Calling all Quants and Curious Investors
Call them factors, smart beta, systematic rules-based strategies, active indices, alternative indices, fundamental indices or simple (circa 1990) quant strategies. In most cases, the basic premise is the same:
Use a set of rules to select stocks and create a portfolio with increased exposure to one or more characteristics (factors) that experience has shown out-perform a market cap index.
The result is a “factor zoo” of hundreds if not thousands of factors. While the debate about what is or isn’t a true factor still rages, most investors and academics seem to agree on the following short list:
- Low Volatility
Where’s growth? After all, if we believe that over the long term stock prices follow earnings, then surely companies that can grow their earnings faster must out-perform?
Unfortunately most of the factor research seems to assume that growth is just the opposite of value (sloppy). Sorry, but this isn’t “growth”, its “expensive” and the two don’t necessarily go together.
Not convinced? I highly recommend reading Corey Hoffstein’s post at Think Newfound entitled Growth is not “not value” explaining the difference between growth and expensive.
So when academic research says that “growth”stocks under-perform, what they’re really saying is expensive stocks “under-perform”. No surprises there!
Why have Academics Avoided Growth?
Because growth is all about the future. Using historical data to identify growth companies is never going to work.
A factor such as value is relatively easy to measure, as it depends on a piece of past information (book value, earnings, cash flow, etc) and a piece of current information (price).
If you believe that value is explained by risk, then cheapness is simply a proxy for higher risk. If you believe that value has a behavioural explanation, then cheapness is a sign of investor over-reaction to a situation that will eventually improve (i.e. reversion to the mean). Either way, future performance correlates fairly neatly with information that you currently know.
Growth is totally different. It depends on things that haven’t happened yet. How do you measure something that takes place in the future? You can’t.
The best you can do is to look for companies that are currently growing fast. For example, you can look for companies with strong sales or earnings growth, high margins and high-levels of profitability. For example, James O’Shaughnessy’s excellent book: What Works on Wall Street, Fourth Edition: The Classic Guide to the Best-Performing Investment Strategies of All Time.
The problem with this approach is that a) historical growth tells you little about future growth and b) such obvious and measures of growth are likely to be already reflected in the current share price.
There may also be a behavioural bias at play. Academics generally dislike publishing papers where their hypothesis is rejected. This is understandable. Its much harder to get funding and recognition for ideas that don’t work.
Solving the Growth Puzzle
All of the above makes designing a systematic rules-based growth strategy an awesome challenge. I’ve been thinking about how I might approach the challenge for a few months now. I wouldn’t say that I’ve cracked it, but I think I’ve identified a set of principles to guide my research:
- Accept that no investor can identify growth in advance because it simply hasn’t happened yet!
- The usual approach of looking for a fairly obvious ratio or financial statement item, such as earnings growth, probably will not work.
- Growth is certainly not guaranteed. But it is logically (i.e. not regression-based using historical data) associated with certain behaviours and conditions that you would reasonably expect of a growing company.
- Select stocks using measures that indicate that company management are behaving in a certain way and/or certain conditions might exist.
- This has to be done progressively as new information emerges (see point 1). Portfolio construction is all about “pruning” away potential growers that didn’t grow while keeping potential growers that are growing.
- This means that most of the usual portfolio construction techniques used by systematic strategies – e.g. equal weighting, periodic rebalancing, etc, probably won’t work.
- Over time, “pruning” (portfolio construction) is likely to be a much larger driver or performance than stock selection .This is because it is the primary mechanism for incorporating new information into the portfolio.
A Simple Example
Imagine that you are the CEO of a company that is profitable and has fantastic long-term growth opportunities. How would you run the business? How would you allocate capital?
If I were such a CEO, I would retain 100% of the companies earnings to fund future expansion of the company. In other words, my divided yield would be zero.
Its also better for the shareholders if the company can fund growth using retained earnings rather than issuing additional equity.
Not every company with a zero dividend yield is a growth company. For example, a company that customarily pays dividends may be forced to suspend its dividend due to falling on hard times. Its likely that these sorts of companies have a lot of debt. Debt holders have the prior claim over the companies assets and need to be paid interest (and sometimes principle) before dividends can be paid.
Avoiding companies with too much debt helps to separate the companies that are retaining 100% of their earnings to fund growth – instead of repaying their creditors.
Growth companies may not have much debt for a few reasons. They include:
- Their value resides primarily in their future potential, so they may not have much in the way of tangible assets to lend against.
- They may also have no revenue or earnings, which makes servicing the debt impossible.
- They may be risky ventures. Consequently, they are better off being 100% equity funded.
- If they are profitable, they may not need the debt as they can self-fund future growth
Selecting profitable companies helps you avoid companies that don’t have any debt because investors aren’t crazy enough to lend them money. It also de-risks your growth strategy by focusing on companies that have a real product or service and are currently making money.
OK, so here we have one behaviour and two conditions that we can logically associate with a profitable company that has great long-term growth prospects. What happens if we create a portfolio using these criteria?
I need to make a couple of things clear before we get to the screen and the results.
I am not a quant. This is just a very simple back of the envelope test to illustrate an idea that needs a lot more work. Please remember that before you start tweeting about the impact of transaction costs, the Fama-French Carhart alpha, factor exposures, out-of-sample evidence etc.
These are all legitimate questions, questions that I ask myself every time a fund managers pitches to me in my role as an institutional investor. But I’m only one person, working on this out of curiosity, and doing it in his spare time, which is almost non-existent as we’re expecting our first child.
I’m planning to do a lot more work on this idea and I welcome your suggestions, comments and genuine offers to collaborate and help.
And of course this is not investment advice, if you try to do this you’re doing it at your own risk! Past results may be very different from future results.
Market Fox Growth Screen (1st draft)
Period = 20 years to 31st December 2016.
Universe = Russell 3000. Includes some smaller stocks but not stocks that are too small for an institutional fund manager to invest in. Avoids tiny micro-cap stocks that are illiquid and expensive to trade.
- Dividend Yield = 0%. Want companies that are retaining 100% of their earnings so that they have maximum growth potential.
- Total Debt-to-Equity <=33%. Avoid companies that a) have no dividend because they’re in financial trouble, or b) can’t self-fund future growth
- Ratio of ROIC:WACC >= 1. Company has to be earning an economic profit otherwise growth destroys value. Avoid companies that have no debt because they have no earnings with which to repay the debt.
- Sales growth >= 6%. Look for companies that are growing faster than the general economy.
- Market capitalization weight each stock. This creates a portfolio that a large institutional investor could reasonably own.
- Rebalancing Frequency = Semi-annual. Hopefully reduces turnover and keep s transaction costs low. The three screening criteria used are unlikely to change too much on a monthly or quarterly basis.
Here’are the results. I’ve split the results into 3 stages to better demonstrate the role played by each of the screening criteria.
- Stage 1 = Companies with zero (0%) dividend yield
- Stage 2 = Companies with zero (0%) dividend yield, total debt-to-equity less than 33% and a ratio of ROIC:WACC greater than 1. This is the model described in the example above (highlighted in red).
- Stage 3 = Stage 2 with the added requirement that the company have sales growth of at least 6%.
|Mean Annual Return (%)||
|Mean Active Return (%)||
|Min Return (%)||
|Max Return (%)||
|Tracking Error (%)||
|Correlation (to Russell 3000)||
The first thing you’ll notice is that adding the sales growth hurdle (Stage 3), a “traditional” growth metric, reduces performance. This is something that I need to investigate further.
The second thing that you’ll notice is that the zero-dividend screen beats the market on its own by a considerable margin. The majority of this performance came in two big lumps, the tech bubble from 1996-1999 and during 2012-2014. Its also an incredibly risky portfolio, both in absolute and relative terms.
There are several academic studies showing that companies with low dividends under-perform. So why the difference? My guess is that it has to do with market-capitalization weighting rather than equal weighting. This means that the screen portfolio is likely to invest more in large non-dividend paying companies, such as:
- Berkshire Hathaway
- Apple (for most of the period it didn’t pay a dividend)
Obviously this is something that I need to do more work on.
The third and most important thing you’ll notice is how the debt and profitability criteria reduce both absolute and relative risk while increasing return. The returns are also far more consistent. The stage 2 portfolio only had 16 periods of negative half-year active performance out of 40. In other words, it out-performed the Russell 3000 in 60% of the time (6-month periods).
The median number of stocks held by the Stage 2 screen was 113 (minimum = 4, June 1997, maximum = 176, June 2004).
The median turnover (semi-annual) for the Stage 2 screen is 26.78% (minimum = 9.79%, maximum = 80.37%).
I’ve used market-capitalization weighting and semi-annual rebalancing because its simple and easy to do ad its good enough for this example. But it’s not what I would do if I were running this with real money.
I don’t have the skills to back test a more complicated methodology. But I do have ideas about what it would be and why I think it would further improve results. I believe that portfolio construction is where the key to unlocking a systematic rules-based growth strategy really lies.
You may be wondering if I data-mined this result by trying several factors over different time periods and markets to see what worked historically. You’ll have to take my word for it, but this is literally the first and only screen that I tested.
What I did “mine” was my intuition – based on years of experience – and logic to realise why the conventional approach to growth can’t work and therefore where I might begin to look for alternatives.
My purpose here isn’t to claim that what I’ve outlined above is THE way to invest in growth using a systematic rules-based strategy. It isn’t. The main points of this post are that:
- Growth is difficult to measure because it’s all in the future.
- Consequently, using historical information can’t work..
- Instead, its more helpful to look for conditions and behaviour that promote growth rather than look for evidence or proof of growth (because it doesn’t exist yet).
- Points 1-3 explain why I believe there’s very little research into systematic rules-based growth strategies.
- Portfolio construction has to do more than just rebalance or refresh a given factor exposure. Common portfolio construction techniques are unlikely to work.
- There has to be a mechanism to incorporate information on growth (or lack thereof) into the portfolio, as it emerges over time.