John M. Griffin
Empirical Methods in Finance
Predictability, Assignment #2:[1]
Prepare a neat and well organized report of your findings with exhibits in the back. Note any assumptions you make as you go.
Predictive instruments are used widely in finance. Your goal is to understand their use. The primary vehicle here will be an understanding of Fama and French (FF) (1989).
Your assignment is threefold:
Obtain the following data:
VWCRSP, EWCRSP, also, the small market cap size decile.
You may divide your labor and pool the results to obtain a common data set of
Dividend yield, DEF, TERM, a January dummy. DOCUMENT what source this came from and who in the class obtained. Find the data back as far as FF.
Although you can ask for help when your stuck, all remaining work in the assignment is individual.
Replicate Table 3 of Fama and French (1989) for the VW and EW indices as dependent variables. Also use the smallest market cap portfolios as used in Keim and Stambaugh (1989).
A.
Replication/Extension: Table 3
1. For the period used in Fama and French (1989)
2. For the entire period (i.e. up to present).
3. For the post-FF period
A1. What is the appropriate way to gauge the model performance?
A2. Which model performs best?
A3. How do the models perform across horizons?
A4. How well do the models do post-FF?
A5. What do you learn?
B.
Other variables
For your favorite specification--perform the same
regression with an individual stock you used in your previous homework. How and
why do you think that the individual stock results are different.
B2. Estimate the same specification for |RVW|. What does this tell you?
B3. Find between 1 and 4 empirical variables of your choosing. Re-estimate both the monthly, 1 yr, and 4 yr, regressions.
C.
Out of Sample
C1. Similar to analysis in Section 7 of FF (1989) generate out of sample forecasts at the 1-month, and 1 yr horizons of your best performing model from A.
C2. Plot the forecasted and actual series? How would this do as a trading rule?
C3. What are the simple correlations between actual returns and out-of-sample forecasts?
C4. Evaluate the out-of-sample performance in any other way you find useful.
D.
Discussion
What have you learned from this exercise? What is surprising?
Do you believe there is real out-of-sample predictability? If you were going to investigate these what might you do?
What are some econometric concerns you might have in your analysis. How could you control for this?