Skip to content

FRM Part II · FRM Exam Part II

Estimating Market Risk Measures: Overview for FRM Part II

This chapter teaches you how to estimate and judge market risk measures: VaR, expected shortfall and spectral risk measures. You compute them from historical data or from a normal or lognormal model, attach standard errors, and check distributional fit with QQ plots. Solve questions by naming the measure, applying the method, then interpreting the result.

What this chapter covers

This chapter is the foundation for market risk in FRM Part II. It starts with Value at Risk (VaR): a loss level that is exceeded with a stated small probability over a stated holding period. You then see how to estimate it, either from past data (historical simulation) or from an assumed distribution (parametric VaR with normal or lognormal returns).

The chapter then moves past VaR. VaR says nothing about losses beyond the cutoff and is not always subadditive. Expected shortfall (ES) averages the losses in the tail, and it is a coherent risk measure. Spectral risk measures generalise this: they take a weighted average of loss quantiles, and the weights reflect risk aversion. Coherence needs weights that do not decrease as losses get worse.

Finally, you learn that every estimate is uncertain. You use standard errors and confidence intervals to see how reliable a number is, and QQ plots to check whether the assumed distribution fits the data. These ideas feed the rest of the paper: later chapters on volatility, correlation, extreme values, stress testing and model risk all use VaR, ES and estimation error. The topics on credit and liquidity risk also borrow the tail-loss language.

Questions on this chapter are usually short and numerical, so they are among the easiest marks in the market risk section if your method is clean. VaR and ES calculations, coherence checks and QQ plot readings are favourites because each has one clear answer. The ideas also recur in later chapters, so a weak base here costs you marks elsewhere. With 80 equally weighted questions in 4 hours, speed on routine calculations leaves more time for long case-style items.

Estimating Market Risk Measures: An Introduction and Overview: topics in the order to study them

  1. 1Value at Risk (VaR) Basics and ParametersStart here: confidence level, holding period and the meaning of a quantile loss are used by every later topic.
  2. 2Historical Simulation VaRIt is the simplest estimate, needing only sorted past returns, and it shows what a quantile means in practice.
  3. 3Parametric VaR: Normal and LognormalOnce you know the data-based approach, learn the model-based one and compare the assumptions.
  4. 4Expected Shortfall and Coherent Risk MeasuresIt builds on VaR by averaging the tail, and it introduces the coherence properties VaR can fail.
  5. 5Spectral Risk Measures and Risk-Aversion WeightsIt generalises ES as a weighted quantile average, so ES must be clear first.
  6. 6Estimating Standard Errors and Confidence Intervals for Risk MeasuresWith the measures understood, you can learn how noisy each estimate is.
  7. 7QQ Plots and Assessing Distributional FitIt ties together the parametric assumptions and the data, so it comes last as a check on everything before it.

How to prepare Estimating Market Risk Measures: An Introduction and Overview

Aim to master the method for each measure, not memorise numbers. Practise until you can state the measure, compute it and interpret it in one pass.

  1. Define VaR in words first: loss, confidence level, holding period. Practise converting between one-day and multi-day VaR, noting the square-root-of-time scaling holds only under stated assumptions such as independent returns with constant volatility.
  2. Do historical simulation by hand: sort the returns, find the cutoff observation for the confidence level, and state the loss. Pay attention to how many observations sit in the tail.
  3. Memorise the parametric normal form: VaR = −μ + σ × z, where z is the standard normal quantile for the confidence level (1.645 at 95%, 2.326 at 99%). Work several examples with and without a mean return.
  4. Learn ES as the average loss beyond VaR. Then list the four coherence properties (monotonicity, subadditivity, positive homogeneity, translation invariance) and know which measure fails which.
  5. Write the spectral measure as a weighted average of quantiles. Practise deciding whether a given set of weights is coherent and what it says about risk aversion.
  6. Practise reading standard errors, confidence intervals and QQ plots. Know how a fat-tailed or skewed sample looks against a normal reference line.
  7. Finish with timed mixed sets. Check each answer for sign, units and the right quantile before moving on.

Common mistakes in Estimating Market Risk Measures: An Introduction and Overview

  • Using the wrong tail or quantile, such as 1.645 for a 99% VaR.

    Fix: Write the confidence level and its one-tailed z-value before you calculate. Remember 95% is 1.645 and 99% is 2.326.

  • Forgetting the mean return in parametric VaR or dropping its sign.

    Fix: Always write VaR = −μ + σ × z first, then check whether the question gives a mean or says to ignore it.

  • Scaling VaR by the square root of time without checking assumptions.

    Fix: Use it only when returns are independent with constant volatility, or when the question tells you to. State this assumption in your reasoning.

  • Treating ES as a number below or equal to VaR, or saying VaR is coherent.

    Fix: Picture the tail: VaR is the boundary and ES is the average of what lies beyond it. Link subadditivity failures to VaR and coherence to ES.

  • Assuming any weighted quantile average is a coherent spectral measure.

    Fix: Check that the weights are non-negative, sum to one and do not decrease as losses become more severe.

  • Misreading QQ plots by focusing on the middle instead of the tails.

    Fix: Look at where the ends bend away from the reference line. Departures at the ends signal fat tails or skewness, which matter most for risk measures.

Last-day revision: Estimating Market Risk Measures: An Introduction and Overview

  • VaR is the loss level exceeded with probability 1 − confidence level over the chosen holding period.
  • Normal VaR = −μ + σ × z; z is 1.645 at 95% and 2.326 at 99% (one-tailed).
  • Historical simulation VaR is a quantile of sorted past returns; it assumes the past sample represents the future.
  • Lognormal VaR keeps asset values positive and works from log returns.
  • Expected shortfall is the average loss in the tail beyond VaR, so it is at least as large as VaR.
  • A coherent measure satisfies monotonicity, subadditivity, positive homogeneity and translation invariance.
  • VaR is not always subadditive; ES is.
  • A spectral risk measure is a weighted average of quantiles; it is coherent when weights do not fall as losses worsen.
  • Heavier weight on worse losses means greater risk aversion.
  • Standard errors shrink as sample size grows; estimates deeper in the tail are less precise.
  • A QQ plot compares sample quantiles with a reference distribution; a straight line suggests a good fit.
  • S-shaped or curving tails on a QQ plot against the normal point to fat tails or skewness.

Estimating Market Risk Measures: An Introduction and Overview practice questions

Estimating Market Risk Measures: An Introduction and Overview in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

Estimating Market Risk Measures: An Introduction and Overview: frequently asked questions

How should I split my time between VaR and expected shortfall?

Give VaR the early practice because the other topics depend on it. Then spend more time on ES, coherence and spectral measures, since they involve concepts as well as calculations and are easy to confuse.

Do I need to memorise z-values for the exam?

Yes, at least 1.645 for 95% and 2.326 for 99% one-tailed. You may be given tables or values in some questions, but knowing these saves time and helps you spot errors.

Is historical simulation or parametric VaR better?

Neither is always better. Historical simulation makes no distributional assumption but depends on the sample window. Parametric VaR is smooth and easy to scale but is only as good as the assumed distribution, which QQ plots help you test.

Why does VaR fail to be coherent?

VaR can break subadditivity, meaning the VaR of a combined portfolio can exceed the sum of the separate VaRs. It also ignores the size of losses beyond the cutoff. Expected shortfall fixes both problems.

How is this chapter linked to the rest of FRM Part II?

It supplies the vocabulary and tools for the whole market risk section. Later topics refine the inputs, such as volatility and correlation, or test the outputs through stress testing and model risk, using VaR and ES as the base measures.