Thanks for the comments so far:-
I have been trying to focus a bit more on building a ranking system that is logical and robust, rather than going for out-and-out performance. I have seen many sims feature out-of-sample performance that quite frankly has no relation to their in sample performance, and this seems to be strongly correlated to the number of stocks they hold - I very much doubt the triple digit gains seen in some 3 stock sims will be replicated in the future.
The more stocks, the harder it is to curve fit.
Ultimately, we are searching for alpha - returns above the benchmark. Some factors provide a source of alpha, and depending on how these are combined, it is possible to achieve synergy and gain further increases in alpha.
One issue I have, is that factors that provided alpha in the past, may fail to do in the future, or may not do for extended periods of time, such as value factors which have clearly been lagging for the past few months, with value-only portfolios doing poorly.
Therefore, having different unrelated sources of alpha is a good thing because only one has to be a source of alpha in the future for your portfolio to have positive alpha.
In each area I will discuss the sources of alpha:
Valuation: - this is obviously the classical version. While most people focus on earnings yield, while this is important it is important to look at cash flow and sales too. Earnings without cash flow are frankly âfakeâ. The price in relation to the top line has also been found to be a very important factor, as discovered by OâShaugnessy. This is interesting - investors canât âeatâ sales. Nevertheless, sales are probably the best economic measure of the size of the business, and arguably they are a more stable measure than earnings. Finally the PE based on next FY projection is important, if only to make sure that the current low valution ratios are not caused by a temporary spike in earnings. Beyond the next FY the accuracy of estimates decreases geometrically, and are probably useless. That is why no other projected figures are used.
I believe the most recent up to date figures (quarterly) are important. This goes somewhat against the Graham approach of averaging earnings, because that ignores the fact that earnings can and do trend higher and lower, the most recent figures are the closest to what is about to be seen. While on their own the quarterly figures do not seem to add more alpha than the TTM figures, when combined with momentum factors the level of synergy is a lot higher, possibly due to the effect of earnings momentum.
Technical - trade with the trend, your trend is your friend, all cliches, but thats because frankly it works. The first thing I used to look at was relative strength, price now vs price 6months or 12 months ago. I noticed, however, that short term RS seems to have little alpha, and very short term RS is a contrary indicator - interesting. A short term price spike is not bullish - probably because it attracts a lot of speculators. Long term investors are less likely to trade on the trend and trade on fundamentals instead. This is probably why a longer term trend is a good indicator - the smart money is moving in based on improving fundamentals. Therefore I use two factors. I use the daily sharpe over the past 6 months instead of 6 month RS because it seems to add more alpha. This may be because the sharpe function punishes volatility - and that makes sense. A very volatile stock can easily make it into the top decline of RS through randomness alone, not because of a true trend.
To guard against recent price spikes, the difference between the 50 day and 200 day simple moving average is used as well. This ensures the higher prices has persisted over a long period of time, not caused by simple speculation, and reduces the risk of simply buying an overbought security.
Short interest - This is perhaps a more unsual source of alpha. One interesting thing is that following average analysts recommendations does not provide any alpha at all, but following the shorts does. I guess it pays to look at opinions where people have put their money where their mouth is! A while ago I read a paper showing that shorts were âsmart moneyâ, more sophisticated than your average investor. It pays to see what they they are up to, and there is a stong correlation between the level of short interest and the future performance. The more the short interest, the worse the performance. A good amount of recent short covering is bullish too. So both factors are considered, the amount of short interest, as a % of shares outstanding, and the 1month percent change.
Efficiency - Warren Buffett has often commented that âROEâ is everything. He believes in investing in high ROE companies. Many people think he is a value investor, but I think he is really a high ROE investor. It has worked for him, he talks about how companies which require little assets to operate (media businesses etc.) can easily grow their earnings year after year just by raising prices.
Now I like to tweak this a bit, and look at ROA instead. The problem with equity is that its âassets-debtâ, so you can have an inflated ROE because of high debt. Of course, Buffett famously prefers companies with little or no debt, so for him ROA=ROE. But since I have found that debt/equity ratios do not appear to be a source of alpha either way, I am indifferent to that ratio. If you test ROA by itself, it is clearly a source of alpha.
Interestingly, related to ROA is the Asset Turnover. Sales/assets. Interestingly it does seem to be an even stronger source of alpha. Again the âyou canât eat salesâ arguement spings to mind, but then again, this may be due to the relative stability of sales compared to earnings (on which ROA is based).
There are other possible measures such as ROI ROC etc. etc. but I think there is little point in putting in extra overlapping factors, just stick to one.
Interestingly, traditional things such as margins are not included as they do not appear to be a signifcant source of alpha!
Growth - it appears that long term growth rates provide little alpha. This is possibly because they are already priced in, possibly because they are irrelevant to future growth rates. However, short term growth rates, specifically year-over-year do provide a strong source of alpha. Comparing the most recent quarter to the prior quarter is less indicative than comparing to the quarter one year ago. I think this is because seasonal effects may cause earnings spikes that distort the true growth trend.
I am interestested in why this is a source of alpha. I believe it is because earnings do have momentum, trends in the underlying economy and business do continue over the medium term, and this is not fully priced in to the market - future positive earnings surprises drive the stock higher.
I look at EPS change, cash flow change, and sales growth over the year, and also the improvement in ROA vs. the long term average. This is probably important because it indicates the company is not having to invest heavily in assets to achieve this growth, this means more returns for shareholders.
Financial Strength - this is born out of âdebt is badâ mantra. The problem is there seems to be little evidence of underperformance of companies with high debt-to-equity ratios etc. and even interest coverage seems to have little effect either way. The only factor I can find that does seem to provide convincing alpha is the coverage of short term liabilities, or EBITA/Current Liabilities. Probably companies that are headed for a hard place start off with a swelling in current liabilities, unable to keep the lid on expenditures. Perhaps the reason long term debt is less important is because - being long term - if the company has survived so far it will survive in the future. The very occasional bankruptcy has little effect on alpha.
So there you go, this is a ranking system built from the ground up designed to capture as many different sources of alpha as possible, and not focus too much on a single source, such as value. Because of the broad spread, it means that in any single area, it is unlikely to get the top ranked stocks, and this means it doesnât have the fireworks performance of some other sims. However, my belief is that it is more robust to some of these sources of alpha from stopping. if you run a deep value portfolio, or a high momentum portfolio, then you are essentially betting on value, or betting on momentum. This is a way to get a more diversified bet, a more conservative portfolio that is more likely to give positive alpha under all market conditions.