Interpreting Fama–MacBeth Regressions for Idiosyncratic Volatility

Summary

The document reports Fama–MacBeth regressions examining whether idiosyncratic volatility predicts stock returns, with beta, size, book-to-market, short-term reversal, illiquidity, maximum return, coskewness, and idiosyncratic skewness considered as controls. Across separate specifications, the estimated volatility coefficient is usually negative but often insignificant; it becomes significantly negative in models that include size and in the full specification. Some controls are significant in selected models, while many are not.

The results illustrate that a coefficient’s sign and statistical significance can change when controls enter a regression. A change after adding size is consistent with shared variation or omitted-variable effects, but the reported results alone do not establish why the change occurs or prove a causal interpretation. The document provides no coefficient estimates, standard errors, sample details, or diagnostics, and it is framed as a request for help rather than a complete analysis. Its findings therefore support cautious interpretation of conditional associations, not a firm conclusion about the return effect of idiosyncratic volatility.

Key ideas

• Fama–MacBeth specifications estimate the association between stock returns and volatility while adding different controls.
• The idiosyncratic-volatility coefficient is generally negative but its significance varies across specifications.
• Adding size coincides with a significantly negative volatility estimate, but this change alone does not identify its cause.
• The reported significance patterns are insufficient to assess effect size, model reliability, or causal interpretation.

Tags

# Fama Macbeth regression results


# Fama Macbeth regression results












I am doing the Fama Macbeth regression analysis for finding the relation between Idiosyncratic volatility and Expected stock return. Stock return is dependent variable and independent variables are: idiovol, beta, size, bm, strev, illiq, max, coskew and idioskew. I have run different models and I get the following results:

- Intercept (insig) + idiovol (neg insig)

- Intercept (sig) + idiovol (neg insig) + beta (insig)

- Intercept (sig) + idiovol (neg sig) + size (sig)

- Intercept (insig) + idiovol (neg insig) + bm (insig)

- Intercept (insig) + idiovol (neg insig) + strev (sig)

- Intercept (insig) + idiovol (neg insig) + illiq (sig)

- Intercept (insig) + idiovol (pos insig) + max (insig)

- Intercept (insig) + idiovol (neg insig) + coskew (insig)

- Intercept (insig) + idiovol (neg insig) + idioskew (sig)

- Intercept (sig) + idiovol (neg sig) + beta (insig) + size (sig) + bm (insig) + strev (sig) + illiq (sig at 10%) + max (insig) + coskew (insig) + idioskew (insig)

A) In the first model idiovol is negative but not significant. In second model when I include beta again both are insignificant. How I interpret this? B) In model third when I include size, idiovol become significant and negative and size is also significant. It mean that earlier (model 1) idiovol was significantly associated with size and reflects the effect of size in addition to its own effect (plus some other unobservables) that is why idiovol was insignificant earlier but now it showing its true impact on stock returns. Am I correct, I read this from some previous post on the forum. C) Now how I interpret the others model and last model in which I have included all the variables.

Please help me in interpreting these results.

Thank you Regards

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.