Group Project
You are an equity analyst group. Your boss asks to you to focus on one stock market sector and provide a model that explains its excess returns. In particular, he lets you choose one among the 12
The sectors contained in this dataset are:
• NoDur: Consumer NonDurables
• Durbl: Consumer Durables
• Manuf: Manufacturing
• Enrgy: Energy
• HiTec: Hi-tech
• Telcm: Telecommunications
• Shops: Wholesale, Retail etc...
• Hlth: Heathcare
• Utils: Utilities
Part 1
You will have to conduct the analysis at the monthly and the annual frequencies and you will have to come up with a linear model that explains the returns in the industry you selected. Your linear model should have the form:
t - Rft) = Xtβ + ut (1)
where Rsec is the return on the sector you chose, Rft is the risk-free rate and Xt is a vector of
variables of your choice (including a constant). For example, you could use simple CAPM regressions, CAPM regressions augmented with the Fama-French factors, CAPM regressions that incorporate the inflation rate etc... Please be creative! Financial markets are always looking for people with new ideas!
The ultimate goal of the exercise is to understand what economic forces explain the excess returns of a given industry.
Your boss wants you to be precise in answering the following questions: 1.Why did you decide to focus on a certain sector?
2. What sample did you use? You do not have to use data from 1926 until today: you can decide on what sample you want to use for your analysis, but you have to explain your choice.
3. What variables did you consider initially? Were they macroeconomic variables? Financial variables? Both? (You should initially consider at least 10-15 regressors). Were the initial regressors you considered correlated among each other?
4. What variables did you discard along the way? Why were they discarded? Were their coeffi- cients insignificant? Did the coefficients make economic sense?
5. Do you find that similar variables explain monthly and annual returns, or do they differ? 6.Are the linear regression assumptions satisfied in your final model? Do the coefficients in your final model make economic sense?
Part 2
Your boss also wants you to come up with a predictive model of the form:
t+1 - Rft+1) = Xtβ + ut+1, (2)
whereby you predict the returns next month (or next year), using financial variables and/or economic variables available today.
1. What variables seem to predict the returns on your sector?
2. What variables did you discard along the way? Why were they discarded? Were their coeffi- cients insignificant? Did the coefficients make economic sense?
3. Do you find that similar variables explain monthly and annual returns, or do they differ? 4.Are the linear regression assumptions satisfied in your final model?
5. How much of the variation in your sector's excess returns' can you explain with your model?
6. Can you make any recommendation as to what is going to be the sector performance next month? How about next year? Should your boss BUY or SELL the stocks contained in the sector you have studied?
Part 3
Please do not start working on this part unless I tell you to... We may not cover enough material for you to tackle the tasks described below.
1. Take the best model you have come up with in Part 2 and estimate the linear regression parameters on the first half of the dataset you are using.
2. On the basis of these estimates compute out-of-sample forecasts for the second half of the dataset?
3. Plot the forecasts against the realizations. Are the forecasts good?
4.What is the R2 of your out-of-sample forecasts?
5. Can you come up with a better model for out-of-sample forecasts? (Hint: you will probably have to reduce the number of variables included in your model)
6. Is it easier to obtain a good out-of-sample performance using annual or monthly data? 7.(Optional) Estimate with a recursive window the coefficients of your linear model. Do the same using a rolling window of your choice. Do you obtain a high R2 out-of-sample?
Data
I will spend some time in Session 5 giving you some hints on where to obtain data that can be useful for this project. I will also get you started with some economic and financial time series.
Computations
I encourage you to use Matlab, but it is up to you and your group to decide what software to use. If you use Matlab, please print the code you used and attach it to the set of slides you hand in.
Deliverables
You will have to hand in a printout of the slides of your presentation and the code. You will have to prepare a 15-20 minutes presentation of your findings. The presentations will take place during the last two classes (October 7th and October 12th), during class time.