Table of Contents

1. Introduction 1

2. Analysis for problems associated with using models 1
2.1. Model error 1
2.1.1. Wrong or simplifying assumptions 1
2.1.2. Over dependence on historical data 3
2.1.3. Black swans 4
2.2. Implementing a model wrongly 4

3. Improvements of the usage of models 5

4. Conclusion 7

1. Introduction

The financial sector plays crucial roles that mobilize savings and allocate credit in economic performance. In recent years, there has been significant technological development within the financial sector, which has enable banks to effectively manage their internal risk through the application of risk models. The use of models to measure risks is the preferred approach by most banks, for example Goldman Sachs applies the Value at Risk model. However, according to Office of the Comptroller of the Currency (2011, p1), “the expanding use of models in all aspects of banking reflects the extent to which models can improve business decisions, but models also come with costs”. Besides, in a recent study (Jorion 2009), it is argued that many financial institutions experienced large losses over the past few decades due to limitations of using sophisticated models. Therefore, it is essential for Andrew Bank Ltd. to have an in-depth understanding of disadvantages relating to using models and solutions to improve these model risks.

2. Analysis for problems associated with using models

“Risks are uncertainties resulting in adverse variations of profitability or in losses”(Bessis 2002, p11). Banks exposes to following risks, credit risk, interest rate risk, market risk, liquidity risk, operational risk, foreign exchange risk and others. As a result, risk models such as the gap analysis, the Value at Risk (VaR) and pricing model are designed to measure and manage risks. Therefore, it is imperative that Andrew Bank executes these models correctly; in order to avoid model error and wrong implementation of the model.

2.1. Model error

A model error is defined as “the model might contain mathematical errors or, more likely, be based on simplifying assumptions that are misleading or inappropriate”(Crouhy 2006, p349).

2.1.1. Wrong or simplifying assumptions

One of the most common and dangerous model risks is based on wrong or simplifying assumptions of risk models. The VaR model is the typical example. Leippold (2004) notes that “The VaR of a portfolio is the minimum loss that a portfolio can suffer in x days in the y% worst cases when the absolute portfolio weights are not changed during these x days”. To calculate VaR, there are three appropriate methodologies for deriving it, the distribution contained the analytic variance-covariance, the historical simulation and the Monte Carlo simulation approach. When Andrew Bank uses the VaR model with the analysis of variance-covariance approach; the bank can deal with “fat tails” problem because there is assumption that the risk factors and the portfolio value are log-normally distributed. Berry (2008) believes that the normal distribution is following a Gaussian distribution, or bell-shaped curve. In other words, “the pattern of normal distribution will cluster around those smaller changes toward the middle of the curve, while the less-frequent distributions will fall along the ends of the curve” (Nocera 2009). Nevertheless, fat tails refers the distribution that there is great number of observations far away from mean. This is shown below in the figure 1.
[Figure 1: Fat tail distribution]
Consequently, fat tails in distributions can show that the extraordinary losses have much higher possible chances than a normal distribution. Lehman Brothers is one of many financial institutions that have applied a quantitative model named VaR. This bank thought the Delta bond market as volatile while the Collateralized debt obligation (CDOs) market was undergoing no volatility and “the Risk management department held to the VaR postion that the CDOs were as safe as government bonds with a AAA rating” (Nocera 2009). In fact, from 2000 to 2006, many mortgage lenders tend to relax their lending standards; as a result there was a bubble in house prices in United States. Based on VaR, Lehman Brothers continuously bought with borrowed funds and increased their leveraged position (McVeigh, Cudmore and Alman). However, in the crisis market conditions, the VaR with normal distribution assumption might be wrong. The Risk Management Department too relied on the VaR number according to their calculations. Consequently, when Lehman Brothers wanted to sell because their loan came due, it become difficult to sell. The bank had to hold the large amount of debt and this is one of considerable reasons why Lehman Brothers, which was the fourth largest investment bank in USA, collapsed on September 15, 2008. Another example of model error caused by wrong assumptions is the failure of a well-established hedge fund operated by Victor Niederhoffer in November 1997. This financial institution had trading strategy based on shaky assumption. Crouhy et al. (2006) state that “the fund had been writing a large quantity of naked, deeply out-of-the-money put options on the S&P 500 stock index, and collecting small amounts of option premium in return”. The Niederhoffer’s strategy assumed that the market would never decrease by more than 5 percent on a given day. But, the stock market actually went down by approximately 7 percent due to the remarkable influence of crisis in the Asian markets. As a result, the firm seriously faced liquidity problems, had to sell at fire-sale prices and even lost the total amount of equity of the fund. Furthermore, Long Term Capital Management also applied the variance-covariance approach for VaR model with simplifying assumption that the market state would remain stable and it aimed at the daily volatility number of $45 million in 1998. In reality, this number could be approximately $90 million because the Russian default leaded to stress market conditions (Feridun 2005). Consequently, LTCM seriously had to deal with liquidity problem. The figures 2 and figure 3 show the losses incurred by LTCM by trade types and by financial institutions.
[Figure 1 and figure 2]

2.1.2. Over dependence on historical data

There are some models that use historical data to produce the future movements, but the problem is that the dramatic financial innovation in recent decades can make historical data become inadequate guide. For instance, to calculate the risk model, it is necessary to predict how the house prices would change over the next year based on real house prices within previous ten years. The modeling exercise might be incorrect because the house prices volatility can be higher or lower in the future than in the past. In other words, the probability of a decrease in prices was remarkably underestimated or overestimated (Stulz 2009). Moreover, only with historical data, it is also really difficult to estimate the impact on the value of assets when the house prices dropped. Therefore, if Andrew Bank excessively relies on the model such as the VaR, the bank can cope with the similar problems with Long Term Capital Management (LTCM). In 1998, the failure of LTCM, which was the largest hedge fund, almost ruined the whole financial system in the world. To forecast and mitigate the risk exposures, LTCM used combination of two main methodologies for VaR model. As Feridun (2005, p4) points out, the VaR analysis noted that the investors could undergo a loss of 5 percent or more in about one month in five, loss of 10 percent or more in about one month in ten and only one year in fifty the firm might lose at least 20 percent of its portfolio. Nevertheless, LTCM claimed that its daily VaR, which means the total amount of loss, was only $35 million in August 1998, whereas it actually went down $550 million in a day (Lowenstein 2008). This failure occurred because LTCM heavily rely on historical information in order to make accurate forecast of future changes. This financial institution believed that thanks to the historical trend of securities movements, can correctly estimate the future price.

2.1.3. Black swans

According to Taleb et al. (2009), studying the past and using historical data cannot certainly help financial institutions manage risks. Also, there is no such matter as a typical failure or success due to the randomness of socioeconomic. In addition, with standard VaR the bank should give more emphasis in what takes place in the other 1 percent instead of the number that loses within 99 percent probability. The VaR cannot measure which particular events could happen in the 1 percent and these events have called “fat tails” or “Black swans” (Taleb et al. 2009).

2.2. Implementing a model wrongly

The second cause of model risk is implementing a model wrongly both by accident and as a fraud. Although the model, which is applied by Andrew Bank Ltd., is correct, it still contains problems caused by inaccurate data and inappropriate length of sampling period. Besides, there are huge number of models which have sophisticated program and calculations, so the process always carries either technical errors or limitations. The failure of the Merrill Lynch is one of considerable example for wrong rate input. The Merrill Lynch started to divide 30-year government bonds into their building-block components and offered to the market as “interest only” (IO) and “principal only” (PO) instruments. The firm used the 30-year par yield to price the IOs and the POs. Thus, the investment firm Merrill Lynch underestimated the value of the IOs and overestimated the value of the POs. The firm sold $600 million of the IOs and none of POs, while its trader hedged the 30-year bonds within about 13 years. Due to the increase of interest rate, the Merrill Lynch had loss of $70 million. Another example of implementing a model wrongly is the failure of Kidder Peabody. Many investment banks tend to create a zero-coupon bond, named a strip. Joseph Jett, who is a trader worked for Kidder Peabody, made a simple trading strategy that he bought strips and sold these bond in the forward market. The Kidder Peabody’s computer system calculated the profit earned from each Joseph Jett’s trade by forward price minus spot price because the forward price was always higher than spot price. As a result, the computer system showed that there was a profit of $100 million; in reality there was a loss of $350 million (Hull 2010). There was difference between the forward price and the spot price just because of the cost of funding when Joseph Jett purchased the strip. Furthermore, the National Westminster Bank is one of typical example of implementation risk. It is crucial for the financial institutions to confirm their volatility estimates as well as principal inputs to a pricing model. In 1997, the traders at NatWest discovered that they had been “selling caps and swaptions in sterling and deutsche marks at the wrong price since late 1994” (Crouhy 2006, p355). Moreover, the firm also bought option priced at remarkable high volatility. Due to wrong principle inputs, the NatWest had lost of $80 million.

3. Improvements of the usage of models

Although applying models has many considerable benefits, these also contain few limitations. Thus, it is really necessary for Andrew Bank to know how to improve the way to use models in order to effectively manage risk. First of all, whatever the method used to calculate model such as VaR, it is important to do back testing. This process can help the bank examine how well the model estimates. For instance, the bank used VaR model to calculate a one-day 99 percent VaR. Back testing can show how often exceptions, which are the real change exceeds VaR, would have appeared. If the exceptions occur on about 1 percent a day, it can be comfortable with the methodology. In contrast, exceptions happen higher or lower than 1 percent, it means that VaR is underestimated or overestimated. Another important risk management method is stress testing. Stress testing aims to evaluate the potential loss associated with particular conditions. Now, the financial institutions recognize how crucial it is to be aware of the financial system’s vulnerability to stress situations. Havro et al. (2011) claim that stress testing is a quantitative method developed to illuminate the vulnerability and forecast the influence on both profit and strength of the financial institutions. When Andrew Bank deals with fat tails or black swan problems, bank stress test, which is stress testing that conducted by financial institutions based on a macro scenario specified by the authorities, can help the bank minimize the model risks of the VaR. According to Havro et al. (2011), the benefit of bank stress test is that the bank can assess how macro scenario impacts the risk associated with each exposure. HSBC, for example, welcomes the publication of the stress test outcomes on European banks by the European Banking Authority. The stress test results show how the current economic conditions in residential and commercial property, sovereign bond and securitized asset markets impact on HSBC (HSBC 2011). Moreover, Nocera (2009, p12) takes a different view that “The math alone was never going to be enough”. Eventhough the quantitative analysis and models play important roles; the qualitative factors such as the macroeconomic also have significant impacts on the risk management. In addition, the bank should not heavily depend on models, especially just a single model. It is essential for the bank to have its own judgment based on the current economic conditions. This is reason why in the summer 2007, Goldman Sachs avoided the large amount of loss, whereas Lehman Brothers, Merrill Lynch and many other institutions eventually collapsed. When Goldman Sachs felt that something was wrong, he called a meeting. After putting the models aside and considering the current economic factors, Goldman Sachs decided to remove the mortgage-backed instruments (Nocera 2009). Finally, the risk managers should invest in reseach to improve models and to develop better statistical tools. According to Crouhy (2006, p354), “An even more vital way of reducing model risk is to establish a process for independent vetting of how models are both selected and constructed”. The vetting team should ask for full documentation of the model consisting assumptions and mathematical statements. In other words, the bank should verify all the components such as parameters and equations, and particular implementation including inputs and outputs. Consequently, the Andrew Bank can mitigate the risks when it applies risk model.

4. Conclusion

It is impossible to argue against the great impacts of risk models on the risk management in banking. However, when Andrew Bank applies the models to help manage risk, the bank might have difficulties, caused by model error and wrong implementation of the model. The recognition of these limitations, the bank can analysis the problem and then improve the way to use these models in order to get rid of inaccurate factors in models and eventually minimize model risks.

(2395 words)

References

BESSIS, Joel (2002). Risk management in Banking. 2nd ed., West Sussex, John Wiley & Sons.

BERRY, Romain (2008). Value-at-Risk: An overview of analytical VaR. [online]. J.P. Morgan Investment analytics and Consulting, 7-9. Last accessed September 2008 at http://www.jpmorgan.com/

CROUHY, Michel et al. (2006). The essentials of Risk management. New York, McGraw-Hill.

Fat tail distribution( 2008). [online]. Last accessed at http://www.fattails.ca/

FERIDUN, Mete (2005). Value at Risk: Any lessons from crash of Long-Term Capital Management? [online]. Journal of Business administration online, 4(1). Last accessed 19 September 2006 at: http://papers.ssrn.com/

HULL, John C. (2010). Risk management and Financial institutions. 2nd ed., Massachusetts, Pearson.

HAVRO, Goril B. et al. (2011). Norges Bank’s stress test in Financial stability 2/10 compared with banks’ projections. [online]. Last accessed 8 November 2011 at http://www.norges-bank.no/

JORION, Phillippe (2009). Risk management lessons from the credit crisis. [online]. European fiancial management, 15(5), 923-933. Last accessed at 21 October 2009 at http://papers.ssrn.com/

LOWESTEIN, Roger (2008). Long-term capital: It’s a short-term memory. [online]. The New York Times. Last accessed 6 September 2008 at http://www.nytimes.com/

LEIPPOLD, markus (2004). Don’t rely on VaR. [online]. Euromoney, November 2004. Last accessed at 18 April 2007 at http://papers.ssrn.com/

MCVEIGH, Natalie, ALMAN, Robyn and CUDMORE, Richard. Lehman Brothers: An exercise in Risk management. [online]. Accessed at: http://www.necb.edu/

NOCERA, Joe (2009). Risk mismanagement. [online]. The New York Times. Last accessed 2 January 2009 at http://www.nytimes.com/

STULZ, Rene M. (2009). Six ways companies mismanage risk. [online]. Harvard business review. Last accessed March 2009 at http://hbr.org/

Supervisory guidance on model risk management. (2011). [online]. Last accessed 4 April 2011 at http://www.occ.treas.gov/

Statement on results of the 2011 EBA EU-wide stress test. (2011). [online]. Last accessed 15 July 2011 at http://www.hsbc.com/

TALEB, Nassim N. et al. (2009). The six mistakes executives make in risk management. [online]. Harvard business review. Last accessed October 2009 at http://hbr.org/

List of figures

Figure 1: Fat tail distribution
[pic]

Source: http://www.fattails.ca/

Figure 2: Losses incurred by LTCM in 1998 by trade types

Figure 3: Losses incurred by financial institutions through collapse of LTCM

-----------------------
Source: Feridun 2005

11

Source: Feridun 2005

12