“Markets for a wide range of securities have de facto disappeared.”
Financial Times, September 20, 2007¹

”Practical men, who believe themselves to be quite exempt from any intellectual influence, are usually the slaves of some defunct economist.”
John Maynard Keynes

It is a truism to observe that financial markets clear at asset prices that balance the demand for reward against the supply of risk that is on offer at any point in time. Textbook models — even the most sophisticated, Nobel prize winning ones — typically begin with the assumption that there is a known underlying distribution from which asset returns are drawn. Given a fixed- and known-return distribution, investors can price options, identify relative values on yield curves, and decide on optimal hedge ratios for currencies.

Of course, in practice the art of investing comes down to making informed judgments about distributions of returns that are not ‘known’ in advance. Indeed, getting the ex ante distribution of returns right is often the most important element of successful ‘risk’ management in actual financial markets, where the distribution of returns is an equilibrium outcome, not a preordained physical constant.

However, under certain circumstances, it is crucially important to distinguish between the risk of undertaking an investment and the possibility of substantial uncertainty regarding the range of possible distributions of investment outcomes that an investor may confront. This distinction is important because financial markets may function very differently in a world of uncertainty than in a world of risk, in which market participants believe they are making informed decisions about the distribution of returns on the investments they make.

Nearly a century ago, the economist Frank Knight (1921) made exactly this point.² He argued that financial markets might operate very differently in a period of uncertainty than they function during periods in which there is broad agreement among investors on the distribution of possible investment outcomes. In the words of a recent paper on ‘Knightian’ uncertainty, Rigotti and Shannon (2005)³ write:

“If risk were the only relevant feature of randomness, well-organized financial institutions should be able to price and market [financial] contracts that only depend on risky phenomena. Uncertainty, however, creates frictions that these institutions may not be able to accommodate. …An event is uncertain or ambiguous if it has unknown probability. Uncertainty and risk are distinct characteristics of random environments, and they can also affect individuals’ behavior very differently…

“Since uncertainty, as distinct from risk, can exert a significant influence on individual behavior, it should also be a significant determinant of equilibrium outcomes [in financial markets]. For example, Knight claims that risk is insurable through exchange while uncertainty is not. [Knightian] uncertainty should arguably lead to two notable departures from standard risk-sharing behavior in [financial markets]. When uncertainty is prevalent, some [financial] markets might break down, resulting in equilibrium with no trade. Moreover, indeterminacy may also arise [and this] can generate excess price volatility.”

The global financial system right now is going through a serious bout of Knightian uncertainty, which may have significant consequences for investing and global financial markets for some time to come.⁴ The proximate cause of this “hard day’s Knight” was the more or less simultaneous realization by millions of global investors that their underlying assumption about the distribution of returns on a wide “variety of asset-backed securities” was fundamentally flawed.⁵ The realization was more or less simultaneous because it was triggered by a common, instantly reported and disseminated event: the July 10^th announcement of imminent downgrade, and a revised methodology for assigning new ratings, by some rating agencies of hundreds of asset-backed securities (ABS), specifically those backed by portfolios of subprime mortgages. The consequences of these downgrades spread far beyond subprime mortgage pools to asset-backed securities secured by portfolios of other assets — corporate bonds, bank loans, automobile loans, and credit cards.

This contagion occurred, at least in part, because these different asset-backed securities were structured and rated by the agencies using a similar methodology. When the agencies downgraded hundreds of subprime asset-backed securities in July, investors in all-asset-backed securities using this methodology realized they were in fact investing in an uncertain market in the sense of Frank Knight, not just a riskier market (though it certainly is also a riskier market than investors were counting on July 9^th). That is, investors began to realize this summer they didn’t know what they didn’t know and this realization fundamentally changed the ways in which the global financial markets clear and price discovery occurs. As predicted by the economics of Knightian uncertainty, many financial markets, in effect, failed to function and price discovery was substantially impaired in those markets that did trade.⁶

In the remainder of this paper, I lay out my argument, and then distill what I think this all means for the global investment outlook over the next several months.

Risk or Uncertainty in Spread Product
An investor who buys a corporate bond, a mortgage, or a sovereign issue knows that there is some likelihood he will not be paid back in full. As a consequence, the investor’s expected return must compensate for that chance of default. If our investor buys one corporate bond that matures in one year and holds it for that one year, one of two things will happen. The investor is either paid back in full, or the borrower defaults and the investor receives a lower — usually much lower — recovery value. The recovery value will depend upon the value of collateral, if any, and the value of collateral will, in part, depend on how many other borrowers have defaulted and thus dumped collateral on the market. Ex ante, when an investor buys the bond, the price he is willing to pay, and his expected return absent default, must compensate him for that event of default – where recovery is much lower and much more uncertain - as well as any additional premium required for taking on that risk. Thus, an investor in a spread product must have a view of the probability distribution of default, in order to decide whether to buy a bond with default risk instead of a government bond of comparable maturity which is backed by the full faith of the issuing government.

There are two ways to do this. The investor can do his own fundamental research on the individual credit and the macroeconomic outlook. He needs to take a stand on the macro outlook as well as the individual credit — its cash flow, balance sheet, payment history — because the likelihood of default may be affected by both. For example, a homeowner with a given credit history as of January 1^st is more likely to default by December 31^st if he becomes unemployed during the year, and he is more likely to become unemployed if the economy goes into recession during that year. This is the way PIMCO invests. Our global credit team does its own fundamental analysis of individual credits — be they sovereigns, corporates, or mortgages — and (this is where I come in) the global macro team does its own macroeconomic analysis — be it G10, emerging markets, or U.S. We then combine our fundamental analysis of individual credits (and portfolios comprised of those credits) with our macro view to assess the richness or cheapness of these securities. Richness and cheapness, in turn, reflect our view of the probability distributions of default and recovery compared with those that are reflected in the market price. We buy (or overweight) credits that are cheap (i.e., those we think have lower default probabilities and/or greater recovery values than those reflected in the market price) and short (or underweight) credits that aren’t (i.e., those we think have higher default probabilities or lower recovery values than reflected in the market price). Of course, this approach is expensive and, as they warn on Mythbusters — young Mathew and Russell Clarida’s favorite TV show — it is something that most people should not be “trying at home.”

But there is an alternative, and historically it has served many investors well. The alternative is to outsource credit evaluation to a rating agency. The rating agencies have large staffs, do serious fundamental analysis, and distill the results of this research into a single index, a rating. Better yet, investors can outsource credit evaluation to the rating agency for free! Rating agencies can offer their ratings as a free service to investors because the agencies charge a fee to those credits that desire or require a rating to place their paper. Many investors decide that, to gain access to credit exposure, and to the higher average returns credit exposure can offer compared to government bonds, they will invest based on the credit rating, often in a portfolio of credits that have similar ratings. Take 100 A-rated credits; combine them into a portfolio, and that is one efficient way to invest in A-rated credits. A portfolio consisting of 100 A-rated credits is an A-rated portfolio, but it avoids the all–or –nothing (technically, all–or-recovery value) aspect of buying a single bond.

But, you may ask, how does a single rating — be it A, AAA, or BBB — convey information about the probability distribution of default that investors need to assess when deciding whether to invest in a security or portfolio of similarly rated securities. After all, distributions are complicated beasts — they have means, variances, skews, and tails that can be skinny or, more often, fat. And did I mention kurtosis, fourth moments, and transition probabilities? How does an investor turn a single rating index — say AA — into the information about the probability distribution of defaults that he needs to make an investment decision?

The answer is track record. The rating agencies have a track record and investors can and do take the historical track record of defaults for credits with a given rating into account to make their assessment of the probability distribution of default for credits with that rating. For example, Figure 2 illustrates the distribution of actual defaults on corporate bonds rated by Moody’s for the period 1983 to 2004.

Figure 3 shows the distribution of ratings transitions for U.S. corporate bonds over the same 21-year period.

These are distributions of default probabilities and ratings transitions that investors have come to expect when they buy corporate bonds or portfolios of corporate bonds with similar ratings. For example, this track record suggests the probability that a Baa-rated security will default in year three is about 1%, and the probability that a Baa-rated security will be upgraded to an A rating in year three is about 3.9%.

ABS Tranche Securities and the Ratings That Make Them Possible
Although the U.K. and U.S. are the global centers for financial innovation, the asset-backed spread products they make possible are distributed and held around the world. While the products differ in some important respects, they share some key similarities.⁷ The portfolios are divided into different tranches, with the riskiest tranches taking the first loss, receiving the lowest credit rating, and offering the highest yield; and with the least risky tranche taking the last loss, receiving the highest credit rating, and offering the lowest yield. The methodology requires the agencies to take a stand not so much on the value of individual credits, but instead to focus on the correlation of defaults among many credits with similar attributes. If the realized correlation of defaults is lower than the rater expected, investors in the more senior tranches would likely be paid back in full, even if defaults occur and the holders of lower-rated tranches take losses. However, if the realized correlation of defaults is higher than the rater expected, some more senior tranches, notwithstanding their higher initial rating, will default. In this way, a portfolio comprised of individual securities with BBB default risk can be packaged to offer senior tranches (those that are senior and take the later or last loss) which receive an A or even AAA rating.

It is my opinion that hundreds of billions of dollars of asset-backed securities were purchased in recent years because investors thought and were encouraged to believe that their ratings offered similar (if not identical) default distributions to those which had for decades been associated with individual corporate bonds. Since the mass downgrade on July 10, investors have concluded they were mistaken, and this rational shift in sentiment has rocked the financial markets. It has placed us in a Knightian world in which investors don’t know what they don’t know. What they do know is that the model upon which they originally based their purchases of these securities has turned out to be the wrong model.

At the 30,000 foot level (which is where I, as a macroeconomist, view such things and often, as I am now, write about them) there is nothing conceptually wrong with this approach to creating asset-backed securities. A tranche that is very senior in the capital structure is less likely to default than the first loss piece and its spread and yield should reflect this. The question now is: how much less likely to default? In other words, what is the distribution of default probabilities for these various and sundry securities? The markets after July 11 are saying, collectively, “We don’t know, but we do think we know they are not the same as the AAA-, A, and even BBB corporate bonds in which we have decades of experience investing.” 

This is a world of Knightian uncertainty, not just risk, and the global ABS market is exhibiting all the symptoms of a serious bout: markets are essentially frozen, bid-ask spreads are wide and ‘indicative’ and many investor portfolios are concentrated in corners, as in “I don’t want any ABS.” Note this is different from a portfolio selection that is long ABS based on the notion that ABS are cheap, or short ABS based on the expectation that ABS are going to get cheaper. Standard textbook models that ignore uncertainty predict that if risk goes up, portfolio composition should change, but should generally not go to zero! In a world of Knightian uncertainty, the best thing to do can be to leave the market! This is not something you learn in Finance 101 where Knightian uncertainty is assumed away.

Investment Implications
A world of Knightian uncertainty suggests several investment implications. First, as Paul McCulley said presciently back in July, these markets will not clear when they are ‘fair’, they will clear when they are ‘cheap’. A real issue with the global ABS meltdown is that fundamental valuation of these securities is exceedingly complex. There is value opportunities today in our Frank Knight marketplace, but you have to work (hard) to ascertain it. First, those firms, such as PIMCO, that have the expertise and resources can and will find value opportunities in this market, precisely because a lot of the players who were in the market on July 10^th have left. Second, as Bill Gross highlighted in his Investment Outlook in September (which was the inspiration for the present article), the realization that complex ABS structures comprised of weak credits that were sold as A-rated aren’t the same as traditional A ratings for individual corporates will have a long lasting effect on investors. The ABS market will not disappear, but going forward it will likely trade with an appropriately healthy risk premium that was all too absent in recent years. This is not a bad thing; indeed it is efficient relative to the alternative of capital being excessively allocated to borrowers that are unable to repay loans. Third, this episode of Knightian uncertainty will eventually pass. While many of these ABS structures are novel and have short track records today, current conditions will provide a track record for future stress testing of ABS tranches. In the future, portfolio managers will likely look back at 2007 as yet another ‘six sigma event’ and will simulate their exposure to see how it would do in a 2007 credit episode.⁸ This, in turn, will enable them to better calibrate risk and reward for their exposures so that today’s 2007 bout of Knightian uncertainty becomes a crucial input to the next episode’s (perhaps in 2011?) risk management exercise. 

---

 

¹ Quoting Marco Annunziata, chief analyst at Unicredit.

² Knight, F. (1921). Risk, Uncertainty, and Profit. Houghton Mifflin Company. For an illuminating discussion of Knightian uncertainty as it relates to finance, see Peter Bernstein’s Against the Gods, especially Section Four.

³ Rigotti, L. & Shannon, C. (2005). Uncertainty and Risk in Financial Markets. Econometrica.

⁴ This paper has benefited from a stimulating discussion I had with Chris Dialynas, PIMCO Managing Director and Portfolio Manager, during our recent Strategy Week.

⁵ A case that has been made for some time by Bill Gross, most recently in: Gross, B. (September 2007). Where’s Waldo? Investment Outlook. See also, Gross, B. (July 2007). Looking for Contagion in All the Wrong Places. Investment Outlook, and Gross, B. (December 2006). Reality Check. Investment Outlook.

⁶ With significant implications for the survival of what Paul McCulley has called ‘the shadow banking system.’ The point being that these opaque ABS structures were held, in part, by special purpose vehicles and conduits that financed the positions by rolling commercial paper. As the world plunged into Knightian uncertainty this summer, the entire business model of the shadow banking system has come under significant stress. This, in turn, has resulted in contagion to the LIBOR market where spreads over policy rates have widened because of the need/desire on the part of banks to hoard cash in anticipation of the collateral of the shadow banking system being returned to bank balance sheets. See Teton Reflections. (August 2007). Global Central Bank Focus.

⁷ For a useful introduction to the wonderful, if uncertain world of structured credit, see Meade, R. (May 2007).Demystifying the Structured Credit Jargon and Identifying the Opportunities. European Credit Perspectives.

⁸ As the saying goes, “On Wall Street, there is a 100-year flood every four years.”