CFA Level I: Common Probability Distributions

Probability distributions are the mathematical engine behind modern finance, allowing you to quantify uncertainty in investment returns and model risks with precision. For CFA candidates and finance professionals, a deep understanding of these tools is non-negotiable, as they form the basis for portfolio theory, risk management, and derivative valuation. This knowledge transforms subjective guesswork into objective analysis, directly impacting investment decisions and strategic planning.

Foundational Distributions: Uniform, Binomial, and Normal

Every probability distribution provides a framework for describing the likelihood of different outcomes. The uniform distribution is the simplest, where every outcome in a defined range has an equal probability. In finance, it's often used in basic simulations or when no prior information favors one outcome over another, such as modeling the initial assumption for an asset's return over a very short interval.

The binomial distribution models the number of successes in a fixed number of independent trials, where each trial has only two possible outcomes. Its properties are defined by the number of trials $n$ and the probability of success $p$ on each trial. You can calculate the probability of exactly $k$ successes using the formula: $$P(X=k)=\left(\genfrac{}{}{0ex}{}{n}{k}\right)p^k(1-p{)}^{n-k}$$. This distribution is crucial for valuing options using simple binomial trees, where each period represents a trial with an "up" or "down" price movement. For instance, pricing a one-year call option might involve a 12-period binomial model, where you calculate the probability of the stock price exceeding the strike price at expiration.

The normal distribution, or Gaussian distribution, is the cornerstone of financial modeling. It is perfectly symmetrical and defined by its mean $\mu$ (the central location) and standard deviation $\sigma$ (the dispersion). Key properties include:

• Approximately 68% of observations fall within $\mu \pm 1\sigma$.
• About 95% fall within $\mu \pm 2\sigma$.
• Roughly 99.7% fall within $\mu \pm 3\sigma$.

This distribution is so prevalent because of the Central Limit Theorem, which states that the sum of many independent random variables tends toward a normal distribution. In finance, it is frequently used as a model for the distribution of returns, though this assumption has important limitations.

The Standard Normal Distribution and Practical Z-Table Usage

To work with any normal distribution, you standardize it. The standard normal distribution has a mean of 0 and a standard deviation of 1. You convert a value $x$ from a normal distribution $N(\mu ,{\sigma }^2)$ into a z-score using the formula $z=\frac{x-\mu }{\sigma }$. This z-score tells you how many standard deviations $x$ is from the mean.

Z-table usage is a fundamental skill. A z-table provides the cumulative probability $P(Z\le z)$ for a standard normal variable. For example, to find the probability that a portfolio return (normally distributed with $\mu =8\%$ and $\sigma =5\%$) falls below 2%, you first calculate the z-score: $z=(0.02-0.08)/0.05=-1.2$. Looking up -1.2 in a z-table gives a cumulative probability of approximately 0.1151, meaning there is an 11.51% chance the return will be less than 2%. A common exam trap is misinterpreting the table output; some tables show $P(Z\le z)$ while others show $P(0\le Z\le z)$. Always know which format your table uses to avoid incorrect probability calculations.

The Lognormal Distribution: Modeling Asset Prices

While returns are often modeled as normal, asset prices are typically modeled using the lognormal distribution. This is because asset prices cannot be negative, and the lognormal distribution is bounded below by zero. If a variable $Y$ is lognormally distributed, then the natural logarithm of $Y$, $\ln (Y)$, is normally distributed. This property aligns with finance theory: continuously compounded returns are normally distributed, so the future price $S_t=S_0{e}^{r_t}$ becomes lognormal. The distribution is positively skewed, meaning it allows for the possibility of extremely high outcomes (large price gains) with a long right tail, which often matches observed market behavior better than the symmetrical normal distribution.

Applied Decision Frameworks: Safety-First and Risk Assessment

Distribution theory moves from description to decision-making through specific criteria. Roy's safety-first criterion helps you choose the portfolio that minimizes the probability of a return falling below a minimum acceptable threshold, or shortfall risk. The optimal portfolio is the one with the highest Safety-First Ratio (SFR): $$SFR=\frac{E(R_p)-R_L}{{\sigma }_p}$$ where $E(R_p)$ is the expected portfolio return, ${\sigma }_p$ is its standard deviation, and $R_L$ is the threshold level. A higher ratio indicates a lower probability of shortfall. For example, if a pension fund must achieve a 3% return to meet liabilities ($R_L=0.03$), it would compare portfolios based on their SFR.

Shortfall risk is the probability $P(R_p<R_L)$ itself, which you calculate directly using the normal distribution and z-tables. This application is vital for portfolio risk assessment, where Value at Risk (VaR) and other metrics rely on understanding the tails of return distributions. In option pricing, the Black-Scholes-Merton model famously assumes that stock prices follow a lognormal distribution, which allows for the closed-form calculation of option prices based on the probability of the stock finishing in-the-money.

Monte Carlo Simulation: A Versatile Application Tool

Monte Carlo simulation is a computational technique that uses random sampling from specified probability distributions to model complex systems. The core concept involves: 1) defining a model with uncertain inputs (e.g., interest rates, volatility), 2) specifying probability distributions for these inputs (like normal or lognormal), 3) randomly generating thousands of possible scenarios, and 4) analyzing the distribution of outcomes. In finance, it's used for tasks where analytical solutions are difficult, such as valuing complex derivatives with multiple sources of risk, assessing the probability of project failure, or performing stress testing on a portfolio under non-normal market conditions. It translates theoretical distribution knowledge into practical, forward-looking analysis.

Common Pitfalls

1. Assuming Normality Without Justification: A major mistake is blindly assuming investment returns are normally distributed. Financial returns often exhibit "fat tails" (more extreme outcomes than the normal distribution predicts) and skew. This can lead to a significant underestimation of the probability of large losses. Correction: Always test distributional assumptions or use more robust models like the lognormal or historical simulation.
2. Misapplying the Lognormal Distribution: Confusing when to apply the normal vs. lognormal distribution is common. Remember, you model returns with the normal distribution (log returns, specifically) and prices with the lognormal distribution. Saying "asset prices are normally distributed" is incorrect and violates the non-negativity constraint of prices.
3. Z-Score and Probability Errors: When using z-tables, a frequent error is miscalculating the area for a two-tailed test or misreading the table format. For example, to find the probability a return is between 5% and 10%, you must find $P(Z\le z_{10\%})$ and subtract $P(Z\le z_{5\%})$, not simply look up one value. Correction: Always sketch a quick diagram of the normal curve and shade the area of interest to guide your calculation steps.
4. Overlooking the Parameters in the Binomial Model: In option pricing with a binomial tree, incorrectly specifying the up-factor, down-factor, or risk-neutral probability will derail the valuation. These parameters are derived from volatility and time, not arbitrary guesses. Correction: Use the standard formulas, like $u=e^{\sigma \sqrt{\Delta t}}$, to ensure the tree correctly models the asset's volatility.

Summary

• Probability distributions, including uniform, binomial, normal, and lognormal, provide the essential language for modeling financial uncertainty and risk.
• The standard normal distribution and z-tables allow for the practical calculation of probabilities for any normally distributed variable, a core skill for assessing return outcomes and shortfall risk.
• The lognormal distribution is specifically appropriate for modeling asset prices due to its lower bound of zero and positive skew, aligning with the reality that prices cannot be negative.
• Roy's safety-first criterion offers a decision rule for portfolio selection based on minimizing the probability of returns falling below a critical threshold, directly applying distribution theory to manager objectives.
• Monte Carlo simulation leverages random sampling from defined distributions to solve complex valuation and risk problems where analytical solutions are not feasible.
• A critical awareness of the limitations and assumptions behind each distribution, such as the fat tails in financial returns, is necessary to avoid misapplication and significant model risk.