The financial system is working through a major shock. It started with problems in the subprime mortgage market but has spread to securitisation products and credit markets more generally. Banks are being asked to absorb more risk – moving off-balance-sheet assets back onto their balance sheets – when their ability to do so is reduced by massive losses. The result is a bank credit crunch as the scarcity of bank equity capital is forcing banks to limit exposure to new risk.
Origins of the turmoil
There are both old and new components in the origins of the subprime shock. The primary novelty is the central role of “agency problems” in asset management.
In previous real estate-related financial shocks, government financial subsidies for bearing risk seem to have been key triggering factors, along with accommodative monetary policy, and government subsidies played key roles in the most severe real estate-related financial crises. While the subsidisation of borrowing also played a role in the current US housing cycle, the subprime boom and bust occurred largely outside the realm of government-sponsored programmes.
Investors in subprime-related financial claims made ex ante unwise investments, which seem to be best understood as the result of a conflict of interest between asset managers and their clients. In that sense, sponsors of subprime securitisations and the rating agencies – whose unrealistic assumptions about subprime risk were known to investors prior to the run up in subprime investments – were providing the market with investments that asset managers demanded in spite of the obvious understatements of risk in those investments.
The subprime debacle is best understood as the result of a particular confluence of circumstances in which longstanding incentive problems combined with unusual historical circumstances. The longstanding problems were (1) asset management agency problems of institutional investors and (2) government distortions in real estate finance that encouraged borrowers to accept high leverage when it was offered. But these problems by themselves do not explain the timing or severity of the subprime debacle. The specific historical circumstances of (1) loose monetary policy, which generated a global savings glut, and (2) the historical accident of a very low loss rate during the early history of subprime mortgage foreclosures in 2001-2002 were crucial in triggering extreme excessive risk taking by institutional investors. The savings glut provided an influx of investable funds, and the historically low loss rate gave incentive-conflicted asset managers, rating agencies, and securitisation sponsors a basis of “plausible deniability” on which to base unreasonably low projections of default risk.
What is the evidence for this? How do we know that asset managers willingly over-invested their clients’ money in risky assets that did not adequately compensate investors for risk?
Detailed analyses by Joseph Mason and Joshua Rosner, by the IMF, and by others describe in detail why the assumptions that underlay the securitisation of subprime mortgages and related collateralised debt obligation (CDOs) were too optimistic. These facts were known to sophisticated market participants long before the subprime collapse.
Consider, for example, rating agencies assumptions about the underlying expected losses on a subprime mortgage pool. They assumed a 6% expected loss on subprime mortgage-backed securities pools in 2006 – a number that is indefensibly low. Expected losses prior to 2006 were even lower. Independent observers criticised low loss assumptions far in advance of the summer of 2007.
The 6% assumption is not a minor technical issue. It was hugely important to the growth of subprime mortgage-backed securities in the four years leading up to the crisis. It goes a long way toward explaining how subprime mortgages were able to finance themselves more than 80% in the form of AAA debts, and more than 95% in the form of A, AA, or AAA debts, issued by subprime mortgage-backed securities conduits.
So long as institutional investors buying these debts accepted the ratings agencies’ opinions as reasonable, subprime conduit sponsors and ratings agencies stood to earn, and did earn, huge fees from packaging loans with no pretence of screening borrowers.
Where did expected loss estimates come from?
How were the low loss assumptions justified, and why did institutional investors accept numbers ranging from 4.5% to 6% as reasonable forward-looking estimates of expected pool losses?
Recall that subprime mortgages were a relatively new product, which grew from humble beginnings in the early 1990s, and remained small even as recently as several years ago; not until the last three years did subprime origination take off. Given the recent origins of the subprime market, which postdates the last housing cycle downturn in the US (1989-1991), how were ratings agencies able to ascertain what expected losses would be on a subprime mortgage pool? A significant proportion of subprime mortgages defaulted in the wake of the 2001 recession. Although the volume of outstanding subprime mortgages was small, a very high proportion of them defaulted; in fact, only in the last quarter has the default rate on subprime mortgages exceeded its 2002 level. The existence of defaults from 2001-2003 created a record of default loss experience, which provided a basis for the 6% expected loss number.
Of course, this was a very unrepresentative period on which to base loss forecasts. Low realised losses reflected the fact that housing prices grew dramatically from 2000 to 2005. In a flat or declining housing market – the more reasonable forward-looking assumption for a high-foreclosure state of the world – both the probability of default and the severity of loss in the event of default would be much greater (as today’s experience demonstrates). The probability of default would be greater in a declining housing market because borrowers would be less willing to make payments when they have little equity at stake in their homes. Loss severity would be greater in a declining housing market because of the effect of home price appreciation on lenders’ recoveries in foreclosure.
This error was forecastable. For the most part, the housing cycle and the business cycle coincide very closely. Most of the time in the past (and presumably, in the future) when recession-induced defaults would be occurring on subprime mortgages, house prices would be not be appreciating. This implies that the loss experience of 2001-2003 (when house prices rose) was not a good indicator either of the probability of foreclosure or of the severity of loss for subprime mortgage pools on a forward-looking basis. Anyone estimating future losses sensibly should have arrived at a much higher expected loss number than the 4.5%-6% numbers used during the period 2003-2006.
Another reason that the expected losses were unrealistically low relates to the changing composition of loans. Even if 6% had been reasonable as a forward-looking assumption for the performance of the pre-2005 cohorts of subprime borrowers, the growth in subprime originations from 2004 to 2007 was meteoric, and was accompanied by a significant deterioration in borrower quality. Was it reasonable to assume that these changes would have no effect on the expected loss of the mortgage pool? The average characteristics of borrowers changed dramatically, resulting in substantial increases in the probability of default, which were clearly visible by 2006 even for the 2005 cohort.
Of course, investors could have balked at these assumptions as unrealistic, precisely because they were based on a brief and unrepresentative period. Why didn’t they? Because they were investing someone else’s money and earning huge salaries, bonuses, and management fees for being willing to pretend that these were reasonable investments. And furthermore, they knew that other competing asset managers were behaving similarly and that they would be able to blame the collapse (when it inevitably came) on a surprising shock. The script would be clear, and would give “plausible deniability” to all involved. “Who knew? We all thought that 6% was the right loss assumption! That was what experience suggested and what the rating agencies used.” Plausible deniability was a coordinating device for allowing asset managers to participate in the feeding frenzy at little risk of losing customers (precisely because so many participated). Because asset managers can point to market-based data and ratings at the time as confirming the prudence of their actions on a forward-looking basis, they are likely to bear little cost as the result of investor losses.
Official input and managerial incentive problems
Various regulatory policies unwittingly encouraged this “plausible deniability” equilibrium. Regulation contributed in at least four ways.
- Insurance companies, pension funds, mutual funds, and banks all face regulations that limit their ability to hold low-rated debts, and the Basel I and II capital requirements for banks place a great deal of weight on rating agency ratings.
By granting enormous regulatory power to rating agencies, the government encouraged rating agencies to compete in relaxing the cost of regulation (through lax standards). Rating agencies that (in absence of regulatory reliance on ratings) saw their job as providing conservative and consistent opinions for investors changed their behaviour as the result of the regulatory use of ratings, and realised huge profits from the fees that they could earn from underestimating risk (and in the process provided institutional investors with plausible deniability).
- Unbelievably, Congress and the Securities and Exchange Commission (SEC) were sending strong signals to the rating agencies in 2005 and 2006 to encourage greater ratings inflation in subprime-related collateralised debt obligations!
In a little known subplot to the ratings-inflation story, the SEC proposed “anti-notching” regulations to implement Congress’s mandate to avoid anti-competitive behaviour in the ratings industry. The proposed prohibitions of notching were directed primarily at the rating of CDOs and reflected lobbying pressure from ratings agencies that catered most to ratings shoppers.
Notching arose when collateralised debt obligation sponsors brought a pool of securities to a rating agency to be rated that included debts not previously rated by that rating agency. For example, suppose that ratings shopping in the first generation of subprime securitisation had resulted in some mortgage-backed securities that were rated by Fitch but not Moody’s (i.e., perhaps Fitch had been willing to bless a higher proportion of AAA debt relative to subprime mortgages than Moody’s). When asked to rate the CDO that contained those debts issued by that subprime mortgage-backed securities conduit, Moody’s would offer either to rate the underlying MBS from scratch, or to notch (adjust by a ratings downgrade) the ratings of those securities that had been given by Fitch.
- Changes in bank capital regulation introduced several years ago relating to securitisation discouraged banks from retaining junior tranches in securitisations that they originated and gave them an excuse for doing so.
This exacerbated agency problems by reducing sponsors’ loss exposures. The reforms raised minimum capital requirements for originators retaining junior stakes in securitisations. Sponsors switched from retaining junior stakes to supporting conduits through external credit enhancement (typically lines of credit of less than one year), which implied much lower capital requirements. Sponsors that used to retain large junior positions (which helped to align origination incentives) no longer had to worry about losses from following the earlier practice of retaining junior stakes. Indeed, one can imagine sponsors explaining to potential buyers of those junior claims that the desire to sell them was driven not by any change in credit standards or higher prospective losses, but rather by a change in regulatory practice – a change that offered sponsors a plausible explanation for reducing their pool exposures.1
- The regulation of compensation practices in asset management likely played an important role in the willingness of institutional investors to invest their clients’ money so imprudently in subprime mortgage-related securities.
Casual empiricism suggests that hedge funds (where bonus compensation helps to align incentives and mitigate agency) have fared relatively well during the turmoil, compared to other institutional investors, and this likely reflects differences in incentives of hedge fund managers, whose incentives are much more closely aligned with their clients.
The standard hedge fund fee arrangement balances two considerations: the importance of incentive alignment (which encourages profit sharing by managers), and the risk aversion of asset managers (which encourages limiting the downside risk exposure for managers). The result is that hedge fund managers share the upside of long-term portfolio gains but have limited losses on the downside. Because hedge fund compensation structure is not regulated, and because both investors and managers are typically highly sophisticated people, it is reasonable to expect that the hedge fund financing structure has evolved as an “efficient” financial contract, which may explain the superior performance of hedge funds.
The typical hedge fund compensation structure is not permissible for other, regulated asset managers. Other asset managers must share symmetrically in portfolio gains and losses; if they were to keep 20% of the upside, they would have to also absorb 20% of the downside. Since risk-averse fund managers would not be willing to expose themselves to such loss, regulated institutional investors typically charge fees as a proportion of assets managed and do not share in profits. This is a direct consequence of the regulation of compensation, and arguably has been a source of great harm to investors, since it encourages asset managers to maximise the size of the funds that they manage, rather than the value of those funds. Managers who gain from the size of their portfolios rather than the profitability of their investments will face strong incentives not to inform investors of deteriorating opportunities in the marketplace and not to return funds to investors when the return relative to risk of their asset class deteriorates.
Propagation of the turmoil
When it comes to the shock’s spread, much is familiar. As usual, the central role of asymmetric information is apparent in adverse selection premia that have affected credit spreads and in the quantity rationing of money market instruments. But there is an important and favourable novelty – the ability and willingness of banks to raise new capital. As of mid-June 2008, financial institutions had raised over $300 billion in new capital to mitigate the consequences of subprime losses.
This novelty is especially interesting in light of the fact that the subprime shock (in comparison to previous financial shocks) is both large in magnitude and uncertain in both magnitude and incidence. In the past, shocks of this kind have not been mitigated by the raising of capital by financial institutions in the wake of losses. This unique response of the financial system reflects improvements in the US financial system’s diversification that resulted from deregulation, consolidation, and globalisation.
Another unique element of the response to the shock has been the activist role of the Fed and the Treasury, via discount window operations and other assistance programmes that have targeted assistance to particular financial institutions. Although there is room for improving the methods through which some of that assistance was delivered, the use of directly targeted assistance is appropriate and allows monetary policy to be “surgical” and more flexible (that is, to retain its focus on maintaining price stability, even while responding to a large financial shock). Unfortunately, in the event, the Fed threw caution to the wind in its Fed funds rate cuts, driving long-term inflation expectations significantly higher over the past year. The Fed could have, and should have, maintained financial stability through surgical interventions and provided less inflationary monetary stimulus than it did in the form of rate cuts.
Near-term implications: monetary policy, regulation, and restructuring
Dire forecasts of the near-term outlook for house prices and attendant macroeconomic consequences of subprime foreclosures for bank net worth and consumption reflect an exaggerated view of downside risk. Inflation and long-term inflation expectations have risen substantially and pose an immediate threat. Monetary policy should focus on maintaining a credible commitment to price stability, which would ensure the continuing stability of the dollar, encourage stock market recovery, and therefore assist the process of financial institution recapitalisation.
Regulatory policy changes that should result from the subprime turmoil are numerous, and they include reforms of prudential regulation for banks, an end to the longstanding abuse of taxpayer resources by Fannie Mae and Freddie Mac, the reform of the regulatory use of rating agencies’ opinions, and the reform of the regulation of asset managers’ fee structures to improve managers’ incentives. It would also be desirable to restructure government programmes to encourage homeownership in a more systemically stable way, in the form of down payment matching assistance for new homeowners, rather than the myriad policies that subsidise housing by encouraging high mortgage leverage.
What long-term structural changes in financial intermediation will result from the subprime turmoil? One likely outcome is the conversion of some or all standalone investment banks to become commercial (depository) banks under Gramm-Leach-Bliley. The perceived advantages of remaining as a standalone investment bank – the avoidance of safety net regulation and access to a ready substitute for deposit funding in the form of repos – have diminished as the result of the turmoil. The long-term consequences for securitisation will likely be mixed. In some product areas with long histories of favourable experiences – like credit cards – securitisation is likely to persist and may even thrive from the demise of subprime securitisation, which is a competing consumer finance mechanism. In less-time tested areas, particularly those related to real estate, simpler structures, including on-balance sheet funding through covered bonds, will substitute for discredited securitisation in the near term and perhaps for many years to come.
1 Of course, either through external enhancement or voluntary provision of support to their conduits, sponsors may still be taking an effectively junior position, and of course, many did so by absorbing losses that otherwise would have been born by other investors.