I85PREP
⌘ K
OSMR study materials
OSMR study resources

OSMR practice questions

One question at a time. Build your OSMR understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1OSMR (domain names not published in source)

In OSMR, what does the term 'marginal risk' most directly refer to?

  1. The risk that a counterparty defaults on a settled transaction
  2. The additional risk introduced by increasing exposure by one further unit Correct answer.
  3. The average risk observed across an entire portfolio
  4. The total risk remaining after all hedges have been removed

Explanation

Marginal risk measures the incremental change in risk caused by a small additional position or exposure. It is used to decide whether adding one more unit of exposure is justified relative to its contribution to overall risk.

Question 2OSMR (domain names not published in source)

An OSMR practitioner is told that a risk measure must satisfy subadditivity. Which property does this guarantee?

  1. The risk of two combined positions never exceeds the sum of their individual risks Correct answer.
  2. The risk measure is unaffected by the confidence level chosen
  3. The risk of a portfolio always equals the sum of standalone risks
  4. The risk measure ignores diversification benefits entirely

Explanation

Subadditivity states that combining two positions cannot produce more risk than the sum of their separate risks, which is the mathematical expression of diversification. Measures lacking this property can penalize diversification, which is why coherent risk measures require it.

Question 3OSMR (domain names not published in source)

A risk manager computes a 99% one-day Value-at-Risk of 5 million for a portfolio. Which statement best describes what this figure conveys?

  1. Expected loss conditional on exceeding the threshold equals 5 million
  2. The portfolio will lose exactly 5 million once every hundred days
  3. Losses are expected to exceed 5 million on roughly one trading day in a hundred Correct answer.
  4. The maximum possible loss over one day is 5 million

Explanation

VaR at 99% indicates a threshold that losses exceed with about 1% probability over the horizon. It says nothing about the size of losses beyond that threshold, which is the well-known limitation of VaR.

Question 4OSMR (domain names not published in source)

Why do regulators often prefer Expected Shortfall over Value-at-Risk for market risk capital?

  1. It requires no historical data and relies only on expert judgement
  2. It always produces a lower capital number than VaR
  3. It captures the magnitude of losses beyond the VaR threshold and is subadditive Correct answer.
  4. It is easier to backtest with a simple pass or fail count

Explanation

Expected Shortfall averages losses in the tail beyond the VaR level, so it reflects tail severity as well as frequency. It is also subadditive, making it a coherent risk measure, unlike VaR.

Question 5OSMR (domain names not published in source)

In stress testing under an OSMR framework, what distinguishes a scenario stress test from a sensitivity test?

  1. A scenario test is always based on historical events while a sensitivity test is always hypothetical
  2. A scenario test applies a coherent set of adverse conditions across many risk factors simultaneously Correct answer.
  3. A sensitivity test measures tail losses while a scenario test measures average losses
  4. A sensitivity test shocks every risk factor at once while a scenario test shocks only one

Explanation

Scenario stress tests impose a joint, internally consistent set of adverse movements across multiple risk factors. Sensitivity tests instead move a single factor in isolation to measure first-order impact.

Question 6OSMR (domain names not published in source)

A bank's model validation team finds that its VaR model produced far more breaches than expected during a calm market. What is the most appropriate conclusion?

  1. The model understates risk and its calibration should be reviewed Correct answer.
  2. The model overstates risk and capital should be released
  3. The confidence level was set too high for the horizon
  4. The model is validated because breaches occurred during calm markets

Explanation

Excess breaches relative to the expected exceedance rate signal that the model's risk estimates are too low. The correct response is to investigate calibration, data, and methodology rather than release capital.

Question 7OSMR (domain names not published in source)

Which input is essential when calculating a historical simulation VaR?

  1. A time series of historical risk-factor changes over a chosen observation window Correct answer.
  2. The current credit rating of each counterparty in the portfolio
  3. The expected recovery rate on defaulted exposures
  4. A set of subjective probability weights supplied by senior management

Explanation

Historical simulation revalues the portfolio using observed past changes in risk factors over a defined window. The empirical distribution of those changes drives the VaR estimate.

Question 8OSMR (domain names not published in source)

An analyst wants a VaR method that explicitly models correlations and can price nonlinear instruments. Which approach best fits this requirement?

  1. Simple historical simulation with linear approximation
  2. Variance-covariance with delta-only approximation
  3. Monte Carlo simulation using a covariance structure and full revaluation Correct answer.
  4. Scenario analysis using a single historical crash

Explanation

Monte Carlo simulation draws correlated risk-factor scenarios from a specified distribution and fully revalues positions, capturing nonlinearity. Linear approximations like variance-covariance struggle with options and other nonlinear payoffs.

Question 9OSMR (domain names not published in source)

In backtesting a VaR model at the 99% level over 250 trading days, how many breaches would be expected if the model is correctly calibrated?

  1. About 125
  2. About 50
  3. About 2 to 3 Correct answer.
  4. About 25

Explanation

A 99% VaR implies a 1% expected exceedance rate, so over 250 days roughly 2.5 breaches are expected. Backtesting compares the observed count against this expectation using statistical tests.

Question 10OSMR (domain names not published in source)

What is the primary purpose of a traffic-light approach in VaR backtesting?

  1. To assign each trade a colour based on its liquidity profile
  2. To determine the optimal confidence level for the model
  3. To rank risk factors by their volatility contribution
  4. To classify the number of breaches into zones that trigger graduated supervisory responses Correct answer.

Explanation

Traffic-light backtesting sorts observed breach counts into green, yellow, and red zones. Each zone maps to progressively stronger supervisory action, such as capital add-ons or model review.

Question 11OSMR (domain names not published in source)

A portfolio holds a long call option. Which risk measure best captures the loss from a large adverse move in the underlying that a delta-based VaR would miss?

  1. Notional-based exposure measurement
  2. Full revaluation or gamma-adjusted sensitivity analysis Correct answer.
  3. Duration-based interest rate sensitivity
  4. Simple linear delta exposure alone

Explanation

Options have convex payoffs, so delta alone understates risk for large moves. Full revaluation or gamma-adjusted analysis captures the curvature that linear delta misses.

Question 12OSMR (domain names not published in source)

Which statement about the difference between risk measurement and risk management is most accurate in an OSMR context?

  1. Measurement is only relevant for regulatory reporting, not for management decisions
  2. Management quantifies exposure, while measurement sets limits
  3. Measurement and management are synonyms and used interchangeably
  4. Measurement quantifies exposure, while management decides how to respond to that exposure Correct answer.

Explanation

Risk measurement produces quantitative estimates such as VaR or stress losses. Risk management uses those estimates, alongside judgement, to set limits, hedge, and allocate capital.

Question 13OSMR (domain names not published in source)

A firm sets a risk limit at 95% of its risk appetite. What is the main benefit of this buffer?

  1. It provides headroom so normal fluctuations do not immediately breach the appetite Correct answer.
  2. It guarantees the firm cannot exceed its risk appetite
  3. It removes the requirement to report to the board
  4. It eliminates the need for stress testing

Explanation

Setting operating limits below the overall appetite creates a cushion, so routine volatility does not trigger breaches. The buffer also gives management time to act before appetite is truly exhausted.

Question 14OSMR (domain names not published in source)

In an economic capital framework, what does economic capital represent?

  1. The capital a firm estimates it needs to absorb unexpected losses at a chosen confidence level Correct answer.
  2. The expected loss already priced into transactions
  3. The accounting equity reported on the balance sheet
  4. The capital required solely by the regulator's minimum ratio

Explanation

Economic capital is an internal estimate of the buffer needed to survive unexpected losses at a target solvency standard. It differs from regulatory capital, which follows prescribed rules.

Question 15OSMR (domain names not published in source)

A risk analyst is aggregating risk across market, credit, and operational categories. Why is simple summation of standalone capital figures often inappropriate?

  1. It double-counts diversification benefits and understates capital
  2. Operational risk cannot be quantified in capital terms
  3. It ignores diversification and can overstate the capital needed Correct answer.
  4. Regulatory rules forbid aggregating different risk types

Explanation

Risks across categories are imperfectly correlated, so the combined capital is generally less than the sum of standalone amounts. Simple summation ignores these diversification effects and overstates required capital.

Question 16OSMR (domain names not published in source)

Which of the following best describes operational risk in an OSMR framework?

  1. The risk that a hedge becomes illiquid in a crisis
  2. The risk of loss from inadequate or failed internal processes, people, systems, or external events Correct answer.
  3. The risk that interest rates move against a bond position
  4. The risk that a borrower fails to repay a loan

Explanation

Operational risk covers losses from process failures, human error, system breakdowns, and external events such as fraud. It is distinct from market and credit risk, which concern price moves and borrower default.

Question 17OSMR (domain names not published in source)

A bank uses a loss distribution approach for operational risk. Which two distributions are typically combined?

  1. A credit migration matrix and a default probability
  2. A yield curve and a spread curve
  3. A normal and a lognormal price distribution
  4. A frequency distribution and a severity distribution Correct answer.

Explanation

The loss distribution approach models the number of loss events with a frequency distribution and the size of each event with a severity distribution. Convolving them yields an aggregate loss distribution for capital estimation.

Question 18OSMR (domain names not published in source)

Which practice most directly supports the 'use test' of an internal risk model?

  1. Documenting the model's equations in a technical manual only
  2. Embedding model outputs into daily limit setting, pricing, and capital allocation decisions Correct answer.
  3. Submitting the model to the regulator once a year
  4. Periodically recalculating the model's parameters without changing any decisions

Explanation

The use test requires that a model genuinely informs business decisions, not merely that it exists. Integrating outputs into limits, pricing, and capital allocation demonstrates real use.

Question 19OSMR (domain names not published in source)

In credit risk, what does Probability of Default represent?

  1. The total exposure at the time of default
  2. The likelihood that an obligor will fail to meet its obligations over a given horizon Correct answer.
  3. The fraction of exposure lost if default occurs
  4. The interest spread charged to compensate for risk

Explanation

Probability of Default is the estimated likelihood of obligor default over a specified horizon, typically one year. It is distinct from loss given default and exposure at default, which together drive expected loss.

Question 20OSMR (domain names not published in source)

Using the standard expected loss formula, what is the expected loss on an exposure of 10 million with a 2% probability of default and a 45% loss given default?

  1. 90,000 Correct answer.
  2. 4,500,000
  3. 450,000
  4. 200,000

Explanation

Expected loss equals exposure at default multiplied by probability of default and loss given default. Here 10,000,000 x 0.02 x 0.45 equals 90,000.

Question 21OSMR (domain names not published in source)

A liquidity risk analyst distinguishes funding liquidity from market liquidity. Which pair correctly matches the terms?

  1. Funding liquidity is measured only by bid-ask spread; market liquidity is measured only by credit rating
  2. Funding liquidity is the risk of default by a borrower; market liquidity is the risk of interest rate moves
  3. Funding liquidity is the ability to trade assets quickly; market liquidity is the ability to borrow cash
  4. Funding liquidity is the ability to meet obligations as they fall due; market liquidity is the ability to trade assets without large price impact Correct answer.

Explanation

Funding liquidity concerns a firm's capacity to raise cash and settle obligations on time. Market liquidity concerns how easily assets can be traded without moving prices significantly.

Question 22OSMR (domain names not published in source)

Which feature most distinguishes a coherent risk measure from an arbitrary one?

  1. It relies exclusively on Monte Carlo simulation
  2. It always produces a smaller number than Value-at-Risk
  3. It satisfies a set of axioms including monotonicity, translation invariance, positive homogeneity, and subadditivity Correct answer.
  4. It can be computed without any historical data

Explanation

Coherence is defined by four axioms: monotonicity, translation invariance, positive homogeneity, and subadditivity. Measures satisfying all four behave sensibly under diversification and aggregation.

Question 23OSMR (domain names not published in source)

An analyst applies a square-root-of-time rule to scale a one-day VaR to a ten-day VaR. Which assumption is critical for this scaling to be valid?

  1. The portfolio contains only options
  2. The confidence level is below 95%
  3. Returns are independent and identically distributed across days Correct answer.
  4. Returns are strongly negatively autocorrelated

Explanation

The square-root-of-time rule relies on independent, identically distributed returns so that variance grows linearly with the horizon. Autocorrelation or volatility clustering breaks this assumption and can misstate longer-horizon risk.

Question 24OSMR (domain names not published in source)

A risk committee observes that its VaR estimate is stable while stress losses rise sharply. What is the most reasonable interpretation?

  1. Both measures are unreliable and should be discarded
  2. Current volatility is low but tail risk has increased, so stress results deserve attention Correct answer.
  3. The VaR model must be correct and the stress test must be wrong
  4. Stress testing is redundant whenever VaR is stable

Explanation

VaR reflects recent volatility and can appear calm while tail conditions deteriorate. Stress tests reveal vulnerabilities that a volatility-driven measure may mask, so rising stress losses warrant scrutiny even with stable VaR.

Question 25OSMR (domain names not published in source)

In a Monte Carlo VaR engine, why might an analyst increase the number of simulation paths?

  1. To guarantee the model captures every possible market scenario
  2. To lower the confidence level of the estimate
  3. To reduce sampling error in the estimated percentile of the loss distribution Correct answer.
  4. To eliminate the need for a random number generator

Explanation

Monte Carlo VaR estimates a percentile from simulated losses, which is subject to sampling variability. More paths reduce this sampling error and stabilize the estimate, though they do not remove model risk.

Question 26OSMR (domain names not published in source)

A risk manager needs a single number that answers 'how much could this portfolio lose in one trading day with 99% confidence?' Which measure directly answers that question?

  1. One-day 99% Value at Risk Correct answer.
  2. One-day 99% expected shortfall
  3. Annualized portfolio volatility
  4. Maximum drawdown over the last year

Explanation

VaR is defined as a quantile of the loss distribution: the loss level that is expected to be exceeded only 1% of the time over the stated horizon. It answers the 'how much at a given confidence' question directly. Expected shortfall answers a different question, namely the average loss conditional on exceeding that quantile.

Question 27OSMR (domain names not published in source)

A risk analyst must state the confidence level and holding period alongside every VaR figure reported to the board. Why is this convention essential?

  1. Because regulators require the same horizon for all trading books
  2. Because confidence level determines the accounting valuation of positions
  3. Because VaR must always be reported on a ten-day horizon
  4. Because a VaR number is meaningless without its quantile and horizon Correct answer.

Explanation

VaR is a quantile tied to a specific probability and a specific time horizon, so the same portfolio can have very different VaR figures at 95% versus 99% and at one day versus ten days. Reporting the number without those parameters makes it impossible to interpret or compare. The convention is about interpretability, not about a universal regulatory horizon.

Question 28OSMR (domain names not published in source)

In a historical simulation VaR model, the simulated profit-and-loss distribution is built by doing which of the following?

  1. Revaluing the current portfolio under past observed market moves Correct answer.
  2. Bootstrapping future cash flows from internal default data
  3. Averaging the worst five losses over the past decade
  4. Drawing returns from a fitted normal distribution

Explanation

Historical simulation takes the portfolio as it stands today and revalues it using each historical set of market factor changes, producing an empirical P&L distribution. No distributional assumption is imposed on returns, which is the method's defining feature. The quantile of that empirical distribution then gives the VaR estimate.

Question 29OSMR (domain names not published in source)

A risk officer is asked to define risk appetite for the firm's trading activities. Which statement best captures what a risk appetite statement does?

  1. It specifies the amount and type of risk the firm is willing to accept in pursuit of its objectives Correct answer.
  2. It lists the individual trades authorized for each desk
  3. It fixes the exact VaR number that must never be exceeded on any day
  4. It records the losses incurred during the previous financial year

Explanation

Risk appetite expresses, at a high level, how much and what kinds of risk the firm chooses to take in order to pursue its strategy. Limits and tolerances are then derived from that statement and cascaded to desks. It is a forward-looking statement of willingness to bear risk, not a transaction list or a historical record.

Question 30OSMR (domain names not published in source)

An operational risk team classifies a loss event caused by a failed internal system upgrade that disrupted client settlements. Which Basel event type does this most directly correspond to?

  1. Damage to physical assets
  2. Clients, products and business practices
  3. Execution, delivery and process management Correct answer.
  4. External fraud

Explanation

Failures in transaction processing and settlement, including system-related processing errors, fall under execution, delivery and process management. This category covers failed transactions, model or system failures in processing, and documentation errors. The other categories address deliberate misconduct, physical damage, or external criminal acts.

Question 31OSMR (domain names not published in source)

Which input is required to convert a probability of default and a loss given default into an expected loss figure for a credit exposure?

  1. The recovery period in days
  2. The risk-free discount curve
  3. The obligor's equity beta
  4. The exposure at default Correct answer.

Explanation

Expected loss is the product of probability of default, loss given default, and exposure at default. Without the exposure amount the other two components cannot be turned into a currency loss figure. The discount curve, equity beta, and recovery timing are not part of the standard three-factor formula.

Question 32OSMR (domain names not published in source)

A model risk policy requires independent validation of every internal model used for risk decisions. What is the primary objective of that independent validation?

  1. To guarantee that the model will never produce an unexpected loss
  2. To replace the model owner as the party responsible for model maintenance
  3. To confirm the model is conceptually sound and performs as intended before and during use Correct answer.
  4. To transfer legal liability for model errors to the validation team

Explanation

Independent validation assesses conceptual soundness, data quality, implementation accuracy, and ongoing performance, giving senior management assurance that the model is fit for its purpose. It supplements but does not remove the model owner's accountability. No validation can guarantee that a model will never produce a loss.

Question 33OSMR (domain names not published in source)

A liquidity desk distinguishes between the ability to sell an asset quickly at close to fair value and the ability to raise cash to meet obligations. Which pair of terms matches these two concepts?

  1. Solvency and liquidity
  2. Market liquidity and funding liquidity Correct answer.
  3. Collateral haircut and margin call
  4. Interest rate risk and credit spread risk

Explanation

Market liquidity refers to how easily an asset can be traded without moving its price, while funding liquidity refers to a firm's ability to meet cash and collateral obligations as they fall due. The two interact, since illiquid assets can impair a firm's funding position. The other pairs describe different risk concepts altogether.

Question 34OSMR (domain names not published in source)

A bank's internal model output is used to set trading limits, price new products, and inform capital allocation. Which term describes this ongoing business reliance on the model?

  1. Use test Correct answer.
  2. Backtesting
  3. Stress testing
  4. Benchmarking

Explanation

The use test asks whether a model is genuinely embedded in the firm's decision-making, such as limit setting, pricing, and capital allocation, rather than built only to satisfy a reporting requirement. Backtesting and benchmarking are validation techniques, not evidence of business integration. Stress testing is a separate risk analysis tool.

Question 35OSMR (domain names not published in source)

A portfolio has a one-day 99% VaR of 4 million. Using the square-root-of-time scaling rule, what is the approximate ten-day 99% VaR?

  1. 12.6 million Correct answer.
  2. 40.0 million
  3. 8.0 million
  4. 1.3 million

Explanation

The square-root-of-time rule multiplies the one-day VaR by the square root of the horizon in days, so 4 million times the square root of 10, about 3.162, gives roughly 12.6 million. The rule assumes independent and identically distributed daily returns. Multiplying linearly by ten would overstate the scaled figure under that assumption.

Question 36OSMR (domain names not published in source)

A desk reports a 95% one-day VaR of 2 million on a portfolio whose daily P&L is roughly symmetric. How should this figure be interpreted?

  1. Losses are expected to exceed 2 million on about one day in twenty Correct answer.
  2. The maximum possible loss on any single day is 2 million
  3. The portfolio will lose exactly 2 million on 5% of days
  4. There is a 95% chance the portfolio will be profitable tomorrow

Explanation

A 95% VaR means that in about 5% of days, or one day in twenty, losses are expected to be larger than the reported figure. It is a threshold on the loss distribution, not a worst-case bound. The size of losses beyond the threshold is not specified by VaR itself.

Question 37OSMR (domain names not published in source)

A 99% one-day VaR model is backtested over 500 trading days and records 3 exceptions. Which conclusion is most defensible?

  1. The model is overestimating risk and should be tightened to generate more exceptions
  2. The number of exceptions is below the expected five, so there is no evidence of underestimation Correct answer.
  3. The model is clearly underestimating risk and must be recalibrated immediately
  4. Backtesting cannot be performed with only 500 observations

Explanation

At 99% over 500 days the expected number of exceptions is 1% of 500, or 5. Observing 3 is fewer than expected, so the test provides no evidence that the model understates risk. A formal statistical test such as a binomial or Kupiec test would confirm that this deviation is within normal sampling variation.

Question 38OSMR (domain names not published in source)

A risk analyst must choose a VaR method for a book of options whose payoffs are highly nonlinear. Which method is most appropriate?

  1. Variance-covariance with delta-normal approximation
  2. Duration-based approximation
  3. Full revaluation Monte Carlo simulation Correct answer.
  4. Simple historical simulation using only linear factor sensitivities

Explanation

Full revaluation Monte Carlo simulation reprices each instrument under many simulated scenarios, capturing the convexity and nonlinear payoff profiles of options. Delta-normal and duration approximations are linearizations that misstate risk for large moves in nonlinear positions. The choice is driven by the need to reprice rather than approximate.

Question 39OSMR (domain names not published in source)

A bank wants to allocate economic capital across business units so that the total equals the firm's diversified risk. Why is summing standalone unit capital inappropriate?

  1. Because business units cannot be assigned individual risk measures
  2. Because diversification always increases total required capital
  3. Because standalone capital ignores diversification benefits between units Correct answer.
  4. Because economic capital must always equal regulatory capital

Explanation

Summing standalone capital ignores that losses across units are imperfectly correlated, so the firm-wide risk is generally less than the sum of parts. Economic capital allocation should reflect that diversification benefit, for example through marginal or incremental contributions. Simple summation would overstate the capital needed for the diversified firm.

Question 40OSMR (domain names not published in source)

A risk committee wants a coherent risk measure that accounts for the severity of losses beyond the VaR threshold. Which measure satisfies this requirement and is subadditive?

  1. Expected shortfall Correct answer.
  2. Value at Risk at 99%
  3. Beta of the portfolio
  4. Standard deviation

Explanation

Expected shortfall is the average loss conditional on exceeding the VaR quantile, so it uses information about tail severity and satisfies subadditivity, making it coherent. VaR can violate subadditivity and ignores the size of losses beyond the threshold. Standard deviation and beta are dispersion or sensitivity measures rather than coherent tail risk measures.

Question 41OSMR (domain names not published in source)

A model validation team finds that a VaR model's exceptions cluster in periods of rising volatility rather than being spread evenly. What does this pattern most likely indicate?

  1. The model's volatility updating is too slow to reflect current market conditions Correct answer.
  2. The exceptions are random and require no investigation
  3. The confidence level has been set too high for the horizon
  4. The historical sample used for estimation is too short to compute a quantile

Explanation

Clustering of exceptions during volatility spikes suggests the model reacts too slowly, for example because it relies on long moving-average windows or stale volatility estimates. A well-calibrated model should produce exceptions roughly independently over time. Clustering is evidence of conditional bias, not merely random sampling noise.

Question 42OSMR (domain names not published in source)

A firm's risk appetite allows a maximum annual loss tolerance of 100 million at a stated confidence. A desk proposes a limit set at 80 million. What is the main purpose of this gap?

  1. To guarantee the desk never records a loss
  2. To eliminate the need for stress testing at the desk level
  3. To leave room for unexpected losses and limit breaches before appetite is breached Correct answer.
  4. To ensure the desk can exceed appetite without escalation

Explanation

Setting operational limits below the overall appetite creates a buffer so that normal volatility or a single limit breach does not immediately breach the firm's stated tolerance. It also provides escalation time for management action. The buffer is a governance mechanism, not a guarantee against loss.

Question 43OSMR (domain names not published in source)

In a loss distribution approach to operational risk, which two distributions are typically combined?

  1. Market volume and bid-ask spread
  2. Interest rate and credit spread
  3. Frequency of loss events and severity of individual losses Correct answer.
  4. Default probability and recovery rate

Explanation

The loss distribution approach models the number of loss events with a frequency distribution and the size of each event with a severity distribution, then convolves them to obtain an aggregate annual loss distribution. Capital is read from a high quantile of that aggregate distribution. The other pairs belong to credit or market risk modeling.

Question 44OSMR (domain names not published in source)

A bank is aggregating market, credit, and operational risk capital. Which approach best reflects diversification across risk types?

  1. Taking the largest of the three standalone figures
  2. Averaging the three standalone figures
  3. Simple arithmetic summation of the three standalone capital figures
  4. Applying a correlation matrix or copula to combine the risk distributions Correct answer.

Explanation

Combining risk distributions with a correlation structure or copula allows imperfect dependence between risk types to be captured, so the aggregate is below the simple sum. Simple summation implicitly assumes perfect correlation and overstates diversified capital. Taking the maximum or averaging ignores the statistical structure of the risks entirely.

Question 45OSMR (domain names not published in source)

A stress test scenario applies a 200 basis point parallel shift to the yield curve. Which limitation should the risk committee keep in mind when interpreting the result?

  1. It captures only one specific scenario and not the full range of plausible outcomes Correct answer.
  2. It is invalid because parallel shifts never occur in practice
  3. It overstates risk because interest rates cannot rise that much
  4. It automatically incorporates liquidity and correlation effects

Explanation

A single deterministic scenario shows the impact of one hypothesized event and says nothing about the probability or the range of other plausible outcomes. Stress testing is therefore used alongside probabilistic measures such as VaR. The scenario's usefulness lies in exploring severe but plausible conditions, not in forecasting.

Question 46OSMR (domain names not published in source)

A risk analyst scales a one-day 99% VaR to ten days using the square-root-of-time rule. Which assumption is most critical to the validity of that scaling?

  1. Returns are negatively skewed
  2. Daily returns are independently and identically distributed Correct answer.
  3. The portfolio contains only linear instruments
  4. The confidence level is exactly 99%

Explanation

The square-root-of-time rule follows from the variance of a sum of independent, identically distributed returns growing linearly with the horizon. If returns are serially correlated or volatility is time-varying, the scaling misstates the multi-day risk. Skewness, instrument linearity, and the exact confidence level are not the assumptions underlying the rule.

Question 47OSMR (domain names not published in source)

A bank's credit portfolio model produces a loss distribution for the coming year. Which point of that distribution is typically used to determine economic capital for credit risk?

  1. A high quantile, such as the 99.9th percentile, minus expected loss Correct answer.
  2. The minimum observed loss in the historical sample
  3. The mean of the loss distribution
  4. The median of the loss distribution

Explanation

Economic capital for credit risk is usually set as an extreme quantile of the loss distribution, often the 99.9th percentile, less the expected loss already covered by pricing and provisions. This captures unexpected loss at a target solvency standard. The mean corresponds to expected loss, and the median or historical minimum do not represent severe tail outcomes.

Question 48OSMR (domain names not published in source)

A bank's VaR model shows a stable estimate while stress losses rise sharply over the same period. What is the most reasonable interpretation for the risk committee?

  1. The stress results must be wrong because VaR is stable
  2. The stable VaR proves the portfolio is well hedged in all scenarios
  3. Both measures should be ignored until they agree
  4. The VaR model is probably not capturing the buildup of tail risk and the stress results deserve attention Correct answer.

Explanation

VaR estimated over a historical window can remain stable while exposures to severe but rare scenarios build up, because recent ordinary volatility dominates the quantile. Divergence between a stable VaR and rising stress losses is a classic warning sign that tail risk is increasing. The committee should investigate exposures and possibly tighten limits rather than dismiss either measure.

Question 49OSMR (domain names not published in source)

A risk team is selecting a copula to combine marginal loss distributions for credit and market risk. Which property most distinguishes a Gaussian copula from a Student-t copula in tail dependence?

  1. The Gaussian copula exhibits positive tail dependence, while the Student-t copula exhibits zero tail dependence
  2. Both copulas exhibit identical tail dependence but differ only in their means
  3. The Student-t copula cannot be applied to more than two risk types
  4. The Gaussian copula exhibits zero tail dependence, while the Student-t copula exhibits positive tail dependence Correct answer.

Explanation

The Gaussian copula has no tail dependence, meaning extreme joint losses become asymptotically independent, while the Student-t copula assigns positive probability to joint extreme events in both tails. This matters when aggregating risks whose severe losses tend to coincide. Choosing the copula therefore materially changes the aggregate tail capital estimate.

Question 50OSMR (domain names not published in source)

A validation team observes that a VaR model passes a standard unconditional coverage test but fails an independence test. What does this combination imply?

  1. The model produces too many exceptions overall and must be recalibrated
  2. The independence test is irrelevant when the coverage test passes
  3. The overall exception rate is acceptable, but exceptions cluster in time, indicating the model does not adapt to changing volatility Correct answer.
  4. The model is correctly specified because the unconditional test passed

Explanation

Unconditional coverage tests only whether the total number of exceptions matches the expected rate, while independence tests examine whether exceptions are spread randomly over time. Passing the first but failing the second means the average rate is right but the model reacts poorly to volatility changes, for example clustering exceptions in turbulent periods. Both dimensions matter for a well-calibrated model.