Thursday, April 14, 2016

Goldies Sampling, and the Silver Settlement

This week has brought with it some significant disclosure on some of the forms of pre-crisis misconduct at the banks, and the first settlement of post-crisis alleged manipulation of the silver markets.

Allegations into silver and gold markets manipulation are among a string of claims made against the big banks for post-crisis benchmark manipulation -- the other well-known cases being the three exchange rate and interest rate manipulation allegations: LIBOR, FX, and ISDAFIX, the last of which survived a motion to dismiss, just two weeks ago.

Deutsche Bank has settled with the class In re: London Silver Fixing Ltd., Antitrust Litigation, 1:14-md-02573, U.S. District Court, Southern District of New York (Manhattan), but those terms have not been disclosed.

Statistical Sampling Techniques ... Gone Wrong

But the Goldman Sachs settlement (a headline settlement amount of $5.06 billion) for what the Justice Department calls "serious misconduct," provides some real insight into the artful approach taken to mortgage sampling for residential mortgage-backed securities (RMBS).

Imagine you wanted to get an idea of the quality of teachers at a school you're considering for your kids.  The school chooses a sample of 100 teachers, removes the 90 worst ones, and presents to you the 10 most credible, as if they were randomly selected to fairly represent the whole group of teachers.

Well that's pretty much what several of the big banks did.  They sampled the mortgage loans under consideration for inclusion in RMBS transactions.  They kicked out some (or perhaps all) of the non-compliant ones in the sample, and then they presented the cleaned-sample as if it represented the entire pool.  

Among the problems was that the samples were not very big.  So if half of the loans are non-complaint (say 50 out of 100), and you sample only ten percent (say 10 out of a 100) ... and then you kick out 5 from those 10, well, there are still 45 non-compliant loans (half of the unsampled 90 loans) going into the pool!

We wrote about this revelation/absurdity back in 2011, just after the Financial Crisis Inquiry Commission released documents showing this incredible pattern.  (You can read our report here, or below.)

The language in Monday's Statement of Facts is additive, and confirms some of the worst suspicions that come from documents provide by Clayton to the FCIC.  Here are some choice snippets: 
"Even when the percentage of loans graded as EV3s and dropped by Goldman from the due diligence samples indicated that the unsampled portions of the pools likely contained additional loans with credit exceptions, Goldman typically did not increase the size of the sample or review the unsampled portions of the pools to identify and eliminate any additional loans with credit exceptions."
"The Mortgage Capital Committee also asked “How do we know that we caught everything?” In response to that question, the Goldman due diligence employee who oversaw the due diligence for one pool of loans purchased from SunTrust Mortgage wrote “we don’t[,] it was sampled w[ith] max at 20%- the drops were a result of timing not systemic issues with SunTrust.” Another Goldman due diligence employee who oversaw the review of a pool of Countrywide loans that Goldman had purchased on March 29, 2006 responded to the same question: “Depends on what you mean by everything? Because of the limited sampling on CW 10-15% we don’t catch everything and the way they [Countrywide] deliver the files we have little chance to upsize. This trade had issues with aged loans and we tried to get pay histories and were told they would not provide them.” In response to the Mortgage Capital Committee’s question “Are these results systemic,” the same employee wrote: “Every trade varies, but typically CW have a very high credit 3 drops on the first review of DD 60% and then clear the docs, so one can assume that the files we are not reviewing would have the same issues.”"
"In April 2006, the Mortgage Capital Committee received a memorandum with a highlevel summary of Goldman’s due diligence results in connection with a proposed Alt-A RMBS offering. The memorandum included aggregate due diligence results for three Countrywide loan pools that Goldman had purchased on March 30, 2006 and indicated that 34.38 percent of the loans in the proposed offering had been drawn from certain of those Countrywide loan pools. The memorandum reflected that Goldman had conducted credit and compliance due diligence on a total of 15.44 percent of the loans in the three March 30 Countrywide pools. ... The memorandum also stated that Goldman had dropped a total of 6.07 percent of the loans in the three Countrywide pools (not the samples) for credit or compliance reasons. Across the three Countrywide pools, Goldman dropped nearly 40 percent of the loans in the credit and compliance due diligence samples for credit and compliance reasons. The memorandum referred to this as an “exceptional drop amount” and stated that “in the case of [Countrywide], an unusually high drop rate for missing or deficient documents resulted in an above average total drop percentage (approximately 33% of the credit drops were due to missing appraisals).” Contemporaneous records reflect that Goldman closed on six Countrywide loan pools on March 29 and 30, 2006, and that Countrywide was struggling with staffing and workload issues that affected its ability to deliver missing documents requested by Goldman for the loans in those six pools."
"Although Goldman dropped 25 percent of the loans in the due diligence sample because they were graded as EV3s, including all the loans graded as EV3s for unreasonable stated income, which comprised at least 2.5 percent of the loans in the due diligence sample, Goldman did not review the portion of the pool not sampled for credit or compliance due diligence, which comprised approximately 70 percent of the total pool, to determine whether there were similar exceptions in the unsampled portion. Goldman subsequently securitized thousands of loans from this pool into one GSAMP transaction. The Mortgage Capital Committee approved the issuance of this offering."

Our report from 2011 can be viewed here:

For an update on the current status of alleged benchmark-rigging antitrust cases, see Alison Frankel's blog, here.

No comments: