MS6217 Statistical Modeling in Economics and Finance
Individual Take-Home Final Project
Prof. Gavin Feng
Due at May 17th 11pm
Guidance
1. Please upload your pdf submission to Cavnas before May. 17th 11pm.
2. Please have your name, student number, and CityU email on top of the first page.
3. If you have any general question, please post it on Canvas discussion before May 15th 11pm.
4. If you have any personal concern about anything, please also email me as early as possible.
Project Question (100 points)
Suppose you are a quantitative fund manager specializing in investing different industries or sectors. You
now have a task to introduce a long-only strategy to your clients for different industry exposures. You need
to provide and justify your portfolio choices for 49 different industries. You are given the below data for
creating the optimal tangent portfolio exercise, which will be evaluated by a hold-out sample.
Please be careful. You can use the risk-free rate for your stock return prediction, but the tangent portfolio
must be formed by these 49 industry portfolios. Hence, the portfolio weights for 49 industries are sum to
100% without any negative weight.
(1) Forty-nine stocks {Ri,t}49i=1.
(2) Ten macroeconmic predictors {Xk,t}10k=1
(3) Twenty stock charactersitics {Zi,j,t}20j=1.
For (2) and (3), the predictors in the hold-out sample are also provided. Your task is to figure out the best
long-only portfolio using assets in (1).
The data provided are calibrated on true predictor-return data. You are provided with a sample 1-500
observations for 49 industries. We are gonig to test your performance with another hold-out 501-1000
simulated observations. Predictors are also given for 501-1000 observations. Hence, as long as your model
learns the sample data well, your portfolio performs well in our hold-out data.
Result Submission
In addition to the report, you need to submit your portfolio choices for the hold-out data contains 501-1000
observations.
Specifically, you need to upload a separate “csv” file with the exact file name “MS6217_studentid”, where
“studentid” should be replaced with your CityU student number. If one’s id is 12345678, the file name is
“MS6217_12345678.csv”. The file contains 240 rows and 49 columns, where corresponds to the stock orders.
Before you submit the “csv” file, please run the program “check_csv.R” by yourself.
1
Please be careful. Any difference in format names causes 10 points for penalties. As long as your output file
can pass the program “check_csv.R”, it should work.
1. Report Writing (30 points)
You are going to prepare an analysis report to the clients, most of whom only expect to see simple statistical
analysis with figures and tables without any coding.
• In your report, please includes useful figures and tables for the empirical illustration.
• Please limit your report in 10 pages.
• There is no need to show us your code.
• You can use any software to prepare for the report. If you are really good in Excel, I don’t mind.
2. Empirical Analysis (40 points)
You are suggested to justify your model choices by following the below question map.
• Why do you choose this model? Any summary statistics for evidences?
• How is the portfolio performance using your model? Create a simple back testing yourseful using the
below three criteria.
• Have you compared your model with some other simple benchmark models?
• If you think you have found a good model, what are the details for your model implementation?
3. Portfolio Performance (30 points)
30% of your grade is based on the class ranking of your portfolio performance in the hold-out data. Investing
is ruthless and winner takes all. In this part, the best performance gets 100% and the worst 0%. The below
three criteria can be used for your model evaluation.
• Sharpe Ratio (10%) – mean(Pt−rft)sd(Pt−rft) . We want to see high a high Sharpe Ratio.
• Correlation with an equally weighted portfolio (10%) – cor(Pt, EWt). We want to see a low correlation
with the EW portfolio.
• Maximum Monthly Loss (10%) – min(Pt). We want to see a small maximum monthly loss.
Pt is the return of your portfolio for 49 industries, and EWt is the return of an equally weighted portfolio. If
your portfolio is very similar to the EW portfolio, there is no need to pay you the management fee :).