Skip to main content
Management in Practice

Inside the CDO Market That Catalyzed the Financial Crisis

Subprime collateralized debt obligations catalyzed the global financial crisis. Where did these toxic assets come from? “Inside the CDO Machine,” a special project from the Yale Program on Financial Stability, explores the first-hand perspectives of market participants. Steve Kasoff, a former equity partner at Elliott Management, one of the world’s largest hedge funds, led the project.

A home in foreclosure in Lithonia, Georgia, in 2007.

A home in foreclosure in Lithonia, Georgia, in 2007.

Chris Rank/Bloomberg via Getty Images
  • Steven H. Kasoff
    Yale School of Management Fellow; Former Equity Partner and Head of Real Estate and Structured Products Investments, Elliott Management Corp.

Steve Kasoff, a 1993 graduate of Yale College, is a Yale School of Management Fellow and former equity partner and head of real estate and structured products investments at the Elliott Management Corp, one of the largest hedge funds in the world.

In conjunction with Andrew Metrick, Janet L. Yellen Professor of Finance and Management, and the Yale Program on Financial Stability’s Lessons Learned Oral History Project, Kasoff developed “Inside the CDO Machine,” a special project drawing on the first-hand experiences of investment professionals who participated in the financial markets tied to subprime mortgage debt, markets which ended up spreading risk throughout the financial system and catalyzed a global crisis.

Q: What is “Inside the CDO Machine”?

Andrew Metrick has an ongoing project to preserve the recollections and impressions of people who were in the public sector and in regulatory positions during the financial crisis to be sure we capture insights that may not be apparent, even with the benefit of hindsight, from just looking at the data.

I had a long career in investing and was very involved in trading subprime. After I retired, Andrew asked me to interview people who had been on the private side—investing, trading, or structuring subprime securities or derivatives—to provide perspectives that are complementary to what he had already been doing on the public side. I was excited to do it.

Q: What did you learn?

“Everybody was doing things that seemed to make sense at the time and were validated by the structure of incentives around them.”

One thing that became clear in talking to individuals in different parts of the subprime CDO market: everybody was doing things that seemed to make sense at the time and were validated by the structure of incentives around them. Not a lot of people really understood the big picture or how it was leading the world down this terrible path.

In terms of what happened, it would be very difficult to say that one thing caused the financial crisis. Like any event of that magnitude, it was a confluence of events, all happening in exactly the wrong combination. But a few things rise to the top of the list as significant contributors—leverage, ratings agencies, and derivatives.

Leverage within the financial system was extremely high. It still is, although it’s better than it was. Even with all the craziness that happened in subprime, when you ask why a relatively small market catalyzed a global financial crisis, leverage is the most significant answer.

With the ratings agencies, it’s clear in hindsight, and it was clear to some at the time, that the ratings were just wrong on subprime products, especially CDOs backed by subprime mortgages.

If ratings are wrong on a certain product, and the yield on that product is higher than anything else with a similar rating, you create a very strong incentive to buy a lot of it, particularly for large regulated financial institutions that need to strike a balance between the ratings on their assets—a proxy for riskiness—and the rate of return.

It’s worth remembering it was a very low-interest environment. Large financial institutions like banks and insurance companies needed to find yield. One of our interviewees described some European banks seeing so much competition that corporate lending, a core business, was being done at unprofitable levels to maintain longstanding relationships. And insurance companies had long-term liabilities that had been created in a higher interest rate environment, so they too were depending on investments for returns.

Because of regulatory limitations, it’s most efficient for banks to invest in AAA-rated products. You could get 10 or 20 basis points [0.1% or 0.2%] more buying a subprime-backed CDO tranche compared to any other AAA-rated product. It may not sound like a big deal, but when you apply the leverage that existed within these institutions, it was a meaningful difference.

Insurance companies tended to buy BBB- or A-rated investments, but the pattern was the same. Yields were higher for subprime backed CDOs versus anything else at those ratings.

Finally, not only did you have a problem with the ratings being wrong, but credit default swaps magnified that problem multiple times. They created new risk linked to these subprime bonds, allowing a small problem to very quickly become a much, much bigger problem.

Q: Let’s take a step back to make sure the terms are clear. What is a CDO?

To explain a CDO, I should take another step back to securitization, which is a financial technology that has existed since the ’70s and ’80s, in its modern form. Freddie Mac, Fannie May, and Ginnie Mae created the early mortgage-backed securities by putting a pool of mortgages into a trust—say a billion dollars of mortgages. Investors could buy a fractional interest in that trust, and they’d receive a share of the cash flow as mortgages were repaid.

Collateralized debt obligations, or CDOs, were the next generation of securitization, where instead of everybody owning a fractional interest in the whole pool, the pool was spilt into tranches. The senior-most tranche gets the first right to cash flows. Junior tranches take more risk through a junior right to the cash flows, so they have a higher expected yield.

It was done with many kinds of debt—corporate bonds, auto loans, student loans. Once the technology established itself, financial engineers responded to the needs of the investment community and structured deals to meet the needs of investors who were more or less risk averse or looking for a higher or lower yield.

At its core, securitization is a very positive thing for the market because it takes illiquid assets and turns them into something that can trade on capital markets, allowing liquidity, price discovery, and a wider range of investors to get exposure to it. There’s nothing inherently wrong with the tranching process either. Collateralized Loan Obligations (CLOs), which uses corporate debt instead of mortgages, performed well through the financial crisis; it remains a vibrant market today.

Mortgage-backed securities broadly have been efficient and lowered the cost of mortgages, making them more accessible to homeowners around the country and around the world. However, CDOs were at the heart of the subprime debacle. Our research focused on a specific type created in increasing volume in 2005, ’06, ’07. They were CDOs created by pooling the mezzanine or middle tranches of subprime mortgage-backed securities.

Q: These are CDOs made up of tranches from mortgage-backed securities?

Yes. It gets complicated. There was a circularity to it, but that’s what was happening.

Q: Why use mezzanine tranches particularly?

“You could take $100 million of BBB tranches of subprime bonds and put them into a CDO. Then, when that CDO issued its own tranches, two thirds would be rated AAA.”

The ratings agencies were fairly transparent about how their methodology worked. So the structurers, the people at the banks creating these different products, were constantly analyzing those rules and looking for whatever edge they could find. They were continually experimenting. It just so happened that when you optimized a CDO backed by BBB-rated subprime tranches, something surprising happened. You could take $100 million of BBB tranches of subprime bonds and put them into a CDO. Then, when that CDO issued its own tranches, two thirds would be rated AAA. Another 10% would be AA rated, and so on down the line.

That process sounds bad. In the context of subprime, it was. The reason that it didn’t work properly is, as you probably would guess, those AAA tranches issued by the CDO didn’t deserve to be AAA.

The demand for AAA-rated product was much bigger than the demand for BBB, so being able to turn BBB bonds into AAA brought tremendous liquidity to the market. The volumes exploded.

Q: Wasn’t there concern with the process that turned BBB tranches into AAA? Wasn’t there concern that the yield was higher than anything else with the rating?

Even after finishing this project, it’s still unclear to me whether buyers saw the significantly higher yield being offered on these CDOs as a red flag. Did they understand that the rating was wrong but that was OK—they viewed it as optimizing around a set of constraints in an economically efficient way? Or did they believe the rating? I imagine among these different institutions there’s a range in answers.

The dealers were all very sophisticated. Yet some bought a whole lot of this risk and lost a lot of money. They succumbed to the same incentives that other buyers did. That’s perhaps a little bit ironic because they had a window into all of the things that were happening in the subprime market.

Some dealers decided, instead of selling all the AAA tranches of the CDOs they were creating to other banks, to keep some on their own balance sheets. They considered the risk from a AAA-rated security as sufficiently remote and the capital required to hold it as minuscule, so it was viewed as an extremely high return on equity.

It didn’t hurt that, since they didn’t have to spend time selling those pieces of the CDO to somebody else, they could move on to the next deal that much faster. The incentives aligned in a way that encouraged that sort of behavior. The banks’ risk management function that should have put a stop to it was not functioning properly.

On the other hand, a couple of banks were shorting subprime; they performed much better through the financial crisis as a result. Some banks managed risk much more successfully than others. That tells you a couple of things. One, it can be done. Two, it’s not easy. Humans are fallible, and the bigger and more complex these institutions get, the harder it is to properly manage them.

Q: You were on the short side, but as you understand it, why was there a market on the long side?

In hindsight, shorting feels like an obvious trade, but at the time I felt a lot of concern. “What if we’re wrong?” A lot of firms were investing large sums on the long side. I wanted to understand the rationale of the people on the other side of the trade.

I remember going to conferences, a couple of times a year, where thousands of people involved in the broader mortgage market would get together. The people who were buying subprime would talk with the people who were shorting it. You’d debate it. It was a very polarized atmosphere because almost everybody had strong views one way or the other.

The buyers understood that the subprime borrowers were not the same credit quality as prime borrowers, but the attitude was “The mortgages might be subprime, but the collateral, the house, is still a house that can be sold if the borrower defaults.”

When you looked at the history of the housing market in the U.S. after the Great Depression, regional pockets had periodically experienced double-digit drops in home prices, but it hadn’t happened on a national level. They had the data, and nobody argued with the data. They showed the pattern held even during periods of stress—the S&L crisis, Texas in the ’80s, California in the early ’90s.

I think the data were the biggest factor for the investors on the long side. Then beyond that I think some people took comfort in the ratings. They felt that the ratings agencies knew what they were doing.

The people on the short side, including myself, believed the risks were highly correlated to one another. In other words, after an unprecedented run-up, if things went wrong, it wouldn’t be a few borrowers defaulting or a regional drop in housing prices. It would be all or nothing.

Q: Was that the investment hypothesis for the short side?

It developed over time. In 2005, what I observed among a number of hedge funds is that they gravitated to the short trade as a hedge for their general activities. Some were actually buying CDOs or other products that had mortgage risk in them and needed a close hedge. Others saw shorting subprime as a good macro hedge on economic weakness.

But as investors realized just how much fraud was happening within the mortgage origination process and how much home prices were rising above what made sense, many who had been shorting for hedging purposes started to think about it as an alpha trade. A hedge is something that you don’t think is going to actually work but protects you if bad things happen in other parts of your portfolio. An alpha short is a trade you just think is going to work. Between 2005 and ’06, a lot of hedge funds transitioned to the alpha short and really built up the trade.

To bring the pieces together, the CDOs structurers had discovered that the ratings agencies’ rules let them turn pools of BBB tranches into AAA, so you started getting a lot of it. In a normal market, that demand for BBB subprime tranches would have caused the yield to drop until it wouldn’t be economical to create those CDOs. In addition, since there were only so many subprime mortgages, eventually you’d run out of mortgages to securitize. Both of those factors should have been constraints on how many CDOs you could create.

What happened instead is the financial tool used to short CDOs effectively removed those constraints.

Q: How does someone short a subprime CDO?

The most efficient way is through a credit default swap (CDS). The CDS was originally created as a sort of insurance. If a manufacturing company has significant exposure to a buyer that pays their bills 90 days after they’re invoiced, the manufacturer could hedge exposure through a CDS. In effect, the manufacturer finds a counterparty and says, “I’ll pay you a certain amount if my buyer doesn’t default, and you pay me a certain amount if my buyer does default.” For the counterparty, it isn’t substantially different than owning bonds of the company that’s buying from the manufacturer.

CDSs became a liquid and easily tradable way to take long or short positions in the credit of a company. CDSs still exists in the corporate world. They were adapted to be used in the subprime market too.

What’s most important in this context is that the CDS technology is a synthetic process. In other words, the constraints on having enough subprime mortgages to turn into CDOs was lifted. You could create an unlimited amount of CDSs as long as you have buyers and sellers on either side. Because you had the hedge funds wanting to short subprime and large, regulated institutions wanting to buy it, the risk posed by the relatively small subprime market multiplied.

Q: Then the housing bubble popped.

An indication of just how bad things had gotten with subprime mortgages is that people began defaulting during the first 90 days after mortgages were originated. They bought counting on flipping the house for a profit, then couldn’t sell.

Mortgage originators, the brokers that normally immediately sell mortgages to banks, are usually required to buy back mortgages that default within the first 90 days. The originators started going bankrupt. That made it harder to get a mortgage. Everything that happened on the way up unraveled on the way down, but faster. Years’ worth of bubble popped in a couple of months.

Q: What happened to the investments in the CDOs and CDSs?

This is where leverage within the financial system came into play. In a well-functioning market, the risk is diffuse. But a handful of banks had bought massive amounts of AAA tranches. There was a concentration of risks in those institutions.

You had lots of winners, and a smaller number of losers who were losing tons of money. Leverage is the difference between an earnings problem and an existential problem for one institution. And it’s the difference between problems in one market and a systemic catastrophe.

The next layer of the cascade happened when large financial institutions, because of this exposure, either went bankrupt or the market viewed them as at risk of going bankrupt. The market always front runs problems when it sees them coming. When the market started to fear that one of these large organizations was in trouble, it almost became self-fulfilling. The institution was unable to roll their short-term debt at levels that are economical. People stopped trading with them. They became pariahs. It’s like a modern version of a bank run, if you want to think of it that way.

And as it unwinds, some people on the right side of a CDS, who should be coming out winners, discover their counterparty isn’t able to pay. That creates new losses which continue the cascade.

Q: The CDO Machine includes the oral histories and two summaries of what was learned, but it also has something truly unusual. Readers get a fly-on-the-wall perspective of a subprime CDO being created with “Anatomy of a Trade.”

CDOs are very complicated structures. We struggled with communicating the key points without getting lost in the details. We wanted to show how a CDO deal actually got put together—what each of the players involved in the process was thinking at the time, how they were motivated, how they were incentivized.

If we had followed the steps of an actual deal, there would have been confidentiality issues. Eventually, we gravitated toward a fictionalized set of scenes. “The Anatomy of a Trade” is an amalgamation of my experiences and observations as well as those of the interviewees.

By necessity, when you fictionalize something like this, it’s not going to be 100% accurate, but I did find that in writing it a lot of memories flooded back. While none of the characters are real people, it was not difficult to create the characters because that period is so vividly etched in my mind. I’m not Michael Lewis; I wasn’t competing with The Big Short. But hopefully I’ve managed to convey the lived reality of that time.