How to lose 2.3 billion USD

Transcription

1 How to lose 2.3 billion USD Last year, a European bank surprised the markets by announcing a trading loss of 2.3 billion USD. It turned out that another rogue trader struck again. History is littered with rogue trading incidents. Of course, we only hear about the ones that went south and the ones with large losses that cannot go unreported. How many rogue trading positions actually made money, or lost small amounts of money? It is remarkable that this incident followed a couple of years after a similar record-breaking rogue trading loss at another European bank. Closer inspection reveals that these losses happened at a comparable delta-one trading desk. As astounding and unbelievable as it might seem to hide such big positions for a significant period, it still happened. This leaves two questions to be answered. Which discrepancies in the bank s operational and risk controls did the trader exploit, and how can financial institutions make sure that similar incidents do not happen again? I will start to describe the typical products that trade at a delta one desk, as well as their operational, and risk characteristics. Next, I will use this knowledge to construct some portfolios for which this type of fraudulent trading can occur. Lastly, I will give my view on how to prevent these types of incidents in the future. Delta-one Trading Delta-one trading desks trade stocks and (linear) derivatives. The name delta one stems from the fact that most of their products have linear pay out profiles. This is a fancy term for saying that most products traded at these desks behave like stocks. If a stock price fluctuates one dollar, a typical delta one derivative also moves by one dollar, or some constant factor times one dollar. This is contrary to options, which have price sensitivities that change when the underlying stock price moves. Typical products that trade at a delta-one desk are: Stocks and (customized) stock baskets If an investor seeks exposure on a stock without buying it, that investor can buy a derivative from the desk that replicates the performance of the stock. An investor can also buy exposure (long or short) to a customized basket of stocks. Either way, the desk hedges this directional risk by buying or selling stocks. A trader cannot hide these transactions if he trades shares that are listed on a stock exchange. Trades typically settle in two or three days. The stocks clear on an exchange and the stocks are kept at a custodian firm (broker dealer). Stocks are paid for in full, or at least the bank has to post margin with the custodian in case the bank borrows money from that custodian (broker/dealer) to finance the stock purchase G.O. Capital Markets Consulting Services Page 1 of 11

2 Futures Delta one desks trade futures to hedge their risk. If a client buys short exposure on the S&P 500 index, the desk sells futures to hedge this position. Desks also routinely engage in proprietary trading where they exploit the spread between futures and the underlying basket or ETF s because the theoretical future price is off by a little amount. Future trades are also hard to hide. Futures clear on exchanges and initial margin is posted the next day. Maintenance margin settles on a daily basis. ETF s ETF s are investment funds that trade on exchanges, and therefore could be classified as stocks. The issuing party typically holds assets and passes the return of these assets minus a fee on to the ETF holders. Delta one desks trade index ETF s as an alternative to futures or basket trades. The SPY ETF, which tracks the S&P500 index is one of the most liquid securities that trade in the U.S. Important to note in this discussion is that ETF issuers in the U.S. at least have to hold 80% of the securities in the funds name. This differs from Europe, where ETF issuers can utilize derivatives to replicate the performance of the reference asset. Total return- or equity swaps These swaps do not trade on exchanges. They trade in the Over-The-Counter (OTC) market. OTC trades are bilateral agreements between two parties. A total return swap pays one party the total return of an asset. Total return for stocks include price appreciation or price depreciation and dividends. The party receiving the total return of the reference asset pays or receives interest. Interest is calculated over the nominal amount of the swap. Interest rates are determined by a reference rate (often LIBOR) plus or minus a spread. The spread is set in such a way that the fair value of the swap equals zero at the time of the initial trade. Therefore, no initial settlement is due, and large trades can be done without a large funding requirement at the beginning of the trade. Settlements typically occur after one or three month periods. Consider a one-year total return swap with three-month resets. After the first three months, the counterparties determine the asset return and the interest payments, which then settle. Interest links to three-month LIBOR. The new rate for the next three months equals LIBOR on that day plus the earlier determined spread. The asset side resets by changing the number of shares, which is equal to the total nominal amount of the trade divided by the share price of that day, or some average share price over some period. Depending on the contract specifications, counterparties pay each other margin before the reset dates if the contract value exceeds predetermined values as a result of market moves. Swaps exist that pay asset returns in other currencies than the reference asset s denomination. An example would be a total return swap, which pays the performance of the S&P500 in Swiss Franc. In this case, interest is also paid in CHF. Investors buy these swaps when they seek exposure in foreign equity markets without facing direct exposure to foreign currencies. These swaps are used by institutions who issue ETF s on foreign markets, which are priced in their home currency. An example of such an ETF is the SPDR EURO STOXX 50 ETF, which tracks the EURO STOXX 50 index (which is a European, EUR denominated index), but trades in the U.S. in USD. Timeline of Events Before I will give my own interpretation, it is helpful to look at the actual events, and the prevailing opinions with regard to what happened. According to the prosecutor, the trader was first asked about his trades by his manager in August. On September 13, his manager inquired for the second time about the nature of the trades. The trader was then asked if he contacted the credit risk department. The trader answered he had not, and was then asked to identify the counterparties of his trades G.O. Capital Markets Consulting Services Page 2 of 11

3 After that, the trader left the office. From home, he sent his manager details about the trades. He then returned to the bank to identify the trading positions. On September 15, the news broke and the trader was arrested. According to the bank, the trader disguised his trades by using forward settling ETF-cash positions on the DAX, S&P500, and SX5E indices. These positions were scheduled to settle on September 22 nd, and involved a large number of transactions. On September 6 th, the trader posted the following comment on his face-book page: need a miracle. On this day, the Swiss central bank intervened in the currency market. This resulted in a 10% depreciation of the Swiss franc. After these events, a lot has been written by news agencies and on forums. I will list those opinions that are relevant. Forums repeatedly report that the trader exploited his previous role as an operations specialist to gain access to the bank s settlement and confirmations systems. His access to these systems had not been revoked after he started to work in the front office. Another theme that comes up repeatedly is the story that it was possible to enter OTC trades in the bank s systems such that no confirmations were sent out or matched, until after settlement. Next, I will list the conditions for which it might be possible to fool the bank s systems, agree with the known events, and hide such a big loss. It is plausible that these trades have a CHF currency component or that the P&L of the trading desk needed to be rebalanced locally to CHF. In August there were large currency and stock market moves, which could have triggered the first round of inquiries. Furthermore, it appears likely that the Swiss bank intervention on September 6 th triggered the second round of questioning, which eventually led to the discovery of the fake hedges. The fake hedges needed to be constructed and booked in such a way that margin calls and counterparty risk could be averted or hidden. The real money loosing trades were listed long positions on the indices mentioned earlier. The purchase of these positions yielded large funding requirements to finance the ETF s and initial margin requirements for futures. Although these funding requirements were massive, they were certainly not unusual for the delta one trading business. The real position needs to have a time horizon of approximately 3 months. The size of the long positions has to reflect a 2.3 billion USD loss, given the stock market move in that time period. An Example of a Possible Trading Position I will use these assumptions to construct a trading position where a real long position is combined with a fake short position. The total position consisted of a large number of trades. Therefore, the following approach is an oversimplification of the real position. I do not claim that this is exactly what happened, although I am confident that the nature of the actual trades will prove to be similar. Let us assume that the position was 33% S&P500, 33% DAX, and 33% SX5E. This yields for the loss of the S&P position: 33% times 2.3 billion, equals 766 million USD. For arguments sake, let the reference date be June 28 th. On that day, the SPY (S&P500 ETF, it s price is the level of the S&P 500, divided by 10), closed at a price of USD. The closing level reference on September 12 th equals USD. If the long S&P500 position consisted entirely of SPY ETF s, the number of ETF s would be: million shares. The average daily volume for the year G.O. Capital Markets Consulting Services Page 3 of 11

4 approximately equals 221 million shares. This is a large position, approximately 8% of total shares outstanding. Hence, the trader probably used a mix of SPY ETF s, other funds, and futures. Prime candidate for the fake hedges could be a range of total return swaps on the S&P500 index. In the following reasoning, the base currency of the desk is USD. Similar P&L evolution appears when the base currency is for example CHF. The trader put these swaps into the system at various dates (coincidently with the purchase of the SPY ETF s, and for the purpose of rebalancing, see below). The total return swaps are quanto CHF, which means that the S&P500 returns are paid out in CHF, not in USD. Therefore the number of swaps is not 1, for each long ETF, but rather 0.85, in case the FX price equals 0.85 CHF for each USD. In this way, measured in USD, the position is completely hedged. To see that this is true, consider a downward move of 1 USD in the SPY. The hedge provides a gain of 0.85 CHF against this move, which equals 1 USD if the Swiss francs are sold against USD. The actual exposure resulting from this position is driven by correlation risk. The book s P&L is very sensitive to stock markets moves that are in- or out of sync with the Swiss Franc exchange rate. A detailed explanation of these risk dynamics is given in the appendix at the end of this paper. As a rule of thumb, the book makes money if the equity markets move in tandem with the CHF exchange rate and the book loses money vice versa. Now the trader can do three things: 1. Leave this position as is, and re-hedge nothing. The position is completely hedged this way if the CHF exchange rate does not move. Moves in the CHF exchange rate affect the position in two ways. 1) The value of the accumulated CHF currency position as a result of market moves is no longer equal to the value of the USD profit or loss. 2) The number of swaps does not cover the long position. The bank starts to have a net long or short market position in the S&P500 index. In case CHF moves down to 0.90 CHF for each dollar, the trader needs to have a short position of 0.90 swaps against 1 ETF. The current number of short swaps equals Therefore, the desk is net long the market. In case CHF moves up to 0.80 CHF for each USD, the desk is net short the market, since the number of swaps (which reflect a short position) is larger (0.85), than is should be (0.80). 2. Adjust the number of swaps for currency moves. The trader adjusts the number of swaps such that the desk does not have an open market position as a result of currency fluctuations. The risk here is that large market moves are accompanied by large currency moves, so that the trader is not able to re-adjust his hedge on time. 3. Trade USD against CHF to convert the swap P&L from CHF to USD. In this way, the desk hedged its currency risk over the accumulated P&L in CHF as a result of the position in the CHF denominated swap. The trader probably booked a combination of the hedging techniques described above. As long as the trading desk remains within their foreign exchange exposure limits, the desk is allowed to have exposures in foreign currencies. The big question here is: why bother booking fake hedges denominated in CHF, instead of just USD hedges, where all this rebalancing because of currency moves is not necessary? I can only think of one reason. In order to prevent 3 rd parties to knock on the bank s door (and vice versa) regarding settlements, confirmations, and margin, the trader needed to book plausible trades with an internal counterparty. It was justifiable to book a TRS to hedge a CHF denominated S&P500 ETF issued by the bank for Swiss clients. The trader knew the operations systems well enough to realize that this trade could be hidden for a significant period of time. External counterparty risk is no factor in this case, and as long as the trade does not settle, margin issues can be manipulated. It can take a long time to sort out the internal confirmations, and these might not be on high alert, since it only involved an internal counter party G.O. Capital Markets Consulting Services Page 4 of 11

5 Figure 1 displays the reported P&L (in USD) for a long SPY and short CHF TRS position as a function of time for the scenario in which the trader rebalances his portfolio on a daily basis. Figure 1 Reported P&L of real long SPY position against fake short CHF denominated TRS position, daily rebalancing Not rebalancing the portfolio at all is not an option for the trader, since this leads to P&L swings that are too big. All goes well, until August 9 th. On that day, the book reports a big loss. Large upward moves in the stock market were accompanied by very large downward moves in the CHF exchange rate. In fact on August 9 th, the Swiss franc trades at its all-time high (0.7209) against the USD. This large gain in CHF is accompanied with very large gains in the equity markets. In reality, the desk made money. The position, including the fake hedges, lost a lot of money because of the correlation exposure described earlier. It certainly would explain the first round of questioning in August that is mentioned by the prosecutor. The worries die out quickly, since the market moves are such that the losses quickly recover. (Aug 11 th : Equities up, CHF Exchange rate up) After the losses disappear, of course it is business as usual. Certainly, when the trader bluffs his way out towards his former peers in operations, who are responsible for the P&L reporting. But the first critical seeds have been planted. This scenario repeats itself when the Swiss national bank intervenes in the Swiss franc on September 6th. Large equity moves go hand in hand with the 10% overnight depreciation in the Swiss Franc (FX rate up). Depending on the exact hedge, large P&L profits are reported. In case of the totally hedged position, the P&L reaches levels of around +100 million USD. More importantly, the P&L levels remain high. At this time, the large profit draws attention. The trader has to justify this large P&L. The USD value of the total return swap position changes by a lot during this day. Therefore, the credit department starts asking question regarding the margin and validity of the counterparties with whom these swaps are traded. Everything starts to unravel at blistering speeds. The bank quickly discovers that the trades in the TRS are bogus. This account exactly matches the narrative of the prosecutor. Figure 2 shows the P&L evolution over time in case of a similar real long SX5E futures position (in EUR) offset by a fake short TRS position (in CHF). Rebalancing is again done on a daily basis G.O. Capital Markets Consulting Services Page 5 of 11

6 Figure 2 Reported P&L of real long SX5E position against fake short CHF denominated TRS position, daily rebalancing We see a similar P&L evolution in case of SX5E and EUR/CHF exposure. It is interesting to see that if the trader actually hedged his fake CHF exposure with real currency transactions, he would have made money on the isolated FX trades. For the total loss to be 2.3 billion USD, the long positions in the market would have to be a bit larger than previously mentioned. Preventing Similar Events in the Future The knee jerk reaction of banks, who experienced rogue trading losses is; to prevent operations (back- and middle office) staff from moving into front-office roles. This is exactly what banks should not be doing. To avoid these type of trading losses a bank must employ highly skilled individuals for their trading operations. Highly qualified operations staff will have the ability to detect anomalies with respect to cash flows, P&L statements, exposures, confirmations, and margin calls. They also can see through dubious explanations from traders who try to cover something up by creating a smoke screen of complicated trading structures they put on the books. The best way to recruit high level employees for bank s operations departments, and maintain a relative low turnover rate, is to offer growth potential into commercial front office roles. Financial institutions will be in a better position to counter rogue trading attempts if they put proper controls in place that prevent former operations staff to access back office systems, and influence current back office staff. Again and again, operations are seen as a cost center, where processing must be done against as little costs as possible. The real costs are presented when this type of scenarios happen. 2.3 billion USD buys a lot of education for a bank s operations staff, Below, I will list some specific operational practices that, when implemented properly, should significantly lower the chance of a rogue trading loss. When OTC trades are booked into the risk management systems, the back office should reconcile their confirmations with the confirmations of the external counter party on that same day or after one business day at the latest. If there is no external trading confirmation after that time, trading has to provide contact details for the counter party so that the back office can immediately follow up with the external financial institution. It is important that back office staff is persistent in their effort to contact the counter party. Failure to do so in any fashion should trigger a red flag. Once the external confirmation is present, all the terms should be checked against the terms of the internal confirmation. It would not be the first time (and this is an understatement) that 2011 G.O. Capital Markets Consulting Services Page 6 of 11

7 deals blow up because of differences in terms between confirmations that come to light in stressed market conditions. Staff (often legal), who perform these checks, should have sufficient technical knowledge of the trades in question. Again, I rest my case about hiring skilled operations staff. Market and foreign exchange exposures should be reconciled with P&L statements and cash balances. Too often, large P&L and exposure discrepancies are not properly and timely addressed. Future settlements (margin and swap resets) need to be tracked and cross referenced against the actual cash flows. It proved to be possible to measure market risk for trades that were booked, but did not exist. The risk system should be designed in such way, that whenever a trader enters an OTC trade, the departments of legal (confirmations), operations (cash flow), and market risk management (market exposure), need to check off this trade within the risk management system. If the trade in question does not have the required checks after one business day, the system should automatically list the trade on a report so that appropriate action can be taken. Guus Oonincx is the founder and owner of G.O. Capital Markets Consulting Services LLC (GO CMCS). He is an expert in exotic and plain vanilla derivatives, trading, and trading operations. Guus has 10+ years of derivatives trading experience in foreign exchange, fixed income, and equity derivatives markets. He combines this experience with a deep academic understanding of financial markets and derivatives in particular. He built and managed a highly successful equity derivatives trading platform for ING Bank in the Americas. He grew this desk from a one person listed options operation to a global hub that manages and prices all US related equity derivatives exposure for the bank and its clients. Products traded at the desk range from simple listed-options to tailor made hybrid OTC structures. He led the cooperation of risk management, legal and compliance, IT, financial and product control, and sales to establish front to back integration of these departments into the trading operation. Guus is a great translator who is able to communicate difficult abstract concepts to all levels of understanding. He often used this ability to educate the back/mid office and other operational groups (market risk management, legal, compliance, IT, financial control) about the trading business and the role these units play with respect to the entire operation. This resulted in a better understanding of the inner workings of the trading desk, and subsequently led to better and more effective support to the trading desk. Guus is a registered series 7 and 63 as well as a registered options principal and general supervisor (series 4 and 24). He holds two masters degrees; one in Mathematical Finance (Columbia University, New York) and one in Theoretical Astrophysics (Utrecht University, the Netherlands). Guus teaching experience started at college, where he was a teaching assistant in subjects ranging from quantum mechanics to computer programming. Apart from running his own business, he is also an instructor in Financial Mathematics/Statistics and Financial Analysis in VBA at New York University. GO CMCS offers a half day work shop/presentation about this subject. For more information, GO CMCS at info@gocapitalmarkets.com, or call us at G.O. Capital Markets Consulting Services Page 7 of 11

8 Appendix: Correlation Dynamics of the Position It is instructive to look at an example to explore the dynamics of a simplified similar trading position. This allows us to gain an understanding of the actual exposure of the trading desk The real market exposure of a trading position, where the desk is long S&P500 and short a number of total return swaps denominated in Swiss francs, is correlation. To be more precise; the desk is exposed to the correlation between the level of the S&P500 and the USD/CHF exchange rate. In this simplified example, the desk purchases 1 listed share at a price of 100 USD. In order to hedge, the desk shorts 0.85 total return swaps on that share. USD/CHF exchange rate is 0.85 at that time. Scenario 1 t=0 (first day): Stock price: 100 USD, USD/CHF rate: t=1 (second day): Stock price: 50 USD, USD/CHF rate: t=2 (third day): Stock price: 100 USD, USD/CHF rate: Table 1 Overview of trading activity of stock, TRS and CHF book scenario 1 t=0 t=1 t=2 USD CHF FX rate Stock price Stock position Stock Value Daily Stock P&L (USD) TRS price (CHF) TRS position TRS position Value (CHF) FX Hedge trade CHF FX Hedge USD side Cash CHF position before FX hedge CHF exposure TRS CHF cash position after FX hedge Total CHF Exposure after FX hedge USD Cash resulting from FX hedge Total P&L Table 1 gives an overview of all cash flows and P&L as a result of rebalancing trades for the stock and TRS positions. Use this table in the following detailed analysis of this hedging activity. At t=0, the desk is completely hedged. With USD/CHF at 0.85, the short 0.85 CHF denominated TRS completely offsets the long stock position. At t=1 the portfolio needs to be rebalanced in order to create a hedged position again. This rebalancing consists of two trades: 1. Buy CHF. Buying 0.20 TRS actually generates a positive cash flow of 10 CHF G.O. Capital Markets Consulting Services Page 8 of 11

9 The desk needs to limit their short position in the TRS because the desk only needs to short 0.65 TRS against the long share. For every 1 USD move in the stock, the TRS moves 0.65 CHF, which is equivalent to 1 USD. 2. Transfer the projected cash flow of the TRS in CHF to USD. With the stock down 50 USD, the desk is poised to receive 0.65*50 CHF = CHF. Because we assume this desk to be a USD trading desk, we need to transfer this projected CHF cash flow to USD, against the prevailing USD/CHF rate of The desk needs to sell CHF. However, the desk also received 10 CHF for the TRS rebalancing in the portfolio. The total currency hedge trade is to sell CHF at 0.65 (buy 42.5/0.65 = USD). At this point in time the position is completely hedged, as long as the currency rate in USD/CHF does not move. At t=2 (third day), the market completely recovers, the stock moves back up to 100 USD and the USD/CHF exchange rate moves back to 0.85, from Again, the desk needs to rebalance. 1. Sell CHF. This trade brings the total TRS position back to -0.85, which is the correct hedge if USD/CHF trades at This trade does not generate any cash flow, since the asset return in the total return swap is back at the original level. The asset side of the TRS therefore has zero value. 2. The projected cash flow of the TRS is again zero CHF. At t=1, the desk sold CHF against USD. 10 CHF of the CHF was generated by a realized profit when the desk bought 0.20 TRS at -50 CHF. This realized 10 CHF was already transferred to USD. The desk therefore only has to buy back the CHF, which was sold against the projected positive cash flow of the TRS. Therefore, the desk buys back and sells 32.50/0.85=38.24 USD. What is the result at t=2 of all this rebalancing? The desk is completely hedged. It has zero Swiss Franc in cash. The net P&L in USD is: The USD equivalent of the realized 10 CHF (10/0.65) = USD. In order to calculate the total P&L, we need to add to that number, the profit that was made by selling and buying CHF. Sell CHF/Buy 50 USD, and buy CHF/sell USD, nets = USD. The total USD P&L after 2 days equals: =27.15 USD. Table 1 provides an overview of these trades. Scenario 2 t=0 (first day): Stock price: 100 USD, USD/CHF rate: t=1 (second day): Stock price: 50 USD, USD/CHF rate: t=2 (third day): Stock price: 100 USD, USD/CHF rate: In this scenario, the stock price, similar to scenario 1, moves down to 50 USD before it recovers again to 100 USD. The difference here is that the USD/CHF exchange rate moves to 1.05 instead of This reversal in the exchange rate hugely impacts the P&L, as we will see below. At t=0, the desk is completely hedged. With USD/CHF at 0.85, the short 0.85 CHF denominated TRS completely offsets the long stock position. At t=1 the portfolio needs to be rebalanced in order to create a hedged position again. This rebalancing consists of two trades: 1. Sell CHF. Selling 0.20 TRS generates a negative cash flow of 10 CHF. We need to extend our short position in the TRS because the desk needs to short 1.05 TRS 2011 G.O. Capital Markets Consulting Services Page 9 of 11

10 against 1 long share. For every 1 USD move in the stock, the TRS moves 1.05 CHF, which is equivalent to 1 USD. 2. Transfer the projected cash flow of the TRS in CHF to USD. With the stock down 50 USD, the desk is poised to receive 1.05*50 CHF = CHF. Because we assume this desk to be a USD trading desk, we need to transfer this projected CHF cash flow to USD, against the prevailing USD/CHF rate of The desk needs to sell CHF. However, the desk already paid 10 CHF for the TRS rebalancing in the portfolio. The total currency hedge trade is to sell CHF at 1.05 (buy 42.5/1.05 = USD). At this point in time the position is completely hedged, as long as the currency rate in USD/CHF does not move. At t=2 (third day), the market completely recovers, the stock moves back up to 100 USD and the USD/CHF exchange rate moves back to 0.85, from Again, the desk needs to rebalance. 1. Buy CHF. This trade brings the total TRS position back to -0.85, which is the correct hedge if USD/CHF trades at This trade does not generate any cash flow, since the asset return in the total return swap is back at the original level. The asset side of the TRS therefore has zero (intrinsic) value. 2. The projected cash flow of the TRS is again zero CHF. At t=1, the desk sold CHF against USD. 10 CHF was paid as a result from realized loss when the desk sold 0.20 TRS at -50 CHF. Total cash position in CHF is CHF. The desk therefore has to buy back against 52.50/0.85=61.76 USD. What is the result at t=2 of all this rebalancing? The desk is completely hedged. It has zero Swiss Franc in cash. The net P&L in USD is: USD received from selling CHF, USD, minus the amount of USD paid when the desk bought back CHF, USD. The total USD P&L after 2 days equals: = USD. Table 2 provides a good overview of these trades. Table 2 Overview of trading activity of stock, TRS and CHF book scenario 2 t=0 t=1 t=2 USD CHF FX rate Stock price Stock position Stock Value Daily Stock P&L (USD) TRS price (CHF) TRS position TRS position Value (CHF) FX Hedge trade CHF FX Hedge USD side Cash CHF position before FX hedge CHF exposure TRS CHF cash position after FX hedge Total CHF Exposure after FX hedge USD Cash resulting from FX hedge G.O. Capital Markets Consulting Services Page 10 of 11

11 The point of this entire exercise is to show that this hedged position can experience large P&L moves, depending on the Swiss Franc exchange rate trading in or out of sync with the stock markets. The relative P&L figures in this example are huge. The examples of market behavior are vastly exaggerated of course to drive this point home. In reality, the stock market did not change by 50%, although the Swiss Franc did show moves of 10% on some days G.O. Capital Markets Consulting Services Page 11 of 11