How does multi-factor regression analysis identify a fund's factor exposures, and how does this affect alpha interpretation?

Question

AcadiFi · Accepted Answer

**Factor exposure analysis** uses multi-factor regression to decompose a fund's returns into systematic factor loadings plus a residual alpha. It reveals how much of a manager's performance comes from exposure to well-known, cheaply replicable risk factors versus genuine security selection skill.

**The Multi-Factor Model**

**(R_p - R_f) = alpha + B_MKT(R_m - R_f) + B_SMB(SMB) + B_HML(HML) + B_MOM(MOM) + epsilon**

| Factor | Measures | Premium Source |
|---|---|---|
| **MKT** (Market) | Market beta exposure | Equity risk premium |
| **SMB** (Small Minus Big) | Small-cap tilt | Small-cap premium |
| **HML** (High Minus Low) | Value tilt | Value premium |
| **MOM** (Momentum) | Trend-following tilt | Momentum premium |
| **Additional:** Quality, Low Vol, Liquidity | Various style tilts | Various |

**Worked Example — Alpha Decomposition**

Grantham Capital claims CAPM alpha of 3.2% per year. A 4-factor regression (60 months) reveals:

| Factor | Loading (Beta) | Factor Return (Annual) | Return Attribution |
|---|---|---|---|
| Market | 1.05 | 7.0% | 7.35% |
| SMB (small-cap) | 0.35 | 2.8% | 0.98% |
| HML (value) | 0.28 | 3.5% | 0.98% |
| MOM (momentum) | 0.15 | 4.2% | 0.63% |
| **Factor-explained return** | | | **9.94%** |
| **4-factor alpha** | | | **0.61%** |

| Metric | CAPM | 4-Factor |
|---|---|---|
| Alpha | 3.2% | **0.61%** |
| t-statistic | 2.45 | 0.48 |
| R-squared | 0.85 | 0.93 |

```mermaid
flowchart TD
    A["CAPM Alpha: 3.2%"] --> B["Small-cap tilt explains 0.98%"]
    A --> C["Value tilt explains 0.98%"]
    A --> D["Momentum explains 0.63%"]
    B --> E["4-Factor Alpha: 0.61%"]
    C --> E
    D --> E
    E --> F["t-stat 0.48 — NOT significant"]
    F --> G["'Alpha' was really factor exposure all along"]
```

**Interpretation**

Grantham's apparent 3.2% CAPM alpha was largely driven by:
- 0.98% from a persistent small-cap tilt (B_SMB = 0.35)
- 0.98% from a value tilt (B_HML = 0.28)
- 0.63% from momentum (B_MOM = 0.15)

These factor exposures can be replicated cheaply through factor ETFs at 5-15 bps. Grantham charges 75 bps for what is essentially 60 bps worth of factor tilts plus 0.61% of non-significant alpha.

**Practical Factor Analysis Steps**

1. **Select factors** — use standard academic factors (Fama-French 5-factor + momentum) or commercial risk models (Barra, Axioma)
2. **Run regression** — minimum 36 months (60 preferred) of monthly returns
3. **Evaluate loadings** — which factors have significant t-statistics?
4. **Assess residual alpha** — is the intercept significant after controlling for all factors?
5. **Stability check** — are factor loadings stable through time? (rolling regression)

**When Factor Alpha Is Genuine**

True alpha exists when:
- The residual alpha is statistically significant (t > 2.0 over 60+ months)
- Factor loadings are stable (not driven by a few outlier months)
- The alpha persists out-of-sample
- The factors capture all known systematic risk sources

**Common Pitfall:** Adding more factors always reduces alpha mechanically. The challenge is choosing factors that represent genuine, investable risk premiums rather than data-mined patterns.

For more on factor analysis, explore our CFA Level III course materials.

How does multi-factor regression analysis identify a fund's factor exposures, and how does this affect alpha interpretation?

Master Level III with our CFA Course

Related Questions

Practice Questions