Study guides / CFA L2: CFA Program Level II / Multifactor models
Multifactor models
CFA L2: CFA Program Level II · quick review
Arbitrage pricing theory and arbitrage opportunities, expected returns from factor sensitivities, macroeconomic, fundamental and statistical factor models, their uses in return and risk attribution and portfolio construction, and active risk and the information ratio.
The rules
- Arbitrage pricing theory (APT): expected return = risk-free rate + Σ factor sensitivity × factor risk premium. It assumes returns come from a factor model, there are enough assets to diversify away company-specific risk, and arbitrage opportunities don't last. Unlike the CAPM, it needs no market portfolio and doesn't name the factors.
- Arbitrage opportunity: a portfolio with zero net investment and zero factor exposure that earns a positive return. If two portfolios with the same factor sensitivities have different expected returns, buy the higher and short the lower. With one factor, combine two portfolios to match a third portfolio's sensitivity and compare expected returns.
- Macroeconomic factor models: the factors are surprises in economic variables (actual − expected), such as GDP growth, inflation or credit spreads. Sensitivities come from time-series regressions, and the intercept is the asset's expected return, because expected surprises are zero.
- Fundamental factor models: the factors are returns to attributes such as size, value, momentum or leverage. Sensitivities are standardized attribute values ((attribute − average) ÷ standard deviation), and factor returns are estimated by a cross-sectional regression each period. The intercept has no expected-return meaning.
- Statistical factor models extract factors from return data alone (factor analysis or principal components). They need few assumptions and explain past returns well, but the factors are hard to interpret and may not persist.
- Return and risk attribution: active return = Σ (portfolio sensitivity − benchmark sensitivity) × factor return, plus security selection. Active risk squared = active factor risk + active specific risk, showing how much tracking risk comes from factor bets versus individual stock bets.
- Portfolio construction: a passive manager matches the benchmark's factor exposures; an active manager tilts toward factors expected to pay off, or picks securities while staying factor-neutral; rules-based strategies target chosen factors systematically.
- Multiple risk dimensions: the CAPM's single market factor misses risks that matter differently to different investors. Someone whose job income falls in recessions should take less business-cycle risk than the average investor, even at some cost in expected return; factor models let a portfolio be tailored to such needs.
- Active risk (tracking risk) is the standard deviation of active return. The information ratio = active return ÷ active risk, the reward per unit of active risk.
Formulas
- APT expected return: E(R) = R_F + λ1 × β1 + λ2 × β2 + … + λk × βk
- Macro factor model: R_i = a_i + b_i1 × F1 + … + b_ik × Fk + ε_i, where each F is a factor surprise
- Fundamental factor sensitivity: b = (attribute value − average) ÷ standard deviation of the attribute
- Active return: R_A = Σ (β_P,k − β_B,k) × F_k + security selection
- Active risk: Active risk^2 = active factor risk + active specific risk
- Active specific risk: Σ (w_P,i − w_B,i)^2 × σ^2(ε_i)
- Tracking risk: TR = σ(R_P − R_B)
- Information ratio: IR = (R_P − R_B) ÷ TR
Common traps
- In a macro model the factors are surprises, not the levels of the economic variables.
- Macro factor sensitivities come from time-series regressions; fundamental factor returns come from cross-sectional regressions.
- An arbitrage portfolio needs zero net investment and zero factor exposure, not just a higher return.
- Active risk squared is the sum of the factor and specific parts; active risk itself isn't their simple sum.
How it's tested
Vignettes give factor sensitivities and premiums and ask for an expected return, whether an arbitrage exists, a return or risk attribution, or which type of factor model is described.
Practise it
The full lesson, worked examples and practice questions are in Tickmark, where every answer choice is explained.
Open the full lesson