Ratios derived from a company’s financial statements together with other information can be used to filter stocks of different companies to identify a handful of good potential candidates for investment. Ratios can be used in screening using both top-down or bottom-up analysis. In a top-down analysis, an analyst first filters industries or geographical segments and then selects individual companies, but in a bottom-up analysis, an analyst applies the filter to all companies in an investment universe. In using ratios for screening, an analyst needs to decide which metrics to use, in what weight and at what cut-off value.
Example of an equity screen
For example, an analyst wants to short-list potential equity investments only if it has ROE of 15%, it is a member of some equity index, its CEO and Chairman are not the same individuals, its market cap is above a certain limit, etc. In selecting filters for screening, an analyst needs to make sure that the screens counterbalance each other. For example, in selecting companies with P/E lower than X, an analyst must assign a minimum profit margin to exclude companies whose P/E is low due to lower profit. Similarly, since criteria are not independent, total firms that would filter through a screen would most likely be higher than the product of the proportion of companies satisfying each filter individually. For example, if an analyst sets two screens, profit margin > 15% (which is satisfied by 50% of the companies in the investment universe) and dividend payout > 40% (satisfied by 30% companies). The ultimate companies that would pass through the filter would most likely be greater than 15% (=50% × 30%).
Types of equity screens
Growth investors may use screens that focus on earnings growth and/or momentum, value investors may assign screens based on some upper limit of valuation ratio (such P/E, P/B, etc.) and market-oriented investors who may use both growth and value screens.
Weaknesses of equity screens
An analyst may back-test an equity screen by applying the screen criteria to an investment universe and finding out how it would have performed historically. This approach is prone to:
- Survivorship bias: If the screen is applied to existing companies and not a complete list of companies that existed at the earliest date under analysis, the screen would exclude performance of companies that have ceased to exist resulting in overstatement.
- Look-ahead bias: If the analysis is based on updated financial data, it would overstate returns because in real-time, the initial financial statement may contain errors and/or lack full disclosure.
- Data-snooping bias: If a model is developed from past data and back-tested against past data, it does not necessarily have any predictive capacity.
If an analyst back tests a model using a database which is updated for restatements, it would most likely suffer from:
- Data-snooping bias
- Look-ahead bias
B is correct. Since the model is being tested using data which would not be available initially (because restatements are made later), it suffers from look-ahead bias.