I’ve been searching for a compelling Microcap strategy for quite some time.
Based on what I’ve learned by researching the market for over 15 years – it is the Microcap (and nano-cap) space where the real opportunities hide. I have thus been wondering how high are the returns that one can achieve by implementing a well-structured microcap strategy.
The fact that the smaller the market cap, the higher the potential average returns are well-supported by the research. Small-size even has its own “Smart Beta Factor”, or “Anomaly”, known as SMB, Small-minus-Big.
And since most of us are individual investors with…less than tens of millions of investment dollars…why not go all the way and employ (at least some of) our capital in high-octane microcap strategy?
As any researcher knows, the best way to start research is to look at what others have already done.
Naturally, my first stop was Ben Graham. Indeed, we covered his net-net strategy, which involves mostly microcaps, in the December newsletter issue.
My second stop was Jim O’Shaughnessy. Jim is the author of the book “What Works On Wall Street”. My Deep Value (VC2) strategy utilizes the VC2 multi-factor, inspired by O’Shaughnessy’s book.
A reminder: VC2 stands for Value Composite 2 – A composite ranking system that averages the ranks for six valuation multiples: P/E, P/B, P/FCF, EV/Sales, EV/EBITDA, and Shareholders’ Yield.
The book, What Works on Wall Street, contains so much more than VC2. It describes tens of quantitative investing strategies, covering almost any stock factor you can think of, from valuation multiples to quality attributes, to momentum and…size. The book provides backtesting results over several decades. It serves as a good starting point for gaining true insight on what works on wall street, as its name implies. It is also a great starting point for designing a new quantitative strategy.
One of the best performing strategies described in the book (4th edition) is “Microcap Trending Value”:
Its characteristics are described as follows:
- Market capitalization between a deflated $50 million and $250 million at the time of purchase.
- Price-to-book (or, book to price, BP) ratio in the three cheapest deciles (i.e., the lowest 30 percent of the universe by price-to-book ratio).
- Three- and six-month price appreciation greater than zero
- The 25 stocks with the best 12-month price appreciation are the ones to buy.
And here are the results:
Several websites and services have recreated the strategy (see here, here and here). However, the most notable person I know who has recreated it is Paul Novell, author of the blog “Investing For a Living”. In his article, Paul re-created the strategy as accurately as he could, with the data he had, beginning of 1999 to October 2015:
The results for 1999-2015 are pretty lame, albeit the extraordinary performance that O’Shaughnessy reported for 1928 to 2009. What is the source for that discrepancy? Probably it has to do with the P/B metric. It is well-known in the quantitative investing milieu that Price/Book has not performed in the last several decades as it did during the first half of the 20th century.
Paul Novell replaced the P/B multiple with Value Composite 2 (VC2) which is comprised of six valuation multiples: Price/Earnings, Price/Book, Price/FCF, EV/Sales, EV/EBITDA, and Shareholders’ Yield. After all, it makes sense. O’Shaughnessy himself has found that VC2 performs better than any single valuation multiple he had tried. Here are his results:
NOW the results are truly extraordinary, with average annual returns of 23.2%, vs. The Russell 2K small-cap benchmark of 7.47% per year.
Notice that no. of positions in the portfolio is lower than 25 between 2008 and 2009. The reason is that the Microcap Trending Value strategy simply did not find enough stocks that meet its criteria to invest in. The rules that are the toughest to meet are those requiring positive returns in the last three months and the last six months. At those dreadful times of 2008 and 2009, there simply weren’t enough stocks with short-term positive returns.
It is my professional opinion that the fact that the strategy is not fully invested at all times impaired somewhat from its statistical validity. The reasons are related to concerns of overfitting and are beyond the scope of this article.
So far Paul Novell’s work.
Microcap Trending Value (Tal’s Implementation)
Next, I tried to replicate the results, which came in very similar (though not identical) to Paul’s. Here are my results for a Microcap Trending Value portfolio, using VC2 for filtering cheap stocks, and using about
the same timeframe as Mr. Novell’s had used:
The results are similar but not identical. I believe that the differences result from the use of slightly different timeframes (quant strategies backtests are very dependant on start date), minor variation in calculating market cap limits, and also simulator updates which were introduced in the time between performing those two simulations.
To further investigate the strategy, and improve it further, I then created the strategy in Portfolio123’s full simulation software. I made several adjustments, in order to apply a bit more rigor and “science” into the experiment.
First, from this point and on, I tested for the timeframe 6/30/1999 to 6/30/2018, a timeframe that is consistent with all the other backtests on TalDavidson.com, and also consistent with the academic research convention. Secondly, for the market cap limits I used: Market Cap larger than the smallest 20% percentile of stocks, but not larger than the 40% percentile ($50M to $215M in today’s prices).
Here are the results:
Putting aside my concern over the fact that the strategy was not fully invested during 2008-2009, the results are pretty compelling. Over 19 years, Microcap Trending Value generated an average annual return of 15.62%, with a standard deviation (a measure of volatility) lower than the benchmark, only 17.94%, vs. 19.24% for the Russell 2000 (and 14% for the S&P 500). It’s Sharpe ratio is high at 0.82. Moreover, Microcap Trending Value performed well recently, beating the Russell 2000’s 11.18% by almost 5% per year, from mid-2015 to mid-2018.
Can it be improved further?
Yes, it can.
I’ve experimented with many ideas – using the academic Momentum factor definitions instead of simply requiring a positive 3-months and 6-months returns, applying hysteresis upon rebalancing, making changes to the Composite Value ranking system and more. In the end, I have found that the best results are achieved when instead of requiring positive 3-months and 6-months returns for the stocks I buy, I require that their 3-months and 6-months returns will be better than those of the S&P 500 during the same period, even if negative. This small change made a big difference, as shown in the results below:
The average annual returns of this modified Microcap Trending Value strategy are, improved by 400 basis points to 19.55% per year. Sharpe has increased to 1.01x, and even volatility is slightly better.
Notice, however, that the asset turnover rate is pretty high, at 82%. It means that upon every annual rebalance, 82% of stocks (20 stocks) are being replaced, on average.
Lastly, I performed a rolling test, to see how sensitive the portfolio is to its starting date, and whether the strategy tends to beat the market over 3-year periods.
Each point in the graph represents either the portfolio’s returns or the benchmark’s returns over 3-years ending in that specific date.
3-years rolling test:
The results are impressive. Microcap Trending Value beats the benchmark in almost any 3-year period, and by a significant margin, often twice or more than the benchmark’s returns.
That’s probably the best Microcap Trending Value implementation that I am able to generate.
I’d like put aside Microcap Trending Value for a while, and start a new microcap portfolio with a variant approach.
We will then compare the two strategies, in an attempt to select the most suitable one for further investigation, and as an addition to our website.
Small Cap Value Momentum Portfolio
I have tried to take a variant approach, to see if it can yield even better results. While Microcap Trending Value first filtered stocks for low valuation using the Value Composite 2 metric, and then selected the stocks with the highest momentum, In the current portfolio, names Small Cap Value Momentum, I’ll take an opposite approach. I will first filter stocks for high momentum, and then select the cheapest stocks among them, using the Value Composite metric.
In constructing the portfolio I am using quite a few learnings and insights I gained over the years. To be frank, I have been working on this portfolio for more than a year, but have not yet revealed it publicly. Now that my confidence in it has risen following the extensive testing I’ve been putting it through, I am ready to share it with you.
Here is how the Small Cap Value Momentum Strategy works:
- Build a universe of stocks with a market cap larger than $30M but lower than $500. Financial stocks are excluded. So are risky stocks with dubious accounting measures and those domiciled in developing countries.
- Remove the very low-quality stocks based on the modified Piotroski score FS_Score (see Glossary for the exact definition).
- Select stock with a yearly momentum (price change in the last 12 months, excluding the most recent month) higher than 50% of the stocks in the universe.
- Do the same using 6-months momentum and 3-months momentum
- Remove stocks with a high short interest
- Sort stocks based on Value Composite 2 and buy the cheapest 25 of stocks
- Rebalance every 3 months. Sell the stocks whose valuation has appreciated, 6-months momentum has faded or quality (FS_score) has degraded.
Without further ado, here are the strategy’s results:
The Small-Cap Value Momentum Strategy achieved an annual average return of 21.56%, compared to the benchmark’s 8.32%. Standard Deviation (volatility) is 17.37%, and the Sharpe ratio is 1.13%. Asset Turnover is 91.77%.
In this case, as well, I had performed a rolling 3-year test. Each point in the graph represents either the portfolio’s returns or the benchmark’s returns over 3-years ending in that specific date.
3-years rolling test:
The figure shows that over 3-year periods, the portfolio beats the benchmark most of the time, yet not as consistently as the Microcap Trending Value strategy. Portfolios ending during the last year has made similar returns as the benchmark.
Comparing The Two Microcap Strategies
Let’s now compare the two strategies. Our first strategy was Microcap Trending Value, suggested by Jim O’Shaughnessy, yet modified by us to use the VC2 multi-factor valuation multiple, and price percentage changes relative to the market. The Second strategy was Small Cap Value Momentum, which flipped the order in which the momentum and value factors are applied.
From a results perspective alone, the latter strategy, Small Cap Value Momentum, performed better. Its average annual returns are 200 basis points higher, 21.56% vs. 19.55% over the same period of 19 years. Its volatility is slightly lower, 17.37% vs. 17.79% for Microcap Trending Value. Those two advantages result in a higher Sharpe ratio, 1.13x vs. 1.01x.
An evident disadvantage for Small Cap Value Momentum is its higher asset turnover rate, 91.77% vs. 82% for Microcap Trending Value. However, that difference is somewhat insignificant.
Another advantage of Microcap Trending Value is its higher recent performance. All other factors equal, we prefer higher recent performance. Microcap Trending Value’s returns during the last 3 years came in 66% vs. Small Cap Value Momentum’s 42%.
Comparing the 3-year rolling tests of both strategies, we like the results of Microcap Trending Value better, as they have been more recently, and overall seem to be more consistent in beating the benchmark, albeit to a somewhat lesser amount.
And the most significant benefit of Microcap Trending Value is the fact that it was tested for 90 years, from 1928 to 2018, the first 70 years by O’Shaughnessy, and the last 19 years by us. Unfortunately, since Small Cap Value Momentum is our own invention, we do not possess such a long term track record for it.