Arbitrage pricing theory
E(Ri) = rf + βi,1 λ1 + βi,2 λ2 + … + βi,k λk
- βs are sensitivities to factors; λs are factor risk premiums.
- Assumptions: returns follow a factor model, enough assets exist to diversify away specific risk, no arbitrage.
- If a portfolio’s expected return differs from what its factor exposures imply, combining it with factor portfolios produces a riskless profit.
Types of factor models
| Model | Factors | Sensitivities |
|---|---|---|
| Macroeconomic | Surprises in GDP growth, inflation, credit spreads, the yield curve | Estimated by time-series regression |
| Fundamental | Returns to attributes: size, value, momentum, sector | Standardized company attributes |
| Statistical | Principal components of returns | Estimated; factors may lack economic meaning |
Analysing active risk
Active return = Σ (portfolio sensitivity − benchmark sensitivity) × factor return + security selection
Active risk² = active factor risk + active specific risk
This shows whether a manager’s results come from style tilts or stock picking.
Uses in portfolio construction
- Factor portfolios: pure exposure to one factor, used to hedge or bet on it.
- Tracking portfolios: match a benchmark’s factor exposures with fewer holdings.
- Risk budgeting: allocate active risk across factors.
- Performance attribution: explain returns by factor.
Factor models also show the “hidden” bets in a portfolio, for example a value tilt that rises in recessions.
Exam focus
Compute expected returns with APT, distinguish macro, fundamental and statistical models, decompose active return and risk, and describe factor and tracking portfolios.