Skip to content

FRM Part II · FRM Exam Part II

Parametric Approaches (II): Extreme Value Theory for FRM Part 2

Extreme value theory (EVT) models the tails of a loss distribution instead of the whole distribution. You use block maxima with the GEV distribution, or losses above a threshold with the Generalized Pareto Distribution, then estimate tail VaR and expected shortfall. Solve questions by identifying the method, the tail index, and the formula.

What this chapter covers

This chapter is about modelling rare, large losses. Normal-based VaR often understates tail risk because real returns have fat tails. EVT fixes this by fitting a distribution only to the extreme observations. You study two routes. The first is block maxima, where you take the largest loss in each period and fit a GEV distribution. The second is peaks over threshold (POT), where you take all losses above a high threshold and fit a Generalized Pareto Distribution (GPD).

The key parameter in both routes is the tail index (shape parameter, often written ξ). A larger positive ξ means a heavier tail. A ξ of zero gives an exponential-type tail, and a negative ξ gives a bounded tail. Most financial loss data show a positive ξ. You then use the fitted tail to get high-confidence VaR and expected shortfall, and you learn the limits of the method: threshold choice, small samples and dependence between assets.

This chapter connects to the rest of Market Risk Measurement and Management. It extends the VaR and expected shortfall ideas you already know, and it pairs with the other parametric and non-parametric approaches. It also supports stress testing, operational risk loss modelling and credit tail risk, where extreme losses drive capital. Expect applied questions that ask you to pick the right method and interpret the result, not only to compute.

EVT questions test whether you understand why normal VaR fails in the tail and how to fix it. They are often conceptual with light calculation, so they are good marks if you know the vocabulary precisely: block maxima versus POT, tail index, threshold, bias versus variance trade-off. The ideas also reappear in operational risk, stress testing and Basel capital discussions, so the effort pays off across the paper. Because the formulas are short, you can learn them once and apply them fast under time pressure.

Parametric Approaches (II): Extreme Value: topics in the order to study them

  1. 1Extreme Value Theory Basics and MotivationStart here to see why tails need their own model and what the tail index means before any formulas.
  2. 2Generalized Extreme Value (GEV) DistributionIt introduces block maxima and the three tail types (Fréchet, Gumbel, Weibull) that the later topics build on.
  3. 3Peaks Over Threshold and Generalized Pareto DistributionPOT is the more data-efficient method and the one used for VaR and ES, so learn it after the GEV logic is clear.
  4. 4EVT-Based VaR and Expected Shortfall CalculationThis applies the fitted GPD parameters to produce risk numbers, so you need the GPD first.
  5. 5Multivariate EVT and Practical IssuesSave this for last: it covers dependence in the tails and the limits of EVT, which make sense once the single-asset method is solid.

How to prepare Parametric Approaches (II): Extreme Value

Treat this chapter as one idea with two methods. Learn the logic first, then the formulas, then practise choosing the method.

  1. Write one sentence on why normal VaR understates tail risk, and one on what EVT does differently.
  2. Learn the two routes side by side: block maxima gives GEV, threshold exceedances give GPD. Make a small comparison note of data used, distribution fitted and main weakness.
  3. Memorise what the tail index ξ means: ξ > 0 heavy tail, ξ = 0 exponential-type tail, ξ < 0 bounded tail. Practise reading this from a stated value.
  4. Learn the EVT VaR and expected shortfall formulas from your GARP reading and practise them with simple numbers until the steps are automatic. Check that ES comes out larger than VaR.
  5. Study the threshold trade-off: a low threshold adds bias, a high threshold leaves too few points and raises variance. Be ready to explain this in words.
  6. Read the multivariate section for concepts only: tail dependence and why extremes tend to occur together. Then do timed practice questions and review every wrong answer.

Common mistakes in Parametric Approaches (II): Extreme Value

  • Mixing up block maxima and POT

    Fix: Remember: block maxima takes one extreme per block and gives GEV; POT takes every exceedance over a threshold and gives GPD.

  • Misreading the sign of the tail index

    Fix: A larger positive ξ means a heavier tail and more extreme losses. Check the sign before you interpret.

  • Choosing the threshold by a one-sided rule

    Fix: Explain the trade-off: a higher threshold cuts bias but leaves fewer observations, so variance rises. A lower one does the reverse.

  • Treating EVT estimates as precise

    Fix: Note that estimates rely on few data points and extrapolate beyond the sample, so they are sensitive to parameters and threshold.

  • Ignoring dependence across assets

    Fix: Remember that tail events can cluster across assets, so combining separate univariate fits can understate portfolio tail risk.

  • Reporting VaR when the question asks for expected shortfall

    Fix: Underline the requested measure first and check that ES is above VaR at the same confidence level.

Last-day revision: Parametric Approaches (II): Extreme Value

  • EVT models only the tail, not the whole distribution.
  • Block maxima uses the largest loss per period and fits a GEV distribution.
  • POT uses all losses above a high threshold and fits a GPD.
  • GEV has three types: Fréchet (heavy tail), Gumbel (light tail), Weibull (bounded tail).
  • Tail index ξ > 0 means a fat, heavy tail; most financial losses show this.
  • POT usually uses data more efficiently than block maxima.
  • Threshold choice: too low gives bias, too high gives high variance.
  • Expected shortfall is larger than VaR at the same confidence level.
  • Normal-based VaR tends to understate losses far into the tail.
  • Estimates beyond the observed data are extrapolations and carry high uncertainty.
  • Extremes in different assets often occur together, so tail dependence matters.
  • Always state the method, the tail index and the interpretation in your reasoning.

Parametric Approaches (II): Extreme Value practice questions

Parametric Approaches (II): Extreme Value in other exams

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

Parametric Approaches (II): Extreme Value: frequently asked questions

Is extreme value theory heavy on calculation in FRM Part II?

Mostly no. Expect questions on method choice, tail index interpretation and threshold trade-offs, with some short calculations using given parameters. Know the formulas, but spend equal time on the concepts.

What is the difference between GEV and GPD?

GEV models the distribution of block maxima, such as the worst loss each month. GPD models the size of losses above a high threshold. Both describe tail behaviour and share the same tail index idea.

Why not just use normal VaR?

Financial returns usually have fatter tails than the normal distribution. Normal VaR can therefore understate the chance and size of extreme losses, especially at very high confidence levels.

How should I study this chapter in the final week?

Revise the comparison of block maxima and POT, the meaning of the tail index and the threshold trade-off. Then do a few timed questions on VaR and expected shortfall and review your errors.