So… I’ve made a few mistakes with this thread which I’ve been called on, so I am going to try to rectify some of them here. In what follows we are going to take an idea and create and backtest the idea to create a fully mechanical trading system based on fundamental analysis. Then as part 2, I am going to run the technical model we have discussed throughout to see if it improves performance.
What follows is a process for creating a fundamental trading system. Much of the work has been put together by Kurtis Hemmerling and he has kindly allowed me to utilise the content. Remember for my own trading as I’ve mentioned earlier I simply follow portfolios put together by people like Kurtis and then apply my trading systems.. This is simple and easy and takes advantage of the skills from people with more expert knowledge than me. However for people who wish to understand how to build a fundamental trading system I hope that what follows is fairly helpful, it also enables us to have a historical record of portfolio constituents we can apply the technical system to.
Before I go further it is important to note that there is no need to apply a technical model to what follows..that is what I like to do, but it is not necessary and it may even turn out not even more profitable, that is a question that can be answered afterwards with additional testing.
So here goes…
(Reference: How to Build a Strategic Model Portfolio – by Kurtis Hemmerling):
To build a sound investment system that generates reliable returns, you first need a thesis or a concept of why and how you will be able to do achieve bigger returns than the broad market. These investment ideas do not need to be new creations from your head - there is already decade’s worth of academic research that you can draw inspiration from. Sinner also kindly been pointed out some resources you could use earlier as well. But, there is plenty of stuff out there so you do need to dig through these papers and abstracts to discover what has already been researched and tested. As you carefully read through the research, try to think about why these rules may translate into market out-performance or a reduction in risk.
Often, the academic will come right out and tell you the rationale in the article summary, but you need to meditate on the line of reasoning to determine whether you comprehend and agree.
Finding the Idea
One comprehensive resource is called Social Science Research Network (SSRN.com). Here you will find hundreds of thousands of papers – many of which relate to market-timing, momentum, capital structure, technical analysis, dividend yields, payout ratios, valuation techniques and much more. It can be a bit overwhelming at first. Start by picking a topic that fits with your investing style and that you already have a little knowledge and experience with. If you are a long-term investor, you may want to start with value premiums and dividends. Or you might simply type in ‘long-term investing’ in the search bar and sort through the results. This process will take some time.
Below are a few papers you may find interesting:
- Adaptive Market Timing with ETFs, 2010, Glenn
- Best Ideas, 2010, Cohen, Polk and Silli
- Enhancing the Investment Performance of Yield-Based Strategies, 2012, Gray and Vogel
- Filter Rules: Follow the Trend, or Take the Contrarian Approach?, 2010, Kozyra and Lento
- How Active is Your Fund Manager? A New Measure That Predicts Performance, 2009, Cremers and Petajisto
- How to Identify and Predict Bull and Bear Markets?, 2010, Kole and Van Dijk
- Insider Trading and Share Repurchase: Do Insiders and Firms Trade in the Same Direction?, 2011, Bonaime and Ryngaet
- Investing in Stock Market Anomalies, 2011, Bali, Brown and Demirtas Is Portfolio Theory Harming Your Portfolio?, 2011, Scott Vincent
- Long-Term Volatility Forecasting, 2012, Reitter
- Market Timing & Trading Strategies Using Asset Rotation, 2010, Schizas and Thomakos
- Market Timing with Moving Averages, 2012, University of Adelaide Business School
- Optimal Portfolio Strategy to Control Maximum Drawdown – The Case of Risk Based Dynamic Asset Allocation, 2012, Yang and Zhong
- Portfolio Diversification Dynamics as a Measure of Market Sentiment, 2012, Roger
- Rebalancing and the Value Effect, 2012, Chaves and Arnott
- Relative Strength Strategies for Investing, 2010, Faber
- Revisiting the Fisher and Statman Study on Market Timing, 2011, Pfau
- Size, Value, and Momentum in International Stock Returns, 2011, Fama and French
- The High Dividend Yield Return Advantage: An Examination of Empirical Data Associating Investment in High Dividend Yield Securities with Attractive Returns Over Long Measurement Periods, 2007, Tweedy, Browne Company LLC
- Timing and Volatility Quantitative Model, 2009, Baryshevsky
- Where the Black Swans Hide & the 10 Best Days Myth, 2011, Faber
There are thousands more of such gems to be found at SSRN.com and these are just a select few to get you started. The important point here is that you need to have a solid concept of what you want to achieve and a rough idea of how to do so before you start hammering out trading rules. Some questions you may want to ask yourself are these:
• Are you looking to lower downside risk in bad markets? Have you considered market-timing? If bear markets produce high price correlation, how can you use this to your advantage?
• Are you looking to lower day to day portfolio price volatility?
• Do you want to ‘beat-the-market’ in that you want stock picks that have larger returns on average?
• Are you looking to compound dividend returns? What role does payout ratio, yield and capital structure play?
• Do you prefer to follow trends or buck them? With the crowd or contrarian?
• How active of an investor are you willing to be?
• What is your preference for market capitalization?
• Are you aware of such effects (more pronounced on smallcaps) such as Post-Earnings-Announcement-Drift, upward EPS revisions, analyst upgrades, momentum, value and short interest?
Once you have your investment idea that harmonizes well with your objectives, goals and level of risk - you can proceed to the next step where we expand a concept into a strategy.
From Concept to Crude Strategy
Over the course of this post, we will be building an investment strategy based on the Tweedy Browne paper of high yield and low payout ratio…which is in turn based on a Credit Suisse report. As well, we will include some complimentary ideas from the Gray and Vogel paper, Enhancing the Investment Performance of Yield-Based Strategies. The underlying concept in these papers is that stocks with higher yields and lower payout ratios, by their very nature, have deep value which investors may not be pricing efficiently. Consider…
A stock offers a 6% dividend yield. But does this stock have good value or not? One cannot tell simply by the dividend yield. On one hand, the company might not evening be generating profit and the dividend comes from cash reserves. This would not be of good value to the shareholder. On the other hand, a company might have so much profit that they pay only 10% of their earnings back to shareholders as a dividend. If the 10% profit represents a 6% dividend yield – this stock has screaming hot value. Of course, this is an extreme example not likely to be found in the market.
Our initial concept will be to trade stocks with higher yields with limits on the payout ratios as it suggests value – which in turn can lead to a rapid share price increase should sentiment improve. After reading the various academic papers, it is good to make a list of concepts that you feel comfortable with. Your system is liable to evolve and change during this process but you need a starting point.
Here is list of some crude investing concepts:
1. High yield
2. Lower payout ratio
3. Low debt and/or paying down on debt
4. Relative strength holdings
Initial Testing
Initially we will utilise a stock screener for testing before moving onto creating a ranking system and portfolio simulation.The stock screener excels at broad strategy testing, its strong point is to test ideas against thousands of stocks. For example, the stock screener can test all value stocks against all growth stocks in seconds with two massive portfolios containing thousands of stocks each. As we progress a portfolio simulation can then be used as the second step where we take the tested strategy and create a portfolio that holds dozens of stocks with real world constraints that mimic the actual brokerage account.
The stock screener layout looks like the picture below. The rules tab is where you enter buying criteria that must be met in order for a stock to be purchased.
The SCREENER layout looks like the picture below. The RULES tab is where you enter buying criteria that must be met in order for a stock to be purchased.
Using the FRank Function
This is where reading the FACTORS and FUNCTIONS really pays off. In our investment strategy we want high-yield stocks. How will we determine what is high yield and what is not since every market is different? Picking an arbitrary number (e.g. four percent) is weak since what is low during one period of time might be high in another. Thankfully, we have a flexible set of instructions that allows us to screen for the highest relative dividend yield.
• Under FUNCTIONS – RANKING & SORTING we find a valuable tool called FRANK (as in Function: Rank). FRANK allows us to sort stocks based on whatever factor we choose and return stocks in a certain percentile range.
We want the top 30% yielding stocks in the S&P 500. While the YIELD rule will return stocks that are above a specified yield, the FRANK function will return the highest yielding stocks regardless of the actual dividend number. The rule we are using looks like this:
• Frank (“Yield”)>70
Creating a Benchmark
Before we test our first rule, we need an appropriate benchmark. Our system is based on the S&P 500 index, therefore, we should use that as our benchmark. Yet, the S&P 500 is a market-cap weighted index and we are using an equal-weighting methodology (where every stock is held at the same weight regardless of how big or small it is). Hence, we will use an equal-weighting of the S&P 500 as our benchmark. Our slippage will be 0% as we are merely testing the validity of our strategy at this point – later we will factor in trading costs and real-world constraints. We will rebalance every four weeks and start the test in January 1999.
The redline is our equal-weight strategy of the S&P 500 with an annualized return of 8.39% (you can also select the “S&P 500 equal-weight index” as the benchmark but it does not include dividends). The blue line underneath is the market-cap weighted S&P 500 index that is the most widely publicized.
The next step is to run our high-yield ranking rule across the index and keep the best 30% of the yielding stocks. Remember that our screener includes all dividend payments – which is vital for a test such as this.
While our total return improves slightly, our drawdown increases somewhat. We might question our initial test results since not all S&P 500 stocks offer dividends. Although we know this is not true, what if the S&P 500 only had 150 dividend stocks? Our rule that keeps the highest-yielding 30% of the index would, in that case, return the entire set of dividend yielding stocks. Our strategy is to find the highest- yielding stocks of ones that pay dividends. This is a different screen altogether and it requires that we modify our universe of stocks.
But before we create an entire universe of S&P 500 dividend stocks when we can simply add a rule that states the yield on our screened stocks must be greater than 0? Wouldn’t that fix the situation? No because one is an ‘after-market’ rule modifying our universe and the other is a changed universe at the source. Consider how it is different…
• In one scenario you have a universe of S&P 500 stocks that pay dividends only. You have modified the universe at the source so only dividend-paying companies are present. Next, you add a ranking rule in the screener to choose the lowest 10% dividend payers. You will get stocks with small yields such as 0.1%, 0.2% and 0.3%.
• In the next scenario you have all S&P 500 stocks in your universe. Your first rule is to have a yield greater than 0. Your second rule is to rank the entire universe of stocks and keep the bottom 10% yields. What will happen? The first rule eliminates the 100 stocks that don’t pay dividends but the second rule ranks the entire S&P 500 universe and finds that the lowest yielding stocks do not pay dividends at all. Thus, the two rules conflict and absolutely nothing turns up on your screen.
It is best to change your universe of stocks at the source since that is what ranking rules evaluate. There is a work-around but it is better to change it at the source so you do not need to worry about it later.
Creating a Custom Universe
What we have done is create a custom universe then added a single rule that states the following:
• Yield>0
We then run a test just to see how many stocks are currently paying a dividend the answer is 405. We select “S&P 500 Yielding Stocks” as the descriptive name for the new universe. What we have accomplished is the creation of a new equity index where the constituents must be a member of the S&P 500 index plus paying a dividend.
Now we return to the SCREENER and re-run the test with this new stock universe that only holds dividend yielding stock in the S&P 500 index. We backtest the ‘top 30% dividend yield’ rule to see what effect this has on my risk and return.
The return improves slightly along with a few other risk/performance statistics although the maximum drawdown during 2008/2009 increases yet again.
Adding In the Other Rules
The next step is to build the rest of my investing rules which includes payout ratio, relative strength and debt ratios.
After some deliberation, it was decided to use an absolute rule for payout ratio. The reasoning? We do not want a company that pays out more than 100% of its profit in the form of a dividend as this is not sustainable. But we neither want to overly restrict my universe of stocks (feel free to modify at will). So the basic payout ratio rule will stipulate that dividends must be less than profit earned.
• PayRatioTTM<100
The next rule either requires there to be low debt or a reduction of net debt. How can you create a rule that allows either one condition or the other?
Creating a Forked Rule
You need to program the system to accept either condition A or condition B – yet you do not have a preference for which one. First, we need to define each rule clearly.
Our first condition is for a low debt-to-equity ratio. We create a simple rule just like our other FRank rules that will limit the debt-to-equity ratio to the bottom 50% of our universe (based on the most recent quarter):
• FRank("DbtLT2EqQ")<50
Our second condition is trickier since we have to dig into the BALANCE SHEETS as we will be comparing quarterly data to determine a change in ratios.
To create this formula, we will use the long-term debt to equity ratio from one year ago (quarterly) and divide this by the long-term debt to equity ratio of the most recent quarter. In this instance, the bigger the number translates into more debt reduction. So the rule will look like this:
• FRank("(DbtLT(4,QTR)/EqTot(4,QTR))/(DbtLT(0,QTR)/EqTot(0,QTR))")>50
I know it looks scary but if you break it down it is simple. This is a ranking rule so we begin with FRank. Next, we take the LT Debt from the same quarter last year (It is number 4 if you count backwards from the most recent quarter being 0) and divide this by the total equity in the same quarter. You divide this by the same formula – only this time you use the most recent quarter (0). Put brackets around the whole string and make it return the highest 50% (which actually means the largest debt reduction).
All you need now is to place an OR operator between the two screening rules and the system will either take the best 50% as regards low debt-to-equity or the best 50% as to reduction of debt-to-equity.
Relative Strength Testing
Our relative strength rule is simple. This rule requires the 52 week performance of our stock to exceed the S&P 500. Better performing stocks have a tendency to do over the following year. This is called momentum much literature has already been written on the subject.
The 4 Rule Investment System
The 4 rules to my investment strategy looks like this:
To Be continued...