AIM 2: Outline the structure of the loss distribution approach discussed in the material.

1、The structure of the loss distribution approach (LDA) would assess capital requirements in cells resulting from grouping the data in a business line/event type matrix. The overall operation risk capital requirement for the firm:

A) would add the capital requirements from all the cells assuming the losses in the cells are uncorrelated.
 
B) would add the capital requirements from all the cells assuming the losses in the cells are perfectly correlated.
 
C) would combine the results in each of the cell in such a way that takes into account the fact that the losses would not be perfectly correlated yet correlated to some extent.
 
D) would take a geometric mean of the capital requirements multiplied by the number of cells.

The correct answer is C

The loss distribution approach (LDA) approach has several steps. One of them is group the data in a business line/event type matrix that corresponds to the model. The overall operation risk capital requirement for the firm would combine the results in each of the cell in such a way that takes into account the fact that the losses would not be perfectly correlated.

AIM 3: Discuss what data sources should be incorporated in a valid loss distribution model.

1、With respect to valid data that can be incorporated in a loss distribution approach model for a given firm, which of the following is most accurate?

A) One firm may not have sufficient internal loss data for modeling a loss distribution, the loss distribution approach typically requires obtaining data from external sources such consortium loss data and data from commercial data bases.
 
B) One firm may not have sufficient internal loss data for modeling; thus, along with the internal actual data, a firm may only use the data from a few select firms that resemble the firm being analyzed. 
 
C) One firm usually has sufficient internal loss data for modeling; therefore, only internal actual data may be used, but under certain circumstances simulated data using the statistics of the internal data can be used.
 
D) One firm usually has sufficient internal loss data for modeling; therefore, only internal actual data may be used.

The correct answer is A

Because one firm may not have sufficient internal loss data for modeling a loss distribution, the loss distribution approach typically requires obtaining data from external sources. External data sources include the consortium loss data from The Operational Riskdata eXchange Association (ORX), loss data from the commercial data base OpVantage, which is a subsidiary of Fitch Risk, and data generated from scenarios specified by experts in divisions, control and support functions, and regions.

AIM 5: Compare various approaches to weigh risk data and scenario information.

1、Bill Stevens, FRM and Vivian Lamb, FRM, are discussing how to use a data set in the loss distribution approach (LDA). They agree that most of the observations will receive an equal weight, but they wish to consider some possible exceptions. Stevens recommends their placing a lower weight on older losses. Lamb recommends their placing a lower weight on external losses obtained from a commercial loss data base. Which of the two recommendations, if either, are acceptable under standard LDA?

A) Both recommendations.
 
B) Neither recommendations.
 
C) The recommendation of Stevens but not that of Lamb.
 
D) The recommendation of Lamb but not that of Stevens.

The correct answer is A

In modeling the distribution, the analyst will assign an equal weight to all data points. There are three exceptions: split losses, old losses, and external losses in the commercial loss data base and scenarios. Old losses should receive a lower weight. One method is to give full weight to losses in the past five years and decrease the weight to zero in a linear fashion over years six to 25. External data and scenarios may require scaling because of inherent biases.

AIM 6: Describe the derivation of frequency distributions in loss distribution models.

1、In the loss distribution approach (LDA), the calibration process for the frequency distributions of losses usually chooses any of the distributions from the following list except the:

A) chi-square (χ2) distribution. 
 
B) negative binomial distribution.
 
C) binomial distribution.
 
D) Poisson distribution.

The correct answer is A

The chi-square distribution is used to test how closely the selected distribution (Poisson, binomial, etc.) fits the actual data.

2、In loss distribution analysis, modeling the frequency distribution requires:

A) less data when compared to modeling the severity.
 
B) more data when compared to modeling the severity.
 
C) the same amount of data compared to modeling the severity.
 
D) either more or less data, it depends upon whether the analyst is measuring frequency and severity in the tails or not.

The correct answer is A

Modeling the frequency distribution requires less data when compared to modeling the severity.

AIM 7: Discuss the role severity distributions play in the determination of loss as well as the practical implementation of these distributions.

1、Which of the following is not a practical consideration in modeling severity distributions in the loss distribution approach (LDA)?

A) Placing too much emphasis on determining the appropriate capital reserve and not on the possibility of severe losses.
 
B) Extrapolating the data beyond those observed in the internal and external data set.
 
C) Explaining the model to colleagues in other departments.
 
D) Obtaining and calibrating external data.

The correct answer is A

A practical consideration is to recognize that the analyst’s ultimate goal is to determine the operational risk capital reserve, and extrapolations could lead to the inclusion of extremely severe losses and an overestimation of the needed capital reserve.

AIM 8: Explain the process of building a piece-wise defined distribution.

1、With respect to the loss distribution approach (LDA), which of the following best characterizes a methodology for building an adequate representation in the tail? Establish a threshold representing an absolute value of losses and then:

A) use only internal data for losses less than the threshold and both internal and external data for losses greater than the threshold.
 
B) use only internal data for losses greater than the threshold and both internal and external data for losses less than the threshold.
 
C) use only internal data for losses less than the threshold and only internal data for losses greater than the threshold but with a higher weight on the latter.
 
D) use only internal data for losses less than the threshold and only internal data for losses greater than the threshold but with a higher weight on the former.

The correct answer is A

To build a model of a distribution that includes adequate representation in the tail, an analyst may wish to piece together the distribution in the following way:

1、For a given cell, model the distribution only using internal data for that cell that are below a given threshold.
2、For the same cell, model the distribution using both internal and external data that fall beyond the threshold.

AIM 10: Discuss various types of dependencies in the model and how to model correlation.

1、In the loss distribution approach (LDA), which of the following within cell dependencies or correlations can affect the probability of loss and the economic capital of the bank? The dependencies between:

severity samples in a cell.
occurrence of loss events.
the frequency distribution and the severity distribution.
A) I, II, and III.
 
B) II and III only.
 
C) I and II only.
 
D) None of these.

The correct answer is A

There can be within cell dependencies which include the dependence between the occurrence of loss events, the frequency distribution and the severity distribution, severity samples in a cell.

AIM 11: Discuss the relationship between loss distribution approach and insurance.

1、With respect to including the role of insurance in the loss distribution, the analyst would:

A) allow for the risk reducing effect of insurance, generally, by reducing the frequency of losses but not their severity.
 
B) allow for the risk reducing effect of insurance, generally, by reducing the severity of losses that exceed a given deductible but not adjusting the frequency.
 
C) allow for the risk reducing effect of insurance, generally, by reducing both the frequency and severity of losses.
 
D) ignore it because the point is to model the losses regardless of insurance.

The correct answer is B

The loss distribution approach allows for a risk profiling of an institution, which can include the risk reducing effect of insurance. Insurance alters the aggregate loss distribution. Typically this is done by reducing the severity of the losses that exceed a given deductible in the insurance policy. In other words, insurance typically lowers the severity but not the frequency.

AIM 12: Explain the approach followed in determining economic capital, including allocation techniques and Monte Carlo methods.

1、The standard procedure for credit and operational risk is to specify the Economic Capital as:

A) Value-at-Risk divided by Expected Loss.
 
B) Value-at-Risk minus Expected Loss.
 
C) Value-at-Risk plus Expected Loss.
 
D) Value-at-Risk.

The correct answer is B

VaR is widely used in this context. In credit risk, Expected Loss denotes the mean of the portfolio loss distribution.

2、When using a Value-at-Risk (VaR) model in the loss distribution approach, a higher confidence level would:

A) increase the importance on modeling the tails and requires the use of both internal and external data.
 
B) decrease the importance on modeling the tails but requires the use of both internal and external data
 
C) decrease the importance on modeling the tails and only allow the use of only internal data.
 
D) increase the importance on modeling the tails and requires the use of only internal data.
 

The correct answer is A

A high confidence level, i.e., 99.98%, places a great deal of importance on modeling the tails of the loss distributions correctly and requires the use of both internal and external data.

3、A bank simulates the distribution of its operational losses. It finds that the loss that corresponds to the 99th percentile of potential losses is $250,000 and the mean of the distribution is $30,000. An estimate of operational risk economic capital is closest to:

A) $30,000.
 
B) $220,000.
 
C) $250,000.
 
D) $270,000.

The correct answer is B

Operational risk economic capital is the difference between the loss at a given confidence level and the expected loss. In this case, $250,000 ? $30,000 = $220,000.

AIM 13: Describe the analysis and validation of loss distribution approaches, including the loss distribution, sensitivity analysis, stress testing, back testing and benchmarking.

1、Which of the following best describes concerns with respect to data and risk-sensitive measures in the validation processes for estimates of regulatory and economic capital using the loss distribution approach?

A) Data problems and the lack of a risk-sensitive measure for operational risk modeling complicate the problem. Thus, there is inherent uncertainty in the validation process, and expert judgment should complement quantitative measures.
 
B) Data gathering techniques have improved, and there is now an established risk-sensitive measure for operational risk modeling. Therefore, validation processes should largely depend upon quantitative measures and not subjective expert judgment.
 
C) There are data problems, but there is an established risk-sensitive measure for operational risk modeling. Thus, there still is inherent uncertainty in the validation process, and expert judgment should complement quantitative measures.
 
D) Data problems and the lack of a risk-sensitive measure for operational risk modeling complicate the problem to such an extent that validation processes are useless.

The correct answer is A

To show that the estimates for regulatory and economic capital are reasonable, there should be a validation process. Given that confidence levels are often high, this can be challenging. Data problems and the lack of a risk-sensitive measure for operational risk modeling complicate the problem. Thus, there is inherent uncertainty in the validation process, and expert judgment should complement quantitative measures.

2、The validation of an operational risk model involves which of the following?

A review of data inputs, model methodology, and model outputs.
An examination of the assumed frequencies, severities, and dependence structure.
An examination of the role of insurance.
An examination of the assumptions in the model such as correlations.
A) II, III, and IV only.
 
B) I, II, and IV only.
 
C) II and IV only.
 
D) I, II, III, and IV.

The correct answer is D

The validation of an operational risk model involves a review of data inputs, model methodology, and model outputs. The review of the model would include examining the assumed frequencies, severities, dependence structure and insurance as well as assumptions in the model such as correlations. The validation would include an assessment of the impact of different modeling assumptions and the sensitivity of capital requirements to parameter changes.

3、In the loss distribution approach, backtesting a model:

A) is recommended; however, it cannot fully validate the model because the capital requirements represent extreme losses.
 
B) is not recommended because it is time consuming, and the models should be forward looking.
 
C) is not recommended because unrealized potential losses are not included in observed data.
 
D) is recommended, straightforward and valid given that it uses observed data.

The correct answer is A

The researcher should backtest the model to see how well it represents what has occurred. Unfortunately, with respect to operational risk, data problems are an issue. In validating a loss distribution model, a good approach is to backtest the capital estimates against actual annual losses; however, this cannot fully validate the model because the capital requirements represent extreme losses.

4、All of the following are acceptable methods for benchmarking operational risk EXCEPT:

A) comparing the advanced measurement approach (AMA) capital charge against the basic indicator approach (BIA) or the standardized approach (SA) capital charges.
 
B) comparing the operational risk capital charge against the LDA Association Capital Index (LDAACI).
 
C) comparing the LDA model outputs against adverse extreme, but realistic, scenarios.
 
D) comparing the operational risk capital charge against the institution’s peers.

The correct answer is B

The LDAACI is a fictitious term.

5、With respect to quantifying operation risk under the Basel II Accord, the loss distribution approach (LDA):

A) is currently disallowed, but its use is under review.
 
B) is the only recommended approach.
 
C) has been permanently disallowed.
 
D) is one of many possible approaches.

The correct answer is D

The Basel II Accord allows several approaches to the quantification of operation risk, and many see the loss distribution approach (LDA) as the natural way to meet the soundness standards for economic and regulatory capital by explicitly modeling the operation risk loss distribution of the bank over a one-year period.