CXO Advisory Group took a look at the stock screens available at The American Association of Individual Investors website. The screens are available if you join the association and pay an annual due (relatively inexpensive). CXOAG finds that trading frictions could significantly reduce performance of the strategies. In addition, the strategies, in general, will be affected by the overall stock market performance.
Two thoughts: I wonder if there is a way to reduce frictions and also add a basic qualifier to some of their strategies to limit exposure (such as the 200 day moving average).
Question for the reader: Would you find it worthwhile for me to track some of the AAII portfolios on this blog?
Here is CXOAG’s full take:
A reader asked:
“The American Association of Individual Investors (AAII) has a lot of strategies they have been paper-trading over the the last 11 years at AAII StockScreens. Have you ever done an evaluation of those performance results? It seem like every strategy builds upon a well-known investing book or otherwise publicized strategy from the last 40 years.”
According to the AAII StockScreens “GettingStarted” introduction, the purpose of these screens “is to provide…access to a wide range of investment approaches. Some approaches follow the methods of well-know professionals, and allow you to implement their ‘style of investing,’ while other approaches implement time-tested techniques used to identify attractive stocks. These approaches run the full spectrum, from those that are value-based to those that focus primarily on growth. Some approaches are geared toward large-company stocks, while others uncover micro-sized firms. Most fall somewhere in the middle.” AAII provides descriptions, characteristics and performance statistics for the screens. What can investors/traders learn from this collection of investment approaches? Using monthly performance statistics for the 59 screening approaches and for various potential benchmarks during the 139 months spanning January 1998 through July 2009 (available from AAII via download), we find that:
AAII cautions that: “Unless otherwise stated, figures do not include dividends or transactions costs.” Since many of the screening approaches involve high monthly turnover of holdings, trading frictions (transaction costs plus bid-ask spread) can materially reduce net returns. Screening approaches that generate a relatively large number of holdings exacerbate the impact of transaction costs because the typical individual investor could take only small positions in each, such that transaction costs would represent a noticeable percentage of buys and sells. Screens that focus on small or micro capitalization stocks tend to further exacerbate the impact of trading frictions, because bid-ask spreads tend to increase as a percentage of stock price as firm capitalization decreases.
We therefore make the following assumptions to approximate the impact of trading frictions on returns for the screening approaches:
- Round-trip trading friction for the typical trade specified by the screening approaches is 1.0%. This assumption may be harsh (generous) for approaches that involve just a few relatively liquid stocks (many small stocks). Actual trading frictions depend on stock-specific bid-ask spreads, actual broker transaction fees and actual position sizes.
- Monthly portfolio turnover is the average monthly turnover reported in the AAII-generated performance statistics. Actual monthly turnover could vary considerably from month to month.
So, for example, if average monthly turnover for a screen is 50%, we debit the monthly return for that screen by 50% times 1.0%, or 0.5%. High turnover screens therefore suffer higher trading frictions than low turnover screens. Average monthly turnover ranges from 7% to 100% across all 59 screens.
We ignore dividends for lack of information. Some strategies involve stocks paying material dividends, and others do not. Portfolio turnover could disrupt qualification for dividends.
We also ignore the tax implications of portfolio turnover. Many of the screens considered would generate predominantly short-term capital gains/losses.
The following chart compares the distributions of average monthly returns for all 59 screening approaches over the entire sample period, before and after debiting for trading frictions. Points of interest are:
- Incorporation of trading frictions moves the distribution to the left (of course), sharpens the peak and suppresses the right tail. Some of the highest-performing screens also have high average monthly turnovers.
- The aggregate (equally-weighted) average monthly return for all 59 screens is 1.26% (0.86%) before (after) incorporation of trading frictions.
- The maximum average monthly return among all 59 screens is 2.79% (2.22%) before (after) incorporation of trading frictions.
- While both distributions exhibit positive skewness (right tail longer than left tail), they have shapes something like bell curves. One thing to consider is whether luck could account for the positions and spreads of the distributions. It is plausible that some or all of the aggregate outperformance implied by the distributions is a result of data mining bias. How many poorly performing screens have been “swept under the rug” in the processes of: (1) developing each of these screens, and (2) selecting this set of screens for consideration?
How sensitive is aggregate average monthly return of the 59 screens to the 1.0% trading friction assumption?
The next chart shows the variation of the aggregate average monthly return for all 59 screens with assumed round-trip trading friction. Included for reference (from the AAII data) are the average monthly returns over the sample period for all U.S.-listed stocks, the S&P Midcap 400 and the S&P Smallcap 600. The chart illustrates the importance of managing trading friction when using strategies that drive substantial portfolio turnover.
Do the best-performing screens exhibit stable performance over time?
The following two charts depict the monthly returns during January 1998 through July 2009 for the two screening approaches with the highest average monthly returns after including approximated trading frictions. The two screens are:
- O’Neil’s CAN SLIM, with average monthly return 2.22% (2.79% without trading friction).
- O’Shaughnessy Tiny Titans, with average monthly return 2.17% (2.59% without trading friction).
Each chart includes a best-fit linear trend line as a rough measure of the trend in monthly returns over the sample period, with trend indications as follows:
The monthly return trend line for the O’Neil CAN SLIM screen is flat, indicating steady performance. However, the performance record of this screen has one extremely influential month (June 2009). Without this one month, the average monthly return for the screen is 1.74% instead of 2.22%, and the return trend line slopes noticeably down over time.
The monthly return trend line for the O’Shaughnessy Tiny Titans screen trends noticeably down over time. The name of the screen suggests high trading frictions (bid-ask spreads). This screen has a drawdown of almost 50% during the three months September 2008 through November 2008.
A downtrending monthly return could indicate that:
- The early part of the sample period contains data mining bias (good luck) due to discovery and selection via testing of many screens on the same data set.
- Increasing use of the screen after discovery has depressed its outperformance as more and more investors/traders share its advantage. In other words, the market has adapted to the screen.
- General market conditions have otherwise changed such that what worked earlier in the sample period does not work later.
It is difficult to test the first two of these potential explanations directly because the screens were likely discovered and promoted at different times. We can test the third one by looking at aggregate data.
The final chart shows the after-trading friction average of the monthly returns by month for all 59 screens, equally weighted, over the entire sample period. The trend line indicates that, on average, the effectiveness of the screens has declined over this time, with late 2008 through early 2009 perhaps decisive.
The trend line for the monthly return of all U.S.-listed stocks is very similar to this trend line, suggesting that general market conditions are decisive in the performance decline for the screens.
In summary, as illustrated using AAII’s StockScreens performance data, investors should consider especially the impact of portfolio turnover (trading frictions) and performance under different market conditions when evaluating stock screens.