Normal Distribution
The normal distribution is a continuous probability distribution characterized by its symmetric, bell-shaped curve, fully specified by its mean (μ) and standard deviation (σ). It is the most foundational distribution in statistics and finance, underpinning risk models, options pricing, and hypothesis testing — though financial returns often exhibit fat tails that deviate from normality.
Key takeaways
- The normal distribution is defined by just two parameters: mean (μ) and standard deviation (σ), with 68-95-99.7% of observations within 1, 2, and 3 standard deviations respectively.
- Asset return distributions typically exhibit excess kurtosis (fat tails) and negative skewness, making pure normality assumptions dangerous for tail risk estimation.
- Black-Scholes assumes log-normal asset prices (equivalent to normally distributed log returns), a simplification that underestimates extreme event probabilities.
- Value at Risk models using normal distribution assumptions systematically underestimate loss frequency in the tails.
- Copulas extend the normal distribution framework to capture non-linear dependence structures between assets, particularly in stress scenarios.
Explanation
The normal distribution, also known as the Gaussian distribution, is characterized by its iconic bell shape — symmetric around the mean, with probability mass decaying exponentially as observations move away from the center. The standard normal distribution has μ = 0 and σ = 1, and any normal distribution can be transformed to standard normal form through the z-score transformation: Z = (X − μ) / σ. Critical values for hypothesis testing and confidence intervals are derived from the standard normal: the 95% confidence interval corresponds to ±1.96σ; the 99% confidence interval to ±2.58σ.
In the Black-Scholes options pricing model, log returns are assumed to be normally distributed with constant volatility — meaning asset prices themselves follow a log-normal distribution (log prices are normally distributed). This assumption generates a tractable closed-form formula for European option pricing and is mathematically convenient, but it systematically underprices far out-of-the-money options and options on assets with jump dynamics. The difference between market-implied option prices and Black-Scholes prices is manifest in the 'volatility smile' — where implied volatility increases for options struck far from the money, reflecting the market's recognition that fat tails exist.
The empirical distribution of daily financial returns exhibits three key departures from normality: leptokurtosis (excess kurtosis, meaning fatter tails than normal — more frequent extreme events), negative skewness (asymmetric tails, with crashes more common than equivalent positive moves), and volatility clustering (variance is not constant but time-varying, as captured by GARCH models). These departures mean that normal distribution-based VaR significantly underestimates tail losses. A one-day 99% normal VaR may be breached not once in 100 days on average, but once in 30–50 days for equity portfolios.
The Law of Large Numbers and Central Limit Theorem provide the theoretical justification for the normal distribution's ubiquity. The CLT states that the sum (or average) of a large number of independent, identically distributed random variables — regardless of their individual distributions — approaches a normal distribution as the sample size grows. This underlies portfolio diversification theory: while individual assets may have fat-tailed, non-normal returns, a diversified portfolio's aggregate return distribution converges toward normality.
Copulas provide a sophisticated extension of multivariate normal analysis. A Gaussian copula models joint distributions using normal marginals but specifies the correlation structure independently from the marginals — allowing analysts to model situations where individual asset returns follow non-normal distributions while capturing their multivariate dependence through a correlation matrix. The Gaussian copula's role in the mispricing of CDO tranches during 2005–2007 — where it failed to capture tail dependence between mortgage defaults — is one of the most consequential modeling failures in financial history.
Formula
f(x) = (1 / (σ√(2π))) × exp(−(x−μ)² / (2σ²)); Z-score = (X − μ) / σ
Example
A risk manager models daily P&L for a bond portfolio with an assumed normal distribution: mean daily return of 0.02% and daily standard deviation (volatility) of 0.85%. The 1-day 99% VaR is: μ − 2.326σ = 0.02% − 2.326 × 0.85% = 0.02% − 1.977% = −1.957%. On a $500 million portfolio, this is −$9.79 million. However, the actual empirical distribution of the portfolio's returns exhibits excess kurtosis of 4.2 (versus 3 for normal) and skewness of −0.8. Backtesting over 500 trading days reveals that the 99% VaR threshold was breached 12 times — a 2.4% exceedance rate versus the expected 1.0%. This backtesting failure would trigger regulatory concerns under Basel III market risk rules and prompt the use of a historical simulation VaR model, which does not rely on the normality assumption.
Related terms
Annuity Backtesting Basel Iii Bond Central Limit Theorem Copula Correlation Correlation Matrix Diversification Equity European Option Fat Tails