Management of Credit Risk - Modern Banking

Background

Market risk has received an inordinate amount of attention in recent years but managing credit risk is the ‘‘bread and butter’’ of most commercial banks. Every commercial bank, by definition, has a loan portfolio. Increases in credit risk will raise the marginal cost of debt and equity, which in turn increases the cost of funds for the bank. Techniques for credit risk management are well known because the banking sector has had a long history of experience in this area. Nonetheless, loan quality problems are an important cause of bank failure. For this reason, all bankers, not just those in a credit risk department, should be aware of the key factors affecting the quality of a loan portfolio, and the methods for managing it.

Credit Risk Decisions: Retail versus Corporate

If a bank is looking to minimise its aggregate credit risk, then good risk management of retail and corporate lending is essential. The approaches taken for retail and corporate loans differ considerably, mainly because a corporation is able to produce a variety of financial ratios, which are not available when the suitability of an individual or a small firm for a loan is being assessed. Most bankers concede that lack of information makes retail lending more difficult than corporate lending. On the other hand, loans to corporates which turn out to be bad can be very serious for the bank because of the large sums involved. Countless cases abound: Maxwell and a number of London-based banks; Schroder and Deutsche Bank; and the collapse of Enron and WorldCom. The number of incidents where retail loan defaults have had serious consequences for a bank is very much lower, and usually occurs if a bank is over-exposed in one area such as mortgages, and property prices collapse at the same time as interest rates rise.

The principles used to model credit risk, and the methods used to minimise credit risk for the retail and corporate sectors, are discussed below.

Minimising Credit Risk

There are five key ways a bank can minimise credit risk: through accurate loan pricing, credit rationing, use of collateral, loan diversification and more recently through asset securitisation and/or the use of credit derivatives. The weights applied to each of the methods will vary, depending on whether the loan is commercial or retail.

  1. Pricing the loan: any bank will wish to ensure the ‘‘price’’ of a loan (loan rate) exceeds a risk adjusted rate, and includes any loan administration costs, that is:
  2. RL = i+ip+fees-----(VII)

    where:
    RL: interest rate charged on the loan
    i: market interest rate, such as LIBOR or an equivalent term42
    ip: risk premium, negatively related to the probability of the loan being repaid (ip = 0 if repayment is certain)

    In the above equation, ip and i are positively related, for one of two reasons. In the case of a variable rate loan, if the market rate rises, so will the interest rate charged on the loan, and the borrower will find it more difficult to repay the loan, so the probability of default increases. Or, at very high market rates, the loan rate will rise, attracting riskier borrowers (due to adverse selection), so the chances of the loan being repaid fall.

    Other factors also influence the loan rate. If there is any collateral or security backing the loan, the rate charged should be lower than in the absence of security. In addition, there are non-price features: the bank may charge a high fee for arranging the loan but the interest rate will be lower. Or, some central banks impose reserve ratios, which means a percentage of a bank’s deposits is held at the central bank, often earning no interest. This is effectively a tax on deposits, which banks will try to make up by imposing higher loan rates.

    Thus, the loan rate should include a ‘‘market’’ rate, risk premium and administration costs. The riskier the borrower, the higher the premium. Should the risk profile of the loan be altered, the rate should change accordingly, though if increased, it should be borne in mind that the potential for adverse selection or adverse incentives is greater. The guidelines may also be difficult to implement in highly competitive markets.

  3. Credit Limits: another method for controlling credit risk. Given the potential for adverse selection, most banks do not rely solely on loan rates when taking a lending decision. Instead, the availability of a certain type of loan may be restricted to a selected class of borrowers, especially in retail markets. Branch managers are given well-defined credit constraints (and checklists – see below), and borrowers usually discover they may not borrow above some ceiling. In retail markets, banks normally quote one loan rate (or a very narrow range of rates) and then restrict the amount individuals or small firms can borrow according to criteria such as wealth or collateral. However, in the United States, legislation prevents banks from discriminating against certain retail customers. The Community Re-investment Act (1977) requires banks to provide evidence to the regulator that loan decisions do not discriminate against the local community. Under the Home Mortgage Disclosure Act (1975) regulators must be satisfied mortgage decisions by banks do not discriminate on the grounds of race, income, age or income status.
  4. By contrast, in the wholesale markets, credit limits are of secondary importance; loan rates (and the risk premium) normally vary from business to business because banks have more information on the value of a firm, such as independent auditor reports on a company’s financial performance.

  5. Collateral or Security: Banks also use collateral to reduce credit risk exposure. However, if the price of the collateral (for example, houses, stock market prices) becomes more volatile, then for an unchanged loan rate, banks have to demand more collateral to offset the increased probability of loss on the credit. Another problem that can arise is if the price of collateral is negatively correlated with the ability of the borrower to repay, that is, as the probability of default among a borrower class increases, the price of the collateral declines. For example, in the 1980s, Texan banks made a large number of loans to firms in the booming oil industry, and the collateral was often the oil well(s) or Texan real estate. When oil prices collapsed, the value of the collateral also collapsed. In the late 1980s, over a quarter of US banks that failed were located in Texas. In Britain, building societies and banks tend to enter into flexible rate mortgage agreements with homeowners, using the property as collateral. In the early 1990s, interest rates began to rise to counter inflation. The housing boom came to an abrupt end, and house prices fell rapidly. Householders who had borrowed up to 100% mortgages found themselves holding negative equity: the value of the house was less than the cost of the total outstanding mortgage. Many households were unable to make the mortgage repayments as recession set in and the unemployment rate rose. The banks/building societies realised they would end up having to dispose of real estate at very low prices. So many accommodated distressed borrowers, by allowing interest repayment holidays and other measures which meant increased arrears. Though costly, losses would have been far higher if houses were repossessed and sold. These examples illustrate an important point. Collateral tends to be a more effective means of managing risk for short-term (e.g. overnight loans) because the risk of its value changing is quite low. In this context, for big corporates and other banks, a bank will often use haircuts: an extra amount of collateral (in the form of a margin) applied to a loan, i.e.
  6. collateral plus a margin. Even with an overnight loan, there is a chance the value of the collateral will decline. A haircut is the amount by which the collateral exceeds the principal of a loan. For example, for an overnight loan the banks may ask for collateral, the value of which is 5% greater than the amount loaned.

  7. Diversification: Additional volatility created from an increase in the number of risky loans can be offset either by new injections of capital into the bank or by diversification.
  8. New loan markets should allow the bank to diversify and so reduce the overall riskiness of the loan portfolio, provided it seeks out assets which yield returns that are negatively correlated. In this way, banks are able to diversify away all non-systematic risk. Banks should use correlation analysis to decide how a portfolio should be diversified. An example of a lack of lending diversification was the US savings and loans sector, or ‘‘thrifts’’ in the 1980s. Regulations required a high percentage of their assets to be invested in home mortgage loans and mortgage-related securities. The thrifts tried to diversify by moving into commercial real estate financing, and later got involved with new financial innovations about which they knew little, resulting in a costly debacle – over 1000 failed. Banks can help to ensure they are properly diversified by setting concentration limits: the bank sets a limit on the amount of exposure in relation to a certain individual or sector.

  9. Credit Derivatives and Asset Securitisation: recall from previous that asset securitisation is a method of reducing credit risk exposure, provided a third party assumes responsibility for the credit risk of the securitised assets. credit risk derivatives can be used to insure against a loan default.

Assessing the Default Risk of Individual Loans

Most banks have a separate credit risk analysis department – their aim is to maximize shareholder value-added through credit risk management. Managerial judgement always plays a critical role, but a good credit risk team will use qualitative and quantitative methods to assess credit risk. The use of different methods will be determined by the information the bank can gather on the individual.

If a bank is unable to obtain information on a potential borrower (using, for example, annual reports), it is likely to adopt a qualitative approachto evaluating credit risk, which involves using a checklist to take into account factors specific to each borrower:

  • Past credit history (usually kept by credit rating agencies).
  • The borrower’s gearing (or leverage) ratio – how much the loan applicant has already borrowed relative to his/her assets.
  • The wealth of the borrower.
  • Whether borrower earnings are volatile.
  • Employment history.
  • Length of time as a customer at a bank.
  • Length of time at a certain address.
  • Whether or not collateral or security is part of the loan agreement.
  • Whether a future macroeconomic climate will affect the applicant’s ability to repay.

For example, a highly geared flexible rate borrower will be hit hard by rising interest rates.

Thus, the credit risk group will have to consider forecasts of macroeconomic indicators such as the interest rate, inflation rate and future economic growth rates.

Along a similar vein, Sinkey (2002) singled out what he calls the ‘‘fives Cs’’ to be used in a qualitative assessment of credit risk.

  • Character: Is the borrower willing to repay the loan?
  • Cash flow: Is the borrower reasonably liquid?
  • Capital: What assets or capital does the borrower have?
  • Collateral or security: Can the borrower put up security (e.g. deeds to a house, share certificates which will be owned by the bank in the event of default)?
  • Conditions: What is the state of the economy? How robust will the borrower be in the event of a downturn?

Quantitative models

A quantitative approach to credit risk analysis requires the use of financial data to measure and predict the probability of default by the borrower. Different models include the following.

Credit scoring. Here, the data from observed borrower behaviour are used to estimate the probability of default, and to sort borrowers into different risk classes. The type of information gathered is listed above but here, a weight is applied to each answer, and a score obtained. The weights are obtained from econometric techniques such as discriminant or logit analysis. Here a large amount of historical data from two populations are obtained, from the population that defaults and a group which does not default.

Discriminant analysis assumes that a borrower will come from one of two populations:

those that default are in one population (P1) and population 2 (P2) consists of firms that do not default. Data from past economic performance are used to derive a function that will discriminate between types of firms by placing them in one of two populations.

Thus, if Z is a linear discriminant function of a number of independent explanatory variables, then

Z = ∑aiXi, i = 1, 2,..., n ----(VIII)

where Xi are the independent explanatory variables, such as credit history, wealth, etc.

Sample data are used to test whether the discriminant function places the borrower in one of the two populations, with an acceptable error rate.

Logit analysis differs from discriminant analysis in that it does not force borrowers into separate populations but instead assumes that the combined effect of certain economic variables will serve to push a borrower over a given threshold. In this case, it would be from the non-arrears group into the arrears group. Note that in logit analysis, the dependent variable is a binary event, and the objective is to identify explanatory variables which influence the event. The logit model may be written as follows:

P{(yit+1) = 1|xit}=[eb+c_xu]/[1+eb+c_xu] ----(IX)

where:

xit: value of the explanatory variable i at time t
P(yit + 1): probability of a firm being in arrears at time t + 1; y = 0 implies the firm is not in arrears

In either the discriminant or logit models, the estimates obtained are used together with out of sample data, to forecast which borrowers will or will not default on their loans. The number of forecasting errors is determined. A type-I error occurs when the borrower is not forecast to go into arrears but does and a type-II error when a borrower is forecast to go into arrears but does not. The average costs of the two types of errors will differ. A type-I error means the value of the lender’s assets will fall, whereas with a type-II error a profitable lending opportunity has been missed – the bank loses in terms of opportunity cost. For this reason, it is normally assumed that a type-I error has a higher average cost for creditors than a type-II error. However, it is necessary to decide where the cut-off is going to be; the optimal cut-off will depend on the value of the cost ratio, defined as:

C = average cost of type-I error ÷ average cost of type-II error

For example, if a bank is very risk averse and puts a high weight on type-I errors, then C will be high (e.g. 2.5) and the bank will require very high scores if an individual or firm is to be approved a loan.

Individuals or corporations can be credit scored, though the variables used to determine the score will differ. For example, an individual will be scored based on age, income, employment and past repayment records, etc. Not all personal loans or firms are subject to credit scoring – it can only be done if there are enough data, which requires a sufficiently high volume of standardised loans that have been granted for some time.

Different financial ratios (such as debt to equity) are used to score corporations, as well as any external ratings of the firm, if is creditworthy enough to be issuing its own securities.

The Altman (1968) Z-score model is derived from discriminant analysis and is used for larger corporations. Based on financial/accounting ratios,44 each firm is assigned a Z score and, depending on that score, either the loan is granted or it is refused. The higher the Z score, the lower the probability of default; If Z is lower than 1.81,45 the default risk is considered too high and no loan is made.

For example, suppose SINCY corporation is applying for a loan. Its credit rating by the Good Rating Company is AB. The bank also requests that it provide extensive financial information: a business plan on what the loan is for, return on assets and equities, the ratio of debt to equity, and so on. This information is fed into a program, where every financial variable is weighted, then summed into a Z score. If it is very high, the loan is granted. If the score is borderline (for example, 2) then the relationship the firm has with the bank may be quite important, as well as Z scores on any previous loans which had been repaid.

Relationship banking also plays its part, especially in countries such as Germany and Japan. If a corporation has done business with a bank for a long period of time, then an application for a new loan will involve a combination of banker judgement and credit scoring.

This may also apply to personal customers with a long-standing credit history at a bank. Small businesses are more difficult. They vary considerably in their activities, and failure rates among small and medium-sized enterprises can be as high as 95%. This makes it difficult to apply credit scoring models to this group.

Problems with Credit Scoring Models. These models are not fail safe. They are only as good as the original specification, and one limitation is that the data are historical. Though the original discriminant analysis may have produced a Z score which was fairly accurate, unless it is frequently updated either the variables or the weights (assumed to be constant over time) make it less accurate. For example, the relevant financial ratios are likely to change, and may even differ depending on the industry being evaluated. The same remarks apply to the weights. This problem can be minimised if the bank keeps records of their type-I and type-II errors, and acts to implement a new model to address any necessary changes. An extensive list of variables must be subject to regular testing in the discriminant model, and any insignificant variables discarded. However, even a comprehensive list cannot take into account variables not easily quantified, such as the length and nature of the relationship between borrower and bank.

A more difficult problem is that the model used imposes a binary outcome: either the borrower defaults or does not default. In fact, there is a range of possible outcomes, from a delay in interest payments to non-payment of interest, to outright default on principal and interest. Often the borrower announces a problem with payments, and the loan terms can be renegotiated. These different outcomes can be included, but only two at a time. For example, the discriminant function can contain the default and no-default outcomes or the no-default and rescheduling outcomes, but not both.

Aggregate Credit Risk Exposure and Management

Up to this point, the discussion has focused on loans to individuals or firms. All banks will want to manage their aggregate credit exposure. A heavy concentration of loans in one sector has the potential of threatening the survival of the bank. There are many examples of banks getting into problems precisely because of over-exposure in one sector. The case of the Texan banks has already been mentioned. Not only did the value of their collateral (oil wells and real estate) fall when the oil industry began to encounter difficulties, but these banks were over-exposed in the oil sector as a whole. In the UK, excessive lending in the commercial property markets resulted in the illiquidity (and later, insolvency, in some cases) of secondary banks.46 In late 1972 secondary banking problems prompted a Bank of England-led lifeboat rescue to prevent a general crisis of confidence. In Japan, the jusenor mortgage corporations collapsed in 1995 after over-exposure in the property markets, five years after they had been instructed by the regulators to curtail their lending to this sector – a directive they ignored at their peril, prompting a public outcry (public taxes were used to fund a government rescue).

When assessing aggregate credit exposure, four factors should be taken into account in any model of credit risk. They are:

  1. Compute the expected loss levels over a given time horizon, for each loan and for the portfolio as a whole.
  2. Compute the unexpectedloss for each loan, i.e. the volatility of loss.
  3. Determine the volatility of expected loss for the portfolio as a whole.
  4. Calculate the probability distribution of credit loss for the portfolio, and assess the capital required, for a given confidence level and time horizon, to absorb any unexpected losses.

In the United States, where many corporations are rated by agencies such as Standard and Poor’s or Moody’s, it is possible to apply standard modern portfolio theory (PT) to get a measure of aggregate exposure. Assume the banks hold traded loans and bonds. The basic principle is diversification: provided returns on loans are not highly correlated, the bank can raise its expected return on a portfolio of assets by diversifying across asset classes. Put another way, suppose a bank has a portfolio consisting of two loans. The bank can achieve the same expected return on its portfolio of assets and reduce its overall risk exposure, provided the returns on two loans are negatively correlated.

Outside the United States, there are not many corporations which are externally rated. Even within the USA, the majority of banks’ main portfolios consist of non-traded loans. New methods are being developed to deal with this problem. Below, two approaches – the KMV model and Credit Riskmetrics – are discussed.

Default Mode Approach

The default model approach draws on modern portfolio theory to measure a bank’s aggregate credit exposure for non-traded assets, such as loans on the banking book. In PT, to obtain a measure of the risk–return trade-off between a portfolio of assets, there must be data on the expected return on the assets, the risk of the asset (measured by the standard deviation), and the correlation between the risks of the assets. If these assets are loans, then this translates into the expected return on each loan, E(Ri), where i goes from 1 to n loans, the risk of each loan σ and the degree to which the risk of each loan is correlated (ρij if there are two loans, i and j).

The emphasis is on loan loss rates. CreditRisk Plus (developed by Credit Suisse Financial Products) uses a default mode model. The ‘‘KM’’ model, developed by the KM Corporation, also takes this approach.48 The advantages over credit VaR (discussed below) are that less data are required, and there is no assumption of a normal distribution, which, as was noted earlier, is unrealistic for a portfolio of non-traded loans.

Some important assumptions are necessary.

  1. Either there is a default on the loan, or there is no default. Thus any migration, which measures the probability of a loan being downgraded, upgraded or defaulted on over a specified period of time, is assumed to be zero.49
  2. When considering a portfolio of loans (e.g. auto, home, personal), the DM approach assumes the probability of one loan default is independent of the probability of default on all other loans.
  3. The risk being measured. Here the focus is on credit risk – the risk of the debt not being repaid as agreed at the time agreed, where all loans are held to maturity. Recall from the discussion on RAROC that the unexpected loss on a loan is how the risk of the loan is measured, and differs from expected loss which is covered in the risk premia charged. Banks can compute the expected loss on a given loan category based on historical data. It will be known from past experience the proportion of borrowers who will default on home loans, car loans, etc. It will also be possible to calculate loss given default, based on past experience. If a borrower either repays the loan, or does not, then the standard deviation or riskof the ith borrower is the square root of the (probability of default)(1 − probability of default). This will be incorporated into equation below.
  4. The holding period. In common with what most banks do, the holding period is defined as one year. While a loan agreement may be for many years, specifying a holding period of a year means the bank can take stock of the status of the loan portfolio on an annual basis, because firms report their annual (sometimes quarterly) performance figures, and the bank can take action should the loan appear to be in trouble.
  5. Loss Given Default (LGD): The amount the bank loses if the borrower defaults. If there is no security on the loan, and the loan is completely written off, this loss will be in excess of the amount loaned if the book value of the loan is less than the current value of the loan, due to compounding. If the bank is able to cash in on collateral, the LGD can be quite low. Some defaults may involve relatively small sums, others involve substantial losses. For example, losses on a personal loan may amount to £5000, but losses when a business goes bankrupt can be in the hundreds of millions, which was the case with the collapse of Maxwell Communications Corporation, LTCM and Enron.
  6. Potential Credit Exposure (PCE): Refers to the amount of credit outstanding at the time of default. If all of the loan is taken at the time it is granted, then the repayment schedule makes it relatively straightforward to compute what the PCE is. It is assumed that 100% of any credit line (or agreed overdraft) has been used at the time of default.
  7. Hence, the PCE is assumed to be fixed over time.50 To compute a loan loss value, it is necessary to compute expected loss (EL) and unexpected loss. Recall from the earlier discussion on RAROC that the risk premium of a loan covers the expected loss, and capital is set aside as a buffer for the unexpected loss:

    EL = AVG(EDF)+AVG(LGD)+AVG(PCE) -----(X)

    where:
    AVG: average of ( )
    EDF: expected default
    LGD: loan loss given default
    PCE: potential credit exposure

    EL, or expected loss, is the average expected loss on a loan or set of loans over a specified length of time. Since EL is an average, averages (AVG) are used for each of the three terms on the right-hand side of the equation.

    For the portfolio of loans as a whole:

    EL(portfolio)= ni=1 ELi ----(XI)

    where ELi is the expected loss on the ith loan, i = 1, . . . , n.

    Recall that unexpected loss will be determined by the volatility (standard deviation) of the expected loss. If PCE and LGD are assumed constant, then the volatility will only depend on the expected default rates, and since there is either default or no default, the unexpected loss is:

    ULi=√ ELi(LGD−ELi) ----(XII)
  8. Correlation between default risks. Suppose a bank has a portfolio made up of car loans and mortgages, or the loans are to firms in different sectors. Then the likelihood of a default occurring at the same time is quite low, unless there is severe depression or recession which affects the majority of firms and individuals.

To obtain the unexpected loss for an entire portfolio, assume the unexpected default on loans will be correlated. Thus, for the portfolio as a whole:

UL(portfolio)= ∑ULiρi -----(XIII)

where:

ρi: correlation between the loss on the ith loan and the loss on the portfolio as a whole
UL(portfolio): standard deviation of the losses on the ith facility
ULi: standard deviation of the loss on the ith loan

Historical data are one way of computing the average correlation of each part of the loan book. For example, a 10-year series on loan losses for each of autos, personal loans, house loans, different categories of commercial loans can be used to estimate the standard deviation of the portfolio’s losses. With knowledge of the standard deviations, and the unexpected losses for each segment, then the average correlation can be obtained, i.e.

UL(portfolio) = √ρs(∑ULi) -------(XIV)

where:

ρs: average correlation of one loan segment (e.g. personal loans) to the whole loan portfolio

This simpler approach assumes that the average correlation remains unchanged over time, and therefore, that the mix of different loan types is unchanged. Nor does it allow for the effects of concentrated risks. If a bank concentrates 90% of its loans in one industry, then its unexpected losses will be far greater than that indicated by the equation above.

The KMV Corporation report that default correlations vary from 0.002 to 0.15. Such low correlation figures mean a bank can reduce their aggregate credit risk by spreading the loans across many firms and individuals.

Example:

Assume:

  • Super-Specialised Bank grants 100 unsecured bullet loans (principal and interest paid in full when the loan matures) of £2 million each, giving a total loan portfolio of £200 million.
  • The probability of default on each loan is 1%.
  • LGD is the value of the loan if a borrower defaults. So LGD will be £20 000 for a given default, rising to £2 million for the whole portfolio if every borrower defaults.
  • Based on KMV estimates, the correlation between each loan default and the portfolio as a whole is 0.02.

Then:

ELi = £20 000 on each loan, i.e. 1% of £2 million LGD = £2 million for the portfolio (1% of £200 million) ULi= √ELi(LGD − ELi) = √(20 000)(2 million−20 000) = £198 997.4874 or about £200 000

The computations show the unexpected loss for the loan portfolio as a whole is roughly £200 000. The correlation between each of the 100 loans and the whole portfolio is 0.02.

Then the risk contribution of one loan to the unexpected loss of the portfolio is £3979.95 or approximately £400052 per loan. For the entire portfolio, the loan loss is £397 995, or approximately £400 000.

The objective is for the bank to hold enough capital to absorb unexpected losses from the loan portfolio. To determine the appropriate confidence interval, the bank’s own rating is normally used. Suppose the bank has an AA rating, and the bank is conservative, with a confidence interval of 99.99%, meaning the bank is focusing on all but the worst outcome 51 KMV Portfolio Manager uses stock price correlation to deduce ρi. KMV was taken over by Moody’s in 2002.

in 10 000. So enough capital needs to be set aside to be 99.99% certain that losses arising from loan defaults will not cause the bank to fail. If the credit losses are normally distributed, then 3.89 standard deviations from the mean are needed for a 99.99% confidence interval.

Then the capital to be set aside for the whole portfolio is:

(£397 995)(3.89) = £1 548 200.55, or about £1.6 million

Remember, however, that for loan portfolios in particular, the assumption of a normal distribution does not hold – see Figure With this type of distribution, the 99.99% confidence interval lies between 6 and 10 standard deviations from the mean. The higher the proportion of commercial loans in a bank’s portfolio, the less likely the returns will approach a normal distribution, because there tend to be very large losses associated with one default, and the probability of default is more closely correlated with the economic cycle, compared with personal loans.

Suppose this portfolio is not too heavily skewed because there are proportionately more personal loans in the portfolio, and the 99.99% confidence interval lies 8 standard deviations from the mean, or 8 standard deviations are needed for the 99.99% confidence level. Then the capital to be set aside is:

Risk contribution × 8 = £397 995 × 8 = £3 183 960, or approximately £3 million

Note the amount of capital to be set aside has risen because the bank, by assuming a skewed distribution, has adopted amore conservative, or it could be argued, amore realistic, attitude.

Credit Value at Risk

Credit Value at Risk

Unlike the KMV and DM approaches, a credit VaR, e.g. Creditmetrics, is a marked to market approach, focusing on a loan loss value and/or a risk–return trade-off for a portfolio of debt. In the VaR approach there is more than one single credit migration. Instead of an asset either being a good asset or in default, in credit VaR, there is the possibility of multiple migration, that is, a range of upgrades or downgrades. Figure above illustrates the difference between credit migration in the two approaches. With multiple credit migration, any upgrade or downgrade will affect the spread changes, which in turn changes the discount rate.

VaR for market risk was covered in some detail, and space constraints preclude an in-depth treatment of credit VaR. Suffice to say that to obtain a credit value at risk for a given portfolio, it will be necessary to address the same list of issues raised in the section on market VaR.

Credit VaR is also the target of criticisms raised with respect to market VaR. Indeed, some of the underlying assumptions, such as a normal distribution, become even more problematic.

Comparison of Credit VaR and Default Mode.

Comparison of Credit VaR and Default Mode.

Financial Innovation and Risk Management

The financial products discussed above are examples of recent financial innovations. Like the manufacturing sector, financial innovation can take the form of process innovation, whereby an existing product or service can be offered more cheaply because of a technological innovation. Product innovation involves the introduction of a new good or service. The new financial instruments discussed above are examples of where technological changes resulted in product innovations.

Silber (1975, 1983) argued that product innovation arises because of constraints placed on a bank – namely, regulation, competition and risk. Kane (1984) thought it important to observe the regulatory and technological factors behind any financial innovation. However, it is more useful to think of financial innovation, regulation and risk management as being interdependent. For example, regulations (such as exchange controls) can be a catalyst for financial innovation which allows bankers to bypass the rules. The eurocurrency markets developed in just this way – US interest rate restrictions and limits placed on foreign direct investment by US multinationals, together with UK exchange controls, created a demand for and supply of an offshore dollar market, the eurodollar market. As technology advanced, this became the eurocurrency market, allowing bankers to trade in all the key currencies outside any domestic regulations. Even though most of the offending regulations have since been relaxed, the market continues to thrive.

Risk management and financial innovation are also interdependent. Financial innovation has made it possible to unbundle the different types of risks which, in turn, has led to measurement of different types of risks. At the same time, financial innovation has forced banks to re-examine their risk management systems, because banks are increasingly exposed to new forms of risk, which are quite different in nature from the traditional credit risk.

For example, Bankhaus Herstatt collapsed in 1974 because of inexperience in dealing with foreign exchange risk. However, to date, very few bank collapses can be said to have been caused by a failure to understand risk exposure associated with a new instrument.


All rights reserved © 2018 Wisdom IT Services India Pvt. Ltd DMCA.com Protection Status

Modern Banking Topics