Common mistake 1 with Multi-factor model

Throwing in many correlated factors and reading individual loadings; multicollinearity makes them unstable and swaps sign easily.

Common mistake 2 with Multi-factor model

Interpreting a high R² as "the model is right". It only means the factors span this asset's returns, not that they are priced risks.

Common mistake 3 with Multi-factor model

Ignoring that the factors themselves are portfolios with their own estimation and data-mining history.

Common mistake 4 with Multi-factor model

Reading alpha as skill without asking what the factors omit. Alpha is whatever the chosen factors do not explain, so adding momentum or quality to the model routinely removes an alpha that looked robust.

EconStatsCraft › Multi-factor model

Multi-factor model

Regresses an asset’s excess return on a set of factor returns - market, size, value, momentum - to see which exposures drive it and whether any alpha survives once they are accounted for.

What it needs

Asset excess return
Factor returns

The intuition

CAPM says one thing moves all stocks. That is plainly incomplete: small companies behave differently from large ones, and cheap companies differently from expensive ones, in ways the market index does not capture. A factor model adds those as extra explanatory returns. What survives as alpha afterwards is much smaller, and much more interesting, than what survived CAPM.

When to use it

Style analysis (what is this fund actually exposed to?), performance attribution, testing whether an apparent CAPM alpha is really compensation for a known factor, or building a factor-replicating portfolio.

Where the estimator comes from

The Fama-French three-factor regression is R_{it}-R_{ft}=\alpha_i+\beta_i(R_{mt}-R_{ft})+s_i\,\mathrm{SMB}_t+h_i\,\mathrm{HML}_t+\varepsilon_{it}. SMB (small minus big) is the return on a portfolio long small firms and short large ones; HML (high minus low) is long high book-to-market and short low. Both are themselves RETURNS on zero-cost portfolios, which is what allows them on the right-hand side of a return regression without a units problem. Carhart (1997) adds momentum, and Fama and French (2015) add profitability and investment for a five-factor version. The estimator is still OLS; what changes is the null, which is again \alpha_i=0 but now against a richer benchmark.

The model

R_{i,t} - R_{f,t} = \alpha_i + \sum_{k} \beta_{i,k}\,F_{k,t} + \varepsilon_{i,t}

What it assumes

The factors are the right ones and are not near-collinear with each other.
Loadings are stable over the window.
The factor returns are measured, not estimated with error - in practice they are published series like Fama–French.
The fund's exposures did not change during the estimation window. A manager who rotated from small-cap to value halfway through gets an average of two strategies, and the alpha absorbs the difference.

Statistical properties

Factors are constructed to be returns, so the coefficients are portfolio weights and dimensionless. A loading of 0.4 on SMB means the fund behaves like 40% of a long-small short-large position.
The factors are correlated with each other, mildly. Reported loadings are partial: the SMB loading holds market and HML fixed.
Adding factors mechanically lowers alpha for most funds. That is the point of the exercise, not a failure of it.
R-squared typically rises from about 0.88 to 0.95 for an equity fund. The gain is largest for funds with a genuine style tilt.
Standard errors ignore that the factors are themselves estimated from portfolio sorts. This is conventional and usually second-order, but it is an understatement.
The GRS test (Gibbons, Ross and Shanken, 1989) tests the alphas of many funds jointly, which is the correct way to ask whether a factor model prices a whole set of assets. A single t-test on one alpha does not.

How to read the output

Each loading is an exposure: a value (HML) loading of 0.6 means the asset behaves like something 60% tilted toward cheap stocks.
The intercept is the multi-factor alpha. If it collapses to zero once the factors are added, the CAPM alpha was style exposure, not skill.
Adjusted R² near 0.9+ for a diversified fund is normal - most of what a fund does is factor exposure.
Compare loadings across funds to see which is really taking the size or value bet it advertises.

A worked reading

Market beta 1.0762 (SE 0.0535), SMB 0.8663 (SE 0.1118) and HML 0.3534 (SE 0.0921), all overwhelmingly significant, on 180 months with an R² of 0.7258. The fund is slightly more than fully exposed to the market, tilted heavily towards small companies and moderately towards value. Alpha is 0.2133% a month with a standard error of 0.2575% - a t of 0.83, indistinguishable from zero. That is the whole argument of a factor model in one line: the fund does have a return the market alone does not explain, and once size and value are priced in, what is left is not distinguishable from luck. Three quarters of the month-to-month variation is accounted for by three factors an index fund can buy cheaply, and the residual volatility of 3.42% a period is what the manager is actually being paid to produce.

Symptom → diagnosis → what to do

If you see this	It means	Do this
Alpha collapses versus CAPM	The fund's edge was style exposure	The honest conclusion, and the common one
SMB loading strongly negative	A large-cap fund	Expected; not a problem
HML loading negative and large	A growth fund	Compare it against growth peers, not the broad index
Loadings unstable across subperiods	The mandate changed, or style drift	Split the sample and report both
R-squared barely moves versus CAPM	No style tilt to find	The single-factor model was adequate here
All factors insignificant but R-squared high	The factors are collinear in this sample	Check their correlations before dropping any

Common mistakes

Throwing in many correlated factors and reading individual loadings; multicollinearity makes them unstable and swaps sign easily.
Interpreting a high R² as "the model is right". It only means the factors span this asset's returns, not that they are priced risks.
Ignoring that the factors themselves are portfolios with their own estimation and data-mining history.
Reading alpha as skill without asking what the factors omit. Alpha is whatever the chosen factors do not explain, so adding momentum or quality to the model routinely removes an alpha that looked robust.

Where to read this properly

Fama and French (1993), "Common Risk Factors in the Returns on Stocks and Bonds", Journal of Financial Economics 33.
Carhart (1997), "On Persistence in Mutual Fund Performance", Journal of Finance 52(1) - momentum.
Fama and French (2015), "A Five-Factor Asset Pricing Model", JFE 116(1).
Gibbons, Ross and Shanken (1989), "A Test of the Efficiency of a Given Portfolio", Econometrica 57(5).

Related Finance & risk

Black–Scholes option pricing · Bond price, duration & convexity · Merton distance to default · EMI / home loan planner · CAPM / beta estimation · Portfolio optimization · Value at Risk & Expected Shortfall · Performance & drawdown · Time value of money & NPV/IRR · Equity valuation (DDM, FCF, residual income) · Cost of capital (WACC) · Binomial option pricing · Option strategies (payoff & profit) · Forwards, futures & swaps · Credit risk & Basel capital · Extreme value theory (tail risk)

All analyses

Data & variables

Each example is simulated from a known model, so you can check whether the estimator recovers the truth - and nothing here is anyone's private data.

Filter variables

Drag a variable into a slot on the right - or just click it.

Choose an example dataset above to begin.

Choose a transformation - it opens in the data editor, with a live preview.

Apply to

Multiply by

Columns ctrl-click for several

Group by ctrl-click for several

Within groups

Statistic

Bands

How many

Cut points

Keep rows where

Value

and

Fill with

Value

Also mark which rows were filled

Rules

Base period = 100

Which part

Percent at each tail

Threshold

Survey weight optional

Direction

Names column

Values column

Largest first

Delete them, not just count them

Read this column as

Tidy the text

New name optional

EconStatsCraft rigorous, plain-spoken econometrics

Econometrics that explains itself.

Regression, panel data, time series and the whole battery of tests. Every result is followed by a verdict you can defend.

Load a dataset to see it here.

Multi-factor model

What it needs

The intuition

When to use it

Where the estimator comes from

The model

What it assumes

Statistical properties

How to read the output

A worked reading

Symptom → diagnosis → what to do

Common mistakes

Where to read this properly

See also

Related Finance & risk

All analyses

Charts

Maps

Explore

Regression

Time series - one series

Time series - several series

Panel data

Multivariate statistics

Marketing research

Causal

Finance & risk

Actuarial

Tests & breaks

Machine learning

Econometrics that explains itself.