John M. Griffin
Empirical Methods in Finance
 

FM Regressions, Assignment #4, Due Monday, Sept. 25 10:30

Prepare a neat and well organized report of your findings with exhibits in the back.

The purpose of this assignment is to gain an understanding of the cross-sectional regression methodology of Fama/McBeth (FM) (1973). Towards this goal I would like for you to estimate cross-sectional regressions similar to Fama and French (1992). Download--25 Portfolios Formed on Size and Book-to-Market (5 x 5).

http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html

Goal: Estimate CS regressions and obtain t-stats on the excess market return.

Step 1. Making the portfolios by pre-ranking at the end of the previous year’s values of size and BE/ME. June rankings of ME are used and December values of BE/ME are used. Note that Professor French has done this step for you!:) Note that portfolios could also be formed on size and pre-ranking values of Beta. This is what FF do in Table 1.

Step 2. The returns are value-weighted portfolio returns in excess of the risk-free rate. Obtain estimates of beta for each portfolio. The original FM methodology estimated OLS regressions for the previous 5 years and used this estimate in the next month CS reg. However, here we will follow FF (92) who follow Chan and Chen (88). They find the betas for each portfolio from time-series regressions and then allocate these betas to the portfolio. (Note: that FF (92) allocate the betas for a stock in a given portfolio to the individual stocks in that portfolio. However, for simplicity and for other reasons that become apparent, here you will just use portfolios.)

Step 3. Save the OLS time-series betas and merge them back to your original return series of size and BE/ME portfolios.

Step 4. Run month by month FM cross-sectional regressions (OLS). Save the coefficients.

Step 5. Decide if these coefficients are significant using t-statistics constructed in the FM manner.

Questions

  1. Program this.
     
  2.  Run market model regressions, save the betas. Report the betas for all portfolios in a table. (Call Table I). Use a format similar to FF 92 for the table. What is the finding?
     
  3. Run second-pass cross-sectional month-by-month cross-sectional regressions and report the average values and FM t-stats on the intercept and beta, and adj. R2. Is beta significant? Intercepts? Discuss these results.

4.      Now include ME in the CS regressions. And report coefficients and t-stats on all variables, and adj. R2. What do you find--discuss.

5.      Re-estimate these regressions now including BE/ME as well. Report and discuss these findings.

6.    How well does Beta do? What do you see as potential limitations or weaknesses with your current analysis.

Use Nice and neat Tables to display your results.