Reference no: EM133254120

Advanced Empirical Finance Assignment

Stock Returns Over the FOMC Cycle

In this assignment, you investigate the behaviour of stock returns in monetary policy announcement- calendar time, as in the recent study by Cieslak, Morse, and Vissing-Jorgensen (2019, Journal of Finance), "Stock Returns over the FOMC Cycle". You will need the following data (available on ADAM):

• Daily CRSP Index returns -> MARKET.mat
- We have prepared this dataset in the beginning of the course
• FOMC announcement dates -> event FOMC.xls
- The dates are taken from:

Preparation
Prepare a dataset that is as comparable as possible to the study by Cieslak, Morse, and Vissing- Jorgensen (2019):

• Load the dataset MARKET.mat and event FOMC.xls and restrict to the sample period 1994

• Add back holidays to the sample with a return of zero.

Question 1

Compute the average return, standard deviation and standard error of average returns for:
• All days
• All FOMC announcement days
• All days, excluding FOMC announcement days
Are FOMC announcement days economically larger than on all other days? Is the difference stat- istically significant?
HINT: convert the excel"date umber" to a Matlab"date number"; then use the function"datefind" to construct an FOMC announcement dummy which is one on FOMC days and zero otherwise.

Question 2
Re-define weeks in FOMC announcement time. Compute a dummy=1 for the days -1 to 3 around an FOMC announcement day (day=0), this is "Week 0". Then compute a dummy=1 for the five days before (days -6 to -2), this is "Week -1". Similarly, compute dummies for the weeks +1 to +6 following "Week 0". In summary: Week -1: Days -6 to -2, Week 0: Days -1 to 3. Week 1: Days 4
to 8. Week 2: Days 9 to 13. Week 3: Days 14 to 18. Week 4: Days 19 to 23. Week 5: Days 24-28.
Week 6: Days 29-33.
There are 8 scheduled FOMC announcement days in each year. Accordingly, between two announcements there are on average 6.5 weeks, or about 30 trading days. However, meetings are not equally spaced over the year. Therefore, you should adjust your dummies such that there is no overlap. Run a "HAC robust" OLS regression of daily stock returns on a constant, the "Week 0", "Week 2", "Week 4", and "Week 6" dummies. Use 10 Newey-West lags.
• How can you interpret the coefficients of this regression?
• Do you find an interesting pattern in the data?
• Explain why you prefer HAC standard errors over "iid" standard errors in this particular regression.
HINT: you need 6 leads and 33 lags of your FOMC dummy. You can generate leads and lags "by hand" or using the function "lagmatrix.m". Use our own regression function or the Matlab build-in function fitlm.

Question 3
Provide a "convincing" figure that illustrates your empirical results from Question 2.