Computing for Investment Analysis

This course is a medium-level introduction to Python and its applications in fundamental analysis, especially estimating the value of companies from fundamental analysis and SEC filings. (If you need a complete introduction, please follow the online "CBS Python Level 1," a prerequisite to this course that I assume all of you have taken. To quote Warren Buffett, we want to automate the work of finding "outstanding companies at sensible prices."

In the first half of the course, we build a quantitative discounted cash flow model in Python to estimate a company's value. We start with a simple one-line formula for the net present value from free cash flow and the average cost of capital. We then add refinements, such as a continuation value, revenue growth predicted from GDP growth, and sensitivity analysis. At the end, we have a tool that takes a stock ticker like "MSFT" and automatically computes a range of intrinsic values for the share price. We apply this tool to US public companies and obtain a list of the most undervalued and overvalued companies.

In the second half of the course, we consider one possible reason for this discrepancy. If a company is under-valued or over-valued, what could the market know that we do not? Is there information that we are missing? Yes, indeed: so far, we have used a company's public disclosures to provide numbers for our quantitative model, but one important piece of information we have not considered is the text. We therefore build a qualitative model from the textual information in a company's public filings. For example, a company may be under federal investigation, and investors are therefore justifiably pessimistic about its future, which could explain a low share price. We therefore apply "text mining" to the risk disclosures in a company's 10-K filing. Our qualitative model identifies new risks within the company and issues a recommendation: "buy" or "sell."

If our quantitative model predicts that a stock is undervalued (the current share price is low relative to fundamentals) and our qualitative model finds no red flags in the risk disclosures, we can recommend the company with reasonable confidence as a "buy." Conversely, if our quantitative model predicts that a company is overvalued and has risk disclosures, we can recommend that the company is a "sell." We then form a long-short portfolio.

Both of these models, quantitative and qualitative, fit into the general 3-part chain below:

- Data (What?): Data gathering,