What Killed the Economy?

“I saw commercial mortgage securitization deals becoming increasingly complex in the years leading up to the financial crisis, and I wanted to figure out why that was the case,” he says. “The question I wanted to answer was: Did underwriters [investment banks] use deal complexity as a way to make it easier to sell lower-quality loans?”

It is a topic of natural interest for Furfine. His experience as an economist for the Federal Reserve and the Bank for International Settlements, combined with his ongoing research in real estate finance, gave him an “inherent interest” in analyzing the complexity of commercial mortgage securitization deals in the run-up to the financial crisis. That led to a specific research question and study.

“The ‘simplest’ deals in my study are still pretty complicated,” he says, “but is the variation in that complexity informative of something?” According to the empirical results Furfine found, the answer is unambiguously yes: the more complex the securitization deal, the more likely the loans within it would go bad.

Furfine emphasizes that his finding is an unusual correlation, not necessarily the solution to an economic whodunit. “A more benign way to state my research question would be: Were loans packaged into more complex deals of lower quality?” he says. “I’ve presented evidence consistent with that being true. But if that’s true, then how did those loans get there?”

His research provided important clues to answer these questions.

The Rise in Complex CMBS Deals

Furfine focused on the complexity of commercial mortgage-backed security (CMBS) deals. “It seemed like everyone thought those kind of securities were just too complicated [in retrospect],” Furfine says. “But I found there was in fact a lot of information that investors could have had about the risks they were taking. So I thought there must be something more to the complexity.”

One of Furfine’s main focuses became the number of AAA-rated tranches contained in each deal’s capital structure. CMBS deals pool many mortgage loans on commercial real estate such as office buildings. Access to the aggregate cash flow generated by the loan payments is divided into segments called “tranches,” which are then rated by ratings agencies based on their perceived risk and sold as bonds to investors. Tranches rated “AAA” represent the highest-quality, lowest-risk segments of the cash flows aggregated in the deal. “AAA means that you have priority over the cash flow [generated by the loans] relative to other bond holders in the deal,” Furfine says. In his analysis, a higher number of AAA-rated tranches was one indication that the CMBS deal was more complex.

But how would that complexity correlate with loan performance?

To correlate CMBS complexity with the performance of loans contained in each deal, Furfine accessed prospectus supplements for 337 such deals offered between 2001 and 2007. As expected, the average number of AAA tranches in a CMBS deal increased significantly during that period, from five to ten per deal. To assess loan performance, Furfine cross-referenced the loans against monthly servicer reports from 2010.

“Every month, the borrowers are making payments and there’s a report generated for every deal that tells me: Did the borrower pay this month? Are they past due? Have I started a foreclosure against the borrower?” he says. “I linked the initial information about each loan when it was made to how it was performing in 2010, and I showed that after controlling for the observable characteristics of the loan, the likelihood of a loan becoming nonperforming is positively correlated with the number of AAA-rated tranches in the deal.”

A Multidimensional Financial Puzzle

These findings, Furfine says, point to a puzzle at the heart of the financial crisis: “There’s no reason why the number of tranches in the deal could cause the loan to go bad. That’s not the source. That means that there’s some characteristic of the loans, unobservable to investors, that is causing them to default more often. What I found most puzzling is that the characteristics of the deal predict behavior of the loan above and beyond the key underwriting characteristics of the loan, like loan-to-value and debt service coverage ratios.”

Furfine notes another element of mystery in the findings, as well. Since the complexity of the deal (e.g., a high number of AAA tranches) is, in fact, observable—even without an intensive empirical study like Furfine’s—the market should reflect this complexity in the pricing of the tranches. “If investors can see a bit of information, they’re going to use it when they decide how much to pay,” Furfine says. “What I’m saying is, here’s a piece of info that’s observable – deal complexity – but I show that it’s not being used to determine price. This means that investors don’t think that deal complexity is relevant. Yet at the same time, complexity correlates with underlying loan performance, which would clearly be useful to know. So that’s a mystery.”

The Closest Thing to a Smoking Gun

Furfine has some ideas about potentially solving that mystery.

“Imagine that investors have a certain, finite amount of time in which to make their decisions,” Furfine says. “If the deal structure were very simple, all of the investors’ time would be spent analyzing the quality of the underlying loans. But if I increase the complexity of the deal, then investors have to divert some of their time to analyzing the deal structure—which necessarily means taking attention away from the loans. All else equal, this gives the underwriter an incentive to place loans of lower quality into the pool.”

As further evidence for this, another empirical result of Furfine’s paper shows that the correlation between AAA tranches and likelihood of loan default exists only in a subset of CMBS deals: deals that contain loans originated by the same underwriter who structured the deal. “In those deals, the more complex the deal was, the more likely it is the loan is going to end up in default,” Furfine says. “For securitizations that are being put together by third parties, this result doesn’t exist. It’s the closest thing to a smoking gun I have in the paper.”

One complication, however, is the quality of the loans originated by the underwriter tended to be of higher quality than other loans in the same pool. “The evidence seems to suggest,” Furfine says, “that the underwriters combined their own high-quality loans—where quality is not readily observable—with lower-quality loans originated by others in more complex securitization deals.” But he points out that the banks did not necessarily structure deals in a certain way to hide information: “The interpretation of my results focused on unobservable measures of loan quality, so something that maybe the underwriters have a sense of, but not something that would be easily transmittable to investors in some sort of table form.”

More specifically, he notes that underwriters, or investment banks, could disguise “soft” information about poor loan quality—so called because it is difficult to transmit in observable financial data like prospectuses—by embedding it in highly tranched deal structures. An example of soft information would be knowledge of whether the major tenants of a commercial office building might intend to vacate the building in the near future.

Moreover, Furfine proposes that the correlation between deal complexity and poor loan performance could be an emergent, unintended consequence of financial innovation. “One possible reason why deal complexity increased in the run-up to the financial crisis was because underwriters were catering to the demand of the investors,” Furfine says. For example, an investor might be more interested in buying a AAA bond that pays back its principal in five years versus nine years, which incentivizes the underwriter to find a way to create that financial product.

But that process could also sweep poorer-quality loans into the deal structure. “There’s a downside,” Furfine notes. “Once I have served my investor community by creating these more complicated securities, it also creates this negative incentive: I am more able to hide things from my investors.”

Once Bitten, Not Shy

Furfine observed that CMBS deals had as many as fifty tranches in the years leading up to the financial crisis. In the immediate aftermath, the first CMBS deal had only four. “This implies that investors had a potentially greater appreciation of this negative aspect of deal complexity,” Furfine says. But that appreciation has been short-lived: “A recent CMBS deal had nineteen tranches,” he says, and “deal complexity is already trending back up. It’s a case of ‘once bitten, then forget that I was bit.’”

According to Furfine, the risk-retention policies implemented in the wake of the financial crisis—those that force underwriters to retain some of the risk of their loans/bonds rather than passing most or all of it to investors—mitigates, but does not eliminate, the incentive for underwriters to tranche deals in increasingly complex ways. “It’s a little bit more costly to tranche, but if I can sell the bulk of my securities for more if I do so, then I’m going to keep doing it,” he explains.

Then again, knowledge of the fact that a negative correlation exists between complexity and loan performance could naturally undermine those incentives. “When an investment bank presents a really complicated deal, investors might push back and say, ‘I like that you’ve carved out these special securities, but on the other hand I appreciate the fact that when you did that, you probably made the loan quality worse in ways that I can’t easily measure,’” he says.

Either way, if investors do not want to get bitten twice when it comes to complex CMBS deals, they need to take the motive of the underwriter into account. “I think moving forward, investors are going to appreciate the fact that deal complexity is a choice being made by the underwriter,” Furfine says, “and as a choice, it might signal something about the underlying loans that are being securitized.”