華爾街-夢之隊-的巨敗之因（LTCM）案例

華爾街-夢之隊-的巨敗之因（LTCM）案例分析（中/英版）ZT 投機市場中不存在百戰百勝的法寶，任何分析方法與操作係統都有缺陷與誤區。 美國長期資本管理公司的故事是最新最有說服力的證據。 一、LTCM營造的海市蜃樓 1.四大天王 美國長期資本管理公司（LTCM）總部設在離紐約市不遠的格林威治，是一家主要從事定息債務工具套利活動的對衝基金。該基金創立於1994年，主要活躍於國際債券和外匯市場，利用私人客戶的巨額投資和金融機構的大量貸款，專門從事金融市場炒作。它與量子基金、老虎基金、歐米伽基金一起被稱為國際四大"對衝基金"。 2.夢幻組合 LTCM掌門人是梅裏韋瑟（Meriwehter),被譽為能"點石成金"的華爾街債務套利之父。他聚集了華爾街一批證券交易的精英加盟：1997年諾貝爾經濟學獎得主默頓（Robert Merton)和舒爾茨（Myron Schols),他們因期權定價公式榮獲桂冠；前財政部副部長及聯儲副主席莫裏斯（David Mullis)；前所羅門兄弟債券交易部主管羅森菲爾德（Rosenfeld)。這個精英團隊內薈萃職業巨星、公關明星、學術巨人，真可稱之為"夢幻組合 "。 3.驕人業績 在1994--1997年間，LTCM業績輝煌驕人。成立之初，資產淨值為12.5億美元，到1997年末，上升為48億美元，淨增長2.84倍。每年的投資回報率分別為：1994年28.5%、1995年42.8%、1996年40.8%、1997年17%。 4."致富秘笈" 長期資本管理公司以"不同市場證券間不合理價差生滅自然性"為基礎，製定了"通過電腦精密計算，發現不正常市場價格差，資金杠杆放大，入市圖利"的投資策略。 舒爾茨和默頓將金融市場曆史交易資料，已有的市場理論、學術研究報告和市場信息有機結合在一起，形成了一套較完整的電腦數學自動投資模型。他們利用計算機處理大量曆史數據，通過連續而精密的計算得到兩種不同金融工具間的正常曆史價格差，然後結合市場信息分析它們之間的最新價格差。如果兩者出現偏差，並且該偏差正在放大，電腦立即建立起龐大的債券和衍生工具組合，大舉套利入市投資；經過市場一段時間調節，放大的偏差會自動恢複到正常軌跡上，此時電腦指令平倉離場，獲取偏差的差值。 5.法寶之瑕 但是不能忽視的是，這套電腦數學自動投資模型中也有一些致命之處：（1）模型假設前提和計算結果都是在曆史統計基礎上得出的，但曆史統計永不可能完全涵蓋未來現象。（2）LTCM投資策略是建立在投資組合中兩種證券的價格波動的正相關的基礎上。盡管它所持核心資產德國債券與意大利債券正相關性為大量曆史統計數據所證明，但是曆史數據的統計過程往往會忽略一些小概率事件，亦即上述兩種債券的負相關性。 6.陰溝翻船 LTCM萬萬沒有料到，俄羅斯金融風暴引發了全球的金融動蕩，結果它所沽空的德國債券價格上漲，它所做多的意大利債券等證券價格下跌，它所期望的正相關變為負相關，結果兩頭虧損。它的電腦自動投資係統麵對這種原本忽略不計的小概率事件，錯誤地不斷放大金融衍生產品的運作規模。LTCM利用投資者那兒籌來的22億美元作資本抵押，買入價值3250億美元的證券，杠杆比率高達60倍。由此造成該公司的巨額虧損。 它從5月俄羅斯金融風暴到9月全麵潰敗，短短的150天資產淨值下降90%，出現43億美元巨額虧損，僅餘5億美元，已走到破產邊緣。9月23日，美聯儲出麵組織安排，以美林、摩根為首的15家國際性金融機構注資37.25億美元購買了LTCM的90%股權，共同接管了該公司，從而避免了它倒閉的厄運。 二、LTCM興衰的啟示 1.投機市場中不可能出現神聖，任何人都會犯錯誤。長期資本管理公司，擁有世界上第一流的債券運作高手梅裏韋瑟和羅森菲爾德，擁有世界上第一流的科研天才默頓和舒爾茨，擁有國際上第一流的公關融資人才莫裏斯。但是這個"夢幻組合"中每個人物都應對LTCM的重挫負有責任。因此，我們股民不應迷信任何人，要有獨立思考的能力。 2.投機市場中不存在致勝法寶，任何分析方法與操作係統都有缺陷與誤區。LTCM曾經以為自己掌握了致富秘笈，在國際金融市場上連連得手，自信滿滿。可是偏偏出現了他們所忽視的小概率事件，使其造成巨額虧損已近破產。因此，我們股民運用任何方法或工具在證券市場上進行運作時，必須認識到它們有時會出錯，會使你錯失一些機會。如果它們出錯隻是小概率事件，正確是大概率事件（例如大於60%），而且比較適合你的個性，你就堅持使用它們，但也要注意下麵的提示。 3.在投機市場上生存與發展，控製風險是永恒的主題。正因為在證券市場上任何人任何方法都可能出錯，所以控製風險是我們股民應終生牢記在心的鐵律。如果你所依賴的方法或工具，在出錯時僅使你錯失一些賺錢的機會，這並不要緊。如果它們出錯時，有可能會令你傷筋動骨、全軍覆沒甚至負債累累，這樣的風險就必須嚴格控製了。為了避免這些悲慘的結局在我們股民身上發生，首先我們不要透支炒股，其次我們不要借錢炒股，再次我們在高位炒股時要注意止蝕，最後我們不要盲目頻繁炒作。 1994: Long-Term Capital Management is founded by John Meriwether and accepts investments from 80 investors who put up a minimum of $10 million each. The initial equity capitalisation of the firm is $1.3 billion. (The Washington Post, 27 September 1998) End of 1997: After two years of returns running close to 40%, the fund has some $7 billion under management and is achieving only a 27% return — comparable with the return on US equities that year. Meriwether returns about $2.7 billion of the fund's capital back to investors because "investment opportunities were not large and attractive enough" (The Washington Post, 27 September 1998). Early 1998: The portfolio under LTCM's control amounts to well over $100 billion, while net asset value stands at some $4 billion; its swaps position is valued at some $1.25 trillion notional, equal to 5% of the entire global market. It had become a major supplier of index volatility to investment banks, was active in mortgage-backed securities and was dabbling in emerging markets such as Russia (Risk, October 1998) 17 August 1998: Russia devalues the rouble and declares a moratorium on 281 billion roubles ($13.5 billion) of its Treasury debt. The result is a massive "flight to quality", with investors flooding out of any remotely risky market and into the most secure instruments within the already "risk-free" government bond market. Ultimately, this results in a liquidity crisis of enormous proportions, dealing a severe blow to LTCM's portfolio. 1 September 1998: LTCM's equity has dropped to $2.3 billion. John Meriwether circulates a letter which discloses the massive loss and offers the chance to invest in the fund "on special terms". Existing investors are told that they will not be allowed to withdraw more than 12% of their investment, and not until December. 22 September 1998: LTCM's equity has dropped to $600 million. The portfolio has not shrunk significantly, and so its leverage is even higher. Banks begin to doubt the fund's ability to meet its margin calls but cannot move to liquidate for fear that it will precipitate a crisis that will cause huge losses among the fund's counterparties and potentially lead to a systemic crisis. 23 September 98: Goldman Sachs, AIG and Warren Buffett offer to buy out LTCM's partners for $250 million, to inject $4 billion into the ailing fund and run it as part of Goldman's proprietary trading operation. The offer is not accepted. That afternoon, the Federal Reserve Bank of New York, acting to prevent a potential systemic meltdown, organises a rescue package under which a consortium of leading investment and commercial banks, including LTCM's major creditors, inject $3.5-billion into the fund and take over its management, in exchange for 90% of LTCM's equity. Fourth quarter 1998: The damage from LTCM's near-demise was widespread. Many banks take a substantial write-off as a result of losses on their investments. UBS takes a third-quarter charge of $700 million, Dresdner Bank AG a $145 million charge, and Credit Suisse $55 million. Additionally, UBS chairman Mathis Cabiallavetta and three top executives resign in the wake of the bank's losses (The Wall Street Journal Europe, 5 October 1998). Merrill Lynch's global head of risk and credit management likewise leaves the firm. April 1999: President Clinton publishes a study of the LTCM crisis and its implications for systemic risk in financial markets, entitled the President's Working Group on Financial Markets (Governance and Risk Control-Regulatory guidelines-president's working group) Lessons to be learned: Market values matter LTCM was perhaps the biggest disaster of its kind, but it was not the first. It had been preceded by a number of other cases of highly-leveraged quantitative firms that went under in similar circumstances. One of the earliest was Franklin Savings and Loan, a hedge fund dressed down as a savings & loan. Franklin's management had figured out that many of the riskier pieces of mortgage derivatives were undervalued because a) the market could not understand the risk on the risky pieces; and b) the market overvalued those pieces with well-behaved accounting results. Franklin decided it was willing to suffer volatile accounting results in exchange for good economics. More recently, the Granite funds, which specialised in mortgage-backed securities trading, suffered as the result of similar trading strategies. The funds took advantage of the fact that "toxic waste" (risky tranches) from the mortgage derivatives market were good economic value. However, when the Fed raised interest rates in February 1994, Wall Street firms rushed to liquidate mortgage-backed securities, often at huge discounts. Both of these firms claimed to have been hedged, but both went under when they were "margin-called". In Franklin's case, the caller was the Office of Thrift Supervision; in the Granite Fund's, the margin lenders. What is the common theme among Franklin, the Granite Funds and LTCM? All three depended on exploiting deviations in market value from fair value. And all three depended on "patient capital" -- shareholders and lenders who believed that what mattered was fair value and not market value. That is, these fund managers convinced their stakeholders that because the fair values were hedged, it didn't matter what happened to market values in the short run — they would converge to fair value over time. That was the reason for the "Long Term" part of LTCM's name. The problem with this logic is that capital is only as patient as its least patient provider. The fact is that lenders generally lose their patience precisely when the funds need them to keep it — in times of market crisis. As all three cases demonstrate, the lenders are the first to get nervous when an external shock hits. At that point, they begin to ask the fund manager for market valuations, not models-based fair valuations. This starts the fund along the downward spiral: illiquid securities are marked-to-market; margin calls are made; the illiquid securities must be sold; more margin calls are made, and so on. In general, shareholders may provide patient capital; but debt-holders do not. The lesson learned from these case studies spoils some of the supposed "free lunch" features of taking liquidity risk. These plays can indeed generate excellent risk-adjusted returns, but only if held for a long time. Unfortunately the only real source of capital that is patient enough to take fluctuations in market values, especially through crises, is equity capital. In other words, you can take liquidity bets, but you cannot leverage them much. Liquidity risk is itself a factor As pointed out in the analysis section of this article, LTCM fell victim to a flight to liquidity. This phenomenon is common enough in capital markets crises that it should be built into risk models, either by introducing a new risk factor — liquidity — or by including a flight to liquidity in the stress testing (see the following section for more detail on this). This could be accomplished crudely by classifying securities as either liquid or illiquid. Liquid securities are assigned a positive exposure to the liquidity factor; illiquid securities are assigned a negative exposure to the liquidity factor. The size of the factor movement (measured in terms of the movement of the spread between liquid and illiquid securities) can be estimated either statistically or heuristically (perhaps using the LTCM crisis as a "worst case" scenario). Using this approach, LTCM might have classified most of its long positions as illiquid and most of its short positions as liquid, thus having a notional exposure to the liquidity factor equal to twice its total balance sheet. A more refined model would account for a spectrum of possible liquidity across securities; at a minimum, however, the general concept of exposure to a liquidity risk factor should be incorporated in to any leveraged portfolio. Models must be stress-tested and combined with judgement www.6park.com Another key lesson to be learnt from the LTCM debacle is that even (or especially) the most sophisticated financial models are subject to model risk and parameter risk, and should therefore be stress-tested and tempered with judgement. While we are clearly privileged in exercising 20/20 hindsight, we can nonetheless think through the way in which judgement and stress-testing could have been used to mitigate, if not avoid, this disaster. According to the complex mathematical models used by LTCM, the positions were low risk. Judgement tells us that the key assumption that the models depended on was the high correlation between the long and short positions. Certainly, recent history suggested that correlations between corporate bonds of different credit quality would move together (a correlation of between 90-95% over a 2-year horizon). During LTCM's crisis, however, this correlation dropped to 80%. Stress-testing against this lower correlation might have led LTCM to assume less leverage in taking this bet. However, if LTCM had thought to stress test this correlation, given that it was such an important assumption, it would not even have had to make up a stress scenario. This correlation had dropped to 75% as recently as 1992 (Jorion, 1999). Simply including this stress scenario in the risk management of the fund might have led LTCM to assume less leverage in taking this bet. Financial institutions must aggregate exposures to common risk factors One of the other lessons to be learned by other financial institutions is that it is important to aggregate risk exposures across businesses. Many of the large dealer banks exposed to a Russian crisis across many different businesses only became aware of the commonality of these exposures after the LTCM crisis. For example, these banks owned Russian GKOs on their arbitrage desks, made commercial loans to Russian corporates in their lending businesses, and had indirect exposure to a Russian crisis through their prime brokerage lending to LTCM. A systematic risk management process should have discovered these common linkages ex ante and reported or reduced the risk concentration.