Assignment Task Introduction and background

Last week your investment company held an all-hands-on-deck meeting to brainstorm new ideas for future investment strategies. Your Chief Investment Officer is particularly concerned that all players in the industry, including your own company, offer very similar products based on a mixture of “classic” (i.e. old school) investment factors such as value, momentum, and profitability…

Another interesting consideration is that the recent raising of interest levels has put on the spot the ability of leveraged companies to generate enough cash flows to be able to service current and future debt. This would be similar, in sprit, to the “safety” component of the quality investment factor, where one would invest in firms with low levels of debt.

As a rising star in the quant team, you are tasked to explore the possible profitability of a strategy based on this idea.

Your report will have the following components:

Statistical analysis of the predictive power (based on the Information Coefficient Analysis) of the three ratios, as well as of a synthetic factor calculated as the average of the A presentation of the results of backtesting of a 130/30 investment strategy based on the “average” synthetic factor over the 2007–2023 period. Optimization of this strategy based on the number of shares over/underweighted and the rebalancing frequency. Address some of the concerns of the CIO of your company Exploration of a Machine Learning approach to this The deliverables for this assignment are:

A Jupyter Notebook with your code A pdf file with a short written There is no formal length requirement for the report. My suggestion is to aim at anywhere between 8-10 pages of “main text” including tables and pictures. You may also add an appendix if you want to add more tables, etc. Please be sure that all the necessary information is in the main body of the report.

Statistical Analysis

In the first part of your assignment, you should present a detailed statistical analysis of the predictive power of these three financial ratios (Interest Coverage Ratio, Cash flow to current liabilities Ratio, and Cash flow to debt Ratio) as well as a syntenic “Average” factor built as the average of these three.. The analysis should be based on the Information Coefficient (you should NOT use Quantile Analysis).

In doing this analysis you should use all the available data (1980-2023) and you should show how the predictive power changes in time . After reading this part the reader should understand:

How strong is the predictive power of these factors over the entire sample? Has this predictive power changed over time ? For the length of the statistical analysis , you should aim at around one-two pages . If you feel the need to add many more tables and graphs, consider the possibility to relegate some of them to the appendix, leaving only your “main narrative” in the body of the report.

Backtesting

For this part of the report, you should create and backtest a long-short 130/30 strategy based on the “Average” factor (you can forget about the three original ratios). Your strategy should:

Overinvest in the top 300 stocks and underinvest in the bottom Have an active percentage of 30% (should, in fact be a 130/30 strategy). Be tested on data between 2007 and Rebalance Assume 2% roundtrip transaction costs. After reading this part of the report your boss should have a clear picture of how the strategy performs over the entire period and whether the performance is focused in a particular period of time (for example the earlier part of the sample, or the covid years…).

The length of this part should be around one-two pages, including comments and graphs/tables .

Optimization

You should find the optimal parameters of the strategy in order to maximize the performance. Here you should focus on the number of shares to over-under-weight and on the rebalancing frequency .

For each parameter you should try three different values. You should use a reasonable range of numbers to capture meaningful differences in the strategy (for example you should not test 301, 302 and 303 shares…).

After reading this part the reader should know what combination of parameters produces the best performance.

Notes:

When performing the backtests for strategy optimization, you should use the same time period and transaction cost assumption as in part If you can, try not to repeat the backtesting function nine times, instead use for-loops. For the length of this part, you should aim at around one page . If you feel the need to add many more tables and graphs, consider the possibility to relegate some of them to the appendix, leaving only your “main narrative” in the body of the report.

Addressing your boss’s concerns

Your boss, the Chief Investment Officer of the company, has two main concerns that would like you to explore.

Question 1: Is this strategy simply a re-skinned version of some other well-known strategy such as value, momentum and/or profitability? In other words, are these “new” investment factors or just a different way of measuring older factors?

Question 2: These new ratios measure the ability of the company to generate enough cash flows to payoff interest on existing debt. It stands to reason that the predictive power of these factors should be stronger when interest rates are higher . Is this true?

It is up to you to decide how to perform your analysis but after reading your report your boss should have clear answers to these two questions.

Machine Learning Approach

Your company has been experimenting with decisions trees, so you decide to round-up your report with a simple analysis of the inclusion of these three signals (you can forget about the “Average” in this part) in a decision tree model.

Build and assess the predictive power of a decision tree with the following characteristics:

Target Variable : the tree should try to predict if the next month return will be positive or negative . Sample : From the beginning of 2007 to the end of Max Depth : 3 Factors : the tree should use data from The three new factors (INTCOV, OCF_LCT, CASH_DEBT). A value factor (EPQ). A momentum factor (MOM12). A profitability factor (GP). A low-volatility factor (TVOL). Accuracy (out of sample) should be measured with a 10-fold Cross Validation model. Your main goal in this section is to assess whether the three new factors:

Significantly affect how the decision tree predicts future Increase the (out of sample) forecasting accuracy of the Data

Together with the assignment, you will find the following data files:

zip contains the monthly data of the Interest Coverage Ratio of US stocks from 1908 to 2023. The ratio is measured as Net Income divided by Interest Expenses . zip contains data on the Cash flow to Current Liabilities Ratio measured as Operating Cash Flow divided by Current Liabilities .

zip contains data on the Cash flow to Debt Ratio measured as Operating Cash Flow divided by Total Debt . zip , MOM12.zip , TVOL.zip and EPQ.zip contain monthly data of traditional investment factors for profitability (Gross Profitability), momentum (12-Months Momentum), low volatility (Total Volatility), and value (Quarterly Earnings to Price). All the factors are already normalized and winsorized (i.e. they are “Ready to go” as the other ones used in class).