Long-Term Capital Management

Analysis and Commentary

Long-Term Capital Management (LTCM) was founded as a hedge fund in 1994 by Salomon Brothers star trader John Meriwether. LTCM enjoyed an impeccable reputation and boasted two Nobel Laureates on staff: Robert Merton and Myron Scholes. The firm primarily invested in risk arbitrage strategies and was well known for its acumen in this area. The firm started with about $1 billion in seed capital and grew quickly from 1994 to 1998. Prior to its collapse in 1998, the firm managed about $4.8 billion in client capital, invested approximately $130 billion in total capital (through borrowing), and controlled over $1.2 trillion in notional amounts of derivatives. The company was also known for receiving favorable terms from lenders and trading partners based primarily on its sterling reputation.

LTCM marketed itself as providing superior returns for risk that was no greater than that found in the US equity market. In 1994, LTCM gained 20% while the S&P 500 was up only 1.3%. In 1995, LTCM gained 42.8% while the S&P 500 was up 37.2%. In 1996, LTCM gained 40.8% while the S&P 500 was up 22.6%. However, the following year, LTCM trailed the market for the first time. In 1997, LTCM gained 17% while the S&P 500 was up 33.1%. In 1998, LTCM was down 84% while the S&P 500 was up 28.3%.

Many suggest, even today, that LTCM was a victim of aberrant market events. According to this theory, the Russian bond default came out of the blue in August 1998, catching the market by surprise, and the subsequent flight to safety was unprecedented. This extreme market movement worked against LTCM’s otherwise intelligent positions and forced both the margin calls and the selling of assets that would cause a systemic risk event for the United States. While the market was indeed caught off-guard, the complete picture is far more complicated. To examine this issue, we first explore the nature of the event and then consider the adequacy of LTCM’s chosen approach to investing.

“To make money they didn’t have and didn’t need, they risked what they did have and did need. And that’s foolish.” (Warren Buffett)

Regarding the nature of the event, if we are to accept the claim that LTCM was blameless and the market’s behavior was simply random, then there are two questions to answer: First, did the Russian bond default, in fact, come out of the blue? And second, why did the Russian bond default have such a large impact on global markets (and was this impact unusual)?

Did the Russian bond default come out of the blue? Quite simply, the answer is no. In the three years leading up to the Russian bond default, there was a series of unfolding events, each of which escalated the risk of a default. In 1995, the US government entered into the so-called Reverse Plaza Accord to increase the value of the US dollar and reduce the value of the Japanese yen. Over the ensuing 18 months (by early 1997), the US dollar had risen by roughly 50%. At the time, Russia maintained an exchange rate band of five to six rubles per dollar. So the rising dollar also lifted the ruble (and Asian currencies pegged to the dollar) by roughly 50%. The rising dollar weakened Asian economies and dampened Asia’s demand for oil, which had been a key driver of global oil consumption—thus resetting the market’s expectations about future oil demand. The Russian economy is highly sensitive to oil specifically and to commodity demand in general. As the US dollar rose and as the Asian countries entered recession in 1997, oil prices declined. Not only was demand for Russian oil faltering, but the tax revenues the Russian government received from oil were falling as well. Russia already had enormous fiscal deficits, equal to 9% of GDP, and now the fiscal problem was getting worse. Many countries (Thailand, Indonesia, Philippines, Korea, Russia, Brazil, etc.) also had substantial amounts of US dollar–denominated debt. Thus, the rising US dollar presented these countries with a stark choice—either devalue their home currency to reestablish an attractive price on the foreign exchange market or live with the higher currency, be forced to repay much greater amounts of foreign-denominated debt, and live with a longer, more severe recession. In July 1997, Thailand succumbed first, breaking the peg on the Thai baht. One by one in rapid succession, other countries with similar financial profiles (Indonesia, South Korea, Malaysia, the Philippines, etc.) also experienced a currency crisis.

Brent crude oil prices fell from a high of $24 a barrel in January 1997 to about $17 a barrel by mid-April 1997 as Asian demand weakened. As of June 1997, the Central Bank of Russia (CBR) had $20.4 billion in foreign currency reserves, but as capital fled Russia, the CBR had to use its reserves to purchase rubles in the currency markets to maintain the peg. The Russian government then derived about two-thirds of its tax revenues from the energy sector. So as the price and volume of oil declined, Russian government tax revenues declined sharply. Realizing that Russia was facing increasing financial duress, the currency markets took their first crack at breaking the peg in November 1997. By this time, Russia’s foreign currency reserves had fallen to $12.2 billion. In December 1997, however, commodity prices began sliding again, with Brent crude prices falling from roughly $19 to $11 a barrel by March 1998. In January 1998, selling pressure on the ruble erupted again. As before, the CBR was able to offset the surge in selling pressure by buying rubles with its foreign reserves and reestablished the trading range of the ruble. But reserves had fallen to $10.5 billion. At this point, the CBR worked behind the scenes to negotiate forward contracts with banks and other private entities to deliver currency. This tactic helped reduce the immediate pressure to sell rubles in the spot market. Rumors of the CBR’s actions—rumors which later turned out to be true—swirled in the market for months. And during the third wave of frenzied selling pressure on the Russian ruble in early August 1998, the CBR’s foreign reserves fell to about $8.2 billion (for a total decline of 60% over 14 months). Both the market and the Russian government understood that the CBR’s foreign currency reserves would not last much longer. These events ultimately culminated in both a default and a currency revaluation on 17 August 1998. In combination, these facts demonstrate that the unfolding events in Russia were public (though the information was not always released in a timely way) and that whole sectors of the market were aware of Russia’s weakening financial position. The likelihood of default and/or currency revaluation was rising materially.

Turning to the next question, why did the Russian bond default have such a large impact on markets (on credit spreads, capital flows, flight to safety, etc.), and was this impact exceptional or unprecedented? This is a more difficult question to answer. Perhaps the market was fatigued by the Asian crisis and the Russian bond default was enough to push market sentiment from optimistic to pessimistic. Perhaps the fact that Russia combined a currency devaluation with a bond default was enough to make market participants view economic conditions as worsening as opposed to improving. Perhaps the Russian bonds were widely held, forcing their owners to scramble for cover.

Whatever the reason, the Russian bond default did act as a powerful catalyst for a flight to safety. But the changes in spreads, the flight to safety, and the spike in volatility had all happened before. The impact was well within the bounds of history. For instance, the yield on 10-year Treasuries decreased from 5.4% on 17 August 1998 to 4.16% by 5 October 1998—a loss of 124 basis points (BPS) in two months. While such a large, rapid repricing of bond yields is not an everyday occurrence, the phenomenon has happened many times before. Looking at 10-year on-the-run US Treasury bond yields from 1926 to 1998, there were 13 such instances in 72 years of data (as measured by a 100 BPS or greater move over two months). Statistically, this means there is an 18% chance of such an event occurring in any given year. However, all of the occurrences happened after 1971, when the United States departed from the gold standard. This is no coincidence. The departure from the gold standard meant that countries were allowed to build up large trade surpluses and their central banks then plowed that money into US Treasury securities (to prevent appreciation of their currencies relative to the US dollar). The difference in yield volatility before and after 1971 is dramatic. Before 1971, the standard deviation of 10-year bond yields was 18 BPS monthly. In contrast, after 1971, the standard deviation was roughly 57 BPS (or 3.1 times greater). So as of 1998, there had been 13 sudden large moves in bond yields in 27 years. Statistically, at the time of the LTCM collapse, a large shift in the yield curve (100 BPS or more in two months) had a 48% chance of occurring in any given year.

The astute observer might point to the widening of arbitrage spreads between off-the-run and on-the-run US Treasury bonds (see graph below). (On-the-run bonds are the most recently issued treasury bonds for a given maturity; for instance 1, 3, 5, 10 or 30 years. As these bonds age, new bonds are issued and take the place of the old bonds as “on-the-run.” The old bonds, in turn, become “off-the-run.”) LTCM bet on the spread between on-the-run and off-the-run UST’s. Historically, on-the-run bonds experience greater liquidity, hence there is always a slight yield spread between on-the-run and off-the-run bonds – even though differences in maturity can be modest. What the spread reveals is that investors of all types value liquidity. This phenomenon is of course magnified during periods of market stress. So, when the Russian Bond Default hit, there wasn’t just a flight to safety, there was a flight to liquidity. Consequently, the spread between off-the-run and on-the-run UST’s widened sharply. Thus, LTCM’s arbitrage positions in UST’s worked against them. In all likelihood, the market’s flight to on-the-run bonds occurred as a simple matter of convenience and liquidity for foreign investors that were desperately searching for safety. This effect is unusual but not at all unexpected in times of crisis. Also, spreads on emerging market bonds spiked sharply in the aftermath of the Russian bond default, but they didn’t even reach the same peak as in 1994! As is clearly illustrated in the graph below, the spike in emerging market bond spreads after the Russian bond default was far from an exceptional event.

Source: Timothy Lane and Steven Phillips. “Moral Hazard: Does IMF Financing Encourage Imprudence by Borrowers and Lenders?” Economic Issues, No. 28 (2002), Figure 2. Copyright 2002 by International Monetary Fund. Reprinted with permission.

So the Russian bond default and the dramatic widening of spreads (the flight to safety) were neither random nor completely unexpected by the standards of market history.

In August 1998, the firm lost more than $1.7 billion—a more than 8.3 standard deviation event, implying that such a loss would occur once every 6.4 trillion years (Dowd, Cotter, Humphrey, and Woods 2008).

This conclusion suggests that we should examine LTCM’s particular approach to investing. LTCM was known for engaging in two types of trades: convergence trades and correlation trades. However, the firm increasingly entered into directional trades, betting on a particular outcome, as demonstrated by the huge losses on swaps and equity options (detailed below). LTCM used a combination of Value at Risk (VaR), stress testing, and scenario analysis to manage risk. And they employed significant resources to do so, as is customary in the hedge fund and investment banking worlds. Each of these approaches has limitations.

In layman’s terms, VaR is a statistical technique to estimate the amount of losses a portfolio might incur over a given day or week (or longer) based on historical price movements. There are many nuanced ways to implement VaR, but they all essentially involve computations of historical volatility, estimates of future volatility, a time horizon, a chosen probability distribution (e.g., normal distribution, Student’s t, chi-square), correlations, and a confidence interval (95%, 99%, etc.). The particular approach to VaR that LTCM used was based on daily standard deviation of security prices and correlations among all securities owned across the entire company over the preceding 500 trading days (estimated) assessed at the 95% confidence interval. In fact, the firm computed a daily VaR figure for the LTCM fund of $45 million, which it believed to be consistent with the US stock market VaR (had LTCM invested all client capital in the plain vanilla S&P 500 Index, for instance). Even though there had been periods historically when crises erupted and spreads widened sharply (such as 1987 and 1994), these periods were excluded from LTCM’s VaR calculations, thus demonstrating the first flaw of VaR: LTCM’s VaR model was applied “out of sample.” The chosen time period for historical data inputs reflects a broad set of conditions that may or may not hold in the future. Moreover, whatever risk events the model does capture (in sample), the model is neutral on timing. The model would view a crisis as just as likely to recur tomorrow as four years from now. Although some financial events are quite random and unpredictable (e.g., Marlboro Friday see Silk and Isaacson 1995) and, as such, do fit well with the VaR model, many risk events are contingent upon the presence of certain conditions and the development of certain factors and so are inconsistent with the VaR model. For instance, if oil prices had surged in 1998 instead of falling sharply, perhaps the Russian government would have escalated tax revenues quickly enough to prevent a default. This example demonstrates that Russia’s default in 1998 was, at least in part, contingent upon the decline in oil prices. If oil prices had not declined, maybe Russia would not have defaulted and the LTCM crisis would never have occurred.

The second and more nefarious flaw of VaR is that it often creates a false sense of confidence that the investor has a handle on risk. Because VaR is focused on volatility of security prices, it is totally independent of fundamental sources of risk. In other words, VaR is neutral on whether an issuer’s debt levels are high or low, spreads are tightening or widening, currencies are pegged or free floating, oil prices are rising or declining, etc. In essence, VaR implicitly trusts that market concerns about risk are embedded in security prices. Clearly, this is not always true, as many crises catch the market by surprise. Yet in the case of Russia, we know beyond any doubt that high debt levels were riskier than low debt levels, widening spreads made it more costly for the Russian government to borrow, a currency pegged to the US dollar meant less financial flexibility for the CBR, the ruble appreciation sent the country’s trade and exports lower, and declining oil prices reduced Russian government tax revenues. Each of these factors intensified between 1995 and 1998. In short, these fundamental risk factors each unequivocally demonstrated that the risk of default in Russia was rising sharply. And this fact remains true regardless of how well or poorly LTCM understood the circumstances in Russia.

In contrast, VaR embraces the randomness of markets, hence the usage of a probability distribution. But probability distributions deal with what should be, while fundamental risk analysis deals with what is. And investors ignore “what is” at their own peril. By recomputing VaR on a daily basis, VaR proponents trust that the market prices will reflect any fundamental risks. But financial history has demonstrated over and over that the market is often caught off-guard (by both fundamental and nonfundamental events), so market prices are often not reflective of the risks (a problem that can be compounded by a poorly chosen sample period). To be sure, the markets also can and do create random price movements based on nonfundamental factors (e.g., rumors, trading momentum, and technical events). But just as clearly and more typically, the market responds to real, fundamental events (such as central bank monetary actions, inflation data releases, GDP growth, recessions, earnings reports, management changes). Fundamental risk events are the result of a sequence of unfolding events at each issuer. Therefore, they are not random, and this fact challenges a fundamental assumption inherent in using VaR.

So, why do firms use VaR? In essence, VaR is a good proxy for what might happen on a day-to-day basis, perhaps 98% of the time. A firm using VaR might happen to choose a particularly good sample period that spans many risk events and cycles. This possibility enables firms to implement a useful technique for most occasions, often lulling VaR users into a false sense of security. But by definition, VaR does not capture risk events that fall outside the sample or the chosen distribution. In other words, it’s good for “normal volatility” but not good for risk. Another reason firms use VaR is that it helps them standardize their approach to risk management across different products, different asset classes, and different portfolio managers. A firm can produce a single VaR figure for the entire firm across multiple divisions, currencies, and regions. This aspect makes VaR a convenient tool for consolidated reporting and capital allocation purposes. But it is not an effective tool for capturing material risk events.

Lastly, the reliance on a select period of historical data or specially chosen volatility inputs necessarily means de-emphasizing actual risk events that are ongoing and developing at various issuers. In other words, despite its positives, VaR is a theoretical exercise. In contrast, true risk analysis is much more difficult and tedious as it entails achieving an understanding of the financial position of relevant issuers and coping with realistic scenarios. But at least, risk analysis deals with the actual, present-day circumstances of flesh-and-blood people and living institutions.

LTCM had targeted a firmwide VaR of $45 million in daily losses. However, in August 1998, the firm lost more than $1.7 billion—a more than 8.3 standard deviation event, implying that such a loss would occur once every 6.4 trillion years (Dowd, Cotter, Humphrey, and Woods 2008). If such an event were indeed random, then perhaps VaR might be appropriate. Yet as is clear from the study of the Russian bond default that precipitated this crisis, the unfolding events were far from random. The series of events that preceded the default were in fact causal in nature. VaR, by virtue of using a probability distribution, treats the markets as a closed system, like a lottery drawing. A typical lottery drawing that chooses 1 of 60 or so balls in a bin is an example of a random, closed system. There are always a certain number of balls to choose from, and the chosen ball is determined through a randomized process. Therefore, the probability of selecting a single ball is exactly the same each time: namely 1 in 60.

In contrast, events in the world of finance are produced by a dynamic, open system of cause and effect. For instance, Russia’s choice to take on more debt or pay down debt is consequential. If oil prices rise or fall, it is consequential. If Asia slips into recession or not, it is consequential. If the prime minister of Russia declares war on a neighboring country, that action introduces new variables into the system. In short, these events have consequences that can be quite different from history. This is risk, and it cannot be fully captured by VaR. Moreover, in 1998, the currency market was clearly anticipating a crisis in Russia, even if LTCM was not. Russia experienced three separate currency market panics in which Russian control over the ruble–dollar exchange rate peg was challenged by the market—in December 1997, in April 1998, and once again in August 1998, ultimately culminating in the Russian bond default on 17 August 1998. The currency market’s behavior demonstrates that the risk of default was nonrandom, conflicting yet again with the assumptions of VaR.

As Russia’s financial condition deteriorated in 1997 and 1998, default grew more and more likely. These risks were brewing for a full three years—since late 1995—and escalated dramatically over time. LTCM (like many others) failed to prepare its portfolio for a crisis in Russia. On the one hand, LTCM’s VaR calculations suggested a nearly impossible event that would happen once in 6.4 trillion years. On the other hand, we have a highly probable event that was well understood by the currency markets (before the default) and whose occurrence was perhaps 60% or 70% likely within, say, three years (if then-current trends continued). Were the currency markets right, or was LTCM? We now know that answer, but this isn’t to criticize LTCM, per se. The Russian crisis reveals the blind spot inherent in VaR, but it also reveals a broader truth about using models. All models fall short of reality at some point. Blind faith in models will lead to crisis sooner or later.

Clearly, LTCM believed in VaR, as the firm’s entire risk management structure was built around it. Faulty beliefs become dangerous weapons even in the hands of intelligent people (and clearly LTCM employed highly intelligent, highly experienced people). The point is that the prospect of default in Russia, or any country for that matter, is a conditional probability. And these were the conditions during the three years leading up to the Russian crisis: a rising US dollar and hence a rising Russian ruble supported by US policy (meaning the rising US dollar was not an aberrant event), falling oil prices, weakening demand in Asia, declining government revenues, and declining foreign exchange reserves. As these conditions intensified, the probability of Russian default grew. But mathematics and economics are two different fields of expertise. In combination, the evidence suggests that the math-heavy principals at LTCM viewed volatility and risk as synonymous. The fact that risk and volatility are often conflated in the academic community also suggests that the LTCM team were blinded by faulty beliefs about how the world works.

History clearly demonstrates that allowances must always be made for the prospect of financial crises. Even today, many of the principals from LTCM describe the 1998 crisis as an “unpredictable” and “unprecedented” market event that caught the firm off-guard. LTCM trader Victor Haghani said, “It was as if there was someone out there with our exact portfolio, only it was three times as large as ours, and they were liquidating all at once” (Lewis 1999). In other words, he diagnosed the problem purely as a liquidity crisis. But this wasn’t just a rumor-fueled, fear-driven event in the markets (which can and does happen from time to time); it was in fact a solvency crisis in Russia that sent markets fleeing for safety, which in turn created a solvency crisis for LTCM. We know definitively that the problem was more than just liquidity.

When an investor’s funds consist solely of equity, the amount of time the investor has for positions to work out as expected is dependent on the market, the client, and the boss.

Though it likely softens the blow to the egos of those involved, the suggestion that LTCM was taken down by an aberrant market event is implausible. By emphasizing the VaR model output, LTCM was underemphasizing, if not outright ignoring, the empirical analysis of actual risks. Though they are often conflated in the curriculum and tools of modern finance, risk and volatility are dramatically different things.

Perhaps LTCM’s VaR calculations used the correct probability distribution but the firm gauged the size of the expected impact wrong? In other words, maybe this was a thousand-year flood? Nope. A look at historical spreads on emerging market bonds, historical shifts in the US Treasury yield curve, and historic volatility demonstrates otherwise. This was far from a thousand-year flood—and certainly not a 6.4-trillion-year flood! It was simply an event that was not well anticipated by the fixed income and stock markets or built into LTCM’s risk models.

Turning now to the unfolding events at LTCM, its liquidity was indeed impaired. On 17 July 1998, Salomon Brothers (Meriwether’s old firm) announced that it was liquidating its risk arbitrage desk and began unwinding many of its trades. This selling probably helped push spreads out, which in turn likely harmed LTCM’s performance in July.

In the months preceding its collapse, managers at LTCM returned approximately $2.7 billion to its investors, leaving $4.8 billion in investor capital. However, through leverage, LTCM invested far more than $4.8 billion. It had borrowed $155.2 billion and controlled an additional $1.2 trillion in notional value of derivative contracts. Using just its debt and equity capital, LTCM’s leverage ratio was 33 to 1. When the firm’s derivative positions are included, its leverage skyrocketed to 283 to 1. In everyday life, this would be the equivalent of buying a $1 million home and only putting up roughly $4,000 as a down payment. The only difference is that a house is a real asset and tends to be more secure as collateral—in comparison to the complex trades in which LTCM engaged.

Over the course of 1998, LTCM lost $4.8 billion, which breaks down as follows (Yalincak, Li, and Tong 2005):

• $1.6 billion in swaps
• $1.3 billion in equity and volatility
• $430 million in Russia and other emerging markets
• $371 million in directional trades in developed countries
• $286 million in equity pairs
• $215 million in yield curve arbitrage
• $203 million in S&P 500 stocks
• $100 million in junk bonds

In interest rate swaps and equity options combined, LTCM lost $2.9 billion. In early 1998, LTCM noticed that equity options were trading at 20% implied volatility according to the Black–Scholes model, but its own work suggested that volatility would decline to 15% (the average standard deviation of the S&P 500 from 1979 to 1997 was approximately 15%). Because a reduction in volatility would lead to a reduction in option prices, LTCM shorted $1.3 billion of option contracts in early 1998. In addition, the firm shorted $1.9 billion of interest rate swaps. In essence, LTCM bet that the spread between LIBOR and government securities would narrow. As markets become more sanguine, such spreads tend to narrow. But when volatility erupts, the spreads tend to spike. LTCM bet that spreads would narrow—just before they spiked. LTCM lost $1.6 billion on interest rate swaps. It should be noted that both the interest rate swaps and the equity volatility trades were directional, speculative trades, not risk arbitrage. Moreover, because LTCM was buying derivatives on margin (similar to putting a down payment on a house), it was able to leverage its capital to buy much more in derivatives than it could have bought of the underlying assets. At its peak in 1998, LTCM owned $1.29 trillion dollars in notional value of derivatives.

Regarding yield curve arbitrage, LTCM identified a 5 basis point spread between 29-year US Treasury bonds (USTs) and 30-year USTs in May 1998. As these bonds were essentially the same security with the same risks, LTCM was expecting these yields to converge, as typically happens between on-the-run and off-the-run bonds (on-the-run bonds are typically more liquid because investors often target specific maturities of 5, 10, or 30 years). So LTCM went long the (higher-yield) 29-year USTs and went short the (lower-yield) 30-year USTs. As markets panicked amid the events unfolding in Russia, global investors sought safety (and liquidity) by pouring capital into on-the-run 30-year Treasury bonds, where LTCM was short, and not into off-the-run bonds, where LTCM was long. Rather than converging as LTCM had expected, this spread blew out to 15 basis points or so, creating $215 million in losses for the firm.

Regarding the convergence-of-yields strategy, LTCM expected global economies to improve and risk levels to decrease. Consequently, the firm went long non-US sovereign bonds (expecting yields to fall relative to USTs) and short USTs (expecting yields to rise relative to non-US bonds). But these expectations were flawed. Markets were becoming more risk averse. A flight to safety occurred, driving up US bond prices and driving down European bond prices. LTCM was wrong.

In equity pairs, LTCM employed convergence strategies against common and preferred stocks of the same issuer. When the prices of the common and preferred stocks diverged widely, LTCM would short the overpriced stock and go long the underpriced stock. In 1998, LTCM lost $286 million in equity pairs.

LTCM had sold itself to clients as a way to get extraordinary returns by taking on risk “no greater than US equities.” However, in 1997, LTCM’s portfolio returned only 17% while the S&P 500 was up 33.1%. According to LTCM, the firm gave the capital back in December 1997 because of declining opportunities to arbitrage the market. LTCM gave back $2.7 billion in capital but retained virtually all of the debt—meaning that the firm raised its leverage ratio from 18 to 1 to 25 to 1 when it returned capital to shareholders. The VaR on the LTCM portfolio was in fact similar to the VaR on the US equity market, yet LTCM collapsed and the US equity market did not (the S&P 500 was up 28% for the year in 1998). Clearly, VaR has material flaws.

Whatever the limitations of VaR, the problems were compounded by LTCM’s use of leverage. High leverage ratios mean that small mistakes in assets yield large mistakes in equity. Moreover, the time horizon of the lender overrides the time horizon of the investment. Even if LTCM were to have ultimately been proven right about its trades, the presence of leverage dramatically shortened the time horizon within which its trades had to work out. Leverage itself impacts the horizon of a trade because adverse market movements can quickly offset the ability of the firm to repay a loan. Moreover, as lenders get nervous, they can accelerate repayment by calling loans, as in the case of margin calls. Many investors overlook the constraints imposed by the time horizons of the critical parties within their supply chain (bosses, consultants, clients, lenders, brokers, etc.), instead emphasizing the rationale for the trade. And a very important constituency, particularly during periods of stress, is lenders. When an investor’s funds consist solely of equity, the amount of time the investor has for positions to work out as expected is dependent on the market, the client, and the boss. Add in leverage, and the amount of time an investor has for positions to work out as expected is also dependent upon the lender’s time horizon and need for collateral. This point cannot be emphasized enough. LTCM gave too little attention to the shifting time horizons of its lenders, given that the firm had levered up to more than 33 to 1. On Tuesday, 21 September 1998, LTCM’s prime broker, Bear Stearns, issued a margin call of $500 million, which appeared to push LTCM over the brink (Shirreff). In the case of margin lending, collateral is entirely a function of market prices. So if an unlevered investor has, say, 12 months for a challenging trade to work out favorably, the levered investor in the same trade may have only days.

The flip side of LTCM’s leverage problem was the perception that the firm was risk free. Many lenders trusted LTCM so much that they provided repo financing with little or no haircut (a haircut is, in effect, the interest rate that a firm is charged to borrow money)—absurdly believing that LTCM was of comparable financial strength to the US Treasury. These small haircuts enabled the firm to lever up more than if it had not received such special breaks. A related problem is that there was apparently no look-through leverage at banks. Many banks lent LTCM money without knowing what the firm’s complete balance sheet actually looked like. Clearly, the lending institutions hadn’t done their homework and lent irresponsibly.

In the midst of the crisis in September 1998, Meriwether and his team still clung to a belief in their models. Meriwether noted in a letter to clients on 2 September 1998 that the fund had lost $2.5 billion or 52% of its value year to date 1998 with $2.1 billion of that loss occurring in August alone (Shirreff). Nevertheless, the LTCM team was still excited about their trades. As LTCM trader Eric Rosenfeld said, “We dreamed of the day when we’d have opportunities like this” (Lewis 1999). They said this just before generating an additional 80%+ loss in the first three weeks of September. Rather than perceiving a flaw in their approach, the team saw the widening spreads as a growing opportunity. There is nothing more dangerous in investing than belief in the wrong convictions. In the final analysis, the principals of LTCM placed their vast personal wealth in the company, levered it up dramatically, and through their own hubris lost it all. As Warren Buffett said about the LTCM founders, “To make money they didn’t have and didn’t need, they risked what they did have and did need. And that’s foolish” (Housel 2012). Considering that these people were already so accomplished and so wealthy, it appears that they were motivated not by money, but by pride.

This crisis is replete with lessons for each and every investor. First, sooner or later, markets will move against you for either good reasons or bad. Be prepared. Second, deploying a highly leveraged position that does not allow for adverse movement in security prices is a recipe for disaster. For unlevered, diversified investors, mistakes are unfortunate, but they can live to fight another day. For levered investors, even small mistakes can easily become a crisis. And if the investment positions are large enough, the mistake may be a systemic risk, as in the case of LTCM. Third, investors must not only assess the probability of being wrong but also assess the cost of being wrong (i.e., the payoff tree). Don’t ever choose to invest when the realization of a loss, however remote the probability is estimated to be, would put the firm/portfolio out of business. Fourth, liquidity must be anticipated. Not only must investors assess current liquidity conditions; they must also assess liquidity in the context of a realistic worst-case scenario (using past crises as guideposts). Fifth, markets ultimately reflect economic activity, but there can be material departures based on changes in investor perceptions or capital flows or central bank actions or liquidity or a host of other factors. Sixth, past-equals-future bias creeps into much analysis. Let go of intellectual anchors and embrace the facts, even if they contradict your existing beliefs. Seventh, diversification should apply not only to types of securities but also to types of strategy.

CFA Institute members: This is eligible for 1.0 CE credit. Click here to record your credit.