Charta is on Google Play: practise on the go. Get the app
Charta

Identify what a bank needs to consider as it appraises whether a VaR model is conceptually sound in the course of validating it.

FRM® Part II · Market Risk Measurement and Management · Validating Bank Holding Companies’ Value-at-Risk Models for Market Risk · FRM2.MR.6.a

The idea

Conceptual soundness asks whether the model, as designed, is a defensible way to measure the risk of this particular portfolio, before any output is compared with outcomes. A validator reviews the documentation and the theory. The review covers the methodology (variance-covariance, historical simulation or Monte Carlo) and why it was chosen, the distributional and independence assumptions, the volatility and correlation estimators, and the length and quality of the data window. It asks whether the risk factors cover every material exposure, including spreads, basis and volatility. It checks how nonlinear positions such as options are valued, by full revaluation or by delta or delta-gamma approximations. It also checks the holding period, the confidence level and any square-root-of-time scaling. The usual mistake is treating a clean backtest as proof of soundness, because offsetting errors or a conservative design can hide flaws. For the exam, tie each assumption to the portfolio where it breaks.

A practice question

From the question bank, with the reason each wrong option is wrong.

In validating a bank's VaR model, which statement best describes a review of its conceptual soundness?

  1. It counts the exceptions between daily VaR and realised P&L over the last 250 trading days and tests whether the count is consistent with the stated confidence level.Why not: Describes backtesting, which is outcomes analysis rather than a review of conceptual design.
  2. It assesses whether the model's design, assumptions, risk factor coverage and methods are appropriate for the portfolio and supported by theory and evidence. - correct
  3. It confirms that the production code reproduces the documented formulas and runs without errors.Why not: Describes implementation or process verification, which tests the code rather than whether the design is sound.
  4. It compares the bank's VaR with the VaR reported by peer institutions with similar portfolios.Why not: Describes benchmarking against other models, which is a different validation component.

Conceptual soundness is about the design: methodology, assumptions, data, risk factors and valuation approach, judged against theory and the portfolio. Backtesting, code verification and peer comparison are separate validation activities.

A harder one

A bank scales its 1-day VaR to a 10-day VaR by multiplying by the square root of 10. Which condition is the validator mainly testing when judging this step?

  1. That daily returns are close to independent and identically distributed and that the portfolio does not change materially over the horizon. - correct
  2. That daily returns are perfectly serially correlated, so that risk grows in proportion to time.Why not: Perfect serial correlation would imply scaling by 10, not by the square root of 10.
  3. That the confidence level is at least 99%, so that tail observations are sparse.Why not: The confidence level does not determine the time-scaling rule; the independence of returns over time does.
  4. That returns are exactly normal with zero skewness at every horizon, since the square-root-of-time rule holds only under normality.Why not: Treats normality as necessary; the rule rests on independent, identically distributed returns and fails for options or dynamic positions even if returns are normal.

The square-root-of-time rule follows from variances of independent returns adding across days, with a constant distribution and a static portfolio. Positive autocorrelation, optionality or active rebalancing break it.

The hardest tier

A bank's VaR model has recorded no backtesting exceptions at 99% over two years (about 500 trading days). A colleague argues that this proves the model is conceptually sound. How should the validator respond?

  1. A clean backtest does not establish soundness: zero exceptions is far below the roughly five expected, so assumptions must still be reviewed. - correct
  2. Agree: zero exceptions is the ideal outcome and confirms that the model's assumptions are correct.Why not: Treats fewer exceptions as always better; at 99% about five exceptions are expected in 500 days, so zero suggests overstatement.
  3. Agree in part: conceptual soundness concerns only the programming of the formulas, which backtesting has already verified, so no further review is needed.Why not: Confuses conceptual soundness with implementation checking; backtesting does not verify the model's design.
  4. Disagree: the confidence level should be lowered to 90% until the model records its expected exceptions.Why not: Tinkering with the confidence level to manufacture exceptions does not assess the model's assumptions or purpose.

Expected exceptions are 1% x 500 = 5. Zero is unusually low, and a clean backtest cannot reveal offsetting errors or untested assumptions. Conceptual review judges design, data and assumptions independently of outcomes.

In the app: 6 practice questions on this outcome, spaced review that brings it back before you forget it.

Start free →

More in Validating Bank Holding Companies’ Value-at-Risk Models for Market Risk