Understand and use single and multifactor models for investment returns

Welcome to Asset Valuations: Factor Models!

Hi there! Welcome to one of the most practical chapters in your CM2 journey. So far, you have looked at how individual assets behave. But have you ever wondered why groups of stocks seem to move together? Or why some stocks crash harder than others during a recession? Factor models are the tools actuaries and investment managers use to answer these questions. Instead of looking at every stock in isolation, we look at the "forces" (factors) that drive their returns. Don't worry if the math looks a bit scary at first—we'll break it down piece by piece!

1. The Big Idea: Why Use Factor Models?

In a perfect world, we would know exactly why a stock price changes. In reality, it's a mix of broad economic trends and company-specific news. Factor models help us separate these. They assume that the return on any asset is sensitive to one or more common factors plus a bit of random noise.

Analogy: Think of a boat on the ocean. The boat's movement depends on the tide (a factor affecting all boats), the wind (another factor), and whether the engine is working (company-specific). Factor models help us measure how much the tide and wind matter versus the engine.

Quick Review: Key Terms
Systematic Risk: Risk that affects the whole market (e.g., a change in interest rates). You can't escape this by diversifying.
Specific (Idiosyncratic) Risk: Risk unique to one company (e.g., a factory fire). This can be removed by holding many different stocks.

2. The Single-Index Model (The Market Model)

The simplest way to model returns is to assume there is only one thing that matters: the overall market. This is often called the Market Model.

The Formula

For a specific asset \( i \), the return \( R_i \) is expressed as:
\( R_i = \alpha_i + \beta_i R_m + \epsilon_i \)

Let's break this down:
1. \( R_i \): The return on asset \( i \).
2. \( \alpha_i \) (Alpha): The "intercept." This is the return we expect if the market return is zero.
3. \( \beta_i \) (Beta): The sensitivity. It tells us how much the asset's return changes for every 1% change in the market return \( R_m \).
4. \( R_m \): The return on the market index (like the FTSE 100).
5. \( \epsilon_i \) (Epsilon): The error term or "noise." This represents the specific risk unique to asset \( i \).

Key Assumptions

For this model to work mathematically, we assume:
- The error terms for two different stocks are not correlated: \( Cov(\epsilon_i, \epsilon_j) = 0 \).
- The error term is not correlated with the market: \( Cov(R_m, \epsilon_i) = 0 \).
- The average (expected) value of the error term is zero: \( E[\epsilon_i] = 0 \).

Memory Aid: Think of Beta as a "volume knob." If Beta is 1.5, when the market turns the volume up by 1, your stock turns it up by 1.5!

Key Takeaway

The Single-Index Model simplifies the world by saying: "I only care about how this stock moves relative to the general market."

3. Multifactor Models

Sometimes, one factor isn't enough. A tech stock might react differently to interest rate changes than a grocery store does, even if the "market" stays the same. This is where Multifactor Models come in.

The General Formula

Instead of just one \( \beta R_m \), we have several:
\( R_i = \alpha_i + \beta_{i,1}F_1 + \beta_{i,2}F_2 + ... + \beta_{i,k}F_k + \epsilon_i \)

Where:
- \( F_1, F_2, ... \) are the different factors.
- \( \beta_{i,1}, \beta_{i,2}, ... \) are the factor loadings (how sensitive the stock is to each specific factor).

Types of Multifactor Models

The IFoA curriculum distinguishes between three main types of models based on what the "factors" actually are:

1. Macroeconomic Models
These use observable economic data. Common factors include:
- Inflation rates.
- GDP growth.
- Changes in interest rates.
Example: An airline stock might have a high sensitivity (beta) to the factor "Oil Prices."

2. Fundamental Models
These use characteristics of the companies themselves. Common factors include:
- Size: Small companies often behave differently than giant ones.
- Value: Comparing the stock price to the company's actual book value (P/E ratios).
- Industry: Is it a tech company or a bank?

3. Statistical Models
These are the "mystery boxes." We use complex math (like Principal Component Analysis) to find patterns in historical data. We don't necessarily give the factors names like "Inflation"; we just know that "Factor A" explains 40% of the movement.

Did you know? The Fama-French Three-Factor Model is a famous fundamental model that uses Market Risk, Size, and Value to explain returns. It’s a classic example of moving beyond the single-index approach!

Key Takeaway

Multifactor models provide a more detailed "fingerprint" of an asset's risk by looking at multiple drivers of return at once.

4. Diversification in Factor Models

One of the most important uses of these models is understanding diversification. Let's look at the variance (risk) of a portfolio return using a factor model.

The total variance of a stock can be split into two parts:
Total Variance = Systematic Variance + Specific Variance

In a Single-Index Model, this looks like:
\( Var(R_i) = \beta_i^2 Var(R_m) + Var(\epsilon_i) \)

Why does this matter?
As you add more and more stocks to a portfolio:
1. The Specific Variance (the \( \epsilon \) part) gets smaller and smaller. It averages out to nearly zero.
2. The Systematic Variance (the \( \beta \) part) does not go away. No matter how many stocks you buy, if the whole market crashes, your portfolio will likely feel it.

Common Mistake: Students often think diversification removes all risk. Remember: Diversification only kills the Specific Risk (\( \epsilon \)). It cannot touch the Market Risk.

5. Step-by-Step: Calculating Expected Return

If you are given a factor model in an exam, here is how you find the Expected Return \( E[R_i] \):

Step 1: Identify the constants. You will usually be given \( \alpha_i \) and the \( \beta \) values.
Step 2: Find the expected value of the factors. \( E[F_1], E[F_2] \), etc.
Step 3: Remember that \( E[\epsilon_i] = 0 \). The noise cancels out on average!
Step 4: Plug them into the formula:
\( E[R_i] = \alpha_i + \beta_{i,1}E[F_1] + \beta_{i,2}E[F_2] \)

Example: If \( \alpha = 2\% \), \( \beta = 1.2 \), and the expected market return is \( 10\% \):
\( E[R] = 2\% + 1.2(10\%) = 14\% \).

6. Summary and Final Tips

Summary Checklist:
- [ ] Single-Index Model: One factor (usually the market). Returns follow \( \alpha + \beta R_m + \epsilon \).
- [ ] Multifactor Models: Multiple factors (Macro, Fundamental, or Statistical).
- [ ] Beta: Measures sensitivity to a factor.
- [ ] Alpha: The return not explained by the factors.
- [ ] Diversification: Eliminates specific risk (\( \epsilon \)) but not systematic risk (\( \beta \)).

Encouraging Word: This chapter is the bridge between simple statistics and real-world finance. Once you master the idea that "Return = Base + (Sensitivity \(\times\) Force) + Noise," the formulas will start to feel much more natural. Keep practicing the variance breakdowns—they are a favorite for examiners! You've got this!