What is a factor-mimicking portfolio and how is it constructed?

Question

AcadiFi · Accepted Answer

A factor-mimicking portfolio (FMP) is a long-short portfolio of individual assets engineered to track a specific factor's return as closely as possible while minimizing exposure to all other factors. It translates an abstract factor into a tradeable sleeve.

Construction methods:

1. Sorting method (Fama-French style):
Rank stocks on the target characteristic (e.g., book-to-market). Long top decile/quintile, short bottom. Earn the cross-sectional spread. Simple but may have unintended exposures (e.g., value FMP often has small-cap tilt).

2. Cross-sectional regression (Fama-MacBeth):
Each period, regress asset returns on characteristic scores. The slope coefficient equals the FMP's return. Weights derived from GLS formula: w = (X'Ω⁻¹X)⁻¹X'Ω⁻¹. Controls for correlated factors.

3. Optimization-based:
Solve min w'Σw subject to w'b_target = 1 and w'b_other = 0 for all other factor loadings. Explicitly pure factor exposure.

Example — Value FMP via regression:
500 stocks, characteristic X = standardized book-to-price. Return on value FMP at time t = coefficient λ_V,t from regression Rᵢ,t = λ₀,t + λ_V,t × Xᵢ,t + ... + εᵢ,t.

Portfolio weights (Fama-MacBeth OLS): wᵢ = Xᵢ / Σⱼ Xⱼ². Sum of weights is zero (long-short), and weights are proportional to characteristic score.

Uses:

1. Isolate factor returns for performance attribution.
2. Construct tradable long-short factor ETFs.
3. Hedge specific factor exposures (short the FMP to neutralize).
4. Test whether a characteristic is priced — if FMP has statistically positive return, the characteristic is a risk factor with a premium.

Limitations: estimated weights are noisy; turnover can be high; OLS weights don't account for transaction costs. Practitioners use regularization (ridge, LASSO) to stabilize weights.

What is a factor-mimicking portfolio and how is it constructed?

Master Level III with our CFA Course

Related Questions

Practice Questions