Does Securitization Affect Loan Modifications?

A few days ago I wrote a long and detailed critique of a Boston Federal Reserve staff study that argued, among other things, that securitization was not a factor in the paucity of loan modifications.  The study reached this conclusion based largely on the similar rate of modifications for portfolio and securitized loans.  Although the study controls for the effect of the modification in terms of monthly payment, it otherwise assumes all modifications are created equal.  But clearly they are not.  There is a significant difference in redefault rates for securitized and portfolio loans, and the securitized loan mods perform much worse.  This is something the Boston Fed's study cannot explain other than if there is (1) unobserved heterogeneity in the loans or (2) differences in the loan mods.  

The nature of unobserved heterogeneity in data is that it can't be observed, so all that can be said of (1) is that it is a possibility.  But assuming that there isn't a heterogeneity problem about the unmodified loans, what about the mods?  Is there heterogeneity problem in mods that makes comparisons of mod rates a poor measure for evaluating the impact of securitization.  It appears that there is. 

The Boston Fed study did not control for the effect of the loan modification on the homeowner's equity. It does have controls for LTV and negative equity, but those don't seem to have been applied to the serviced/portfolio distinction, at least in the paper.  I'm not sure whether there is sufficient data to do this, but what the study could have controlled for, but did not, was whether the modification involved a reduction in the unpaid principal balance.  In this aspect, there is a significant difference between portfolio and securitized loans.  

OCC/OTS Mortgage Metrics Data for the first quarter of 2009 indicates that very few loan modifications have involved principal balance reductions.  In fact out of 185,186 loan modifications in Q1 2009, only 3,398 (1.8%) involved principal balance reductions.   All but 4 of those 3,398 principal balance reductions were on loans held in portfolio.  The other 4 are quite likely data recording errors.  This means that there is heterogeneity in loan mods between securitized and portfolio loans.   

The difficulty in doing principal reduction mods for securitized loans is quite important because to the extent that negative equity is driving foreclosures (and there is significant evidence that it is), principal reduction modifications are the tool for eliminating negative equity (with an shared appreciation clawback or not).  The quality of loan modifications matters, and securitization affect the quality.  

There is also a major difference in the ability of portfolio lenders and Fannie/Freddie/Ginnie servicers to extend the term of a mortgage that private-label servicers don't have.  Not all securitization is the same.   Private label servicers can usually stretch out the term of a loan by no more than a year or so because the servicing contracts prohibit the extension of the term beyond the last maturity date of any loan in the pool, and pools are usually of similar vintage and duration loans.  Fannie/Freddie/Ginnie loans can be bought out of a pool and modified, making them more like portfolio in this regard.  Thus 49.2% of portfolio loan mods, 50.8% of Fannie, 61.2% of Freddie, and 17.2% of Ginnie mods involved term extensions, but only 3.9% of private-label securitization mods.  

Quite likely there is other heterogeneity that cannot be as easily discerned.  This makes sense--portfolio lenders are much less constrained in modifications than securitization servicers.  Attempts to quantify servicers' constraints by looking at contract language are inherently limited, as there are structural and functional constraints that are not apparent from an examination of the face of the servicing agreements. Moreover, securitization servicers are adverse to principal write-downs because that affects their compensation far more than an interest rate reduction.  The agency problem just doesn't exist for portfolio loans.  

Finally, the study has a very strange observation that there is a moral hazard problem in principal balance reductions, but apparently not for interest rate reductions:  "Balance reductions are appealing to both borrowers in danger of default and those who are not."  Therefore, borrowers might default to get principal reductions.  Sure, that's right, but everyone would also like a lower interest rate too.  I don't see why a principal reduction presents a different level of moral hazard from an interest rate reduction.  In terms of net present value, principal and interest rate are interchangeable (yes, there's an interest deduction, and a principal reduction changes the ability to refinance, but that's not the distinction at issue).  The bigger factor pushing against principal reductions (other than servicer compensation) is an accounting issue.  A principal reduction shows up on the balance sheet immediately.  A reduction of interest just reduces future income.

The take-away here is that even if the Boston Fed staff is right that securitization doesn't affect the prevalence of loan modifications, it clearly affects the quality of those modifications, and that is every bit as important, not least because the performance of past modifications is the basis for servicers' calculation of the redefault risk that the Boston Fed staff emphasizes as constraining modifications.  If servicers do bad mods and have high redefaults, that will make them more adverse to doing mods in the future because they will think that the mods don't work. 

Comments

[nick]

I think some of the reluctance to do principal write-down modifications in securitized deals comes from the servicers' fear of litigation. I belive that you can interpret the PSAs to prohibit these types of modifications; even aside from the issue of how to allocate the losses within the trust, which is another big hurdle that could bring in law suits. I think that's why they wanted the servicer safe harbor put into the mortgage legislation. However, once it was in there, they got criticized for trying to insulate themselves too much and for conflicts of interest (I'm not a lawyer but it did seem to be written very broadly). People thought that an unqualified safe harbor would give the servicers leeway to do either no mods at all, or mods that were negative NPV and there would be no recourse. If this was the reason for the absence of principal write-downs, we may see more in the future now that the bill including the servicer safe habor was passed.

[Blossom]

Professor,
There certainly is a significant difference between portfolio and securitized loans. Presumptions of homogeneity issues within securitized and portfolio loans that preclude comparisons of mod rates miss the point as does the notion that negative equity drives foreclosures. They are misguided and divert public attention away from real causes. This is why I had serious issues with FRBB mod report and frankly came to view its authors as spin doctors, obfuscating hard realities of their subject matter.
Just as you have written "Foreclosure is frequently more profitable to servicers than loan modification", we need to keep asking the question "Who profits?" Securitization offers so many more avenues of profit than portfolio loans. When these profit sources occur high up in the food chain, they dictate and often control what happens in sectors lower down such as subsidiary servicers. One of these profit sources, Markit Partners, Ltd. is presently under DOJ investigation. Markit, an oligopoly owned by a handful of big banks, Wall St. firms and hedge funds runs varous indexes upon which credit default swaps are "speculated". Markit's ABX.HE Index, the creation Wall St. traders was packed with subprime REITs in which mortgage servicing fraud by investment bank subsidiary and contract servicers was rampant to the extent that existence of firewalls between trading desk and subsidiary servicers have been questioned. Investment banks directed servicers to manufacture bogus defaults so they could profiteer with insider CDS bets. Markit's owners, identified in the past as ABN AMRO, Bank of America, Barclays Capital, Bear Stearns, BNP Paribas, Citigroup, Credit Suisse, Deutsche Bank, Goldman Sachs, HSBC, JPMorgan, Lehman Brothers, Merrill Lynch, Morgan Stanley, UBS are the real masters that servicers of securitized loans serve. Higher numbers of re-defaults in securitized loans are direct result of these 'higher ups' calling the shots for servicers to deliberately manufacture defaults for greater credit default swaps profit. Mod studies such as recent one by FRBB not only fail by not taking into account these egregious aspects of the big picture but they are seriously misleading to the public.

[Rick Arvielo]

Thanks for the information

[Mike Dillon]

"If servicers do bad mods and have high redefaults, that will make them more adverse to doing mods in the future because they will think that the mods don't work."

This is beautiful, Professor. I think I'm going to stick with:

"Servicers don't CARE if they do bad mods because SERVICERS MAKE MONEY ON THE MOD (from both the borrower AND NOW, if I understand correctly, from the federal government to the tune of potentially as much as $4500.00 but at least $1000 per mod), MONEY EVERY MONTH THAT A BORROWER REDEFAULTS (late fees), MONEY ON BPOS REGARDLESS OF WHETHER THE BPO IS ACTUALLY PERFORMED OR NOT, MONEY ON THE FORECLOSURE OF THE PROPERTY ITSELF AND NO ONE ELSE IN THE TRUST CARES BECAUSE THE OUTSTANDING BALANCE OF THE LOAN IS COVERED BY THE **NOTE INSURANCE** COVERING THE POOL THAT THE NOTE WAS SECURITIZED INTO."

I'm sorry... That was out loud, wasn't it. That banging sound you may have heard in the distance was just me decorating the walls of my office with more "forehead art".

Oh, by the way Professor, good luck with your Judiciary testimony tomorrow.

[Blossom]

Professor,
In no way do I wish to derail your continued analysis of FRBB report however it is of vital importance to realize there is a 900 lb gorilla in the room controlling the entirety of mortgage defaults, foreclosures, and modifications, all the while using insider information gleaned for its own gain in not so speculative CDS bets.
For those curious about which banks are the biggest Markit shareholders, Bloomberg has the breakdown. JPMorgan is Markit’s largest shareholder, with at least 1.67 million ordinary voting shares out of a total of 14.38 million, according to filings at U.K. Companies House. Bank of America Corp. is the second-largest, with more than 1.52 million shares held through its own units and those acquired in its purchase of Merrill Lynch & Co. last year. Royal Bank of Scotland Group Plc owns at least 1.35 million shares after its purchase of ABN Amro Holdings NV, while Goldman Sachs Group Inc. has about 1.11 million shares, the filings show.
http://www.bloomberg.com/apps/news?pid=20601087&sid=ahcdJoyFIDsk
Though PSAs would have us believe servicer fiduciary is to REIT investors, in practice this is not the case. As I said, these are the real masters securitized loan servicers serve, many of them subsidiaries of these very same firms.

[Doug]

Did you receive/hear anything from the Boston Fed?

[Loan Modification CA]

Hello,
I assume that some of the lack of enthusiasm to do principal write-down modifications in securitized deals comes from the service’s' terror of proceedings.

[Kris Gerardi]

Adam, I'm one of the authors of the "Boston Fed Study" (by the way I'm not at the Boston Fed). It's too bad I didn't notice this blog post sooner. Maybe you should read the paper a little more carefully next time before you go spouting off about it. We look explicitly at whether there are differences in redefault rates (over a 6-month horizon) between securitized and portfolio loans in order to address the possibility of differences in the quality of modifications being performed. There aren't any differences in redefault rates. As for the observation regarding principal reductions, I think you need to do a little more reading. There are currently 2 relatively large servicers of private-label mortgages (Litton and Ocwyn) who are engaged in mass principal reduction modification programs. Unfortunately, they are not in the LPS dataset that we are using in our study, which is why we don't pick up many principal reductions.

[Adam Levitin]

Kris, I can tell you're upset, but I'm not really sure about what. I know you're with the Atlanta Fed, but the paper came out originally, at least, as a Boston Fed working paper. I've read your paper more carefully and more times than I care to admit. There's a lot of really interesting stuff in the paper, and I hope I accurately described it. I disagree with some of your conclusions; I don't think the evidence you marshall allows for them.

Your main contention seems to be that I didn't register that you looked at redefault rates by securitized/portfolio status over a 6 month horizon to address differences in the quality of modifications. Actually, I just think that metric presents a very incomplete test. First, the 6 month horizon is too short to produce much. You would expect all the marginal borrowers to redefault in the first 6 months regardless; loan performance differences, if any, would likely appear over a longer horizon. Moreover, a lot of mods have an initial 3-month trial period--if the mods fail during that time, they aren't treated as mods. If they succeed, the mod is backdated to the beginning of the trial. That means we're often talking about a 3 month horizon, not 6 months.

Second, it's not surprising that redefault rates would be similar, because most mods have been similar, regardless of whether they are securitized or portfolio. There's a very interesting issue there, but it won't do to say that securitization doesn't matter. While most of the mods are similar, the OCC/OTS Mortgage Metrics data shows that there are some major differences. Almost all of the principal mods and term extensions have been on portfolio loans. The real question is whether those loans are performing better and if that performance is because of the modification type. The first issue is easy enough to answer, and hopefully OCC/OTS will have that in their future reports. The second issue, though, causality, or even a meaningful correlation, is a bear because (1) there aren't a lot of principal reduction mods to look at and (2) they are likely a heterogenous lot, where LTV might have been particularly high. It's also worth noting that not all principal reductions are equal. There's a world of difference between putting a borrower into positive equity and simply reducing the depth of negative equity. Almost all principal modifications still leave the borrower with negative equity, lest the borrower then refinance.

Regarding principal reductions, I suggest you undertake some more research on Ocwen and Litton; they don't prove your case at all. Ocwen and Litton have done some principal reductions, but they are red herrings. They are relatively small operations. As of 2008, Litton was the 17th largest servicer by dollars of mortgages outstanding, and Ocwen was the 24th. Together they had 1% of the market. Today it is probably less because of the high run-off rates on their portfolios (due mainly to foreclosures). Ocwen and Litton also have a history as long-time subprime servicing specialists. Their particular value-added is on the default servicing side. Ocwen also had some unique funding factors that might have encouraged it to do principal mods. Ocwen and Litton can do principal mods because their PSAs are either silent or vague on the issue. But they are not silent or vague on term extension, and I do not believe they are extending terms beyond what the PSAs explicitly allow. (They can still run out amortization periods in most cases, however.) In short, Ocwen and Litton are pretty unique within the servicing industry. I would be surprised if their inclusion would change the data significantly, if only because of their size.

Bottom line is that there are a lot of factors at play affecting loan mods. Some relate to the legal language of PSAs. Others relate to the incentives and business structures created by securitization. And some have nothing to do with securitization per se. I don't think your data supports the conclusion that securitization doesn't matter. Rather, it is consistent with securitization being one of multiple factors involved.