Explain the Fama-MacBeth two-step procedure

Q: Explain the Fama-MacBeth two-step procedure

Fama-MacBeth (1973) estimates risk premia in two steps:\n\nStep 1 (time-series): For each asset i, regress asset excess returns on factor returns over the full sample: r_{i,t} - r_f = alpha_i + beta_i * F_t + e_{i,t}. This estimates each asset's factor betas.\n\nStep 2 (cross-sectional, repeated monthly): At each date t, regress excess returns on estimated betas: r_{i,t} - r_f = lambda_0,t + lambda_1,t * beta_hat_i + u_{i,t}. The average lambda_1 across months estimates the factor risk premium. Standard error is computed from the time series of {lambda_1,t} using Newey-West adjustment for autocorrelation.\n\n```mermaid\nflowchart LR\n S1[Step 1 time-series] --> B[Betas]\n B --> S2[Step 2 cross-section]\n S2 --> L[Lambda per period]\n L --> R[Risk premium]\n```\n\nKey advantage: standard errors naturally account for cross-sectional correlation in residuals at each date (all stocks share macro shocks). Pooled regression assumes independence across (i,t) pairs, understating SE.\n\nLimitation: errors-in-variables from using estimated betas. Shanken (1992) correction adjusts SE upward by factor (1 + market risk premium^2 / market variance). For CFA Level II context, Fama-MacBeth validates CAPM/Fama-French by testing whether lambda_market, lambda_size, lambda_value are significantly different from zero.\n\nEmpirical results are mixed: market beta often shows insignificant premium post-1970s (beta puzzle), while size and value show stronger premia. Modern practitioners use Fama-MacBeth routinely as the workhorse for asset pricing tests, including in factor investing due diligence.

Question

AcadiFi · Accepted Answer

Fama-MacBeth (1973) estimates risk premia in two steps: Step 1 (time-series): For each asset i, regress asset excess returns on factor returns over the full sample: r_{i,t} - r_f = alpha_i + beta_i * F_t + e_{i,t}. This estimates each asset's factor betas. Step 2 (cross-sectional, repeated monthly): At each date t, regress excess returns on estimated betas: r_{i,t} - r_f = lambda_0,t + lambda_1,t * beta_hat_i + u_{i,t}. The average lambda_1 across months estimates the factor risk premium. Standard error is computed from the time series of {lambda_1,t} using Newey-West adjustment for autocorrelation. ```mermaid flowchart LR S1[Step 1
time-series] --> B[Betas] B --> S2[Step 2
cross-section] S2 --> L[Lambda
per period] L --> R[Risk
premium] ``` Key advantage: standard errors naturally account for cross-sectional correlation in residuals at each date (all stocks share macro shocks). Pooled regression assumes independence across (i,t) pairs, understating SE. Limitation: errors-in-variables from using estimated betas. Shanken (1992) correction adjusts SE upward by factor (1 + market risk premium^2 / market variance). For CFA Level II context, Fama-MacBeth validates CAPM/Fama-French by testing whether lambda_market, lambda_size, lambda_value are significantly different from zero. Empirical results are mixed: market beta often shows insignificant premium post-1970s (beta puzzle), while size and value show stronger premia. Modern practitioners use Fama-MacBeth routinely as the workhorse for asset pricing tests, including in factor investing due diligence.

Explain the Fama-MacBeth two-step procedure

Master Level II with our CFA Course

Related Questions

Practice Questions