Estimating Monthly Factor Returns with Cross-Sectional Regressions

Summary

The document explains how a factor model can produce a time series of returns for a selected characteristic, such as dividend yield. The key idea is to run a cross-sectional regression at each date: asset returns are related to the characteristic of interest and to other risk characteristics, and the estimated coefficient for the target characteristic is recorded over time. Repeating this process yields the sequence of monthly factor returns shown in the paper.

A complementary interpretation is portfolio based. The pure factor return corresponds to the return of a characteristic portfolio designed to have exposure to the target factor while neutralizing exposure to the other modeled factors. Holding that portfolio over a month gives the period's return. The discussion is brief and provides no derivation, portfolio construction details, data, or empirical results. It also does not specify how the regression handles weighting, estimation choices, or practical constraints, so those details must be obtained from the paper's methodology.

Key ideas

• Estimate factor returns by running a cross-sectional regression at each point in time.
• The target factor's regression coefficient forms a time series when recorded across dates.
• A pure factor return can be interpreted as the return of a portfolio with target exposure and zero exposure to other modeled factors.
• Holding that characteristic portfolio for a month produces the corresponding monthly return.

Tags

# pure factor return for factor model


# pure factor return for factor model












I am reading a paper. The authors use the multivariate regression to calculate the pure factor return $\beta_F$ using the following equation: $$Return_{t+1}=\beta_F f_F + \beta_{RF_1} f_{RF_1} +⋯+ \beta_{RF_N} f_{RF_N} +\epsilon$$

where

- $\beta_F$ = pure factor return for the desired return factor,

- $\beta_{RF_i}$ = pure factor return for Risk factor $i$,

- $f_F$ = evaluated factors (ex: Dividend Yield)

- $f_{RF_i}$ = risk factor (ex: size)

After that, they produce a monthly factor return as the graph here. I just wonder from the pure factor return for the desired return factor, how they can get the return over months like the graph?

## Answer by lebelinoz (score 5)

https://quant.stackexchange.com/a/36119

From that picture you took, it looks like the $\beta_F$ is a time series. It's computed by doing a cross-sectional regression at each point in time.

## Answer by user36511 (score 1)

https://quant.stackexchange.com/a/42585

In the cross sectional framework, the "pure factor return" is actually the return of the characteristic portfolio, which has unity exposure to the underlying factor and zero to others. If you hold the portfolio for one month, you get the monthly return.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.