A mid-sized Sydney prop desk ran parametric VaR daily. Their model assumed normally distributed returns. In August 2015, the Chinese market shock produced return distributions with fat tails the model never anticipated. Their reported 1-day 99% VaR was $380,000. Actual loss: $2.1 million. The method wasn't wrong — it was wrong for that portfolio.

Value at Risk quantifies the maximum expected loss over a defined period at a given confidence level. Three dominant methodologies exist: historical simulation, parametric (variance-covariance), and Monte Carlo simulation. Each carries structural assumptions that make it powerful in specific contexts and dangerous in others. Selecting blindly is how risk managers end up explaining themselves to boards.

CONCEPTVaR method selection is a structural decision — change the method, change the number, change the risk profile entirely.
WARNINGParametric VaR systematically underestimates tail risk in portfolios with options, skewed assets, or crisis-regime correlations.
KEY IDEANo single VaR method dominates across all portfolio types — infrastructure, asset class, and horizon determine fit.

Historical simulation replays actual past returns against today's portfolio. No distribution assumptions. No correlation matrix. If your lookback window is 500 days, you get 500 P&L scenarios drawn from reality. The weakness is obvious: it can only surface risks that already happened. A volatility regime absent from your sample window simply does not exist in your model.

VaR Estimate by Method ($000s)0200400600380560640450720800ParametricHistoricalMonte CarloNormalParametricNormal→StressHistoricalMonte CarloStressStress Regime →

Parametric VaR is computationally elegant: multiply portfolio volatility by a z-score (2.33 for 99% confidence), scaled by position size. A $10M book with 1.2% daily volatility produces a 1-day 99% VaR of $279,600. It runs in milliseconds. But it assumes normality and static correlations — both assumptions that collapse precisely when you need the model most.

Monte Carlo generates thousands of synthetic return paths using specified statistical processes. It handles options Greeks, non-linear payoffs, and custom volatility surfaces. A well-constructed Monte Carlo with 10,000 simulations on an options book will surface tail scenarios parametric VaR literally cannot compute. The cost is model risk — garbage parameters produce confident-looking garbage numbers. Calibration discipline is non-negotiable.

Practitioners typically layer all three: parametric for real-time intraday limits (speed matters), historical simulation for end-of-day regulatory reporting (Basel III prefers it), and Monte Carlo for structured products and stress testing. Backtesting each method against actual P&L — a process well-documented at Investopedia's VaR reference — reveals which model is systematically over or underestimating. The theoretical underpinnings of all three trace back to Value at Risk on Wikipedia, while the original RiskMetrics framework that formalised parametric VaR is detailed under RiskMetrics on Wikipedia.

The method that fits your infrastructure is the one you can build, calibrate, backtest, and explain under pressure. A VaR number nobody trusts is just a number.

This content is for educational purposes only and does not constitute financial product advice. Past performance is not indicative of future results. Profit Logic Ltd (ACN 688 669 936) accepts no responsibility for errors or omissions in this content or anywhere on this website. Always seek advice from a licensed financial adviser before making investment decisions.