60 percent in 2006. By 2007, that defect rate had risen to 80 percent.
By applying a simple formula to these numbers, we can arrive at a back of the envelope calculation for Citi's potential repurchase exposure from the Delegated Flow sub-channel.
Using the 60 percent defect rate Bowen cites from 2006, and based on the figure of $40 billion a year in mortgages sold to investors, I come up with a $24 billion unadjusted potential repurchase exposure for 2006.
Using the 80 percent defect rate from 2007, and applying the same underlying $40 billion a year resale volume, I reach a unadjusted potential repurchase exposure to Citigroup of $32 billion.
Citi claims a 22 percent indemnification on their Serviced But Not Held (SBNH) portfolio, by dollar value.
If we discount the figures from 2006 and 2007 by this rate, we get indemnification adjusted potential repurchase exposure values of about $19 billion and $25 billion respectively. (The indemnification discounts I've applied assume an equally weighted indemnification rate across loan acquisition channels and loan vintages.)
Summing the indemnification-discounted potential exposure figures from both years brings us to a total of $44 billion.
In an official comment on a previous story, Citi pointed out that approximately 50% of prior claims to date required repurchase or make-whole payments. Discounting the $44 billion figure by half, to account for low-severity claims, would bring Citi's total potential exposure to $22 billion.
It is critical to note, however, that these numbers represent calculations of potential repurchase exposures—not projections or estimates of potential dollar losses to Citigroup.
Citi reports a total repurchase reserve balance for 3rd quarter 2010 of $952 million. Based on my final $22 billion figure, Citi's potential exposure to existing reserve ratio would come in at a little over 23 to 1. And that ratio represents only a portion of Citi's SBNH portfolio — a single loan channel over a two year period that does not include the 2008 vintage.
Perhaps Citi is being criticized unfairly for its attempts at transparency. As John Carney and I wrote in a joint byline post on the subject earlier in the week, Citi may be "'best in class' when it comes to mortgage put-back exposure." After all, while Citi did acquire loans in bulk, and buy individual loans it didn't underwrite, it did not acquire a mortgage lender on the verge of bankruptcy, unlike some other big banks on Wall Street. But that doesn't diminish the potential significance of these calculations. It just makes Citi sound better by comparison - and that is a truly frightening proposition.
Citi declined to provide a comment for this story. Instead, a Citi spokesman directed us to the statement provided for our story on Tuesday.
Questions? Comments? Email us atNetNet@cnbc.com
Follow NetNet on Twitter @ twitter.com/CNBCnetnet
Facebook us @ www.facebook.com/NetNetCNBC