FRM 9: Learning from Financial Disasters

Across the last century, interest rate risk has taken down individual firms and, on at least one occasion, most of an industry inside the financial services sector. Controlling it means arranging the balance sheet so that whatever an interest rate movement does to the value of the assets stays highly correlated with what it does to the value of the liabilities. That correlation has to hold in volatile rate environments and not only in calm ones, which is where most attempts at it come apart. Classical duration matching gets a firm part of the way there.

More sophisticated approaches reach for interest rate derivative products: futures, forwards, swaps, caps and floors. Each lets a firm change the interest rate sensitivity of a position without changing the underlying business, which matters when asset maturity is dictated by what borrowers want rather than by what the treasurer would prefer.

Riding the yield curve

When the yield curve slopes upward, long-dated lending pays more than short-dated funding, and a lender that borrows short and lends long books the gap as profit. The banking industry calls this riding the yield curve, and nothing about it is improper so long as everyone understands that the earnings compensate for a repricing mismatch.

The savings and loan industry in the United States, the S&Ls, prospered through most of the twentieth century on two supports: regulation of what could be paid on deposits, and a yield curve that sloped upward, so the rate a borrower paid on a ten-year residential mortgage sat above the rate paid on the short-maturity savings and time deposits that funded it.

Regulation Q restricted interest payments on deposit accounts from 1933 until 2011, and banks could pay nothing at all on demand deposits. The Monetary Control Act of 1980 set up the Depository Institutions Deregulation Committee, and rate regulation was then withdrawn across a six-year period running from 1980 to 1986.

How the spread disappeared

The mortgage product of the era was the fixed rate mortgage. An S&L originated a 30-year fixed rate mortgage, kept it in its own investment portfolio, and funded it by borrowing short. Rising inflation in the late 1970s pushed the Federal Reserve into a restrictive monetary policy and short-term interest rates climbed sharply. Removal of the deposit rate ceiling then forced the S&Ls to pay market rates in order to compete with the newly created money market fund industry. Funding costs rose, the spread the industry lived on was wiped out, and many long-term residential mortgage portfolios began generating negative net interest margins.

The second mistake

Through the 1980s the S&Ls tried to rebuild their balance sheets through new business activities and higher-margin lending, which meant riskier lending, and lost still more money through credit and business risks they controlled poorly. Between 1986 and 1995, 1,043 S&Ls out of a total of 3,234 in the United States either failed or were absorbed, and the number remaining eventually fell below 2,200. The clean-up required what was then one of the most expensive banking system bailouts in the world, USD 160 billion, paid for by American taxpayers.

During this period the industry learned to manage the interest rate risk, and the credit risk, in its mortgage portfolios by issuing mortgage-backed securities. Those products, issued from 1969 onward with government agency backing, did not remove the underlying mismatch of borrowing short while lending long, but they did give the mortgage portfolios of the S&Ls some liquidity.

Lesson from the savings and loan crisis

An unmanaged repricing mismatch can destroy an entire industry without a single borrower defaulting, and a lender that tries to earn its way out of a rate loss by taking credit risk it does not understand usually compounds the damage.

Funding liquidity risk is the risk that an institution cannot raise the cash it needs to meet obligations as they come due, at a price it can bear. It has two sources. One is external, the state of the market: a crisis closes a funding channel for everybody at once. The other is internal and structural, a feature of how the balance sheet has been assembled. Most of the time the answer is both, because an external shock does its damage by finding the structural weakness that was already there.

Market liquidity risk is a different thing and the two are easily confused. Market liquidity risk concerns an asset: whether a position can be sold near its marked value. Funding liquidity risk concerns the institution: whether it can borrow.

Three failures show the loop turning at different speeds. Lehman Brothers, brought down at the peak of the 2007-2009 financial crisis, and Long Term Capital Management a decade earlier, both show funding crises set off by unexpected external conditions that then exposed vulnerabilities built into the business model. Continental Illinois shows an internal credit problem creating the crisis. Northern Rock shows a funding structure that was fragile before any bad news arrived.

Through the late 1990s and the early 2000s the investment bank Lehman Brothers invested heavily in the securitized real estate market of the United States. The 150-year-old institution had pioneered an integrated model: it sold mortgages to residential customers, packaged those loans into highly rated securities, and sold the securities to investors. Much of that paper was backed by subprime mortgage loans rather than by government-backed and prime loans, and the firm had bought several mortgage lenders early in the millennium, among them the subprime lender BNC Mortgage.

Housing in the United States began to sour in 2006 and prices fell after a long boom. Lehman kept building the business anyway, and kept increasing the mortgage-related assets it held as longer-term investments on its own account rather than acting as a middleman. Those assets rose from USD 67 billion in 2006 to USD 111 billion in 2007, according to the Financial Crisis Inquiry Report, and the firm also placed outsized bets on commercial real estate.

The funding strategy that turned a bad position into a fatal one

Banks are leveraged by nature, funding themselves with debt rather than equity, and Lehman took leverage to an extreme. It reached an assets-to-equity ratio of approximately 31:1 by 2007 and borrowed enormous sums on a short-term basis, including daily borrowing in the repo markets, to hold illiquid long-term real estate assets. Under regulatory scrutiny in 2007 and 2008 the temptation to look less leveraged proved strong, and the bankruptcy examiner found that Lehman understated its leverage through Repo 105 transactions, an accounting manoeuvre that moved assets off the balance sheet before each reporting date.

Bear Stearns went first: in July 2007 it had to support two of its own hedge funds after steep subprime losses, and by March 2008 its repo lenders and bank counterparties had lost confidence in its ability to repay. J.P. Morgan bought it for a fraction of its former market value.

Investors then turned to how accurately Lehman had valued its real estate-based assets. Market confidence, the load-bearing element of the funding strategy, drained fast. Major counterparties demanded more collateral for funding transactions, others reduced their exposure, and some refused outright to deal with the firm. Attempts to organise an industry rescue or a sale to another large bank failed, and in the small hours of September 15, 2008, Lehman Brothers filed for bankruptcy, setting off months of panic in global financial markets.

Lesson from the Lehman collapse

Funding that has to be renewed daily is not funding at all once the assets behind it cannot be sold quickly, and leverage of that order converts a question about valuation into an immediate question about solvency.

Continental Illinois Bank ran the sequence in the opposite order from Lehman Brothers. Trouble began inside the credit portfolio, and weaknesses in the funding strategy then turned a bad loan book into a full liquidity crisis.

Once the largest bank in Chicago, Continental began an aggressive growth push from the late 1970s. Its commercial and industrial lending jumped from USD 5 billion to over USD 14 billion in the five years before 1981, and total assets grew from USD 21.5 billion to USD 45 billion.

Penn Square and the loans that came with it

The first crack was the closure of Penn Square Bank in Oklahoma. Penn Square had lent to oil and gas producers across Oklahoma through the boom of the late 1970s, and whenever a loan exceeded what it could service it passed the exposure to a bigger institution such as Continental Illinois. Prices for oil and natural gas fell after 1981, and some of those firms defaulted. Penn Square became insolvent in 1982 and regulators closed it. By then Continental held more than USD 1 billion of loans to Penn Square’s oil and gas customers, and its losses as defaults rose were heavy.

Many other banks took credit losses in the same period. What made Continental unusual was the liability side. It ran only a tiny retail banking operation and held a relatively small amount of core deposits, so it funded its lending mainly with federal funds, a form of interbank lending, and with large issues of certificates of deposit.

Once Penn Square failed, Continental found it increasingly difficult to fund itself in the markets of the United States and began raising money at much higher rates in foreign wholesale money markets, including Japan. In May 1984 rumours about the bank’s worsening condition spooked the international markets, foreign investors pulled their deposits, and depositors withdrew USD 6 billion in only ten days. Regulators intervened to prevent a domino effect that they feared could put the entire banking system of the United States at risk.

Lesson from Continental Illinois

Wholesale funding is the most confidence-sensitive money a bank can hold, and a credit problem large enough to reach the newspapers will remove it. Without a core deposit base there is no cushion between a credit loss and a funding run, and buying replacement money above the market advertises the problem to the lenders the bank needs to keep.

The failure of the mortgage bank Northern Rock in 2007 is the clearest example of liquidity risk created by a structural weakness in a business model. Heavy reliance on short-term financing for long-term assets, combined with a sudden loss of market confidence, produced a funding liquidity crisis that ran to disaster in weeks.

Northern Rock was a medium-sized and fast-growing mortgage bank in the United Kingdom specialising in residential mortgages, and it had been growing assets by around 20% per year over several years. It was still expanding aggressively in the first quarter of 2007. Behind that growth was a funding strategy unusual among banks in the United Kingdom: an originate-to-distribute approach, raising money by securitizing mortgages, selling covered bonds and using the wholesale funding markets. Retail deposits mattered far less to Northern Rock than to its peers, and wholesale investors mattered far more. The bank tried to address the weakness by diversifying its funding geographically, tapping markets in continental Europe, the Americas and the United Kingdom, and that diversification proved worth much less than its executives believed.

When every channel closes at once

After several years of strong growth and rising house prices, doubts about mortgage-related assets surfaced among investors early in 2007. Rising default rates in the subprime mortgage market of the United States triggered them, and the doubt spread outward: to asset-backed securities as an investment class, then to the institutions that had invested in or depended on those securities, and finally to the interbank markets. Once the interbank funding market seized up in early August 2007, every one of Northern Rock’s global funding channels seized simultaneously, a scenario its executives later called unforeseeable. Earlier that summer the bank had announced increased interim dividends after regulators in the United Kingdom approved a Basel II waiver allowing advanced approaches for calculating credit risk, which looked likely to cut its minimum regulatory capital requirements.

With interbank funding gone, the authorities began discussing ways to relieve the difficulty. News that the Bank of England planned a support operation leaked, and a run on deposits followed in mid-September. Panic was worsened by the compensation rules then in force, under which private depositors had a full guarantee only up to £2,000, plus a guarantee of 90% on sums as far as a ceiling of £33,000. Calm returned slowly, and only after the authorities publicly promised that deposits would be repaid. Northern Rock accepted emergency government support and was then taken into public ownership.

Lesson from Northern Rock

Diversification of funding sources is worthless when the sources share a driver, and a model that depends on wholesale markets staying open is a bet on market conditions rather than a funding plan. A partial deposit guarantee, meanwhile, gives retail depositors a rational reason to queue.

After the 2007-2009 crisis the Federal Reserve began requiring liquidity stress testing programmes at the largest banks, designed to confirm that a bank’s liquidity and funding strategies can survive a system-wide stress scenario rather than an idiosyncratic one. That is the regulatory answer. The management answer is that funding liquidity risk is not eliminated, it is priced, and every choice about it costs something elsewhere.

Much of the work sits in optimising the composition of a bank’s borrowing sources, usually by managing the contractual maturities of assets and liabilities. Those maturities can be adjusted directly, by changing what the bank issues, or synthetically, using derivatives such as interest rate swaps. Like most complex decisions, asset/liability management decisions come down to trade-offs, and two of them matter most.

The two trade-offs

The first is between funding liquidity risk and interest rate risk, and it is unavoidable. When the funding liabilities have shorter duration than the loan assets, the bank carries less interest rate risk and more funding liquidity risk. Lengthen the liabilities relative to the assets and the exposures swap places. There is no arrangement that removes both.

The second is between cost and risk mitigation. In a positively sloped yield curve environment an institution can mitigate funding liquidity risk by extending the maturity of its funding liabilities, and doing so plainly costs more than cheaper short-duration funding. Working from the other side, a bank can shorten the maturity of assets such as commercial loans, but that is often not available, because asset maturity is driven by what borrowers want, by the nature of the business and by the competitive environment.

Because liquidity cannot be coordinated perfectly, firms also need emergency liquidity cushions. A larger and better quality cushion lowers the risk and lowers earnings too, since highly liquid and marketable assets yield less than illiquid ones, and credit lines command a cost even when nothing is drawn.

Funding liquidity risk management, interest rate risk management, product pricing, profit planning and capital management are all linked, and an asset/liability management policy that treats them separately will optimise one at the expense of the others.

Building an effective hedging strategy is useful and difficult, and the difficulty is not confined to banks and other financial institutions. Non-financial firms face the same problem whenever a commercial contract fixes a price they do not control. Whoever is given the job needs access to relevant information, market data and corporate information among it, and often needs statistical tools that are at least appropriate to the exposure and sometimes need to be advanced.

The tactical choice

Whether to hedge statically or dynamically is a key tactical choice. A static hedging strategy buys a hedging instrument that closely matches the position being hedged and typically holds it for as long as the underlying position exists, or at least for a set period. It is relatively easy to implement and easy to monitor, and its attention is on the result at the horizon.

A dynamic hedging strategy instead adjusts the hedge through a series of ongoing trades, recalibrating the hedge position continuously or frequently as the underlying exposure changes. It demands more managerial effort to run and to monitor, and rebalancing may carry higher transaction costs. Where a static approach cares about the outcome at the end, dynamic hedging tries to rebalance over short intervals, in some cases daily.

Firms that run dynamic hedging strategies must have the models and the market expertise to trade and monitor the positions. Having them does not prevent mistakes in the implementation of a risk management strategy, nor in the communication of it, and the case that follows failed on both counts.

MGRM, Metallgesellschaft Refining & Marketing, Inc., was the subsidiary in the United States of Metallgesellschaft AG, an industrial conglomerate based in Frankfurt, Germany. During 1993 MGRM signed long-term contracts at fixed prices to deliver oil products, mainly gasoline and heating oil, to its end-user customers. Prices could not be changed once those contracts were signed, which left the firm exposed to rising energy prices.

No liquid market existed in long-term futures contracts of the matching maturity, so MGRM built a dynamic hedge out of short-dated energy futures contracts. The hedging instruments had to be rolled forward each month as they expired, and the derivative position was adjusted each month to reflect the changing amount of outstanding contracts to be hedged and preserve a one-to-one hedge. Culp and Miller argued afterwards that the strategy was neither inherently unprofitable nor fatally flawed, provided top management understands the programme and the long-term funding commitments needed to make it work.

Backwardation, contango and the roll

This construction is known as a rolling hedge, and its economics depend on the shape of the price curve. It pays when assets for immediate delivery are priced above assets for future delivery, a shape called backwardation, because the expiring contract is then sold at a price above that of the longer-delivery contract replacing it and each roll produces a rollover profit. In contango, the opposite shape, every roll produces a loss.

MGRM was therefore exposed to curve risk, the risk that the price curve shifts between backwardation and contango, and to basis risk arising from deviations between short-term and long-term prices. Both arrived. Spot oil prices fell heavily during 1993, from nearly USD 20 a barrel mid-year to less than USD 15 a barrel by year-end, producing margin calls of USD 1.3 billion on the long futures positions of MGRM that had to be met in cash. The firm held unrealized economic gains on the original short forward contracts it had written, but gains are not cash, and the negative cash flow was substantial even though temporary. The price curve then changed shape, moving from backwardation to contango.

The parent company had been told that the position was hedged and did not expect a negative cash flow at all. In December 1993 it ordered the hedges liquidated, which turned large paper losses into large realized ones.

Lesson from Metallgesellschaft

A position can be hedged in economic terms and still fail, because a hedge that is right about value can be wrong about the timing of cash. A dynamic hedging programme creates a funding commitment, and the board has to understand and approve that commitment before the first margin call arrives.

The horizon over which a hedge runs is a decision in its own right. Horizons can be fixed, at a quarter-end or a year-end, or they can roll, and whichever is chosen, performance evaluations and investment horizons need to be aligned.

Accounting can undo an economic hedge

Accounting and tax consequences have to be settled while the strategy is being designed rather than after it is running. The rules covering derivatives and hedging are complex and they change. A derivative and the position it is meant to hedge have to be matched precisely, on dates and on quantities, before the two can be reported together in operational profit with no accounting profit or loss to show. Without that match, International Financial Reporting Standards require the mark-to-market profit or loss on the hedge to be recorded. Where the hedge is at least 80% effective, that profit or loss can go into operational or gross profit. If not, the financial position counts as a financial expense while the underlying one counts as an operational expense, and the two land in different places.

How derivatives are accounted for therefore affects the quarterly and annual financial reports and the profit and loss statement directly. The MGRM case is the standing illustration of the gap between economic hedging and accounting hedging, and of the difference between hedging the profit and loss statement and hedging the cash flows. MGRM was close to fully hedged in economic terms and fully exposed in accounting terms, and it had made no preparation for absorbing the liquidity risk that followed.

Tax treatment varies with the instrument and by country, so the cash flows of a firm can turn on which derivative was used and where, and multinationals sometimes find it advantageous to hedge a position relating to business in one country using derivatives written in another.

Implementation is part of the strategy

Markets move while a programme is being put on, so a hedge that looked attractive at the design stage can be unattractive by the time it is executed. Firms have to adapt during implementation with the same care they gave the original design, and once the hedge is on, the positions need monitoring for their fit with the overall strategy and for their continuing effectiveness.

Sophisticated financial products are priced by valuation models, some theoretical, such as the capital asset pricing model, and some statistical, such as a fitted term structure of interest rates. Model risk is the exposure a firm takes on by relying on those models. It can come from using the wrong model, from specifying a model incorrectly, or from feeding a sound model with insufficient data and poor estimators.

The most dangerous version is a flawed assumption about the process being modelled, for the underlying asset price or for interest rates. A bond pricing model might assume a flat yield curve while the actual curve is upward-sloping and unstable. This kind of error is common, severe and among the hardest to detect.

The Niederhoffer put options

Victor Niederhoffer was a star trader running a successful and well-established hedge fund. One of its strategies was to write large quantities of uncovered, or naked, put options that were deep out-of-the-money on the S&P 500 Index, collecting the premiums. Because the options were deep out-of-the-money, each premium collected was small. Behind the strategy sat an assumption that any one-day market decline beyond 5% would be rare, and under normally distributed returns a fall of that size would be virtually impossible.

The assumption was tested in October 1997, when the stock market fell by over 7% in a single day. A steep overnight decline in the Hang Seng Index came first, itself the product of a developing crisis in Asian markets. Liquidity in the markets dried up on the back of the shock, the fund could not meet over USD 50 million in margin calls, and its brokers liquidated the positions for pennies on the dollar, wiping out the fund’s equity.

Lesson from the Niederhoffer case

An option strategy can be constructed that yields a small profit over a long stretch of time while carrying a small probability of a very large loss. The distribution of results is not the same thing as the average of them, and competitive financial markets rarely offer a free lunch.

The collapse of Long Term Capital Management in August and September of 1998 stands out both for the size of its exposures and for the pedigree of the people running it. LTCM was founded in 1994 by John Meriwether, and its principals included the former Federal Reserve Board Vice-Chairman David Mullins, the Nobel laureates Myron Scholes and Robert Merton, a number of world-renowned academics, and experienced traders from the bond arbitrage desk at Salomon Brothers. Before it failed the fund had USD 4.8 billion in equity supporting USD 125 billion in assets, a leverage ratio of 25-to-1.

The trigger came in August of 1998, when the government of Russia declared a debt moratorium and devalued the ruble. Those actions drove the value of LTCM’s holdings down by over 40%, which was a loss of nearly USD 2 billion. Concerned about a systemic crisis, the Federal Reserve Bank of New York brokered a rescue in which a group of banks put USD 3.5 billion into the fund, taking a 90% equity stake and management control in exchange.

Why one country’s default mattered so much

LTCM traded on a market-neutral basis, also known as relative-value trading, buying one asset and simultaneously selling another to exploit a perceived mispricing between the two. The fund judged that spreads between sovereign and corporate bonds across various countries were too wide and would revert to normal levels, so it would buy corporate bonds in the United Kingdom and short the appropriate government bonds there. Other trades anticipated the convergence of sovereign bond yields as several European countries prepared to join the European Economic and Monetary Union, for instance buying Spanish or Italian government debt while selling German bunds. Provided the yield spread narrowed, the position paid regardless of absolute prices.

Returns from low-risk strategies of this kind are thin, and they thinned further as more traders arrived to chase the same opportunities. To lift the return on equity LTCM used leverage, helped by its ability to obtain very large financing collateralized by the bonds it held, which lenders granted partly because the strategies were widely perceived as low-risk in nature.

Models and strategies built on relationships observed in benign conditions are vulnerable exactly when conditions stop being benign. Events in Russia during August 1998 made market participants fear further sovereign defaults, and those fears drove an exodus of investors from emerging markets and other risky assets into liquid, less risky assets such as the government debt of the United States and Germany. This flight to quality pushed spreads apart sharply: safe haven assets such as treasuries against emerging market bonds and high-yield corporate bonds, German debt against Italian debt, and credit spreads over a range of asset classes.

As spreads widened, relative-value trades lost money and lenders demanded more collateral, so hedge funds had to sell assets at fire-sale prices or abandon their arbitrage plays. Liquidity drained out of many markets and volatility increased. The breakdown of the historic correlation and volatility patterns assumed in LTCM’s models caused most of its losses, through three effects. Interest rates on treasuries in the United States fell in tandem with stock prices, because investors deserted the stock market for government bonds, whereas in normal markets the two are negatively correlated. Liquidity vanished in many markets simultaneously, which made unwinding positions exceedingly difficult. Portfolios that appeared well diversified across markets began to behave as though they were concentrated in a single market, and market-neutral positions became directionally exposed, usually to the wrong side.

What the risk model missed

For risk control LTCM leaned heavily on a Value-at-Risk (VaR) model. VaR measures the worst-case loss for a set of investments under normal market conditions, over a specified time horizon at a given confidence level, and every one of those qualifiers did work here. The episode showed how poorly the assumptions behind regulatory VaR transfer to a hedge fund. The horizon for economic capital ought to be the time needed to raise new capital, to liquidate positions in an orderly manner, or the span over which a crisis scenario unfolds, and on the LTCM experience ten days is far too short. Liquidity risk does not enter traditional static VaR models at all, since they assume normal conditions and perfect market liquidity. Correlation and volatility risks, meaning the risk that realized correlations and volatilities deviate significantly from expectations, can be captured only through stress testing, which was probably the weakest point of the LTCM system.

The president of the New York Federal Reserve, William McDonough, told Congress that stress testing was a developing discipline and that adequate testing had clearly not been done on the conditions that precipitated the problems at Long-Term Capital. The numbers make the same point. In the run-up to the collapse the fund experienced daily volatility of more than USD 100 million, over twice the level it had envisioned, and while it estimated its ten-day VaR at USD 320 million, its losses exceeded USD 1 billion.

Lesson from Long Term Capital Management

Its strategies were valid in the medium term, and the banks that took the fund over realized substantial profits once the crisis ended. Being right eventually is worthless without the funding to survive being wrong temporarily.

J.P. Morgan Chase lost several billion dollars over the first six months of 2012 on a massive credit derivatives portfolio run out of its London office. What follows draws on the 300-page report produced by the investigation of the United States Senate. With USD 2.4 trillion in assets, it is the biggest financial holding company in the United States, the world’s largest derivatives dealer and the single largest participant in the world’s credit derivatives markets, and it had consistently portrayed itself as an expert in risk management running a fortress balance sheet. Its Chief Investment Office, charged with managing USD 350 billion in excess deposits, placed a massive bet on a complex package of synthetic credit derivatives that lost at least USD 6.2 billion in 2012. The bank’s Chief Executive Officer initially dismissed the episode as a tempest in a teapot, and the losses then doubled and tripled in a relatively benign credit environment.

How the portfolio grew

The Chief Investment Office approved a proposal to trade synthetic derivatives in 2006, and by 2008 was calling the business the Synthetic Credit Portfolio. In 2011 its net notional size jumped from USD 4 billion to USD 51 billion, a more than tenfold increase, and late that year it bankrolled a USD 1 billion credit derivatives bet that gained approximately USD 400 million. In December 2011 the bank instructed the office to reduce its Risk Weighted Assets so that regulatory capital requirements could come down. Rather than dispose of the high risk assets, which is the most typical way to reduce Risk Weighted Assets, in January 2012 the office bought additional long credit derivatives to offset its short positions. That trade increased the size, the risk and the Risk Weighted Assets of the portfolio, and by taking it into a net long position it eliminated the hedging protection the portfolio was supposed to provide.

Valuation, limits and the VaR model

The portfolio opened 2012 with losses, and across January, February and March there was not a single day in the black. To minimize reported losses the office stopped marking credit derivatives at or near the midpoint of the daily bid-ask range, which had complied with the requirement to use prices most representative of fair value, and began assigning itself more favourable prices from within that range. By March 16, 2012 it had reported year-to-date losses of USD 161 million, and at midpoint prices they would have swelled by at least another USD 432 million to a total of USD 593 million. Two business lines inside the same bank now held different values for identical holdings, and once counterparties learned of the difference, collateral disputes peaked at USD 690 million. In May the Deputy Chief Risk Officer directed the office to mark its books as the Investment Bank did, using an independent pricing service, which ended the mismarking.

Limit breaches were routinely disregarded, risk metrics were frequently criticised or downplayed, and risk evaluation models were targeted by staff seeking artificially lower capital requirements. The office used five key metrics and limits to control its trading, Value-at-Risk (VaR) among them, and the portfolio breached all five during the first three months of 2012. From January 1 through April 30, 2012 the office breached its risk limits and advisories more than 330 times. The breaches were reported to management, risk personnel and traders, and were largely ignored or ended by raising the limit that had been breached.

The model itself was then adjusted. After analysts concluded that the existing VaR model was too conservative and overstated risk, an alternative was rushed into use in late January 2012 while the office was breaching both its own VaR limit and the bankwide one, and without the approval from the Office of the Comptroller of the Currency that should have been obtained. The new model immediately lowered the reported VaR by 50%, which ended the breach and left room for substantially more risky trading. Months later the bank determined that it had been improperly implemented, requiring error-prone manual data entry and containing formula and calculation errors, and on May 10 it revoked the new model and reinstated the prior one.

Lesson from the London Whale

A risk limit that is raised whenever it binds is not a limit, and a model replaced because its output is inconvenient is no longer a control.

Profit is normally treated as good news, particularly inside a financial firm. The collapse of Barings Bank, brought about by Nick Leeson, is the standing argument that outsized profits can equally signal risk nobody has recognised, and that they deserve as much curiosity as celebration.

Leeson moved to Singapore in 1992 as local head of operations for Barings, a centuries-old British institution founded in 1762, and part of his role was executing client trades on the Singapore International Monetary Exchange. His responsibilities widened and he was authorised to run an arbitrage strategy exploiting price disparities between Nikkei futures contracts listed there and those listed on the Osaka Securities Exchange. Arbitrage of that kind requires offsetting trades in the two markets. Leeson did not place them: he bought in one market, held the contracts, and built speculative positions that quickly generated huge losses.

He also controlled the Singapore back office, and the combination of the two roles made concealment possible. Using a reconciliation account he turned a real 1994 loss of GBP 200 million into a reported profit of GBP 102 million, and he arranged for that account to be excluded from the reports sent to head office in London.

By late 1994 the size of the reported profits had drawn the attention of the Barings risk controllers. Their questions went to Leeson’s superiors and were rebuffed on the grounds of the bank’s unique ability to exploit this arbitrage, and the extra bonuses those superiors collected on the back of the reported profits may have clouded their judgement. Suspicions surfaced again in January 1995 after another implausibly large weekly profit, and were dismissed again. Simple calculations would have settled the matter, since profits of that size in the manner claimed would have required Leeson to trade a multiple of the entire weekly volume in Nikkei futures on both exchanges combined. By the time Barings discovered the rogue trading, the accumulated losses were too large to absorb and the bank had to be liquidated. The Dutch bank ING acquired it for a nominal sum.

Part of the difficulty is structural. Large trading volumes and revenues produce large bonuses for senior managers, which encourages them to trust the traders reporting to them, and a trader can use superior knowledge of pricing models to confound internal critics. The antidote is scepticism about any strategy promising above-market returns and insistence that every model be transparent and independently vetted.

Lesson from Barings

Reporting and monitoring of positions and risks, which is back office work, must be separated from trading, which is front office work. Outsized or strangely consistent profits, the kind later associated with Bernie Madoff as well, should be independently investigated and rigorously monitored to confirm that they are real, that they were generated in accordance with the firm’s policies and procedures, and that they do not result from nefarious or unacceptably risky activity. A risk manager also owes the firm a judgement about whether reported business profits look logical given the positions held. Barings would have been caught by rules introduced a few years later, since the Basel Committee, alongside capital adequacy requirements for market risk, set limits on concentration risk that oblige banks to report large exposures relative to capital and forbid positions beyond a stated share of it.

Forwards, swaps and options are the building blocks from which financial engineering works. Each can hedge a specific risk on its own, and they can be combined into complex structures that meet a client’s needs. Derivatives let an institution break risks apart and treat each separately, and they equally allow several risks to be managed jointly.

Financial engineers devise these instruments to satisfy the risk and return appetites of clients. That is not by itself risk management, and in derivatives markets the line between hedging and speculation blurs easily. Firms are tempted into complex transactions that enhance immediate portfolio returns, and enhancing returns almost always means taking more risk somewhere, often as a loss that is unlikely but potentially severe. Too often the embedded risk is not properly understood by the firm entering the transaction, or is never communicated to senior managers and other stakeholders.

Bankers Trust, Procter & Gamble and Gibson Greetings

In the early 1990s Bankers Trust proposed that two corporate clients, Procter & Gamble and Gibson Greetings, enter complex leveraged swaps in order to lower their funding costs. In the swap with Procter & Gamble, Bankers Trust would pay a fixed rate for five years while the company paid a floating rate, which came to the commercial paper rate minus 75-basis points as long as rates stayed stable. A complex formula sat behind that floating leg, and it made the rate rise steeply if rates rose during the period: a rise of 100-basis points in rates generated a spread of 1,035-basis points over the commercial paper rate.

In 1994 the Federal Reserve raised the federal funds rate by 250-basis points. The losses to both companies were colossal, and both sued Bankers Trust for misrepresenting the risk embedded in the transactions. Bankers Trust never fully recovered from the reputational damage and was eventually acquired by Deutsche Bank.

Lesson from Bankers Trust

A structure that lowers a funding cost in the base case has usually sold an option to do so, and the buyer of that structure owns the tail. Selling complexity that the client cannot value is a reputational exposure for the dealer that can outlive the trade by years.

Repurchase agreements, or repos, are a way of borrowing cash by selling securities to a counterparty and agreeing to buy them back shortly afterwards at a slightly higher price. They let an investor finance a large part of a portfolio with borrowed money, which multiplies both profit and loss, so even a small move in market prices matters.

Leverage obtained through repos was central to the undoing of Orange County in California. In the early 1990s the county treasurer, Robert Citron, had borrowed USD 12.9 billion in the repo market, which let him accumulate around USD 20 billion of securities although the fund he managed held only USD 7.7 billion in invested assets.

Citron put the borrowed money into complex inverse floating-rate notes, instruments whose coupon payments fall when interest rates rise, the opposite of a conventional floater. In the favourable upward-sloping curve environment of the years before 1994 the structure worked, and he increased the return of the fund by 2% compared with similar pools of assets.

Across 1994 the Federal Reserve pushed interest rates up by 250-basis points. The market value of the positions dropped substantially, generating a loss of USD 1.5 billion by December 1994, and at the same time several of the fund’s lenders stopped rolling over their repo agreements. Orange County was forced into bankruptcy. The cause was excessive leverage combined with a risky interest-rate bet embedded in the securities the fund had bought, a bet that turned out to be wrong. Citron later admitted that he had understood neither the position he had taken nor the risk exposure of the fund.

Other asset managers who bought inverse floaters made the same mistake about embedded leverage. An inverse floater could carry a duration of 15 to 25, yet in several administrative hearings at the Securities and Exchange Commission the managers of limited duration funds whose portfolios were destroyed by these holdings testified that the durations they ran were between 1 and 3.

Lesson from Orange County

Firms have to understand the risks that sit inside their own business model, and senior management has to put robust policies and risk measures in place that tie risk management, and the use of derivatives in particular, to the risk appetite and the business strategy that stakeholders have been told about. Management and boards should keep asking where the risks are hiding and under what circumstances they would produce a loss.

Some of the largest buyers of subprime securities from the United States before the 2007-2009 financial crisis were European banks, among them the publicly owned German banks known as the Landesbanken. Those instruments offered an attractive risk premium, and they also required understanding and pricing expertise.

The Landesbanken had traditionally specialised in lending to small and medium-sized companies in their own regions, which is a different discipline entirely. During the boom years some of them opened overseas branches and built investment banking businesses, and the most notorious example was Sachsen Landesbank, based in Leipzig.

Sachsen opened a unit in Dublin whose job was to set up vehicles holding large volumes of highly rated mortgage-backed securities from the United States. Those vehicles were technically off the parent bank’s balance sheet, and they carried the guarantee of Sachsen itself, which is the detail that mattered. The operation was highly profitable and it was simply too large relative to the size of the Sachsen balance sheet. When the subprime crisis arrived in 2007, the rescue wiped out the capital of the bank and Sachsen had to be sold to Landesbank Baden-Wurttemberg, another German state bank.

Reading the 2007 annual report of Sachsen suggests its risk management systems did not treat the funding liquidity commitment to these vehicles as a credit or liquidity risk at all. They booked it under operational risk, on the argument that only an operational failure could cause the loan facility to be drawn down, and it attracted little or no capital.

Lesson from Sachsen Landesbank

A guarantee to an off balance sheet vehicle is an on balance sheet exposure that has not been measured yet, and a high credit rating on a tranche is not a substitute for the ability to price what sits underneath it. Classifying a funding commitment into the risk category that attracts the least capital does not make the exposure smaller.

A firm’s reputation rests on a belief that it will keep the promises it makes to counterparties and creditors, and that it deals fairly and follows ethical practices. Concern about reputation risk has grown with the spread of social networks, since a rumour can travel across the internet and destroy a reputation within hours. Companies also face growing pressure to demonstrate commitment to environmental, social and governance best practice.

The scandal that hit the German carmaker Volkswagen concerned regulatory testing. In September 2015 the Environmental Protection Agency of the United States announced that Volkswagen had programmed some of the emissions controls on its diesel engines to activate only during regulatory testing and not during real-world driving. Nitrogen oxide levels therefore met the standards of the United States while a car was being tested, and greatly exceeded them in ordinary driving. From 2009 through 2015 Volkswagen installed this programming in over ten million cars worldwide, 500,000 of them in the United States alone. Executives in Germany and in the United States admitted the deception on a conference call in September with the Environmental Protection Agency and officials from California.

The damage to the world’s biggest carmaker was heavy. As the scandal unfolded the share price of the company fell by over a third, the firm faced billions of dollars in potential fines and penalties, and numerous lawsuits were filed. Its reputation took a severe hit, particularly in the important market of the United States. The effect reached past the company as well: German government officials worried publicly that the value of the label Made in Germany would be diminished by what Volkswagen had done.

Lesson from Volkswagen

Reputation risk is not a soft category sitting next to the real ones. A deliberate breach of a rule can convert into share price losses, fines, litigation and lasting commercial damage, and the damage is not contained by the boundaries of the firm that caused it.

Enron was formed in 1985 by the heavily leveraged merger of InterNorth and Houston Natural Gas. Deregulation then stripped the firm of the exclusive rights to its pipelines, and to survive it devised a new strategy, the so-called gas bank: buying gas from a range of suppliers and selling it on to a network of consumers at guaranteed volumes and prices, charging fees in return for assuming the risks in between. Along the way it created a market for energy derivatives where none had existed.

Fortune named it America’s Most Innovative Company in 1995, and it won that award for six consecutive years. Its shares were worth almost USD 90.56 at their peak in August 2000, a year in which the firm employed 20,000 people and earned revenues of nearly USD 101 billion.

Enron pushed constantly for deregulation of energy markets, and California was the prominent example. California had capped retail electricity prices after a shortage it attributed to market manipulation. Enron could drive power prices up by as much as 2,000% by taking plants offline at times of peak demand, which squeezed revenue margins across the industry and led to the 2001 bankruptcy of Pacific Gas and Electric Company, among the largest power companies in the country. Enron filed for bankruptcy itself in December 2001, the largest corporate bankruptcy in the history of the United States at the time.

Agency risk, a passive board and fraudulent accounting

Many in senior management acted in their own self-interest and against the interests of shareholders, which is agency risk. The chairman and chief executive Ken Lay was charged with falsifying the publicly reported financial results of Enron and making false and misleading public representations about its business performance and financial condition. The board failed its fiduciary duties too: it was aware of and allowed an arrangement under which the chief financial officer became the sole manager of a private equity fund that did business with Enron, and that fund turned out to lack economic substance.

Most damning was the accounting. Enron transferred its own stock to a special purpose vehicle in exchange for cash or notes, and the vehicle recorded the stock as an asset while Enron guaranteed the vehicle’s value to reduce its credit risk. Since the vehicle was capitalized entirely with Enron stock, a fall in that stock would raise the credit risk of the vehicle exactly when Enron could least support it, and the firm failed to disclose adequately that the relationship was not at arm’s length. A second device was more direct: the firm would construct a physical asset and immediately declare a projected mark-to-market profit on it, before earning anything, and if actual revenue came in below the projection it transferred the asset to a special purpose vehicle so the loss went unreported. Unprofitable activities could be written off without touching the bottom line.

The audit function had been outsourced to Arthur Andersen, then one of the Big Five accounting firms, which either failed to catch or explicitly approved many of the fraudulent practices. Once the scandal came to light, Andersen surrendered its accounting licences to the Securities and Exchange Commission, which was effectively a death sentence for the firm.

Aftermath and the lesson from Enron

The Sarbanes-Oxley Act of 2002 was the key legislative reform to come out of the affair, together with associated changes in stock exchange and accounting rules. It created the Public Company Accounting Oversight Board, which sets auditing standards, holds the power to investigate, and has taken an important role in promoting good corporate governance and financial disclosure. Boards and audit committees now rely increasingly on the chief risk officer to integrate corporate governance responsibilities with existing risk management responsibilities. Enron is the poster child of corporate governance failure and poor risk management, and the lesson is that no amount of risk measurement protects a firm whose board does not enforce independence, whose auditor is not independent either, and whose accounting can move losses out of sight.

Cyber risk has turned into a critically important consideration. Bank systems can be hacked, cash machines can be used to steal money and client information, and customer identities can be stolen and misused. Financial institutions spend billions of dollars every year on making their systems safer, against the outside world and against internal misuse alike. Threats from cyberattacks worry the Bank for International Settlements and the International Monetary Fund as much as they worry local regulators.

SWIFT is the leading system in the world for transferring funds electronically between banks, processing billions of dollars of transactions every day. It is regarded as so reliable that transactions which would ordinarily take days, precisely so that fraud can be prevented, are instead completed in seconds.

In April 2016 an article in The New York Times revealed that hackers had exploited the SWIFT network to steal USD 81 million from the account held by Bangladesh Bank, which is the central bank of Bangladesh, held at the New York Fed. The theft used malware that sent unauthorized SWIFT messages directing funds to an account the hackers controlled. The malware then deleted the database record of the transfer and disabled the transaction confirmation messages that would otherwise have exposed the theft.

Lesson from the SWIFT case

Speed and trust in a payment network are the features an attacker uses, since a system built to settle in seconds leaves no interval in which a human being can query anything. The attack also went after the audit trail rather than only the money, which is the part worth remembering: a control that reports on itself can be silenced by whoever has already got inside.

Ordinary business factors cause large losses too: adverse macroeconomic conditions, tougher competition, evolving technology. Every loss in this lesson came from beyond that. Understanding the mistakes other people have already made is one of the few reliable ways to design better risk management systems across an enterprise, and that work matters more than calculating a regulatory or economic capital number.