Capital Inadequacies the Dismal Failure of the Basel Regime of Bank Capital Regulation by Kevin Dowd, Martin Hutchinson, Simon Ashby, and Jimi M

No. 681 July 29, 2011

Capital Inadequacies The Dismal Failure of the Basel Regime of Bank Capital Regulation by Kevin Dowd, Martin Hutchinson, Simon Ashby, and Jimi M. Hinchliffe Executive Summary

The Basel regime is an international sys- ness is regulatory capture. tem of capital adequacy regulation designed to The Basel regime is powerless against the strengthen banks’ financial health and the safe- endemic incentives to excessive risk taking that ty and soundness of the financial system as a permeate the modern financial system, particu- whole. It originated with the 1988 Basel Accord, larly those associated with government-subsi- now known as Basel I, and was then overhauled. dized risk taking. The financial system can be Basel II had still not been implemented in the fixed, but it requires radical reform, including United States when the financial crisis struck, the abolition of central banking and deposit and in the wake of the banking system collapse, insurance, the repudiation of “too big to fail,” regulators rushed out Basel III. and reforms to extend the personal liability of In this paper, we provide a reassessment of key decisionmakers—in effect, reverting back to the Basel regime and focus on its most ambi- a system similar to that which existed a century tious feature: the principle of “risk-based regu- ago. lation.” The Basel system suffers from three The Basel system provides a textbook exam- fundamental weaknesses: first, financial risk ple of the dangers of regulatory empire build- modeling provides the flimsiest basis for any ing and regulatory capture, and the underlying system of regulatory capital requirements. The problem it addresses—how to strengthen the second weakness consists of the incentives it banking system—can only be solved by restoring creates for regulatory arbitrage. The third weak- appropriate incentives for those involved.

Kevin Dowd is a visiting professor at the Cass Business School, City University, London, and an adjunct scholar at the Cato Institute. Martin Hutchinson is a journalist and a former investment banker who writes the weekly “Bear’s Lair” column at www.prudentbear.com. Simon Ashby is a senior lecturer at Plymouth Business School in the United Kingdom, and Jimi M. Hinchliffe is an investment banker based in London. Kevin Dowd and Martin Hutchinson are coauthors of Alchemists of Loss: How Modern Finance and Government Intervention Crashed the Financial System. At the dawn of Introduction community has been to patch up the system the crisis, the as quickly as possible. A new, improved Basel One of the most important and distinc- system is already agreed in principle, with ne- big banks in the tive features of modern central banking is gotiations underway regarding its implemen- United States and the growth of bank capital adequacy regula- tation. Its designers tell us that this new Basel Europe were fully tion—the imposition by bank regulators of regime takes on board the lessons of the crisis minimum capital standards on financial in- and the weaknesses of its predecessors, and Basel-compliant stitutions. The main stated purpose of these assure us that it will deliver a safer and more and more than regulations is to strengthen institutions’ fi- stable financial system in the future. nancial health and, in so doing, strengthen Let’s see if we understand this correctly. adequately the safety and soundness of the financial The regulators spent most of a fairly quiet capitalized. system as a whole. decade producing Basel II—which, in essence, Until the second half of the 20th century, is just thousands of pages of regulatory gob- capital regulation was fairly minimal and bledygook—designed to make sure that the often quite informal. In the 1980s, however, international banking system is safe. The the major economies attempted to harmo- banking system then collapsed shortly after- nize and expand the scope of bank capital ward. In the resulting panic, they rushed out regulation under the auspices of the Bank thousands of pages of new draft rules, plus for International Settlements’ Basel Com- many more pages of discussion and con- mittee. The defining event was the 1988 Ba- sultation documents. Indeed, the deluge of sel Accord, now known as Basel I. This “Basel regulatory material was so great that by the regime” was a rapidly expanding work-in- spring of 2010 observers were jokingly refer- progress as the regulators attempted to keep ring to it being tantamount to a Basel III. up with developments in banking, finance, But by the fall, the joke—Basel III—had be- and financial risk management. The most come a reality. But surely the real jesters were significant overhaul was Basel II, published those telling us that the solution to all that in 2004, which was the subject of protracted gobbledegook was now to have even more of negotiations around the turn of the millen- it, crafted on the fly under crisis conditions nium. Member countries had either just ad- by the very same people who had gotten it opted Basel II, as the European Union had wrong before. Now these same people ask us in 2007, or, in the case of the United States, to believe that they have gotten it right this were preparing to adopt it when the finan- time around, and never mind all those previ- cial crisis hit. ous failures. At the dawn of the crisis, the big banks With a track record like this, it must be ob- in the United States and Europe were fully vious that what is needed is not some rushed Basel-compliant and, as far as Basel was con- patch-up of the Basel system, but a serious cerned, more than adequately capitalized. reassessment from first principles. That is The crisis then revealed the true weakness what this paper seeks to provide. In particu- of the banks in the starkest possible terms. lar, it aims to offer a (fairly) readable guide to Many of the biggest banks failed, and most the policy economics of the Basel system: the of the banks that have survived are likely to underlying issues and objectives of the main remain on government life support for years players, the policy issues Basel presents, the to come. The collapse of the banking system effectiveness of the system, and how it might is, of course, the clearest imaginable evi- be reformed. It seeks to outline the big pic- dence that Basel has not worked as intend- ture and, as far as possible, avoid the vast and ed: not to put too fine a point on it, but ev- oppressive minutiae that render this subject ery ship in the fleet passed inspection—and almost impenetrable to the outsider.1 At the then most of them were lost at sea. same time, it highlights the most innova- The knee-jerk reaction of the regulatory tive and ambitious feature of the Basel sys-

2 tem: the principle of “risk-based regulation,” are reminiscent of King Canute ordering the which attempts to build a capital adequacy incoming waves to stop rising. In the case regime on firms’ own risk models using the of the modern financial system, the waves rapidly evolving practices of modern finan- are the endemic incentives to excessive risk cial risk modeling. taking, particularly those associated with We suggest that the Basel system suffers government-subsidized risk taking such as from three fundamental weaknesses. First, deposit insurance, the lender of last resort, financial risk modeling provides the flimsi- and the now-established doctrine of “Too est basis for any system of regulatory capital Big to Fail.” These waves are now so strong requirements. This is, in part, because of the that even another massive overhaul of Basel intrinsic weaknesses of such models, in part would still leave the financial system open because physical science models do not easily to being overwhelmed when the next wave carry over to social and economic problems, of crisis occurs, which will probably happen and in part because of basic economics—that sooner rather than later. is, neither the models nor the regulators’ use This said, the financial system can be fixed, of those models allow for the bankers’ incen- but to do so would require much more radi- tive to produce low-risk estimates to obtain cal reform. This would include the abolition low regulatory capital charges. Taken to- of central banking and deposit insurance, the Basel capital gether, these factors suggest that risk-based repudiation of “too big to fail,” and the estab- rules are the regulation is fundamentally unsound even lishment of reforms to extend the personal primary factor in principle, let alone in practice. liability of key decisionmakers—in effect, re- The second weakness consists of the incen- verting back to a system similar to that which driving the tives that Basel creates for regulatory arbitrage, existed a century ago. At the same time, the securitization typically taking the form of securitizations restoration of a sound monetary standard—or bonanza of the designed to produce lower regulatory capital at the very least, the abandonment of mone- charges. Indeed, it is no exaggeration to say tary policy activism typified by successive Fed- last two decades. that the Basel capital rules are the primary fac- eral Reserve chairmen Alan Greenspan and tor driving the securitization bonanza of the Ben Bernanke—would put a stop to the cen- last two decades, the main consequence of tral bank stoking the highly damaging boom- which has been to greatly weaken the financial bust cycles of recent years. The combination system by depleting it of much of its capital. of extended liability, market forces, and sound The third weakness is regulatory capture. money would then force the banks to become The key players in the modern banking sys- safe and stable again: their capital levels would tem have little interest in maintaining high rise and excess risk taking would be reined in levels of capital or even practicing serious risk because the risks would again be borne by the management. Both are drags on the short- risk takers and not innocent parties—most ob- term profit-making (via excessive risk tak- viously, the taxpayers. Within that context, the ing) that really drives the system. Again and capital adequacy “problem” would disappear; again, they have been able to overwhelm the capital adequacy regulation would become re- feeble attempts of the regulatory system to dundant and could be safely scrapped. control them, capturing the regulatory system and manipulating it for their own ends. We suggest that this weakness is simply in- A Tale of Two Banking tractable and that any worthwhile reform of Systems: American financial regulation needs to start from the premise that capture will be inevitable. Banking Then and Now These weaknesses suggest that Basel III has much the same chance of success as its The issue of bank capital adequacy predecessors—that is to say, none. Its rules is best understood in historical context.

3 American banking a century ago was very the threat of a run kept the bankers in check different from what it is today.2 In those and kept the banks themselves strong. days, there were no systems of deposit in- The system did experience occasional surance or capital adequacy regulation, and crises, but as time passed the banks evolved no central bank. The preeminent financier ways of responding to them. This private- of the time, J. P. Morgan, chose to operate sector resolution of financial crises centered his firm as a partnership with unlimited on bankers’ clubs, which would issue emer- liability. This meant that with every deal gency loans to banks in difficulty, provided his firm undertook, Morgan put his own those banks were considered sound, while personal wealth, and not just his share- throwing the unsound ones to the wolves. holding, on the line. This liability made The most famous example occurred in 1907, for conservative banking: risk taking was orchestrated by Morgan from his personal limited and credit was extended cautiously; library with no federal government involve- relationships were carefully cultivated with ment. a focus on long-term profitability and not The earlier American banking system short-term results. Even where banks oper- differed from its modern counterpart in a ated with limited liability, bank sharehold- number of other notable respects: ers were often still subject to double liability—they were responsible in bankruptcy to ●● In lending, the “originate to hold” a liability of twice their shareholding—and/ model predominated—banks would or shareholdings were often still only par- make loans from their own deposits, tially paid up, meaning that shareholders and then would hold those loans until could be required to pay up their remain- maturity. Banks focused on their core ing shareholder obligations in bankruptcy. function of credit assessment: by keep- These practices of unlimited and extended ing loans on their books, banks were liability and partially paid-up sharehold- liable for their own credit mistakes ings gave partners and shareholders strong and hence were incentivized to be incentives to exercise tight corporate gov- careful with the credit they extended. ernance, which led to careful risk taking. It was in the mutual interest of all con- Banks were also very highly capitalized by cerned to build up long-term relation- modern standards—American banks oper- ships. Therefore, relationship banking ated with capital ratios not far short of 20 was paramount, and “name”—whether percent—and, although capital was always it was the good name of the bank or expensive, there was no perception of a bank of the borrower—was all-important in Unlimited capital adequacy problem or of any need for maintaining the trust on which depos- and extended extensive systems of capital adequacy regu- itor confidence and access to future lation along modern lines. credit depended. liability and This conservatism in banking practice ●● Banks’ positions were relatively un- partially paid-up was also due, in part, to the absence of de- complicated and fairly transparent, posit insurance and any official lender of given the limited disclosure practices shareholdings last resort as they exist today in the Federal of the time. A bank depositor could gave partners Deposit Insurance Corporation and the easily assess the financial health of his and shareholders Federal Reserve, respectively. A bank always bank and decide whether to withdraw had to operate subject to the threat of a run his deposit. Similarly, a bankers’ club strong incentives in the event that it lost the confidence of its could easily assess whether a bank ask- to exercise depositors. Therefore, a bank had to culti- ing for assistance in a crisis was funda- tight corporate vate and maintain that confidence by reas- mentally sound or not. suring its depositors that it was well capital- ●● Banks did not get involved with deriva- governance. ized and not taking excessive lending risks: tives or complicated securitizations

4 that were difficult to value and even Modern American Banking The investment more difficult to risk-manage. Finan- Fast forward to the eve of the recent finan- banks were cial product innovation was a slow cial crisis and the system is transformed com- and steady process: new products were pletely. The practices of extended liability and severely usually developed by investment banks partially paid capital had fallen into disuse, undercapitalized and only became widely accepted after and the old system of shareholder capitalism just as the they had gone through the crucible of based on strong individual shareholdings had having survived a market downturn in given way to managerial capitalism, giving se- financial crisis reasonable shape. Also, there was little nior managers the upper hand. The remain- was about to hit. need for financial engineering and ing investment-bank partnerships started to (thankfully, given its recent achieve- disappear in the 1980s, led by Salomon Broth- ments) only limited ability to do it be- ers under CEO John Gutfreund. Converting fore the advent of the computer. the partnerships to joint stock form allowed ●● There was no need for a separate “risk the principals involved to get their hands on management” function in the mod- the accumulated capital that the investment ern sense of the term. In essence, risk banks had built up over many decades. The management was the assessment and CEOs of other firms on Wall Street voiced management of the risks of individual their disapproval, but then gradually fol- loans or underwritings, a task that the lowed suit themselves—with Goldman Sachs bankers of the day performed very well being the last to make the transition in the and much better than many of their late 1990s. These changes led to a significant modern counterparts. weakening of the corporate governance mech- ●● Trading activities and bonuses were anisms: senior management was no longer limited and of little significance. incentivized to take the long-run perspective ●● Last but not least, government involve- and shareholder capital was now seen as so ment was extremely limited. Not only much more “dumb money” that they could was there no central bank and no de- manipulate for their own ends. These changes posit insurance, there were also no also combined to create and intensify a ma- equivalents of government-sponsored jor disconnect between the interests of share- enterprises (GSEs), such as Fannie Mae holders and those of senior managers, whose or Freddie Mac, offering government remuneration skyrocketed. guarantees to bundles of securitized Leverage ratios soared and investment mortgages. There also were no govern- banks had average leverage ratios of 30 to 1 ment policies to push homeownership by the end of 2007.3 These were exceedingly and there was no expectation of a feder- high by traditional standards, and especially al bailout if banks got into difficulties. so in speculative boom conditions, but even these ratios greatly understated banks’ true This model worked because the prin- leverage because of the hidden risks involved ciple decisionmakers—those who extended in their off–balance sheet and derivatives po- credit—bore the consequences of their own sitions. The investment banks were now se- mistakes in a context of well controlled con- verely undercapitalized just as the crisis was flicts of interest. about to hit. For their part, the commercial It is important to note the basic logic banks were in little better shape, with average of the system: extensive liability leading to leverage ratios of about 15–20 to 1, which tight corporate governance, in turn leading also greatly understated their true leverage to conservative risk taking and maintenance because of their myriad off-balance-sheet ve- of adequate capital—in an environment sans hicles. central bank and with minimal government These incentives led to the most distinc- intervention. tive features of the modern system, including

5 a huge amount of trading activity and a shift banks could also regard themselves as being in focus toward short-term results and the “too big to fail,” thus further encouraging next bonus. Concurrently, within the banks, their irresponsible risk taking. In addition, there was a major power shift away from tra- accommodative monetary policy, especially ditional bankers and corporate financiers the Greenspan-Bernanke “put,” artificially toward the traders, whose own remunera- inflated asset prices and helped create a se- tion sometimes vastly surpassed that of the ries of asset price boom-bust cycles from the senior management, who were now grossly mid 1990s onwards. In short, government overpaid themselves. Additionally, there was and central bank interventions were now major growth in derivatives, securitizations, encouraging banks’ excessive risk taking and financially engineered products gener- and further destabilizing the financial sys- ally. Many of these “innovations” involved tem. Smugly, if candidly, Salomon Brothers’ new forms of creative, and often hidden, risk Gutfreund summed up the modern system taking, but they were also very remunerative three years ago: “it’s laissez faire until you for their designers. Some prominent exam- get in deep [trouble],”5 at which point it is ples were mortgage-backed securitizations no longer your problem, but the taxpayers.” based on subprime mortgages and credit de- The recent bailouts, of course, then made American banks’ rivatives such as credit default swaps (CDSs), these incentive problems very much worse. capital ratios fell both of which proved to be highly toxic in The modern system is therefore rather dif- sharply in the the recent crisis: the latter in particular are ferent from its pre-1914 predecessor. Today, the archetypal financial “weapons of mass weak corporate governance—itself a product aftermath of the destruction.” In short, there was a major of state interventions to limit the liability of establishment shift towards short-term risk taking, the key decisionmakers—leads to out-of-control profits of which were privatized, whereas the moral hazard problems, excessive risk tak- of deposit risks and losses were passed on to others and ing, ineffective risk management, inadequate insurance. effectively socialized. capital, and increased systemic instability. Unlike the early 20th-century model, the Instead of bank capital requirements be- modern system was characterized by exten- ing determined by the banks themselves, sive state intervention, including the estab- motivated by the desire to remain financial- lishment of a central bank in 1914 and fed- ly strong over the long term, we now have a eral deposit insurance in 1934. Over time, situation where key decisionmakers simply the former took on an ever expanding “lend- see capital requirements as a constraint on er of last resort” role and the latter offered short-term profitmaking. This creates the banks protection against bank runs. Both potential for a “race to the bottom,” in which of these developments enabled banks to the competition for profits leads banks to take more risks than they would otherwise ever lower capital standards—and, hence, an have taken and reduce both their liquidity ever weaker banking system. and their capital ratios below the levels they It was against this secular backdrop that could have successfully maintained before. the capital adequacy issue gradually rose up It was noteworthy, for example, that Ameri- the regulatory agenda. can banks’ capital ratios fell sharply in the aftermath of the establishment of deposit insurance—from 14.1 percent in 1934 to American Capital only 6.2 percent in 1945.4 In later years, after Regulation the bailout of Continental Illinois National Bank in 1984, and even more so after the American banks have been subject to bailout of Long-Term Capital Management capital regulation in one form or other since in 1998, banks could also anticipate bail- the antebellum period. This took a variety of outs if they got into difficulties. The larger forms, but one of the most significant was

6 the use of maximum leverage ratios. How- member countries to discuss banking su- ever, these were crude metrics, not least be- pervisory matters, and its initial focus was cause they made no allowance for the risks simply to establish rules for bank closures. inherent in banks’ positions. In the early 1980s, however, the committee Various attempts were made over the years became increasingly anxious about the capi- to improve these metrics. Perhaps the most tal ratios of the major international banks, important was the development of what which were deteriorating at the same time as later became known as the “standardized ap- the international environment was becom- proach” (also sometimes known as the “risk- ing more more risky. The committee sought bucket” or “building-block” approach) by to reverse this deterioration and strengthen officials in the New York Fed and the Board the banking system while working toward of Governors in the 1950s. The idea was to greater convergence across different coun- group assets by their level of riskiness. Assets tries’ national capital requirements. There- in each group would be given a fixed “risk after, the committee experienced one of the weight” and the minimum capital require- most remarkable cases of mission creep in ments were calculated by multiplying the size history. Over time, the Basel system trans- of each asset class by its risk weight. These formed into a vast transnational regulatory weights included 0 percent for assets classi- empire that spawned a huge cottage indus- fied as “riskless” (e.g., cash or short-term U.S. try of parasitic “Basel specialists” whose sole debt); 5 percent for “minimum risk” assets purposes were to interpret and implement (e.g., longer-term U.S. debt); 12 percent for the ever-expanding Basel rulebooks. This “normal risk” assets (e.g., performing loans); Basel empire is still growing strongly and, 20 percent for “substandard” assets; and 50 thanks to its own repeated failures, is likely percent for “workout” assets (i.e., those that to expand much further yet. can only be recovered by foreclosure).6 The metric on which these original Basel This new system was far from perfect, capital requirements were to be based was the as both the risk classifications and result- standardized approach, which by this time ing risk weights were crude, but it did take was already being used by a number of cen- account of the broad levels of riskiness of tral banks to determine their own national different assets based on past experience. capital standards. The new system, the Basel The system also worked tolerably well for a Capital Accord, or “Basel I,” was approved af- long time because it produced conservative ter much negotiation in 1988 and was imple- capital requirements, banks’ positions were mented in 1992. Under this system, any asset straightforward, and economic conditions on a bank’s balance sheet would be given a were fairly stable until the 1970s.7 This stan- highly judgemental risk classification and, dardized approach was to become the model dependent on that, an arbitrary fixed-risk for the later Basel system. weight between 0 and 100 percent. (The original Basel Accord dealt only with credit risk, so “riskiness” in this context referred to the The Basel empire The Basel Regime: risk of default by the issuer.) At one extreme, is still growing Basel I and Basel II a risk weight of 0 percent would be given to the safest assets, such as the government strongly and, The origin of the Basel system8 can be bonds of countries in the Organisation for thanks to its traced to the aftermath of the serious dis- Economic Co-operation and Development. turbances to banking and currency markets At the other extreme, the supposedly riskiest own repeated that followed the Herstatt bank failure in assets (e.g., corporate bonds) would be given failures, is likely 1974.9 The resulting Basel Committee on a weight of 100 percent. Assets of intermedi- Banking Supervision was to provide a co- ate risk would be given weights of 10 percent, to expand much operative forum for the central banks of 20 percent, and so on. “Risk-weighted assets” further yet.

7 The regulators’ were then obtained by multiplying the size pline of financial risk management, and its acceptance of the of each asset by its risk weight. Basel I then guidelines about good practice were readily specified minimum capital requirements for accepted by industry and regulators alike. principle of risk- core capital (“Tier I” capital, consisting of eq- They were so well received, in fact, that they based regulation uity, preference shares, and some debt-equity soon found their way into the regulators’ created an hybrids) and supplementary capital (“Tier own supervisory manuals. The relationship II” capital, consisting of subordinated debt between the regulators and the firms they obvious moral and other debt/equity hybrids), both equal regulated was, to say the least, uncommonly hazard. to 4 percent. Thus, total capital—Tier I capi- cozy, and not least because of the revolving tal plus Tier II capital—was to be at least 8 doors operating between larger institutions percent of risk-weighted assets.10 and the regulators.11 Various amendments were issued over This was also the peak of the “VaR Revo- subsequent years, the most significant of lution,” when VaR models were all the rage which was the Market Risk Amendment and only a few people recognized their weak- (MRA) in 1996. The MRA specified capital nesses. In this atmosphere, the industry had requirements for banks’ market-risk posi- no difficulty persuading the Basel Commit- tions such as positions in equity, foreign tee that such models would provide a better exchange, traded debt, commodities, and basis for regulatory capital requirements many derivatives. It also allowed approved than the standardized approach, whose banks the option of having their capital re- weaknesses were already becoming appar- quirements determined by their own inter- ent. Nonetheless, the regulators’ acceptance nal risk models or, more precisely, by mod- of the principle of risk-based regulation cre- els that used the Value-at-Risk (VaR) risk ated an obvious moral hazard from their measure that was then taking the financial point of view: it gave institutions an ad- world by storm. The adoption of the MRA ditional incentive (as if they did not have now signified that bank regulators had ac- enough already!) to produce models that cepted the principle of risk-based regula- delivered low VaR numbers. tion; thereafter, they were hooked on it. At the time, VaR models were still re- These developments took place against stricted to the estimation of market risks, the background of extensive lobbying by the but they were soon to be applied to credit industry. At the time, there was mounting and operational risks as well. Consequently, concern about financial derivatives because once the regulators had bought into the of a number of recent scandals—Procter & principle of risk-based regulation, it was Gamble, Orange County, and others in the only a matter of time before it would be early 1990s—and influential calls for greater extended to other types of financial risk as regulation of financial markets. The indus- well. In any case, the regulators had already try responded to these calls with a highly ef- accepted that the regulatory system needed fective lobbying campaign that, despite the to keep up with developments in financial occasional derivatives scandal, left the in- risk management and to provide incentives dustry more or less free to police itself. for future improvements. However, the only Apart from lobbying influential politi- realistic way to encourage a bank to choose cians and use of effective public relations, to have its capital requirements determined the industry also produced a number of re- by its risk models is if doing so leads to ports about best practices in financial risk lower capital requirements than the alterna- management—how to manage the risks of tive—after all, few institutions would freely derivatives, and so on—the most important opt for an internal model option that would of which was the Group of 30 Report in penalize them by delivering higher capital 1993. This highly influential report became requirements.12 In short, everyone, includ- the handbook of the newly emerging disci- ing the regulators, was now pushing for risk-

8 based regulation—but only when it would air—and achieving a strong and stable finan- lead to lower estimates of financial risk!13 cial system. For market risks, Basel II involved no ma- Basel II jor fundamental changes to Basel I. However, By the late 1990s, it was being argued that whereas Basel I offered a single approach to the Basel accord needed a major overhaul. calculating regulatory capital for credit risk, The banks were becoming more sophisticat- Basel II offered banks a choice of up to three ed and risk management and risk modeling different approaches, depending on the level were rapidly evolving—especially in the area of sophistication of the bank concerned. (The of credit risk. Accordingly, in 1999, the Basel term “level of sophistication” is just a euphe- Committee issued a proposal for a new capi- mism for “how big.”) The first of these is a tal adequacy framework, the main stated ob- revamped standardized approach, which has jectives of which were to better align regula- the novel feature that, for some types of cred- tory capital with banks’ risk taking and take it risk (notably, loans to sovereigns, corpora- better account of recent financial innovation tions, and banks), the risk weights are tied (e.g., in securitization), while continuing to to the borrowers’ external credit ratings; for incentivize improvements. The underlying other types of risks, however, the risk weights reality, however, was that the path to Basel are fixed. In addition, approved banks could Banks approved II was really determined by the large banks, also use their own internal models to deter- by Basel II could who wanted recognition for their models so mine their credit-risk capital requirements. use their own they could reduce their capital requirements This Internal Ratings Based approach (IRB) and further increase their profit margins.14 comes in two forms: a foundation approach, internal models After a very long and highly politicized in which banks would estimate their posi- to determine process—and a lot of industry lobbying— tions’ default probabilities and a supervisor Basel II was finally published in June 2004. would set values for loss given default, expo- their credit- The new framework involved major changes sure at default, and loan maturity; and an risk capital to the credit risk models and, for the first advanced approach for the sophisticated (i.e., requirements. time, brought operational risks within the larger) banks, in which all these parameters Basel ambit. would be estimated by a bank’s own model. Basel II was based on a new three-pillar For operational risks, Basel II also offered principle: a choice of up to three approaches, depending on the bank’s level of sophistication: ●● Pillar 1: minimum regulatory capital requirements based either on stan- ●● a basic approach in which operational dardized approaches or on banks’ own risk charges are equal to a proportion internal risk models; (usually 15 percent) of the average of ●● Pillar 2: supervisory review of institu- a bank’s annual gross income over the tions’ capital adequacy and internal previous three years; risk assessment processes; and ●● an intermediate approach in which ●● Pillar 3: market discipline to make capital charges are allowed to depend banks safe and sound, aided by the ef- on the gross income by line of busi- fective use of disclosure. ness; and ●● an Advanced Measurement Approach As with its predecessor, Basel II’s focus (AMA) in which qualifying banks are was on capital adequacy, but with the stat- allowed to use their own internal oper- ed intent of encouraging good risk-man- ational risk models to determine their agement practices—seeking to encourage a operational risk capital charges. strong risk-management culture throughout the organization, and so much other hot Thus, there is the public face of Basel II, the

9 Dr. Jekyll who offers superficially plausible in which markets operate is itself always principles about the need to build on market changing. This absence is also partly due to developments, especially in modeling and se- the variety of ways in which people respond curitization, and stresses the need to incen- to the environment and interact with each tivize capital requirements. At the same time, other. Pricing relationships fluctuate with there is also the hidden side of Basel II—the supply and demand conditions, for instance, Mr. Hyde—who represents the unscrupulous and other relationships are simply the tem- self-interest of the big banks and is solely porary consequences of unusual conditions concerned with gaming the system to achieve or the result of policies (e.g., exchange-rate lower capital charges, higher upfront profits, policies) that are apt to change. There is, and higher remuneration for himself. relatedly, a great danger of identifying spurious, but superficially plausible, patterns that are little more than accidental and have Modern Financial no serious predictive value. Risk Modeling In addition, the phenomena measured in physics typically do not change with the As we have already noted, one of the measurement itself or with the ways that the main objectives of Pillar 1 is to encourage observer uses those measurements. The well- the larger institutions to use their own risk known exception, the Heisenberg Uncertain- models to determine their regulatory capital ty Principle, is a feature of subatomic particle requirements and, in doing so, build capital physics, but it does not affect cosmology or regulations on evolving best practice in risk those problems where Newtonian physics modeling.15 gives good answers. In finance, by contrast, Yet risk modeling offers a very shaky foun- the financial equivalent of the Heisenberg dation for either capital adequacy or good principle is much more prevalent. The act risk management. Indeed, it has a number of of modeling a financial process over time— fundamental problems. such as the movement of a stock price—will often lead observers to react in ways that af- Assumption of Stable “Laws of Motion” fect the process itself—for example, by adopt- One such problem is the maintained ing a particular risk-management strategy. If belief that quantitative methods from the enough risk managers adopt the same strat- natural sciences, particularly physics, can be egy, however, then that strategy will likely af- applied mechanically to social and economic fect the dynamics of the stock price itself. problems.16 This belief is naive for a number Nonetheless, one feature that one can of reasons, but one of the most obvious is confidently identify in financial markets is that the processes governing the operation the apparently random oscillation between of financial markets (and more generally, “normal” periods, in which markets are sta- any social system) are not immutable “laws” ble, and “crisis” periods, in which markets are comparable, say, to the laws of physics. Any volatile. Most of the time, markets are fairly social system is changing all the time—as the stable: volatilities and correlations are low, Greek philosopher Heraclitus put it, you can pricing relationships are steady, markets are Risk modeling never step into the same river twice. In social liquid, credit is both cheap and easily avail- offers a very systems, including financial markets, time- able, and returns are good. However, once invariant phenomena, if they exist at all, are in a while, a crisis emerges and all the above shaky foundation the exception rather than the rule. This is one phenomena disappear: volatilities rise, corre- for either capital of the reasons why financial “rocket science” lations radicalize, relationships break down, adequacy or good is much more difficult than the real thing. credit and liquidity dry up, risk-manage- The comparative absence of stable “laws” ment strategies that had previously worked risk management. is, in part, because the broader environment well unravel, and financial institutions suf-

10 fer large losses. Financial markets have fluc- is relevant, to a large extent, begs the ques- Many tuated between these alternate states since tion at issue: after all, one does not know in correlations their inception, and one cannot predict advance whether future conditions will be what will happen in the one state from what stable or not. between happened in the other. A good analogy may This also means that if one estimated al- financial returns be with fluid dynamics: markets generally most any model using data from the “Great are merely follow a pattern of streamlined flow, obey- Moderation” period from the early 1990s to ing one set of equations, with only local in- early 2007, then that model’s calibration is temporary stances of turbulence where those equations unlikely to provide an accurate forecast for phenomena and break down. But in extreme circumstances, conditions in the recent financial crisis. A such as those of 2007–2008, the turbulence model of a portfolio of subprime mortgages often a product spreads throughout the markets, causing a calibrated on Great Moderation data would of policy. general breakdown of systems. suggest a much higher value and much less A final key difference between physical risk than was subsequently revealed in the and social models of behavior is that physi- crisis. cal models ignore the ways in which think- Among the most difficult parameters to ing agents react to, and try to anticipate, forecast are the correlations between finan- each other. From this perspective, the basic cial returns. These are important for risk physical model can be described as a “game models because correlations take account of against nature”: the intelligent human agent how the risks diversify within a portfolio— interacts with nature, and nature responds the more they diversify, the lower the port- predictably (and unintelligently) in accor- folio risk. Most of the time, in normal mar- dance with its own laws. However, the physi- ket conditions, estimated correlations tend cal “game against nature” is a poor analogy to be moderate (and, even then, they still for many important problems in econom- tend to move around quite a bit) and sug- ics, finance, and society in general. Instead, gest both a fair degree of diversification and many of these problems are characterized by a fairly low risk. However, under crisis condi- strategic interaction or economic games and tions, correlations will often suddenly radi- can only be understood if one takes account calize: portfolio diversification then evapo- of how intelligent agents interact. rates and the estimated risks rise sharply. The portfolio can experience losses that are Parameter Problems orders of magnitude larger than earlier risk Then there are the problems of calibrat- forecasts had anticipated. We saw this with ing the parameters of any model, and those the hedge fund Long Term Capital Manage- of financial markets are notoriously difficult ment (LTCM) in the summer of 1998 and to calibrate. The biggest problem, by far, is many times during 2007–2008. that parameters are usually calibrated by es- A related problem is that many correla- timates based on historical samples, and are tions are merely temporary phenomena and almost always very sensitive to the choice of often a product of policy. The standard exam- sample period, which is itself arbitrary. So if ple is the correlation between two exchange the historical sample period is one in which rates, which depends on the relevant coun- markets were quiet, then estimates will re- tries’ central banks’ policies: an exchange flect that quietness and any resulting risk es- rate can be very stable for years and then timates will be low. Conversely, if the sample suddenly jump when a central bank changes period is a volatile one, risk estimates will be its policy, which is usually in response to a high. The problem for risk modelers is that crisis. This has happened repeatedly with the they need to choose a sample period that is Mexican peso, for example. believed to be relevant for the horizon over Yet the problems of estimating correla- which they are trying to forecast, but what tions between financial returns pale beside

11 those of estimating the correlations between in that tranche losing all his investment is defaults, and these correlations are the key 0.00000004 percent, which also looks very parameters governing credit-risk portfolios. safe. The more senior tranches are appar- To start with, one has to estimate default ently even safer. So it is no wonder that the probabilities. How do you estimate the prob- upper tranches of these securities seemed to ability that a particular firm, which (typi- be so safe that their originators described cally) has not yet defaulted, is likely to do them as “supersenior”—even safer than U.S. so over the next year or so? You might, per- government debt. haps, collect data on the default histories of Second, the reality is that the very cir- other, apparently similar firms, but no two cumstances that would lead one mortgage firms are the same and often such data are to default are likely to lead others to default very limited anyway. Even if you have the as well: defaults might be related to the state luxury to choose, what is the relevant his- of the economy or housing market. To deal torical period? Defaults tend to follow the with this and be on the safe side, we might business cycle, and if you choose a relatively assume that the default correlation is 1. In benign historical sample period, your results this case, if one mortgage defaults, then they will give you little sense of what to expect if all default, and so the probability that all of The very the economy tanks. About the only sensible them default is 4 percent. Moreover, since we circumstances thing you can do is try to build a model of probably did not take account of downturn that would lead how defaults—including the possibility of conditions, the true probability of them all multiple defaults—might be related to rel- defaulting is more likely to be 20 percent or one mortgage to evant factors such as the business cycle, but more if we are talking subprime, in which default are likely you still have the problems of finding suit- case the probability that all the tranches will to lead others to able data, choosing a relevant sample period, default, including the most senior, is also 20 and so on. percent or more. In this case, what appeared default as well. Unfortunately, financial valuations and earlier to be supersenior is actually super- risk assessments are very sensitive to these toxic. elusive parameters. To give an example, In the one case, the CDO seemed to have which also illustrates what went wrong with a high value and be extremely safe, but in the collateralized debt obligations (CDOs) in other, its value is much smaller and its true 2007–2008: suppose we have a portfolio of riskiness is revealed: it all depends on what 100 mortgages, all roughly comparable, and we assume about default probabilities and, we think the probability of default on any especially, default correlations. one of them is, say, 4 percent a year. We build We see here the dangers of mark-to- a CDO with a set of tranched claims against model valuation methods, which can pro- it, with the lowest tranche absorbing the first duce notional valuations with little or no 10 defaults, and so on. Now consider two al- connection to market reality. We also see ternatives. the impact of modern financial alchemy, in First, if we now make the all-too-con- which underlying garbage can be magically venient assumption, commonly made in transformed into securities of the highest years before the recent crisis, that defaults quality: we take a bunch of subprime mort- are independent—that is, the correlation be- gages with poor credit ratings, build a CDO tween defaults is zero—then the probability and, voilà, we generate tranched securities of the lowest tranche experiencing 10 de- that are even safer than U.S. government faults (and hence of an investor in the first debt—at least until the market collapses tranche losing all his investment) is 0.22 and reality reasserts itself. percent. This looks pretty safe. Similarly, if These problems go to the heart of what the second tranche is exposed to the next 10 was wrong with the calibration of credit de- defaults, then the probability of an investor rivatives models: it is no wonder credit de-

12 rivative portfolios collapsed in value when ation engine for CDOs encouraged them to the crisis hit! design more “sophisticated” products based on the plain-vanilla CDO model. These in- Problematic Models cluded synthetic CDOs (in which pools of The subprime mortgage products also corporate bonds were replaced with credit de- point to deeper issues with the design, cali- fault swaps) and “CDO-squareds” (in which bration, and use of modern financial models. bond pools were replaced with CDOs). These In the early years of CDOs, these products products involved one very risky structure were hampered by the absence of a suitable built upon another, and the risks involved model, and practitioners hankered after some were often hidden. In short, the development equivalent to the famous Black-Scholes- of the model itself led to a spurt of highly de- Merton option pricing model of the early stabilizing financial “innovation” that could 1970s, which had kick-started the growth not otherwise have occurred.17 of options markets. The industry needed a plausible credit-risk model that took into ac- Gaussianity count the way defaults were correlated. The Another problem involves the common industry’s dreams seemed to come true in assumption that risks can be modeled by a 2000, when Wall Street statistician David X. Gaussian distribution, often known as a nor- Li proposed such a model based on a statisti- mal distribution or bell curve. The Gauss- cal approach known as a “Gaussian copula.” ian distribution (illustrated in Figure 1) is a This model could be calibrated using what- very convenient distribution and is a good ever data were available and then used to approximation to many real-world distribu- value CDOs. The industry seized on the Li tions. Unfortunately, it also provides a very model with great enthusiasm and the market poor fit to the tails of the distributions in for CDOs and CDO-derivative products took which risk modelers are (or should be) main- off. By the end of the 2007, these had grown ly interested, and there is abundant evidence to about $35 trillion in notional principal. to indicate that financial returns are far from Without going into detail, the model it- Gaussian. self contributed to the financial crisis in at A colorful example of the inadequacy of least two different ways. First, it took lim- the Gaussian occurred in August 2007, when ited account, at best, of the factors driv- Goldman Sachs’ CFO David Viniar admitted ing defaults—such as the state of the busi- to being puzzled by a series of “25–standard ness cycle—and was calibrated using data deviation moves” hitting his institution, im- from the Great Moderation period. Unfor- plying that Goldmans was very unlucky—as tunately, since such data could give no in- opposed to the more obvious explanation dication of how bad things might get in a that Goldmans was merely incompetent.18 major downturn, the models were blind to His comments were widely ridiculed and a the real risk exposures. For example, AIG number of commentators rushed to point Financial Products modeled its CDOs us- out that a single 25–standard deviation event There is ing a copula calibrated with whatever data was likely to occur only once in 10,000—or they could get—and then, to be on the safe even 100,000—years. abundant side, they stress-tested their valuations. AIG However, even these estimates were way evidence to thought they were being conservative, but off the mark. Under a Gaussian curve, a it turned out that their stress tests were not single 25–standard deviation event would in indicate that nearly stressful enough, and the model’s in- fact occur in one day in every 10137 years.19 financial returns adequate valuations contributed directly to To give a sense of magnitude, 10137 is equal are far from AIG’s ruin and to the subsequent crisis. to the number of atoms in the universe Secondly, the fact that the Li model gave (1079) multiplied by the nanoseconds since being Gaussian product designers a purportedly solid valu- the big bang (1026) times all the cells in the distributions.

13 Figure 1 Standard Gaussian and Standard Cauchy Distributions

0.4 Gaussian 0.35 x

0.3

0.25

0.2

0.15

Probability density of 0.1

0.05 Cauchy

0 −8 −6 −4 −2 0 2 4 6 8 x

bodies of all the people on earth (1023) mul- sumptions that, if biased, are biased on the tiplied by one billion . . . more or less.20 Such conservative side. After all, it is better to be an event is about as likely as Hell freezing careful a hundred times than to get killed over. The occurrence of even a single such just once. The most conservative of the event is therefore conclusive proof that fi- stable Paretians is the Cauchy distribution, nancial returns are not Gaussian—or even which is illustrated below, along with the remotely so. Gaussian for comparison. To be plausible, risk models need to be What is most striking about the Cauchy based on alternative distributions to the is its long drawn-out tails, implying that ex- Gaussian. Among the more promising distri- treme losses are much more likely than under butions are the stable Paretians: these have the Gaussian. To illustrate, under the Gauss- certain desirable theoretical properties, but ian shown in the figure, a loss of x = 4.47 (a are also plausible because the interconnect- 4.47-sigma event under the Gaussian) would edness of network interactions often gives occur 1 day in just over 1,000 years. Under rise to distributions that have stable Paretian the Cauchy, by comparison, we would expect characteristics. Many crises are characterized the same loss to occur in just over 14 trad- by a system that is stressed by a failure in a key ing days, and the waiting time for a 25-sigma node that then produces a cascade of further event is a mere two and a half months. This failures in its wake. Much like power outages is almost certainly too extreme—25-sigma and fires caused by hot dry weather, we often events do not occur that frequently—but the To be plausible, see the same effect in financial systems: the main points here are simply that (1) the like- entire system gets stressed and a key element lihood of extreme outcomes is itself extreme- risk models fails, starting off a chain reaction of further ly sensitive to the assumed distribution; and need to be based failures. In the financial context, a stable Pa- (2) the Gaussian is extremely implausible. on alternative retian model will often provide a good esti- There are further complications. Up to mate of the losses involved. this point, we have implicitly assumed that distributions to In risk management there is also a pre- we are dealing with a single type of position the Gaussian. mium on prudence: it is better to make as- whose losses have a particular statistical dis-

14 tribution. But, actually, we are often dealing These problems have been recognized The continued with many different positions, and each of since the mid 1990s. For the most part, how- widespread those has its own distribution. We can then, ever, the financial risk management profes- perhaps, estimate each of these distributions sion has ignored these problems and the VaR use of VaR and and then estimate their risk measures—but, remains the dominant financial risk measure the Gaussian in practice, we would dearly love to be able in use. One can only speculate that the con- distribution to aggregate these risk measures to produce tinued widespread use of VaR and the Gauss- an estimate of our overall risk. This turns ian distribution might have something to do might be because out to be something of a holy grail and, as with the fact that they produce low estimates they produce low so often in the risk-modeling area, we have of risk and therefore serve the interests of an answer in theory that does not really those who want low-risk estimates. estimates of risk work in practice. This answer is provided by Fortunately, there are good alternatives and therefore the statistical theory of copulas, such as the to the VaR: serve the Gaussian copula discussed above: we take each separate distribution, fit a copula (or a ●● There is the expected shortfall, which interests of those correlation structure) across them, and then shows what one can expect to lose if who want low- obtain a multivariate distribution from a tail event occurs—what to expect on which we can estimate our overall risk mea- that one bad day. This measure has risk estimates. sure. Unfortunately, this only works well in been used by actuaries for many years. toy examples. We might be able to aggregate ●● There is the probable maximum loss, over most of our market risks, for example, the biggest loss from a set of simulated but to estimate the overall risk of our market loss scenarios. and credit risk positions still remains elu- ●● There is the loss from a stress test, or sive. It follows, then, that although we can “what if?” scenario, which will often estimate the risk of this or that portfolio on reveal problems that no other method a stand-alone basis, and can occasionally ag- could identify, such as taking account gregate certain risks, we cannot reliably esti- of what might happen in hypothetical mate total risk exposures at an institutional bad-case scenarios. level and, realistically, will never be able to. These risk measures are much better than The Inadequacies of Value at Risk the VaR, but they also tend to produce high- Another major problem is that most risk er risk estimates. This is good for risk man- models use the VaR, which is a particularly agement, but will not appeal to firms that inadequate measure of risk. The VaR itself is wish to use their risk models to determine simply the maximum likely loss, where the their regulatory capital requirements. term “likely” is to be understood in probabi- listic terms. For example, the VaR at the 99 Inaccuracy percent probability (for short, the 99 percent Many risk models are also highly inaccu- VaR) over a horizon of a day would be the rate, even under ideal circumstances where highest loss we would expect on 99 otherwise risks are believed to be estimable in princi- similar days out of 100. Put another way, the ple. Suppose we wish to estimate the 99 per- VaR gives us the worst we can do on the 99 cent daily VaR on a portfolio and we suspect best days out of a 100. It does not, however, that the true distribution is somewhere be- tell us anything about how badly things tween Gaussian at one extreme and Cauchy might go on the remaining bad day. This is a at the other. Let us assume a daily volatility most unfortunate property for financial risk of 2 percent, which is not unreasonable for management, where it is the tail events—the most days. If we assume a Gaussian, then we very large losses—that matter. After all, it is would estimate the VaR to be 4.65 percent of the large losses that can be bankrupting.21 the value of our portfolio, but if we assume

15 a Cauchy, we would estimate the same VaR type of problem by imposing position limits to be 63.64 percent of the value of our port- on traders, usually determined by risk-mod- folio. Thus, we think that the true VaR is el assessment. From the traders’ perspective, somewhere between 4.65 percent and 63.64 however, those position limits are a nuisance percent—in other words, we have no real idea to be circumvented. Traders typically deter- what it is. mine where the system is weak and “game” It is therefore not surprising that VaR it accordingly: they build up positions models can sometimes fail to give the re- whose risks are underestimated or missed motest indication of the true risks to which completely by the risk model. One way to a firm is exposed. A case in point is LTCM: do this is to “stuff risk into the tails,” for its risk model estimated its daily 99 percent example, by traders selling options that are VaR in August 1998 to be about $30 million, very unlikely to pay off. Many risk systems implying that there was a 1-in-100 chance fail to pick up the risks involved and traders of a loss in excess of that amount, and yet can then sell them with impunity. A similar LTCM’s average daily losses in that month problem arises with financial engineers who were about three times that figure. are incentivized to produce highly complex This case might be extreme, but the prob- structures whose risks are incomprehen- Models can be lems with VaR systems are certainly wide- sible, but which are highly remunerative for undermined by spread and well documented, even in the their designers. the ways in which relatively benign (and hence predictable) Risk models and risk-management strat- conditions that existed prior to the current egies can also be undermined by the way in people respond crisis. To give just one example, a study by which users utilize them. A common mis- to them. Berkowitiz and O’Brien in 2001 examined take is to design a risk-management strategy the forecasting performance of sophisticated on the assumption that market prices are (and expensive) VaR models used by leading independent of what one is doing. This is a U.S. commercial banks. It found that they major problem with the big players in the were highly inaccurate and were outper- market. For example, in its last year, LTCM formed by the forecasts of simpler rule-of- had grown to dwarf its competitors. In late thumb models. August 1998, when it desperately needed to sell positions to cover its mounting losses, it Endogeneity Problems could not feasibly do so because its own size Yet another problem is that models can would have moved market prices against it be undermined by the ways in which people and so precipitate more of the losses that respond to them. Any risk-management sys- were already pushing it into its death spiral. tem involves an attempt to control intelli- A slightly more subtle mistake, made by gent, self-interested agents who will respond LTCM and also those following portfolio in- to any control system by exploiting its weak- surance strategies in the October 1987 stock- nesses. So if traders are remunerated by the market crash, is to implement a risk-manage- profits they make, and if the only way to ment strategy that fails to take account of make large profits is to take large risks, then how other parties are behaving. Ignoring how they will take large risks—regardless of the others will behave is like assuming that you interests of their employers. If the risks pay can safely get to the theater exit in the event off, they make a nice bonus; and if they do of a fire, without realizing that everyone else not, it is only the bank’s money that they will be running for the exit as well. The un- have lost and they can always get another derlying problem here is the assumption that job elsewhere. They therefore have an incen- you are in a “game against nature,” overlook- tive to take more risks than the employing ing the point that you must consider not just bank would (or at least, should) like. “dumb” nature, but other self-interested ac- Banks traditionally responded to this tors who think the same way that you do.

16 Let us say you have come up with a VaR- we do not know. But there are also unknown based risk-management strategy that calls unknowns—the ones we don’t know we don’t on you to reduce your VaR by selling risky know.”23 Mr. Rumsfeld’s splendid com- positions in the event of a crisis. This makes ments were widely lampooned, but showed sense in a game against nature, when you are an excellent grasp of statistics. the only person implementing this strategy, This is a particular problem with opera- and everyone else carries on as before. If ev- tional risks—the risks of loss due to people eryone sells in a crisis, however, then the col- or systems failures.24 They range from the lective reaction will itself exacerbate the fall risks of running out of paperclips, at one ex- in prices and create the danger of a positive treme, to the risks of bankruptcy-inducing feedback loop in which the crisis grows as events such as uncontrolled rogue trading, it feeds off itself. Some initial trigger leads which have brought down numerous finan- prices to fall and VaRs to rise. The increased cial institutions and severely damaged many VaRs then generate further sales as risk man- more, at the other. The risks that can be agers struggle to get their own VaRs back predicted are those that occur frequently— down, and the new sales cause further price running out of stationery, turnover of staff, falls and even higher VaRs. The collective at- and so on—and their probabilities and as- tempt to get individual VaRs down destabi- sociated losses are fairly straightforward to lizes the market and, paradoxically, increases estimate. The ones that matter—the prob- everyone’s VaR in the process, inflicting the abilities and losses of extreme operational high losses that the risk-management strat- risk events, such as those associated with egy was meant to avert.22 rogue trading—are much more difficult to estimate and are, in fact, unknown.25 Unknowable Risks Nassim Taleb gives a nice example in his In this particular case, if we have no idea book The Black Swan, which gives a good sense how many other investors might be follow- of the limits to quantitative risk modeling. ing the same strategy as we are, then our He describes an unnamed casino in Las Ve- risks are unknowable. In such cases, the saf- gas.26 As one would expect, the management est assumption, but one seldom made in the of the casino had a good handle on their core financial markets, is to assume the worst business risks: they understood gambling case. So, in the earlier theater example, it is odds, diversified risk across tables, and had probably best to assume that if a fire breaks good systems to counter cheating. So where out, then everyone else will probably notice did they experience their main losses? Their it, too—and we should expect a stampede for worst loss was when their star performer, the exit. In the financial context, this sug- who had a popular tiger act, was incapacitat- gests that we should seek to avoid following ed after his (hitherto friendly) tiger attacked the herd and that the best strategy is a con- him. Their second-largest loss occurred when trarian one: buying when everyone is selling a disgruntled former contractor decided to in a panic. This, of course, is easier said than settle scores with the casino by dynamiting done, because most investors lack the acu- it. The casino’s third-largest loss was caused men and resources of Warren Buffett. by a clerical employee who failed to file tax Obviously, risk can also be unknowable reports on customer winnings over a long pe- even in the absence of strategic interactions, riod of time, which exposed the casino to a such as other parties implementing the same major fine and nearly cost it its license. Their actions. To quote the former U.S. secretary of next-worse losses included, among others, The risks that defense Donald Rumsfeld, “There are known the kidnapping of the casino owner’s daugh- can be predicted knowns; there are things we know we know. ter, which led the owner to violate gambling We also know there are known unknowns; laws by digging into the casino’s coffers to are those that that is to say we know there are some things pay her ransom. The point is that none of occur frequently.

17 Financial these particular events could have been antic- spoken. The [Bank] officials in the modeling is ipated in advance: the “unknown unknowns” room were aghast.27 will always be there and we can never measure highly imperfect them, by definition. To draw this discussion together, finan- and very cial modeling is not some grounded process judgmental. Undermining by Senior Management like rocket science, where you can learn the Perhaps the most difficult and intrac- mechanics, build the model, and then send a table institutional problem with good risk man to the moon. On the contrary, financial modeling—and one of fatal significance for modeling is highly imperfect and very judg- the effectiveness of risk-based regulation—is mental. Many important risks cannot be the fact that risk managers report to senior reliably estimated and risk models are open management and, in most modern financial to all manner of manipulation and abuse. institutions, senior managers have an inter- Consequently, risk modeling cannot provide est in risks being underestimated. This issue a solid foundation for capital regulation. In- was highlighted by a revealing anecdote told stead, it is a game like fiddling one’s tax re- by Andy Haldane, the Bank of England’s di- turns, a process in which integrity is under- rector for financial stability, in early 2009: mined by self-interest—and in the modern financial system, it pays to underestimate A few years go, ahead of the pres- your risks. In short, the very principle of ent crisis, the Bank of England and risk-based regulation—the use of firms’ own the Financial Services Authority [the risk models to set regulatory capital require- UK’s financial services regulator] comments—is unsound, in theory as well as in menced a series of seminars with finan- practice. cial firms, exploring their stress-testing practices. The meeting of that first group sticks in my mind. We had asked Weaknesses of the firms to tell us the sorts of stress which Basel System they routinely used for their stress tests. A quick survey suggested these were Leaving aside the issues associated with very modest stresses. We asked why. risk-based regulation per se, there are also Various lame possibilities were other major problems28 with the Basel sys- discussed, but eventually the real tem.29 explanation came out: one of the bankers blurted out that: The Process that Produced It “There was absolutely no incen- The first problem is that the Basel system tive for individuals or teams to run is the result of a highly politicized interna- severe stress tests and show these to tional committee process, involving many management. First, because if there arbitrary decisions, irrational compromises, were such a severe shock, they would and horse-trading. Basel insiders talk of tight very likely lose their bonus[es] and deadlines, jetlag, stress, intense political pres- possibly their jobs. Second, because sures, and stand-up shouting matches.30 in that event the authorities would Such a process almost inevitably produces have to step in anyway to save a a groupthink mentality that leads to poorly bank and others suffering a similar thought-out rules, a compliance culture, and plight.” no thought given to the costs involved. Un- All of the other assembled bank- der pressure from the industry, it also led to ers began subjecting their shoes to an obsession with risk modeling and capital intense scrutiny. measurement, with those involved becom- The unspoken words had been ing so bogged down with the risk metrics

18 that they lost all sight of the risks. Over time, sel II rulebook at a public conference, sighed it also led to ever-longer rulebooks that at- and said, “It does read a bit as if it has been tempt to standardize practice in an area written without adult supervision.”31 Such where practice is always changing and where comments by the very people who write the the development of best practice requires rulebooks make external criticism redun- competition—not standardization—in risk- dant. management practice. Instead, the product is an inflexible and hard-to-change rulebook The Standardized Approach that is out of date before it is even published, Then there are problems with the stan- not to mention its onerous implementation dardized approach in the Basel system, and costs. Given the history of the Basel system, in particular with its arbitrary risk weights. we are tempted to say that the only quantita- In the original Basel Accord, the debts of all tive rule that withstands serious scrutiny is OECD governments were given a risk weight that the effectiveness of capital regulation is of zero, implying that all OECD government inversely proportional to its quantity. debt (including, say, Greek government Another problem is that the outcome of debt) is in fact perfectly safe. The original ac- such a process is inherently unpredictable. cord also gave weights that implied that all Indeed, at one point in the tortuous and corporate debt is equally risky, so corporate In the original prolonged Basel II gestation process, there AAA is just as risky as junk: the effect of this Basel Accord, was so much uncertainty about the outcome was to encourage banks to get into junk and the debts of that some banks had created a new risk cat- get out of high-quality assets. Another dam- egory called “Basel II risk” to deal with it! aging anomaly was the arbitrary assignment all OECD Once the Basel II juggernaut was let loose, no of only a 50 percent weighting to loans se- governments one could anticipate where it would eventu- cured by first mortgages; this led to an exces- ally end up. sive concentration by the banking system on were given a risk It is also a curious paradox that although largely unproductive and highly risky real- weight of zero. the regulations are approved by the commit- estate lending. Such anomalies encouraged tees that produce them, individual members the very risk taking that the system was sup- of those committees are notoriously reluc- posed to counteract. tant to defend them when speaking on their Furthermore, some very obvious key own account. It is as if each member under- weaknesses survived into Basel II. One was stands that the rules are indefensible and the assignment of unduly low risk weights is too embarrassed to defend them, but he on liquid assets; this encouraged banks to still feels obliged to stand up for the group- become overexposed to low-maturity sources think process that produces them—suggest- of funding and leave themselves with liquid- ing that committee members are so worn ity problems in a crisis. This happened dur- down by the pressures involved that getting ing the Asia crisis and again in the prelude it agreed was more important than getting it to 2007–2008. Another weakness was the as- right. One of us also fondly recalls the time signment of very low risk weights for some when a very senior Basel official, who shall positions that were in fact very risky (e.g., be nameless, gave a seminar on Basel, and credit derivatives)—as mentioned already, then privately admitted afterwards that all this was a bad anomaly under Basel I that this financial regulation would be unnec- was made very much worse under Basel II. essary if we had free banking and held the There was also much criticism of the bankers personally accountable for their de- “adding-up principle,” by which risk-adjust- cisions—but she could not say this in pub- ed assets were simply added together with lic. Then there is the story of another very no allowance for the possibility of diversifi- senior regulatory official who, in looking cation. Critics had a field day poking holes over the hundreds of pages of the new Ba- in this, not least because it was inconsistent

19 with the most basic principles of portfolio poor ratings was overwhelming. This was es- theory. However, in retrospect, this limi- pecially so for newer products such as CDOs, tation is more theoretical than real; other where the absence of a long track record things being equal, adding up risk-weighted made such inflation easier to hide.33 assets will overestimate the total risk expo- However, in the longer term, competition sure because it implicitly sets correlations among ratings agencies can only lead to rat- to their worst-case values. However, this can ings deterioration if the investors are also be defended as a prudent approach when complicit in this process. If the investors re- the true correlations are unknown. Further, ally wanted reliable ratings, and issuer-paid- from a risk-management perspective, risk for ratings are unreliable, then they would measures should be biased on the conserva- start to pay for them again, and the practice tive side because it is better to overestimate of issuers paying for rating would disappear them than to do the opposite—especially as those ratings gradually lost their credibil- when portfolio diversification can vanish ity and their market. overnight in a crisis. So why should investors, of all people, be complicit in ratings deterioration? The an- Reliance on External Ratings swer is that most of the large institutional Another major weakness relates to the investors suffer from much the same short- reliance of the system on external ratings. term moral-hazard problems as the banks. The conventional wisdom about the ratings Their managers, too, were getting rich from agencies is that ratings are reliable because managing other people’s money, and they the ratings agencies’ long-term livelihood have every reason to go along with the pre- depends on their reputation and, with each tense of good ratings. The last thing any rating they give, they put their reputation investment manager wants to do is turn lu- on the line.32 Competition then ensures crative business away, and the fact that the that the industry is fairly efficient, quality is ratings seemed good and everyone else was good, and fees are reasonable. doing the same thing gave them plausible The reality is a little different. To begin deniability.34 with, the old practice of investors paying for Other factors helping to explain why rat- ratings has largely given way over the last ings might be unreliable are the models the 40 years to ratings being paid for by issu- agencies were using and the uses to which ers. Whereas investors want honest ratings, they were put. In the run-up to the finan- the issuers want favorable ones, so this shift cial crisis, virtually everyone was using poor puts pressure on the agencies to accommo- models. The agencies were therefore making date their clients. A ratings agency that is the same mistakes as other people, assuming too strict will lose business to more agree- that defaults were independent and, at best, able rivals. It also became common for is- using copula models that were unreliable Adding up risk- suers to tell the ratings agencies the ratings because they were calibrated using historical they were looking for and have the agencies data from an unusually stable economic pe- weighted assets suggest how the securities could be tweaked riod. They were also making a few additional will overestimate to get the desired rating. Such practices fur- mistakes of their own. To give an example, the total risk ther intensified the pressures on the agen- just as the real estate market was approach- cies to lower their standards. ing its peak, Standard and Poor’s was pric- exposure because Over time, the agencies’ business grew ing mortgage securities using a model that it implicitly sets enormously and, by the early years of the assumed that real estate prices could only go correlations to new millennium, they were rating hundreds up—in other words, S&P was using a risk of thousands of securities and their individ- model that ignored the main risk involved!35 their worst-case ual tranches. By this point, the evidence of A third factor is the government—or values. ratings inflation and sometimes downright rather, its agencies.36 In 1973, the Securities

20 and Exchange Commission (SEC) revised its tors are silent on the problems of Gaussian- The SEC’s capital rules for broker-dealers, and the new ity for much the same reasons that they are attempt to rules were dependent on risk assessments silent on the inadequacies of the VaR itself. that reflected external ratings. The SEC then Another problem is that, for market protect ratings worried that ratings agencies might start sell- risks, the VaR-based capital requirements quality had the ing favorable ratings, so it decided that only are based on a 99 percent VaR with a 10-day opposite effect. Nationally Recognized Statistical Rating forecast horizon. This, in itself, produces a Organizations (NRSROs) would have their risk estimate that, if the model is correct, ratings recognized by the SEC. The NRSROs should be exceeded on one 10-day period effectively became an SEC-regulated cartel: out of 99 such periods—a little more than the existing agencies were grandfathered in, once every three years. If regulatory capital with access to a now-captive market; new en- requirements were made equal to such a risk trants were deliberately excluded; and there measure, then an institution holding only were the predictable effects on innovation, that portfolio would expect to be bankrupt- quality, and fees. The SEC’s attempt to pro- ed about every three years. Such a capital lev- tect ratings quality had the opposite effect. el is of no effective use if we wish to set capi- If the SEC and its counterparts really want tal requirements for solvency purposes, and to enhance ratings, they could start by dis- the avoidance of bankruptcy is—or at least mantling the NRSRO cartel and establishing should be—the main concern of any capital a free ratings market.37 requirements, regulatory or otherwise. The Basel regulations seek to get around Regulatory VaR this problem by imposing an arbitrary mul- Another glaring problem is the regula- tiplier on the VaR, in which the capital re- tors’ continuing encouragement of the dis- quirement should be a multiple of the 99 credited VaR risk measure, whose only pos- percent 10-day VaR. The value of this mul- sible rationale is that it is very convenient tiplier, affectionately known as the “hysteria for financial institutions, which want their factor,” is determined by a bank’s regulatory regulatory capital requirements to be as low supervisor based on his or her assessment as possible. It is also striking that regulato- of the quality of the bank’s risk models, and ry documents are deafeningly silent on the is set at some value between three and four. weaknesses of the VaR, and there seems to be So, in the worst case, the bank’s regulatory virtually no pressure by banking regulators capital requirement would be equal to four to replace the VaR with a more suitable (i.e., times the 99 percent 10-day VaR. Unfortu- more conservative) risk measure for regula- nately, this “fix” gives us no idea of how safe tory purposes. We can only speculate that the resulting capital requirements might this is due to the influence of the industry, turn out to be: this in fact depends entirely which itself continues to make widespread on the probability distribution. use of the VaR. Once again, a comparison of the Gauss- As an aside, there also seems to be little ian and the Cauchy is very informative: regulatory pressure to do away with woeful- ly inadequate distributional assumptions, ●● Under a Gaussian, wiping out this level particularly the assumption of Gaussianity. of capital would require a 9.305-sigma If the VaR and Gaussianity are each likely to 10-day loss event—an event so remote produce major underestimates of true finan- it would never happen. The capital re- cial risk when used on their own, the com- quirement then appears to be conser- bination of the two is especially dangerous vative to a paranoid degree. and, in many cases, is virtually guaranteed ●● Under a Cauchy, on the other hand, to massively underestimate true risk expo- the probability of the capital being sures. We can only speculate that the regula- wiped out in any 10-day period is 3.41

21 percent. This suggests that we could in advance and adjust capital requirements expect to see the capital wiped out accordingly. However, even the best market more than once a year, and is anything practitioners have difficulty anticipating but conservative. market turnarounds, and their own track record suggests that the regulators them- Consequently, the regulatory use of the selves are usually well behind the market.39 99 percent 10-day VaR is highly unsatisfac- Attempts by regulatory authorities to cycle- tory over and above the weaknesses of the adjust risk assessments have also proved to VaR itself. The root problem here is that one be less than satisfactory.40 cannot extrapolate from moderate one-in-a- From a risk-modeling perspective, per- hundred type of risk events to the extreme haps the best practice is to take a long-term one-in-tens-of-thousands type of risk events view and estimate risks (whether those be that are relevant for solvency. As Riccardo Re- ratings or other risk measures) while paying bonato nicely put it, no study of the height particular attention to what might happen of hamsters, however good, will ever give us on the downside of the cycle, and then try much sense of the height of giraffes.38 to make worst-case assumptions. Applied to capital charges, regulatory or otherwise, The Basel system Procyclicality this would lead to capital requirements that produces capital Ideally, capital requirements should be were fairly high and steadier across the cy- requirements anti-cyclical. They should rise as the econo- cle.41 Some degree of anti-cyclicality could my booms (and so help to counter the boom also be achieved by making capital charges that are as it approaches it peak) and fall as the econ- dependent on easily observed variables that pro-cyclical. omy goes into recession. Unfortunately, the move with the cycle, such as reported profits. Basel system produces capital requirements Beyond that, we would caution against that are pro-cyclical. This procyclicality is sophisticated attempts to resolve this prob- driven by the procyclicality of the risk esti- lem, especially by regulators. The real answer mates themselves—whether those of ratings, to this problem, to the extent there is one, is credit-risk models, or risk models. When the to be found in economics rather than statis- economy is booming, markets appear less tical forecasting. Key decisionmakers should risky and perceptions of risk typically dimin- be incentivized to take responsibility for ish as the economy approaches its peak. The what might happen in a downturn, rather implication is that capital charges will be than kick the can down the road. We will lowest and lending highest just at the point come back to this later. when the danger to the economy is greatest. This problem affects any form of risk- Systemic Instability related capital charging and applies both to Another problem is that the Basel sys- regulatory capital charges and to the inter- tem tends to promote systemic instability.42 nal charges that firms might apply them- This is, in part, because any weaknesses in selves. From the perspective of Basel, this the Basel rules have systemic potential. Any problem goes directly against the main pur- weakness in those rules is likely to affect all pose of the system, which is to make the fi- banks at much the same time, whereas the nancial system more stable. same weakness in any one institution’s risk The usual response to this problem is to management will be unlikely to have any suggest that capital requirements should be systemic impact. Thus, Basel imposes its designed to move counter to the cycle. This, own weaknesses upon everyone. however, is much easier said than done. Pro- However, the problem goes deeper. If we ponents of capital regulation often glibly were dealing with a single institution, we suggest that some supposedly “wise” regula- might advise it to move out of risky posi- tor would see the turning points in the cycle tions when a crisis occurs. This might make

22 sense for a single player, but the market as though the bank’s exposures were the exactly a whole cannot divest itself of risky posi- same as before. In effect, he argued that relations—someone has to hold them. There is, beling his positions warranted lower capital consequently, a fallacy of composition in requirements—and amazingly, the regula- which the individual institution can sell, but tors bought the argument. Frank Partnoy the market cannot. The collective attempt wryly commented that this was like the own- to dump such positions then sends prices er of a three-story house persuading the local down sharply and creates the vicious spiral municipality that it was really three separate we discussed earlier, in which the collec- pieces and that only the ground floor should tive attempt to move out of risky positions be subject to property tax.43 makes those positions even riskier. The This pathbreaking innovation paved the fundamental problem is, then, that the en- way for traded securitizations: you take any couragement (by regulators or anyone else) portfolio (of bonds, mortgages, etc.) and se- of a single risk-management strategy can it- curitize them through a special-purpose ve- self destabilize the market. Market stability hicle. The tranching will ensure that all but instead requires players who pursue differ- the most junior tranches obtain investment- ent strategies: when many firms are selling grade ratings, and the more senior tranches in panic, we need other institutions willing can then be sold off. Often the recipients will to move in and buy. In short, any regulatory be other banks, and the high credit ratings risk-management standard is inherently sys- on these securities will ensure correspond- temically destabilizing. ingly low regulatory capital requirements. The net result is that the same underlying as- Manipulating and Gaming the Rules sets have been reallocated among the banks A final problem with the Basel system is to obtain lower capital requirements. Any as- the way in which it is open to manipulation sets (including previously securitized assets) by regulated financial institutions. Institu- can be packaged up and recycled in this way, tions will lobby for changes in the rules and with each successive securitization leading to leniency on new products, and they have the further “capital release” and large profits for influence and resources to ensure that they those involved, and never mind what might often get their way. More often than not, happen down the road. The regulatory capi- manipulation of the system takes the form tal system thus helped create a securitization of regulatory arbitrage, in which institutions bonanza that lasted a quarter of a century play the existing rules to their own advan- and depleted the banking system of much of tage. This typically takes the form of regula- its capital.44 tory-driven securitizations leading to lower Another example was J. P. Morgan’s now capital requirements, higher short-term infamous BISTRO securitization, an early Manipulation profits and big bonuses for those involved, credit default swap, in 1997.45 The team that and a hidden transfer of risk to other parties, put this together argued that it was “super- of the system usually without them even realizing it. safe,” and regulators eventually agreed, cut- takes the form An early example is the very first CDO in ting the capital requirement on a $9.7 bil- the late 1980s, carried out under First Execu- lion issue from $700 million to a mere $160 of regulatory tive Life Insurance Company’s Fred Carr. His million—from 8 cents on the dollar to 1.6 arbitrage, “problem” was that the regulators insisted cents—on what was, in fact, a very risky se- in which that First Executive’s junk-bond portfolio curity. To give them their due, the regulators be backed by the usual full capital require- insisted that the bank prove the risk was re- institutions play ments. To solve this problem, Carr securi- ally negligible and ensure that the securitiza- the existing rules tized the lot, kept them all on his books, and tion get a AAA credit rating, but these super- then argued that only the bottom tier war- ficially demanding conditions were easy to to their own ranted the full capital requirement—even meet given the inadequate models everyone advantage.

23 Regulatory was using. This transaction saved the bank Damaging securitizations have also con- arbitrage is $540 million in capital and set the pattern tinued apace in the last couple of years. for the enormous wave of similar securitiza- Amongst these are “failed sale” and “covered the primary tions that followed. By the eve of the crisis, bond” securitizations, both of which typi- factor driving the notional principal in the CDS market cally have the effect of secretly pledging bank securitizations had grown to over $60 trillion. Consequent- assets. Bank counterparties then enter into ly, the Basel system was a key driver behind arrangements with banks, not realizing that and the growth of the CDS bonanza as well. the prime assets that appear to buttress their credit derivatives. A third example is “pig on pork.” In a re- balance sheets are already pledged to other cent article, financial engineer Graham Kerr parties. In the short term, such securitiza- discusses how, around the turn of the mil- tions give banks better access to finance (be- lenium, one of his colleagues working for a cause of the higher collateral involved), but as British bank came up with a clever plan that banks’ positions become ever more opaque generated completely illusory profits by ex- and the pool of pledgeable assets diminishes, ploiting the weaknesses of the regulatory their longer-term impact is to undermine capital system.46 Their bank had a portfolio confidence in the banking system and de- of commercial bonds worth some $4 billion, stroy banks’ abilities to raise finance. In addi- and the capital requirement on this portfo- tion, central bank assistance programs to the lio was 8 percent, or $320 million. The bank banks—especially quantitative easing—have then entered into two credit derivatives spawned a new and highly profitable securi- transactions involving an American insurer tization bonanza in which the banks package and a European bank. These transactions up their most toxic assets into asset-backed were a scam,47 but had the effect of enabling securities and sell them to their local central the bonds to be reclassified as a form of cred- bank. Many of the most toxic assets end up it derivative, so lowering their capital charge on the central banks’ balance sheets, where from 8 percent to 0.5 percent. This freed up they pose a potentially lethal threat to the $300 million of capital, which could then be central banks, some of which (especially the booked as instant profit. In addition, under Fed and the European Central Bank) are al- the lax accounting rules in place, the expect- ready leveraged like aggressive hedge funds. ed profit on the bond book over the next 30 years could be classified as an additional instant profit; this further boosted reported Supervisory Weaknesses profits and the bonuses for those involved. This transaction was widely copied, and We should also consider that all modern Kerr was later left wondering why it took so financial regulatory systems are at a number long for the system to collapse. of disadvantages relative to the institutions It is widely accepted by finance indus- they regulate and, as a consequence, are try insiders that regulatory arbitrage is the prone to intimidation and capture: primary factor driving securitizations and the growth of credit derivatives. The result, ●● Financial firms have vastly greater re- therefore, is that in a context where deci- sources, so they can hire the best tal- sionmakers are already incentivized to take ent and assemble expert teams in rel- excessive risks: the regulation itself pushes evant areas (e.g., financial modeling, bankers to plunder their own banks, reduce accounting, law, etc.), giving them the their capital standards, and take additional ability to outgun the regulators, es- risks that are passed on to unsuspecting pecially on complex technical issues. third parties. In this respect, Basel is a spec- This helps them to “blind the regula- tacular example of a policy that achieved the tors with science” and set the agenda direct opposite of what it set out to do.48 on those issues. By comparison, regu-

24 latory bodies are often short-staffed lined tripartite system. Different tasks were and have inadequate research support. assigned between the Treasury, the Bank Consequently, regulatory officials are of England, and a new monolithic finan- often outnumbered and outgunned cial regulatory body, the Financial Services in key meetings. They are also ham- Authority (FSA), which would focus on the pered by a steady exodus of their staff, day-to-day regulation and supervision of fi- who take their skills and institutional nancial institutions operating in the United knowledge to the private sector. Kingdom. The idea was that the Treasury, ●● Firms are often able to hold carrots in the Bank, and the FSA each had their own front of individual regulators with the assigned responsibilities, but they would co- prospect of much-better-paying future ordinate harmoniously with each other in a jobs in the private sector. As a result, spirit of selfless dedication to their shared regulators are often reluctant to chal- task. The reality was rather different: each lenge firms for fear of jeopardizing party loathed the others and their “coordina- their own future prospects. tion” was reminiscent of the Three Stooges ●● Financial firms wield big sticks: they on a bad day.Within this context, the FSA’s have great influence and powerful main policy was essentially one of industry- friends who can (and sometimes do) friendly “light-touch” regulation—in short, The British bring pressure to bear and, if neces- give them what they want for fear that they experience sary, intimidate individual regulators might sue us or relocate elsewhere. offers a classic who get in their way. Their greater re- The FSA senior management took the sources also allow firms access to su- line that the models needed to be approved case study of perior legal firepower, which means, in with few questions asked. FSA supervisors the weaknesses practice, that they can often get their and senior management would then often of financial way merely by threatening the regula- defend the models and silence the concerns tors with legal action. of their own risk specialists who understood regulation. ●● There is the remarkably cozy relation- the models much better than their superiors. ship between the financial industry FSA officials often behaved this way in the giants, the regulatory system, and the model-review meetings, thereby publicly un- government. Key players move back dermining their own experts in front of the and forth between all three, leading to very people that they were meant to stand up industry capture, not just of the finan- to. (The incentive to stand up to the industry cial regulatory system, but of the po- was also undermined further, it later turned litical system, too.49 out, because the FSA was using feedback from the industry in the performance assess- United Kingdom Experience ment of its supervisory staff, incentivizing its Among the worldwide regulatory com- own staff to curry favor with the people they munity, the United States and, supposedly, were meant to stand up to!) This was in spite the United Kingdom, are often said to have of the fact that most FSA senior management the best regulatory institutions. Yet the Brit- had a very poor understanding of the funda- ish experience offers a classic case study of mentals of risk modeling and still struggled the weaknesses of financial regulation, the with the subject even after remedial training dangers of regulatory capture, and the re- sessions had been put on for them. As well as sults of establishing a self-serving regulatory being undermined, FSA risk specialists who empire that grows from its own failures. voiced concerns risked becoming pariahs, When it took office in May 1997, the new with supervisors sometimes requesting that Labour administration replaced the previous certain specialists not be assigned to review byzantine and highly dysfunctional system their firms’ models in case they raised too of financial regulation with a new stream- many inconvenient questions.

25 This culture invited institutions to treat The weakness of this system was high- the review process with barely disguised con- lighted by the FSA’s handling of the big tempt: in one widely discussed case, a team bank, HBOS. In the years before the crisis, from a major U.S. investment bank was its head of regulatory risk, Paul Moore, had asked about their contingency plan in the warned his bosses—including then-CEO event that the FSA refused to approve their James Crosby—that the bank was heading model. After some inhaling of breath—the for problems. The bank was, he said, was “go- firm was not used to such impudence!—the ing too fast, had a cultural indisposition to most senior bank official replied that he challenge, and was a serious risk to econom- would phone the FSA chairman to take it ic stability and consumer protection.”51 He up with him. The risk specialist who asked subsequently likened his experience to being the question was well known for challenging “like a man in a rowing boat trying to slow firms’ approaches and was quickly labeled down an oil tanker.” His superiors dismissed a troublemaker, overlooked for promotion, his concerns, although they turned out to be and eventually left. amply justified. He also raised his concerns In another case known to us, a bank had with the FSA, but they apparently wanted an lobbied for FSA approval on some dubious easy life and did nothing. HBOS senior man- new innovation, and the FSA sent along a agement eventually decided that Mr. Moore single inexperienced official to the review “didn’t fit in”—he clearly didn’t—and fired meeting. For its part, the bank sent in an him; for his part, Crosby was subsequently experienced team whose members were de- rewarded with a knighthood for his services termined to obtain approval; they wore her to the finance industry and became a key fi- down, keeping her late and overwhelming nancial adviser to the government. her until she finally broke down in tears and Bad as the HBOS case was, the FSA’s han- was too exhausted to resist any further. dling of Northern Rock beggars belief. This With this sort of model review process, was an institution that was traditionally re- it was inevitable that some very dubious garded as solid, even staid, but it changed fo- models would be approved. In this context, cus and grew very rapidly around the turn of it is telling that none of the major problem the millennium, specializing in the British banks in the UK—Northern Rock, HBOS, version of subprime and using an extreme Lloyds TSB, and RBS—had any problems ob- business model that relied more heavily on taining regulatory approval for their internal access to wholesale funding and securitiza- risk models not long before they hit the buf- tion for its financing than any other major fers themselves. British bank. The FSA, however, seemed The FSA performance on Basel II’s Pillar stuck in a time warp and was oblivious to 2, regulatory supervision, was equally un- Northern Rock’s new aggression. The bank impressive. Many sensible regulatory insid- was supervised by insurance regulators who ers had felt that this should have been even knew little about how a mortgage bank op- more important than Pillar 1, because it was erated; apparently, the FSA wanted to give under Pillar 2 that supervisors could weigh them some work experience on a safe bank The FSA busied in and demand capital charges beyond Pillar where nothing much could go wrong. Only itself with 1. But neither supervisors nor the industry eight supervisory meetings were held be- really wanted that. Instead, the FSA busied tween 2005 and August 9, 2007—mostly in- producing itself with producing enormous rulebooks volving low-level FSA staff. Of those meet- enormous and sundry other useless activities, and Pillar ings, five were held over just one day and rulebooks and 2 became a box-checking exercise.50 Natu- two by telephone, and—on the one occasion rally, everyone paid lip service to the need for when paperwork would actually have been sundry other good risk management, but the reality was helpful—the supervisors did not bother to useless activities. that the industry did whatever it wanted. take notes.52

26 In February 2007, Northern Rock’s share in its resources and staffing. Its budget in- By the eve of the price started to deteriorate and, as the year creased by 36 percent during 2009 and by crisis, all major progressed, concerns about mortgage de- 9.9 percent in 2010.53 Naturally, there were faults were rising. Although it was obvious the usual reassurances that the FSA now fi- international to the market by this point, it still had not nally had its act together (of course!), and banks were Basel occurred to the ever-vigilant FSA regulators the FSA’s new chairman, Lord Turner, was compliant, with that Northern Rock might be in any danger now telling bankers to “be afraid, be very or that it might want to suggest that the bank afraid” of the new get-tough FSA.54 The era capital ratios stress-test its liquidity exposure. Instead, the of light-touch regulation was over, we were between one FSA’s response was to approve a dividend assured, and the FSA’s new policy was now payment and fast track the approval process one of “intensive supervision”55 and, im- and two times for its models. Northern Rock hit the rocks plicitly, covering its back.56 Then came a their minimum shortly afterwards: on September 13, 2007, general election in May 2010, and at the end requirements. it acquired the ignominious distinction of of the year, the new coalition government being the first English bank to experience a announced its solution—yet another reshuf- run since Overrend and Gurney in 1866. The fling of the regulatory pack: it intended to Treasury Select Committee’s subsequent break up the FSA, moving some of it back report about the fiasco was scathing in its to the Bank of England and the rest to a new criticism of the FSA’s handling of the case— Prudential Regulation Authority. It remains “asleep at the wheel” being the gist of it. to be seen whether the new regulatory struc- The FSA chief executive Hector Sants ture will turn out to be any better than its then offered an apology of sorts. “We’re predecessors, but we are not optimistic; the sorry that our supervision didn’t achieve all more things change, the more they stay the it could have done.” He was certainly right same. about that and had the decency to forgo his own bonus for that year. That said, he was still happy to defend FSA bonuses increas- Basel and the ing 40 percent the year after the British Financial Crisis banking system collapsed. He also did himself no favors by claiming that the North- Returning to Basel, by the eve of the cri- ern Rock fiasco was unpredictable. In fact, sis, all major international banks were Ba- it later came out that, as early as 2003, the sel compliant, with capital ratios between British authorities had anticipated a sce- one and two times their minimum require- nario remarkably close to the one that oc- ments. The crisis itself resulted in the col- curred, correctly identifying Northern Rock lapse of the banking system: Basel had failed as a potential problem. They also drew an spectacularly. The inadequacies of the Basel important conclusion: there was a need for rules were sometimes highlighted very strik- a fast-track bankruptcy procedure to handle ingly, too: five days before its bankruptcy, such a problem. But having identified both Lehman Brothers boasted a Tier 1 capital the problem and the solution, they then ratio of 11 percent—nearly three times its did absolutely nothing about it. The buck- regulatory minimum requirement—which passing that followed was straight out of Yes, would indicate strong capital health by Ba- Minister. sel standards. There followed much soul-searching, Yet there were warning signs that some- numerous reports, and one reorganization thing was wrong even before the crisis. For after another. Yet the FSA managed to ride example, under Basel II’s most conservative out withering public criticism of its mis- method, banks were allowed a maximum takes and argued successfully for both an of 10 times the leverage in equities and 50 extension of its mandate and a large increase times that in AAA bonds. Both of these rep-

27 resent very high levels of risk. The equity banks’ risk disclosure practices were woeful- maximum meant that a 10 percent fall in ly inadequate. Such reporting has to be seen equities would wipe out the portfolio. The in the context of the huge growth in all man- bond maximum meant that an increase of ner of financial “disclosure” leading to ever- 0.25 percent in risk spreads on a portfolio longer reports with ever-increasing data, of 10-year 4 percent bonds would wipe out but less and less real information. By 2008, even this purportedly stable portfolio. The for example, HSBC’s annual report had be- most elementary “what if?” would therefore come so heavy that the British Post Office have revealed that the banks were very vul- refused to deliver it on the grounds that it nerable. posed a health and safety threat to its staff. Some major banks had core capital equal As for banks’ risk disclosures, studies have to 3 percent or less of their assets, which is shown that most so-called risk disclosure is very low by traditional standards, and in- no more than boilerplate in nature, typically formed observers were already warning that consisting of bland statements along the banks had been lending too much, particu- lines of “We do this and that” and providing larly to customers with poor credit ratings, VaR information that would give little real and that the credit boom was unsustain- indication of banks’ risks or their ability to Market discipline able. Then there was the impact of the many withstand severe events.59 It was therefore is undermined by hidden risks banks were taking—the impact no surprise that banks’ reports gave no in- pervasive state- of off–balance sheet securitizations, credit dication of the banks’ actual risk exposures derivatives, inadequate risk models, and in the run-up to the crisis. In addition, risk created incentives poor risk management—indicating that the disclosure was not so much undermined as to take excessive banks were even more vulnerable than they made completely impossible by the com- appeared to be.57 plexity of banks’ positions and, of course, by risks. Despite these warning signs, it is remark- the fact that most bankers themselves did able how clueless the bankers themselves not know the risks they were running. The seemed to be about their own vulnerabil- unfortunate timing of the implementation ity. A well-cited case was the chairman of of the new “Fair Value” accounting stan- Citigroup, Charles “Chuck” Prince, who on dard, FAS 133, did not help either.60 July 10, 2007, famously brushed off con- These problems had fatal implications cerns about growing liquidity problems by for Basel II’s Pillar 3, which relied on dis- explaining that: “When the music stops, in closure to produce market discipline: it was terms of liquidity, things will be complicat- therefore always naive to expect Pillar 3 to ed. But as long as the music is playing, you’ve work. In any case, typical Pillar 3 reports got to get up and dance. We’re still dancing.” were themselves mindless compliance ex- This state of denial continued into the fall. ercises consisting of basic balance-sheet The music finally stopped when analyst material and uninformative twaddle about Meredith Whitney announced on October risk-management systems that only worked 31 that Citi was in such a mess that it would on paper. More fundamentally, market dis- have to either slash its dividend or go bust. cipline is undermined by pervasive (and in Her announcement took almost everyone by the final analysis, state-created) incentives surprise and financial stocks instantly fell to take excessive risks: “market discipline” $369 billion in value. Prince was chucked in the modern financial system therefore out of his job four days later. Citi’s case was amounts to a license for risk takers to take not atypical, however. Going into the crisis, excessive hidden risks, knowing that they there were few, if any, senior bank manage- will reap the benefits but other people will ments who understood how vulnerable their bear most of the downside. institutions really were.58 To be fair, the regulators were quick to It was also apparent before the crisis that respond to the possible lessons to be learned

28 from the crisis and there has been a flurry of is also interesting that the Basel Committee analysis and reports.61 These culminated in ducked the awkward question of how this Basel III, which was unveiled in September should be implemented by passing it back to 2010. Its main proposals were: the national regulatory bodies. Beyond that, Basel III retains many of ●● The definition of core capital is to be the weaknesses of its predecessors: its reli- tightened up, with more emphasis on ance on a highly gameable weighting sys- common equity. There is to be a mini- tem with risky positions still receiving very mum common equity (or core Tier 1) low weights, its reliance on banks’ own risk requirement of 4.5 percent, and that models, and so forth. In any case, under pres- for Tier 1 more generally is to be 6 per- sure from the banks, the proposed higher cent. On top of this, there is a “conser- requirements for larger banks were quickly vation buffer” of another 2.5 percent, watered down and the implementationof its taking the minimum core Tier 1 and key proposals were postponed, allowing the more general Tier 1 requirements to 7 banks to carry on as before. Liam Halligan and 8.5 percent respectively, and 10.5 gave Basel III its epitaph before the ink on it percent if one includes Tier 2 capital. was barely dry. “In truth,” he said, the new Banks that fall below this minimum Basel Accord “had been eviscerated by the will be unable to pay dividends. all-powerful banking lobby.”63 So much for ●● There is to be a countercyclical capi- Basel III! tal buffer of up to 2.5 percent, which is meant to counter the procyclicality of capital requirements and ensure Conclusion higher capital requirements when the economy is booming. Any effective reform needs to go much ●● There is to be higher capital require- further, and a good guiding principle is ment on “systemically important” (i.e., that any solution should address underly- bigger) banks. ing causes. As we discussed at length above ●● There is to be a minimum liquidity in the section on financial risk modeling, standard—including, in particular, the one of Basel’s most serious weaknesses is its requirement to maintain enough li- most distinctive innovative feature—and the quidity to withstand the impact of a foundation on which much of it is built: the 30-day freeze in market liquidity.62 principle of risk-based regulation. Once we ●● There is to be a maximum leverage accept the need to abandon risk-based regu- ratio to provide an additional capital lation as a failed experiment, three possible constraint. ways forward then suggest themselves.

These amount to a significant tighten- Abandon the Internal Model Approach ing—and most of these proposals are wor- The least radical approach (we may as thy enough, as far as they go. The greater well call this Basel IV!) would be for the Ba- emphasis on common equity helps amelio- sel Committee to scrap the internal model rate the problem of banks bolstering their approach and revert back to simpler, more core capital with dodgy debt-equity hybrids, robust, and less gameable metrics that gen- and the introduction of liquidity and lever- erate higher capital charges, while patching age requirements provide additional use- up other loopholes as best it can. The key Basel III retains ful safeguards. However, the proposals for features would be an updated, higher-charg- many of the a countercyclical capital requirement are so ing version of the standardized approach vague that they amount to little more than and (as in the Basel III proposals) a greater weaknesses of its a restatement of the problem to be solved. It emphasis on severe stress tests and maxi- predecessors.

29 Simply sending mum leverage ratios, especially those focus- ●● Profits reflect risk taking and, other the Basel ing on core capital,64 reinforced by a paral- things being equal, rise when banks lel system of liquidity regulations to protect take more risks. A profit-based capital Committee back banks’ liquidity. charge would therefore be indirectly to the drawing The implementation of the standardized risk based, but without requiring any board and telling approach could also be improved by estab- dubious risk models. lishing a standing committee that would pe- ●● A profits-based charge would build it to “get it right riodically revise risk weights in the light of in an element of anti-cyclicality, lead- this time” is evolving experience, so making the weight ing to capital charges that rise as the revision process more flexible and (hope- economy booms and fall when the plainly naive. fully) more timely. economy goes into recession. The biggest technical problem faced by this committee would be to assess the capi- The charges would, however, need to be tal requirements of complicated products carefully specified to protect them from ma- such as CDOs, CDSs, and their progeny. We nipulation by creative accountants playing would suggest that they base these on se- with the profit figures—a requirement easier vere risk assessments, in particular the use said than done. of Cauchy and fuzzy-logic “litmus tests” If suitably implemented, a revamped Ba- that would flag up the true risks of such sel approach along these lines would help pathological products.65 This would reflect limit model abuses and lead to higher capi- a hard-line policy on them. Naturally, the in- tal requirements, especially on very risky po- dustry would object vociferously and argue sitions. Such reform also has the practical at- that this would make them uneconomic, traction that its successful implementation but we would counter that this is exactly the does not require a whole range of other com- point: high-risk products need high capital plementary reforms, desirable as those might requirements, and if this helps them to ex- be. Instead, it could be usefully implemented tinction, so much the better. Their extinc- almost on a stand-alone basis and with relation would make the system much safer and tively little danger of being undermined if also go some way toward helping make dis- other financial and banking reforms are seri- closure effective again. ously botched—which is a major risk in the We would also suggest that the Basel current frenetic policy environment. Committee abandon attempts to impose However, such reform is still not a sub- capital requirements on operational risk. stitute for effective risk management and Operational risk models have shown them- only goes a small way to counter the current selves to be useless in the face of the opera- system’s endemic incentives toward exces- tional risks that matter—such as out-of- sive risk taking. Even worse, simply sending control senior management. Banks would the Basel Committee back to the drawing still be allowed to model operational risks if board and telling it to “get it right this time” they wished to, but such models would have is plainly naive, not least because it begs the no regulatory status. question of why so many well-known weak- That said, there is one feature of the Ba- nesses in Basel II made it through so easily sel operational risk system that could serve into Basel III. And we all know the answer to a useful purpose: the basic version involves that: the process was captured by the banks. the imposition of capital charges on firms’ Consequently, there is no reason to believe profits. We would suggest that this is poten- that a future Basel IV would be much better tially useful, albeit stripped of its redundant than Basels I, II, or III. operational-risk overcoat. Indeed, in some respects, bank profitability is quite an at- An Insurance-based Approach tractive basis for capital charges: A second possibility is to seek a new ba-

30 sis upon which to establish capital adequacy own risks properly. regulation. One possibility sometimes sug- ●● Even ignoring these problems, any in- gested is to require banks to take out insur- surance “solution” would merely trans- ance against certain risks such as insolvency fer the burden from the banks’ risk or illiquidity risks. This sounds plausible at models to the insurers’ risk models. first, but in reality it boils down to passing This just boils down to another form of the buck to the insurance industry, which is risk-based regulation—one that failed itself in no position to carry the burden. spectacularly in the case of AIG. Leaving aside such obvious questions as ●● Leaving aside other objections, there is which particular risks should be insured and no reason to think that an insurance- how the insurance would work, any insur- based regulatory system would be any ance-based approach raises major feasibility more capture-proof than the Basel problems: regulatory system.

●● Insurance companies have limited In short, there is no insurance-based so- capacity. Their already-overstretched lution either. Indeed, we need to acknowl- capital is much smaller than that of edge that there is no regulatory “solution” at banks, and they certainly do not have— all, because any regulatory apparatus would There is no and are unlikely ever to have—the re- be captured. regulatory sources to underpin the capital of the “solution” at banking system. Free Banking ●● An “insurance solution” to the bank There is, however, one solution that all, because capital adequacy problem raises the would work if the political will were there to any regulatory questions about insurers’ own solvency implement it. This would deal with both the risk and the potential for systemic in- major problems of the Basel system—risk- apparatus would stability. No one wants a repeat of the based regulation and regulatory capture—by be captured. AIG fiasco, where one firm becomes restoring appropriate incentives and abol- the dominant seller of some form of ishing the regulatory system. The solution corporate insurance and then goes is free banking or financial laissez faire. The belly-up in a crisis—with potentially state would withdraw entirely from the fi- systemic consequences. nancial system and, in particular, abolish ●● Any insolvency insurance has the capital adequacy regulation, deposit insur- drawback that it does not avoid the ance, financial regulation, and the central need for risk modeling banks’ posi- bank, as well as repudiate future bailouts tions, but rather passes that task to (and especially the doctrine of Too Big to the insurer. Furthermore, the insurer Fail).66 We would also suggest that the es- is less well placed to handle that task tablishment of a free banking system re- than the bank it is insuring, and any quires major corporate governance reforms such insurance is likely to create major and the reintroduction of extended liability moral hazard problems. Insured banks for senior officers and shareholders. would be able to cut corners on their Such systems have worked well in the own capital and risk management, al- past, and reforms along these lines would lowing them to free ride on the insur- take the United States a long way back to its ers’ guarantee and pass all manner of banking system of a century ago, in which hidden risks back to the insurer. These banks were tightly governed and moral haz- considerations suggest that any insur- ards and risk taking were well controlled ance solution would be inferior to an because those who took the risks bore their alternative in which banks are given consequences. Such incentives would re- adequate incentives to manage their store effective risk management and lead to

31 It is possible to banks recapitalizing themselves and return- 2. For more on the U.S. financial system and how it compares to its modern successor, see, for sort out the Basel ing to a state of true capital adequacy. Dan- example, John C. Bogle, Enough: True Measures of gerous financial products would disappear, Money, Business, and Life (New York: Wiley, 2009); mess, put the banks’ positions would simplify, disclosure or Kevin Dowd and Martin O. Hutchinson, Alche- would be restored, and the financial system mists of Loss: How Modern Finance and Government financial system Intervention Crashed the Financial System (Chiches- back on its feet, would become safe and stable again. ter, UK and New York: Wiley, 2010). This would, however, have to take place in and restore the a context where the problem of accommoda- 3. The leverage ratio is the ratio of assets to tive monetary policy had also been resolved— share capital: the higher the leverage, the great- integrity of the er the shareholder return if the market goes the capitalist system ideally by the restoration of a gold standard. right way, and the greater the shareholder loss if Within this context, there would no longer it does not. Our main concern is with the risk of itself. be a capital adequacy problem; capital regu- the shareholder being wiped out. So, for example, lation would have become redundant and at a leverage ratio of 10 to 1, it would take a 10 67 percent fall in assets to wipe out a bank’s equity, could then be safely abolished. Banks whereas at a leverage ratio of 30 to 1, a mere fall would then be left free to determine their of 3.3 percent would accomplish the same result. own capital requirements based on whatever models, financial ratios, or any other rules of 4. Richard M. Salsman, Breaking the Banks: Cen- tral Banking Problems and Free Banking Solutions thumb they consider relevant. (Great Barrington, MA: American Institute for And so we have a delightful good news/ Economic Research, 1990), p. 56. bad news situation worthy of the best an- cient Greek morality plays. The good news 5. Quoted in Michael Lewis, “The End,” Port folio.com, November 11, 2008, http://www.port is the solutions to our problems exist—it is folio.com/news-markets/national-news/portfo possible not only to sort out the Basel mess, lio/2008/11/11/The-End-of-Wall-Streets-Boom. but also, much more importantly, to put the Gutfreund’s poignant remarks are a perfect de- financial system back on its feet and restore scription of modern crony capitalism. the integrity of the capitalist system itself. 6. Howard D. Crosse, Management Policies for The bad news is that we have to do so in a Commercial Banks (Englewood Cliffs, NJ: Prentice- context where the policy community—and Hall, 1962), 169–72. the political system more broadly—are hell- 7. The biggest problem facing United States’ bent on pursuing more of the same policies capital adequacy regulation over much of the that got us into this dreadful state in the 20th century was that of forbearance: banks that first place, and where most of the propos- struggled to meet their requirements were poten- als currently offered threaten to make a very tial liabilities of the regulatory bodies, whose own resources were limited. This made the authori- bad situation much worse. ties reluctant to take over troubled banks. The regulators were also subject to lobbying from the banks themselves, which often led to pressure for Notes forbearance from the political system. These factors led to a long-term decline in both U.S. capi- We thank a number of readers, and most espe- tal standards and in the effectiveness with which cially, Gordon Kerr, for their helpful comments. they were imposed. These problems are discussed further in Salsman. 1. We cannot even begin to do justice to the in- comprehensible jargon, innumerable acronyms, 8. For more on the history of the Basel system, and mind-numbing institutional material that see, for example Daniel K. Tarullo, Banking on Ba- plague this subject, and trust that the reader will sel: The Future of International Financial Regulation thank us for not attempting to do so. We also fo- (Washington: Peterson Institute for Internation- cus on the banks (and so ignore insurance com- al Economics, 2008); and Basel Committee on panies and insurance regulation: e.g., Solvency Banking Supervision, History of the Basel Committee I and II), and stay clear of vast areas such as the and its Membership (Basel: Bank for International politics of Basel; international harmonization; Settlements, 2009). harmonization across commercial banks, investment banks, and insurance companies; and the 9. Herstatt was a medium-sized German bank, legal issues involved. the failure of which would have been insignifi-

32 cant in ordinary circumstances. However, the (supposedly) safe positions—typically, highly rat- German authorities foolishly closed the bank ed positions involving credit derivatives—the ex- while the New York Stock Exchange was still trad- traordinarily low capital charge of 0.5 percent. By ing. This led to major settlement problems that, the turn of the millennium, the manufacture of given the unsettled conditions already existing, ingenious securitizations that met this require- destabilized international markets on a scale not ment had become the biggest growth industry seen since the failure of the Creditanstalt in 1931. in town, thus drastically reducing banks’ capital The Herstatt failure made it obvious to bank su- charges; the capital “released” in this way could pervisors and central banks that they needed to then be used to pay bonuses to the financial engi- be more coordinated—not that the resulting co- neers involved and their managers. The reality, of ordination was of any use when the next major course, was that risks were merely being shuffled cross-country bank disaster occurred, involving around out of sight of the regulators, and many the Bank of Credit and Commerce International, of these “safe” positions later turned out to be in 1991. very risky indeed. Yet instead of closing this loophole down, Basel II expanded it, and spurious se- 10. It is worth comparing these Basel I require- curitizations flowered like never before. ments to those of U.S. banks in the 1950s. For example, under Basel I, normal commercial assets 15. The discussion in this section draws heavily would merit a minimum capital requirement (in- from Dowd and Hutchinson, Parts 2 and 3. cluding Tier I and Tier II) of 8 percent regardless of their quality, whereas under the earlier regime 16. This has been known to be a problem for a normal commercial assets would merit a mini- long time. It is discussed, for example, in F. A. mum capital requirement of 12 percent and im- Hayek, “The Use of Knowledge in Society,” Ameri- paired assets would merit a minimum of up to 50 can Economic Review 35, no. 4 (September 1945); percent. The Basel requirements were therefore Richard Hoppe, “VaR and the Unreal World,” much less stringent. Risk 11 (July 1998); and Nassim Nicholas Taleb, The Black Swan: The Impact of the Highly Improb- 11. A leading example is William J. McDonough, able (London: Penguin, 2007). The response of the previously of First Chicago and president of the econophysics crowd to these profound criticisms New York Fed from 1993–2003, who led the Basel has simply been to ignore them. A good example committee while it was drafting Basel II. He took is the otherwise good book on econophysics by a very industry-friendly view and was also a lead- Jean-Philippe Bouchaud and Marc Potters, Theory ing exponent of the unspoken but understood- of Financial Risks: From Statistical Physics to Risk Man- by-all theory that the biggest banks were some- agement (Cambridge: Cambridge University Press, how “superior” to the smaller ones and should 2000). therefore qualify for regulatory breaks denied to the latter. 17. Nor were CDOs the most damaging of these financial innovations. That particular honor 12. This was evident when the internal-model goes to credit default swaps, whose notional prin- approach was first adopted in the 1990s. This cipal by the end of 2007 had grown to about $62 problem was highlighted more recently in the trillion, well in excess of annual world product. UK, when most of the banks that applied for their The dangers these products—which are essen- operational risk capital charges to be determined tially just bets on corporate default—posed to the by the Basel II Advanced Measurement Approach system as a whole were highlighted for all to see dropped out when they realized that the models in the AIG debacle. For more on these, see, for ex- would probably lead to higher capital charges ample, Dowd and Hutchinson, pp. 190–97. than under the simpler approaches. Reassuring- ly, the banks involved subsequently claimed that 18. “We were seeing things that were 25- regulatory capital was not an issue for them and standard-deviation moves, several days in a that they were only interested in good risk man- row.” David Viniar, quoted in the Financial Times, agement. August 13, 2007.

13. At the risk of belaboring the obvious, it is 19. Estimates based on those in Kevin Dowd, exactly when risks are high that risk modeling John Cotter, Chris Humphrey, and Margaret would be most useful. Hence, the Basel system Woods, “How Unlucky is 25-Sigma?” Journal of managed to deprive itself of the benefits of risk Portfolio Management 34, vol. 4 (2008). modeling in exactly those circumstances where it would benefit most from them. 20. Matthew Tagliani, The Practical Guide to Wall Street: Equities and Derivatives (New York: Wiley, 14. The big banks also sought to reduce their 2009). capital requirements by pushing to expand a key loophole in the Basel I regulations: this allowed 21. Here we gloss over the awkward point that

33 the forecasting performance of VaR models is ers that there really would be no bailout if they very difficult to assess. These problems also get got themselves into trouble. No private-sector worse as the VaR probability increases: as we go banker really believed this, and quite rightly, too: further out into the tail, we encounter fewer ob- when the crisis came, the government duly rode servations and any estimates become less precise to the rescue just as the unnamed banker and his and harder to verify. Needless to say, these prob- colleagues had anticipated. lems make solvency modeling a real headache. See, for example, Kevin Dowd, Measuring Market 28. The list that follows is by no means exhaus- Risk, 2nd ed. (Chichester, UK and Hoboken, NJ: tive. For example, there were also problems with Wiley, 2005), chap. 15. the definitions of capital used by Basel—in particular, Tier 1 capital allows banks to include 22. There can also be related problems: since es- debt-equity hybrids, and led to all sorts of creative timated VaRs are often very volatile themselves, manipulation. there is the danger that some sudden short-term VaR spike, of no importance in itself, might trig- 29. Among the many existing critiques of the ger a crisis as the risk management models kick in Basel system, we would particularly recommend: and force everyone to sell, causing the lemmings David Jones, “Emerging Problems with the Basel to run over the cliff. There is also the danger of Capital Accord: Regulatory Capital Arbitrage and unexpected ricochet effects. A prime example oc- Related Issues,” Journal of Banking and Finance 24 curred during the Asia crisis, when South Korean (2000); Edward I. Altman and Anthony Saunders, banks had loaded themselves with all sorts of “An Analysis and Critique of the BIS Proposal on emerging market bonds: Brazilian Brady bonds, Capital Adequacy and Ratings,” Journal of Banking Ukrainian bonds, and so on. When their VaR and Finance 25 (2001); Jón Daníelsson et al., “An numbers shot up, they responded by trying to Academic Response to Basel II,” LSE Financial reduce their positions—but the only bonds they Markets Group Special Paper no. 130 (London: could really sell were the Brazilian Bradys. Unfor- London School of Economics, 2001); L. Jacobo tunately, they also held a substantial proportion Rodríguez, “International Banking Regulation: of the market in those bonds, and their attempts Where’s the Discipline in Basel II?” Cato Insti- to sell them had a substantial negative impact on tute Policy Analysis no. 455, October 15, 2002; that market; the net results was that the Brazilian and Marc Saidenberg and Til Schuermann, “The currency, the real, then took a hit out of nowhere. New Basel Capital Accord and Questions for Re- search,” Wharton Financial Institutions Center, 23. Department of Defense briefing, February DP 03-14 (Philadephia: The Wharton School, 12, 2002. University of Pennsylvania, 2003). It is noteworthy how many problems were identified early on 24. There are also other problems, like the con- and how little was done to remedy them: the Ba- ceptual belief that operational risk can be treated sel Committee’s imperviousness to known prob- as some add-on risk factor, as opposed to a risk lems is a recurring theme in its history. that permeates everything an institution does. For more on this, see, for example, Michael Pow- 30. Some of these “tensions” are also recorded, er, “The Invention of Operational Risk,” Review of for example, in Elroy Dimson and Paul Marsh, International Political Economy 12, no. 4 (October “Capital Requirements for Securities Firms,” 2005). Journal of Finance 50, no. 3 (1995): 831–33; and Tarullo (see note 8 above). 25. A similar problem is how to handle a bank’s strategic risks, which are fundamentally impor- 31. Riccardo Rebonato, Plight of the Fortune Tell- tant because it is the bank’s strategy that ulti- ers: Why We Need to Manage Financial Risk Differently mately determines what other risks it will even- (Princeton: Princeton University Press, 2007), p. tually expose itself to. However, from a modeling xxiii. perspective, the risks associated with a firm’s basic strategy are unknown and certainly beyond 32. See, for example, Richard Cantor and Frank the capacity of its risk models. It does not help Packer, “The Credit Rating Industry,” Federal Re- that, in practice, the strategists typically know serve Bank of New York Quarterly Review 19, no. 2 nothing about risk modeling and the risk man- (Summer–Fall 1994). agers know virtually nothing about strategy, and the two parties rarely communicate. 33. The basic economics of the ratings are ex- plained, for example, in Charles W. Calomiris, 26. Taleb, pp. 129–30. “The Debasement of Ratings: What’s Wrong and How We Can Fix It,” mimeo, Columbia Business 27. The fact that the Bank officials were sur- School, 2009; and Alan D. Morrison, “Ratings prised is itself revealing. So, too, were their ear- Agencies, Regulation and Financial Market Sta- nest attempts to persuade the commercial bank- bility,” in Verdict on the Crash: Causes and Policy Im-

34 plications, ed. Philip Booth (London: Institute of useless. Economic Affairs, 2009), pp. 117–28. 39. An honorable exception is Claudio Borio 34. In the wake of the financial crisis, the sto- from the Bank for International Settlements ries are now coming out. See Simon Johnson and (BIS), who persistently warned about the dan- James Kwak, “The Hearing: Why Companies Do gers of a credit crunch years before the crunch Stupid Things, Credit Rating Edition,” Washing- actually occurred, and he was ignored by his own ton Post, October 20, 2009, for a report on how colleagues. More typical is Brian Quinn from Moody’s, the most reputable and conservative of the Bank of England, who was the bank official the agencies, stacked its compliance department responsible for the bank’s handling of BCCI. In with people who awarded the highest ratings to the aftermath of that fiasco, he told a stunned mortgage securities that were soon revealed as audience in London that the bank realized there junk, while punishing executives who attempted was a problem with BCCI when they learned that to protect the intregrity of the firm’s ratings. people in the city were openly referring to it as the Bank of Conmen and Cocaine International. By 35. Lewis. this point, “respectable” bankers had been avoid- ing BCCI for years. 36. For more on this story, see James Alexander, “Regulatory Arbitrage and Over-Regulation,” in 40. The Bank of Spain introduced just such a Verdict on the Crash, pp. 89–90; and Mark A. measure in the late 1990s. This took the form of Calabria, “Did Deregulation Cause the Financial a capital requirement on loans that took account Crisis?” Cato Policy Report 31, no. 4 (July–August of expected losses at the time the loans are en- 2009): 7–8. tered into. These expected losses were calibrated on past recession loss experience. The Spanish ex- 37. Naturally, the ratings also involve regulatory periment was not too successful—they took data arbitrage problems. For example, under then- from the past three wimpy recessions, and the current Basel proposals, unrated firms get bet- latest one blew through past parameters—but it ter treatment than some junk-rated firms. This was still better than what other regulatory bodies encourages very risky firms to avoid being rated, were doing. and means that capital charges for such firms will, in some cases, be considerably lower than 41. We could also imagine the furor from the they would otherwise be. See Rodríguez, 13. bigger banks if regulators attempted to use such an exercise to increase capital charges, especially 38. Rebonato, p. 180. We gloss over the many as the boom approaches its peak. Unlike the problems involved with credit-risk modeling, mythical central banker of old, who saw his main and list just three: (1) the Basel regulations job as being to take away the punchbowl just as specify an arbitrary 99 percent credit VaR over a the party is getting started, the modern finan- one-year horizon, which tells us nothing about cial regulator has the much more difficult job of credit risks in a downturn; (2) credit-risk models taking away the punchbowl when everyone, the are (for reasons discussed in the text) much more regulator included, is already drunk and rowdy. difficult to calibrate and to back-test than mar- It is therefore little wonder that regulators rarely ket risk models, so it is very difficult to know if a try to do so. particular model is any good; and (3) even now, credit-risk modeling is still highly problematic, 42. Recall, too, that we have already discussed and the lack of good data is a major problem. We how systemic problems can arise from Basel, ar- also skip over the problems involved in the Basel tificially boosting risky positions by giving them approach to operational risk modeling: see Imad very low risk weights. Again, AIG and credit de- A. Moosa, “A Critique of the Advanced Measure- fault swaps come to mind. ment Approach to Regulatory Capital against Operational Risk,” mimeo (Melbourne, Austrai- 43. Frank Partnoy, Infectious Greed: How Deceit la: Monash University, 2008), which has a devas- and Risk Corrupted the Financial Markets (New York: tating critique on the difficulties of estimating Owl Books, 2004), p. 78. operational risk using the AMA approach. The problems of operational risk modeling were also 44. The securitization alchemy was, of course, highlighted by a number of notable operational- especially profitable with low quality assets. As risk experts at a risk conference one of us attend- we discussed earlier, a portfolio of, say, subprime ed in 2008: they voiced relief that they lacked the mortgages would be converted into a series of data to do op-risk modeling “properly,” as this tranched securities, most of which are rated AAA. forced them to rely on scenario analyses that can It was therefore hardly surprising that banks were look forward, rather than be pushed by the Ba- so eager to seek out new low-quality asset classes. sel system to use backward-looking risk models based on historical data, which would have been 45. Gillian Tett, Fool’s Gold: How Unrestrained

35 Greed Corrupted a Dream, Shattered Global Markets 50. The Prudential Sourcebook for Banks, Build- and Unleashed a Catastrophe (London: Little, Brown, ing Societies and Investment Firms is the main FSA 2009) provides a very readable account of this epi- rulebook for the Basel rules and is nearly 1,000 sode. pages long. One risk management expert, Philip Booth, bravely attempted to go through it, but 46. Gordon Kerr, “How to Destroy the British even he could only get so far. He was also unable Banking System: Regulatory Arbitrage via ‘Pig on to find anything in it on liquidity, which is what Pork’ Derivatives,” Cobden Centre, mimeo, 2010, brought down Northern Rock, but (ever the op- http://www.cobdencentre.org/2010/01/how-to- timist!), felt it must be in there somewhere, even destroy-the-british-banking-system/. though he could not be sure because the file was too big for him to download. See Philip Booth, 47. The scheme worked as follows: the bond “More Regulation, Less Regulation or Better Reg- portfolio was already insured by the American ulation?” in Verdict on the Crash, pp. 159–60. Then insurer via a CDS. The first transaction then in- there was the General Prudential Sourcebook Instru- volved the European bank guaranteeing the Brit- ment, which contains additional capital rules ish bank’s bonds, with the beneficiary being the and is 261 pages long; and Senior Management British bank if the American insurer did not pay Arrangements, Systems and Controls, which covers up, and the American insurer otherwise. The sec- further rules on risk management and is another ond transaction involved the American insurer 221 pages long. And all this was before the flurry insuring those same bonds again—so guarantee- of new rules that came in the wake of the post– ing what it had already guaranteed, hence “pig Northern Rock reorganization and, after that, on pork”—with the beneficiary being the Euro- Basel III and the major new shakeup of UK finan- pean bank. The combination of the two guaran- cial regulation announced in late 2010. The Brit- tees then allowed the British bank to reclassify ish regulators liked to brag that they had wisely its bonds as a credit derivative and so reduce its followed principles-based regulation instead regulatory capital. Apart from benefitting from of the mindless rules-based regulations of their the arrangement fees involved, neither of the American counterparts, but as Alan Cathcart, a other two parties was materially affected: in the senior FSA official, once publicly admitted, they event of default on the bonds, the European bank certainly had a lot of principles. would have to pay the British bank, but would be recompensed by the American insurer, and the 51. Unless otherwise stated, the remaining American insurer was not affected because its quotes in this article are taken from Dowd and additional insurance payment to the European Hutchinson. bank would be offset by the payment from that bank. 52. For more on the Northern Rock case, see also Alex Brummer, The Crunch: The Scandal of Northern 48. For their part, the regulators would re- Rock and the Escalating Credit Crisis (London: Ran- spond to these problems by sporadically patch- dom House, 2008). ing up the odd loophole at a snail’s pace. For example, the rules now require that a securitiza- 53. Rosali Pretorius and Ming D. Wang, “FSA tion that qualifies for capital reduction involve Business Plan 2010/11 Intensive Supervision,” some risk transfer, and this eliminates some of Complinet.com, April 7, 2010. the simpler, spurious securitizations such as that by Fred Carr. Yet many known loopholes 54. Such threats are frankly risible, however. Un- still persist, and institutions continue to come like its counterpart in the United States, the ju- up with ever more ingenious ways around the dicial system in the UK is hopelessly inadequate system. when it comes to major white-collar crime. Con- victions are rare, and we are not aware of a single 49. The most damaging aspect of the crisis, case in modern UK history in which a convicted by far, is the takeover of the political system by major white-collar criminal served a lengthy jail Wall Street, which amounts to nothing less than sentence. a coup d’etat. The situation is comparable in a number of other countries too: Iceland, Ireland, 55. A key feature of this is to develop the FSA’s and the United Kingdom come to mind. For more own inhouse risk-modeling capabilities so they on this, see Kevin Dowd, “Lessons from the Fi- can replicate the banks’ own risk models. This, nancial Crisis: A Libertarian Perspective” Economic however, is a very expensive indulgence and a dis- Notes 111 (2009), http://www.libertarian.co.uk/ traction. That said, other features of the new pol- lapubs/econn/econn111.pdf; Simon Johnson, icy—such its emphasis on corporate governance, “The Quiet Coup,” The Atlantic, May 2009; Luigi “living wills,” and stress testing—are to be com- Zingales, “Capitalism after the Crisis,” National mended. Affairs 1 (Fall 2009); and Kevin Dowd and Martin Hutchinson, Alchemists of Loss, chap. 14. 56. An example of this is the abuse of Section

36 166 or “skilled persons reports,” in which the FSA model-based valuations, which bear no relation- would instruct a regulated firm to engage an out- ship to market reality. So, for example, when FAS side expert to review a particular area of concern 133 came into operation in late 2008, Goldman (e.g., risk-management systems) at the firm’s ex- Sachs reported Tier 3 assets (based on model val- pense. Prior to 2007, the FSA rarely issued these uations) of over $70 billion, or more than double reports, not least because they were expensive at its core capital. Most of these assets were believed an average cost of £500,000 each. After the cri- to have market values of much less than this sis, however, they are being thrown around like amount. These model-based “valuations” thus confetti; one insider estimated that there were ap- allow banks to create fictitious capital and so proximately 100 issued in 2010 and perhaps 700 hide their true financial health; they allow banks would be issued in 2011. The “skilled persons” to manufacture fake profits and then distribute consulted were typically ex-FSA people who had those as dividend and bonus payments, and so gone to the major accountancy firms, who would give bankers yet another means to denude their give the FSA what they and the banks wanted— banks of their capital and dump their losses on that is, something not too critical. This was a the taxpayer. cozy arrangement for all concerned: the accountancy firms made easy money and the prospect of new audit business, the banks benefited because 61. In some cases, regulators were too quick. the FSA could not reasonably justify further ac- An example was the hapless FSA, accident-prone tion, and the regulators could cover their backs as ever. The FSA moved quickly and introduced by pointing to all the (other people’s) money new liquidity rules for banks operating in the UK. spent on expert advice. These included waivers for subsidiaries or branch- es of non–European Economic Area banks. Hav- 57. By this point, most “risk management” was ing set up this new system, however, the FSA then merely window dressing that provided no real discovered that the draft new EU Capital Require- protection, and the occasional risk manager who ments Directive due to be implemented across recognized and spoke out about the approach- the EU does not allow such waivers. One insider ing problem was usually ignored or fired. As one described this as a potentially very embarrassing frustrated risk manager later summed it up, “Risk mess. management discipline rarely made it into the chief executive’s office or the boardroom, or into 62. It is nice to see that the Basel Committee now day-to-day business decisionmaking. We had the takes liquidity risk seriously. For those watching appearance of risk management without the real- the markets, this has been a clear problem since ity of risk management so when we really needed the stock market crash of October 1987, and a it, it wasn’t there.” See Dan Borge, “A Return to blindingly obvious one since 1998. One might Crediblity,” Risk (June 2009): 56, http://www.risk. also note that it was the failure to take account of net/risk-magazine/opinion/1497734/a-return- liquidity risk that led to the failure of Overrend credibility. and Gurney back in 1866.

58. Citigroup found itself holding $55 billion in 63. Liam Halligan, “Bankers Regain Power as toxic CDOs when the music suddenly stopped, Davos Summit Ends with a Big Fudge,” Daily in addition to having to buy back another $25 Telegraph, January 29, 2011. billion on which it had unwisely written liquidity puts. Its then-chairman, ex-treasury secretary 64. There is also an argument for the use of hy- Robert Rubin, later admitted that he had never brid instruments as supplementary regulatory even heard of these instruments at the time. He capital, such as subordinated debt, sometimes had to resign, too. David Viniar’s immortal line referred to as contingent convertible or “CoCo” about 25–standard deviation events (quoted capital. This would be automatically converted above) suggests that Goldman’s team was not too into equity when the Tier I capital hit a particu- well informed either. lar threshold. Such instruments have the advantage of allowing for an automated recapitaliza- 59. See, for example, Margaret Woods, Kevin tion of a bank when it gets into difficulties. We Dowd, and Christopher Humphrey, “The Value are somewhat skeptical of this proposal, how- of Risk Reporting: A Critical Analysis of Value-at- ever; in an environment where banks are already Risk Disclosures in the Banking Sector,” Interna- weak, there is the danger that the triggering of tional Journal of Financial Services Management 8, no. such conversions could, in itself, destroy confi- 1 (2008): 45–64. dence and trigger the very runs they are meant to prevent. There is also the danger of investors in 60. The implementation of FAS 133 (the ac- such instruments buying insurance “wrappers,” counting standard for financial derivatives) offloading the risk to another counterparty in created valuation chaos at a very difficult time. much the same way as AIG became a dumping Perhaps the biggest problem is that it allows ground for the tail risks that no one else wanted.

37 65. For more on these and how they might be produced IFRS, is on record as saying that if you used, see Dowd and Hutchinson, chap. 15. think you understand IFRS, then you haven’t read it properly. 66. It is also important to mention, at least in passing, the need for major reform of U.S. 67. There is a caveat. This first-best solution in- accounting standards. Currently, the United volves not just a single reform but a package of States is moving from Generally Accepted Ac- major reforms that form a coherent whole. Our counting Principles (U.S. GAAP) to Internation- recommendation to abolish capital adequacy al Financial Reporting Standards (IFRS). How- regulation is contingent on it being part of that ever, IFRS is a very poor accounting standard. package. Given that the incentives to excessive Amongst its other hideous features, it throws risk taking that pervade the modern financial out the old principle that accounts should be system were created by misguided government prepared prudently; allows banks to manufac- policy, the removal of capital adequacy regula- ture fake profits and then distribute them as tion, while leaving the rest of the system intact, dividends and bonuses; allows banks to inflate would mean moving from weakly controlled to capital and hide expected losses; and seriously almost uncontrolled risk taking. The financial weakens the audit function in its key task of system would soon blow up again, even sooner challenging management. In short, IFRS gives than it otherwise would, and the bankers would no indication of banks’ true financial positions doubtless be demanding yet more massive bail- and provides a perfect smokescreen for bankers outs from the taxpayer or else civilization will to plunder their own banks. It is also incompre- end. Consequently, unless we go for the whole hensible, even to experts. Indeed, its chief ar- package, it would be better to tighten capital ad- chitect, Sir David Tweedie, the chairman of the equacy regulation until the rest of the mess has International Accounting Standards Board that been sorted out—assuming it ever is.

38 RECENT STUDIES IN THE CATO INSTITUTE POLICY ANALYSIS SERIES

680. Intercity Buses: The Forgotten Mode by Randal O’Toole (June 29, 2011)

679. The Subprime Lending Debacle: Competitive Private Markets Are the Solution, Not the Problem by Patric H. Hendershott and Kevin Villani (June 20, 2011)

678. Federal Higher Education Policy and the Profitable Nonprofits by Vance H. Fried (June 15, 2011)

677. The Other Lottery: Are Philanthropists Backing the Best Charter Schools? by Andrew J. Coulson (June 6, 2011)

676. Crony Capitalism and Social Engineering: The Case against Tax-Increment Financing by Randal O’Toole (May 18, 2011)

675. Leashing the Surveillance State: How to Reform Patriot Act Surveillance Authorities by Julian Sanchez (May 16, 2011)

674. Fannie Mae, Freddie Mac, and the Future of Federal Housing Finance Policy: A Study of Regulatory Privilege by David Reiss (April 18, 2011)

673. Bankrupt: Entitlements and the Federal Budget by Michael D. Tanner (March 28, 2011)

672. The Case for Gridlock by Marcus E. Ethridge (January 27, 2011)

671. Marriage against the State: Toward a New View of Civil Marriage by Jason Kuznicki (January 12, 2011)

670. Fixing Transit: The Case for Privatization by Randal O’Toole (November 10, 2010)

669. Congress Should Account for the Excess Burden of Taxation by Christopher J. Conover (October 13, 2010)

668. Fiscal Policy Report Card on America’s Governors: 2010 by Chris Edwards (September 30, 2010)

667. Budgetary Savings from Military Restraint by Benjamin H. Friedman and Christopher Preble (September 23, 2010)

666. Reforming Indigent Defense: How Free Market Principles Can Help to Fix a Broken System by Stephen J. Schulhofer and David D. Friedman (September 1, 2010) 665. The Inefficiency of Clearing Mandates by Craig Pirrong (July 21, 2010)

664. The DISCLOSE Act, Deliberation, and the First Amendment by John Samples (June 28, 2010)

663. Defining Success: The Case against Rail Transit by Randal O’Toole (March 24, 2010)

662. They Spend WHAT? The Real Cost of Public Schools by Adam Schaeffer (March 10, 2010)

661. Behind the Curtain: Assessing the Case for National Curriculum Standards by Neal McCluskey (February 17, 2010)

660. Lawless Policy: TARP as Congressional Failure by John Samples (February 4, 2010)

659. Globalization: Curse or Cure? Policies to Harness Global Economic Integration to Solve Our Economic Challenge by Jagadeesh Gokhale (February 1, 2010)

658. The Libertarian Vote in the Age of Obama by David Kirby and David Boaz (January 21, 2010)

657. The Massachusetts Health Plan: Much Pain, Little Gain by Aaron Yelowitz and Michael F. Cannon (January 20, 2010)

656. Obama’s Prescription for Low-Wage Workers: High Implicit Taxes, Higher Premiums by Michael F. Cannon (January 13, 2010)

655. Three Decades of Politics and Failed Policies at HUD by Tad DeHaven (November 23, 2009)

654. Bending the Productivity Curve: Why America Leads the World in Medical Innovation by Glen Whitman and Raymond Raad (November 18, 2009)

653. The Myth of the Compact City: Why Compact Development Is Not the Way to Reduce Carbon Dioxide Emissions by Randal O’Toole (November 18, 2009)

652. Attack of the Utility Monsters: The New Threats to Free Speech by Jason Kuznicki (November 16, 2009)

651. Fairness 2.0: Media Content Regulation in the 21st Century by Robert Corn- Revere (November 10, 2009)