Interpreting Residuals and Market Returns in the Single Index Model

Summary

The document presents the single-index, or market, model for a security’s return: an intercept plus its market return exposure, represented by beta, and a residual term. It asks what the residual represents, how it is computed, which probability distribution might be assumed, and why its expected value is often set to zero. In a fitted linear regression, the residual is the part of the observed stock return not explained by the fitted intercept and market component; a zero conditional mean is a modeling assumption, not a claim that each residual equals zero.

It also raises an empirical choice about constructing the market-return input, contrasting a simple annual mean with an exponentially weighted alternative. The document contains questions rather than answers, so it supplies no preferred estimator, distribution, or evidence comparing market-return calculations. Model users must specify the return frequency, estimation window, and assumptions appropriate to their application, and should not treat the residual distribution as settled by the equation alone.

Key ideas

• The model decomposes a security return into an intercept, market-linked component, and residual.
• A residual is the difference between observed return and the model’s fitted return.
• A zero expected residual is an assumption about the unexplained component, not a value for every observation.
• The document asks how to estimate market returns but offers no empirical recommendation.

Tags

# Deriving Single Index Model (Market Model)


# Deriving Single Index Model (Market Model)












$R_{it}=\alpha_i+\beta_i\cdot R_{mkt}+\epsilon_{it}$

- $R_{it}$ is the return of the stock of observation

- $R_{mkt}$ is the return of the reference market

- $\beta_i$ is the regression coefficient between the observed stock and the reference market

- $\alpha_i$ is the regression intercept between the observed stock and reference

- $\epsilon_{it}$ is the error (a random variable with expectation zero and finite variance)

First of all can someone help me to understand what $\epsilon_{it}$ represents in the model and how one should compute it. Considering it is a random variable which distribution is often used? Why expectation zero?

Secondly from the empirical point of view which is the best practice in computing $R_{mkt}$ in relation to this model? (i.e. simple annual mean, annual ewm etc.)

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.