Introduction: Why Random Variables Matter
Welcome! If you’ve ever wondered how risk managers predict the likelihood of a stock market crash or the number of loan defaults in a month, you are in the right place. In the world of Finance, we can’t predict the future with 100% certainty, but we can use Random Variables to model the possibilities. This chapter covers the most common "building blocks" of probability that you will use throughout your FRM journey. Don't worry if the formulas look intimidating at first—we'll break them down into simple, real-world ideas.
1. The Big Picture: Discrete vs. Continuous
Before we dive in, remember there are two main types of random variables:
1. Discrete: Things you can count (e.g., the number of days a bank's system goes down).
2. Continuous: Things you measure (e.g., the exact percentage return on a portfolio).
Think of discrete like steps on a ladder and continuous like a smooth slide.
2. Common Discrete Distributions
The Bernoulli Distribution
The Bernoulli distribution is the simplest of all. It models a single trial with only two outcomes: Success (1) or Failure (0).
Example: A single borrower either defaults on their loan or they don't.
• Probability of Success: \( p \)
• Probability of Failure: \( 1 - p \)
• Mean (Expected Value): \( E(X) = p \)
• Variance: \( Var(X) = p(1 - p) \)
The Binomial Distribution
If you take a bunch of independent Bernoulli trials and add them up, you get a Binomial distribution. It tells us the probability of getting exactly \( x \) successes in \( n \) independent trials.
Example: If you have 10 independent loans, what is the probability that exactly 2 will default?
• Key Parameters: \( n \) (number of trials) and \( p \) (probability of success).
• Mean: \( n \times p \)
• Variance: \( n \times p \times (1 - p) \)
Common Mistake: Students often forget that for a Binomial distribution to work, the trials must be independent. If one loan defaulting makes another more likely to default, the Binomial model isn't the best fit!
The Poisson Distribution
The Poisson distribution is used to model the number of events occurring within a specific time or space interval.
Example: The number of cyber-attacks on a bank's server per week.
• Key Parameter: \( \lambda \) (Lambda), which is the average rate of occurrence.
• Unique Property: The Mean and Variance are both equal to \( \lambda \).
Quick Review: If you see a question about "the number of events in an hour," think Poisson!
Key Takeaway: Discrete distributions help us count outcomes. Bernoulli is for one "yes/no" event, Binomial is for multiple "yes/no" events, and Poisson is for the frequency of events over time.
3. Common Continuous Distributions
The Continuous Uniform Distribution
In this distribution, every value within a certain range \( [a, b] \) is equally likely. The probability density function (PDF) is a flat horizontal line.
• Mean: \( \frac{a + b}{2} \) (The exact midpoint).
• Variance: \( \frac{(b - a)^2}{12} \)
The Normal Distribution
This is the "King" of distributions—the famous Bell Curve. It is symmetrical and defined by its mean \( \mu \) and standard deviation \( \sigma \).
• Skewness: 0 (It's perfectly symmetrical).
• Kurtosis: 3 (or "Excess Kurtosis" of 0).
Did you know? In risk management, we often use the Standard Normal Distribution (\( Z \)), which has a mean of 0 and a standard deviation of 1. You can convert any normal value to a Z-score using: \( Z = \frac{X - \mu}{\sigma} \).
The Lognormal Distribution
If the natural log of a variable is normally distributed, the variable itself is Lognormal.
Why it matters: We use Lognormal for asset prices (like stocks). Why? Because stock prices cannot be negative, and the Lognormal distribution is bounded at zero and skewed to the right.
Memory Aid: Returns are often assumed Normal; Prices are assumed Lognormal.
Student’s t-Distribution
The t-distribution looks like the Normal distribution but has "fater tails." This means it predicts a higher probability of extreme events (outliers).
• Parameter: It is defined by "Degrees of Freedom" (df).
• Rule of Thumb: As degrees of freedom increase, the t-distribution starts looking exactly like the Normal distribution.
Why we love it: Financial markets have more "crashes" than the Normal distribution predicts. The t-distribution captures this "tail risk" better.
Chi-Squared (\( \chi^2 \)) and F-Distributions
These are mainly used for statistical testing rather than modeling prices.
• Chi-Squared: Used to test variances and goodness-of-fit. It is always positive and skewed to the right.
• F-Distribution: Used to compare two different variances. It's what you use when you want to know if one portfolio is significantly more volatile than another.
Key Takeaway: Continuous distributions model measurements. Normal is the standard, Lognormal is for prices, and the t-distribution is for when we expect more extreme "surprises."
4. Central Limit Theorem (CLT): The Magic of Statistics
The Central Limit Theorem is one of the most important concepts in the FRM. It states that if you take a large enough sample from any distribution (even a weird, non-normal one), the sum or mean of that sample will follow a Normal Distribution.
• Why it's cool: It allows us to use the Normal distribution's tools even when the underlying data is messy, provided the sample size is large (usually \( n > 30 \)).
5. Summary Checklist for Success
Quick Review Box:
• Bernoulli: One coin flip (1 or 0).
• Binomial: Multiple coin flips; trials must be independent.
• Poisson: Counting arrivals/events over time.
• Normal: Bell curve, Mean=0, Std Dev=1 for "Standard Normal."
• Lognormal: Right-skewed, no negative values (Stock Prices).
• Student's t: Fat tails, used for small samples or risky assets.
• Chi-Squared/F: Used for testing variances.
Don't worry if this seems tricky at first! The key to mastering this chapter is recognizing which distribution fits a specific scenario. When you see "number of defaults," think Binomial. When you see "stock prices," think Lognormal. When you see "fat tails," think Student's t. You've got this!