The backtest covered in your book buys X stocks and holds them for Y, correct? Or does your system buy the top ## names?
As discussed in the book, Quantitative Value, the system assesses the universe of firms on June 30 of each year.
The general outline of the process is as follows:
An example of how this might work on a universe that begins with 1000 names:
- Avoid stocks that cause a permanent loss of capital—>1000 to 900
- Discover the deeply undervalued stocks—>900 to 90 (top 10%)
- Find stocks with the highest quality—> 90 to 45 (top 50%)
In the example above, the portfolio would hold 45 names for the next year. At the next rebalance period on June 30, the process would be repeated.
Historically, the names in the portfolio ranges between 30 and 40 names.
Looking at the list, it seems there is a high concentration of stocks in the Technology sector. There are 6 stocks of the current Top Ten based in the Technology sector. Do the algorithms favor technology stocks? Isn’t this a concern regarding diversification and risk management?
The screens do not control for industry–this is by design. If a screening technology is forced to buy certain portions of the portfolio in sectors that are expensive, the portfolio will not perform as well. The key insight from our research is that cheapness matters above all else. Splitting your portfolio between 20 cheap technology firms is better than putting your portfolio in 10 cheap names in the technology sector and 10 names in the consumer discretionary sector (assuming all firms in this sector are generally not considered “cheap” on an absolute basis.)
Empiritrage produces detailed research on this subject.
One thing you will notice is that various data providers provide different numbers for similar calculations (partly due to differences in underlying calculations, for example, EBIT can be calculated any number of ways). The ultimate source to assess data discrepancies is sec.gov. In our experience, data is typically reliable for firms with market caps above $1B.
So what is one to do?
We do what we can to eliminate obvious data errors, but nothing is foolproof. When using data systems (not just ours), one should only screen for stocks above a reasonable threshold (e.g., $1B market cap). Otherwise, there is a high likelihood there will be a “garbage-in, garbage-out” problem.
The data comes from a variety of free financial websites, such as SEC.GOV, Google Finance and Yahoo Finance.
We constrain the quantitative value screen results to 15 firms over $10B.
Wes Gray, Ph.D. manages assets in strategies related to the screen discussed in Quantitative Value. Due to the massive traffic to the site, Dr. Gray and his team at Empiritrage, LLC determined that it was in the best interests of their clients to only share the most liquid names.