Study guides / CFA L2: CFA Program Level II / Market risk
Market risk
CFA L2: CFA Program Level II · quick review
Measuring and managing market risk: value at risk by the parametric, historical simulation and Monte Carlo methods, its strengths, limits and extensions, sensitivity and scenario measures, risk constraints, capital allocation, and the measures different institutions use.
The rules
- Value at risk (VaR) is the minimum loss expected over a stated period with a stated probability, under normal market conditions. A 5% one-day VaR of $2 million means a 5% chance of losing at least $2 million in a day. It isn't the worst case.
- Parametric (variance–covariance) VaR assumes normally distributed returns: VaR = (z × σ − expected return) × portfolio value, with z = 1.65 at 5% and 2.33 at 1%. It's simple and fast but understates fat tails and handles options poorly because their payoffs aren't linear.
- Historical simulation applies past returns to today's positions and reads off the loss at the chosen percentile. It needs no distribution assumption and captures fat tails and option payoffs (positions are fully revalued), but it assumes the historical period represents the future.
- Monte Carlo simulation draws many random scenarios from assumed distributions and revalues the portfolio in each. It handles complex, nonlinear portfolios and any distribution, but it's computationally heavy and only as good as its assumptions.
- VaR's strengths: a single number that's easy to grasp, comparable across asset classes and desks, usable for capital allocation and regulation, and testable by backtesting. Limitations: it depends on subjective choices (confidence level, horizon, method), says nothing about the size of losses beyond it, underestimates how often extreme events happen, ignores liquidity and breaks down when correlations jump in a crisis.
- Extensions: conditional VaR (expected shortfall) is the average loss when VaR is exceeded; incremental VaR is the change in VaR from adding, removing or resizing a position; marginal VaR is the change in VaR per small change in a position, showing each position's contribution; relative VaR (ex ante tracking error) measures risk against a benchmark.
- Sensitivity measures: beta for equities; duration and convexity for bonds; delta, gamma and vega for options (vega measures exposure to volatility). They show how value changes for a small move in one factor, but give no probability and miss large or simultaneous moves.
- Scenario measures: historical scenarios replay past crises on today's portfolio; hypothetical scenarios design plausible extreme events; reverse stress tests look for the scenarios that would cause unacceptable losses. They capture extremes and shifting correlations, but history may not repeat and hypothetical scenarios are subjective with no stated probability. Use them alongside VaR and sensitivities.
- Constraints and capital: risk budgeting divides total risk among units or strategies; position limits cap the size of exposures (simple, but ignore correlations); scenario limits cap the loss in a defined scenario; stop-loss limits force a position to be cut after a set loss. Risk measures also guide capital allocation: activities get capital in line with the risk they take and are judged by the return on that capital.
- Users differ: banks watch liquidity gaps, VaR, leverage, sensitivities and economic capital; asset managers watch volatility, the probability of loss and the probability of underperforming a benchmark (tracking error), with hedge funds also watching leverage, drawdowns and scenarios; pension funds measure asset/liability mismatch and surplus at risk; property and casualty insurers use sensitivities and exposures to keep allocations in range, economic capital and VaR to size a catastrophic loss, and scenarios that stress market and insurance risks together; life insurers measure the exposures of the investment portfolio and the annuity liability, the asset/liability mismatch and potential stress losses.
Formulas
- Parametric VaR: VaR = (z × σ_p − E(R_p)) × portfolio value; z = 1.65 at 5%, 2.33 at 1%
- Daily from annual figures: σ_daily = σ_annual ÷ √250; E(R)_daily = E(R)_annual ÷ 250
- Multi-day VaR (mean near zero): VaR(T days) ≈ VaR(1 day) × √T
- Two-asset portfolio risk: σ_p^2 = w1^2 × σ1^2 + w2^2 × σ2^2 + 2 × w1 × w2 × ρ × σ1 × σ2
- Conditional VaR: Average loss across outcomes worse than the VaR
- Equity sensitivity: %ΔP ≈ β × market return
- Bond sensitivity: %ΔP ≈ −ModDur × Δy + ½ × convexity × (Δy)^2
- Option sensitivity: ΔV ≈ delta × ΔS + ½ × gamma × (ΔS)^2 + vega × Δσ
Common traps
- VaR is a minimum loss at the stated probability, not the maximum possible loss.
- Square-root-of-time scaling assumes independent returns and a mean near zero.
- Delta-only parametric VaR understates option risk; gamma and non-normal returns matter.
- Conditional VaR is always at least as large as VaR at the same confidence level.
How it's tested
Vignettes give positions and statistics and ask for a parametric VaR, the best VaR method for a portfolio, the meaning of CVaR or marginal VaR, a sensitivity estimate, or which constraint or measure suits an institution.
Practise it
The full lesson, worked examples and practice questions are in Tickmark, where every answer choice is explained.
Open the full lesson