Research Skills Part 5

Question 1

5 Step Fama-MacBeth Approach

Answer 1

1. Estimate beta of each firm i using a time-series regression
2. Run a cross-sectional regression of returns on estimated betas in each period t. This leads to T estimates for the risk premium (beta coefficient in cross-sectional regression).
3. Compute the coefficient of interest (risk premium) by averaging all risk premier over time
4. Compute the S.E. of the avg risk premium by computing the SD/SQRT(T)
5. Compute the t-stat = avg risk premium / S.E.

Variant: include observed characteristics in step 2, in addition to, or instead of estimated betas. E.g. firm size, B/M ratio.

Question 2

Name the benefits and drawbacks of Fama-MacBeth

Answer 2

Benefit: corrects for cross-sectional correlation in panel
Benefit: allows for arbitrary number of stocks (i) in each period t
———————————————
Drawback: does not account for serial correlation
Drawback: relies upon T being sufficiently large
Drawback: it is sensitive if the true coefficient varies much over time

Question 3

When do you use binary choice model?

Answer 3

When the dependent variable is a dummy variable

Question 4

What is a probit/logit model?

Answer 4

Probit/logit is non-linear estimation method (in contrast to OLS) → fitted probability is non-linear function of x.

Probit employs a normal distribution function to transform. Logit employs a logistic distribution function. The scaling of coefficients differs, but marginal effects are similar.

For testing purposes, we can use standard t-tests and F-tests.