THE DEATH CROSS DOES NOT LIVE UP TO ITS HYPE
Tom McClellan – McClellan Market Report
Chartists are all excited about the August 2011 “Death Cross” that appeared in charts of several market indices. This is an event where the 50-day simple moving average (50MA) crosses below the 200-day (200MA). It is also variously referred to as a “Dead Cross”, but the meaning is the same.
The opposing condition is known as a “Golden Cross”, which is when the 50MA crosses above the 200MA.
The idea behind this is that the two moving averages are sensitive to different length trends and cycles in price movements. The 50MA is more sensitive to price changes, and so when it crosses the longer moving average it is saying that a move is getting under way. Or perhaps more accurately, it is announcing that a move has already gotten under way, since it will lag the turns in prices.
I have to tip my hat to whoever came up with the jazzy names for these signals. They join others like the Miekka “Hindenburg Omen“, the Ohama “Titanic Syndrome”, the Zweig “Breadth Thrust“, and others whose catchy names capture the public’s attention. Dow Jones has even created an index based on switching in and out based on using a 50MA and 200MA crossover system.
And the Direxion family of ETFs has filed paperwork with regulators for a launch of the Direxion Large Cap Tactical Advantage Shares, which intend to shift asset allocations based on these moving average crossing signals for that Dow Jones index.
The Death Cross is clearly a concept which has captured the public’s attention. In terms of being a great signal, though, Golden Crosses and Death Crosses do not have a stellar long term track record. The chart below shows a hypothetical equity curve if one had traded the SP500 based on these signals.
This study assumes that one could “own” the SP500 if its 50MA was above its 200MA, or exit out to cash if the opposite condition occurred, using the SP500 Index and Cowles Index data. This quick look is not meant to be a thorough analysis of an actual trading system, which would incorporate additional factors like transaction costs, dividends, interest on cash, and other factors which would affect actual returns. But creating a hypothetical equity curve like this is a good way to investigate whether there is really any merit to a trading signal one might be contemplating.
What this quick study shows is that trading the 50MA and 200MA crossovers does beat buy and hold for the 73 year period studied, but not by all that much. It softens some drawdowns, but misses efficient reentries.
Exiting on a Death Cross would have kept you out of the ugly declines in 1930-32, 1974, and 2008-09. So if you believe that you are in an environment where lengthy declines of more than 30% are going to be the order of the day, it can be helpful.
But someone trading based on these rules would have ridden down the 1987 crash, and then sold on Nov. 5, 1987. Reentry did not come until June 28, 1988, when the SP500 was already 9% higher than the exit point. Similarly, one would have ridden down the Flash Crash decline in 2010, and sold on July 2, 2010, the exact day of the bottom.
The August 2011 Death Cross may be another example of this very principle that Death Crosses quite often mark important bottoms. It takes lengthy and protracted declines in the market to make the exits profitable. Declines that arrive too quickly result in whipsaws, and missing out on the reentries until well after the rebound has started, which hurts system efficiency.
The point here is that we should not get drawn into the crowd behavior of getting all excited about the latest jazzy-sounding name for a technical phenomenon. Focusing just on the trend change aspect of a Death Cross signal can prevent us from seeing how it can do a good job of marking nice intermediate term bottoms.
Related Charts
Jul 15, 2011 Long Term Haurlan Index Divergence |
Jul 01, 2011 A Breadth Thrust Signal |
May 27, 2011 Commercial Traders Foretell Market’s Movements |
Mar 11, 2011 DJIA Price Oscillator |
I could be off, but to the naked eye, it looks like the volatility of the “Cross” performance is lower than the “S&P” line.
Given the research in behavioral economics, maybe risk-adjusted returns are a more appropriate comparison?
Mac, I think you’ve made an excellent point.
Tom, did the comparison included careful accounting of dividend and transaction cost?
As I mentioned in the article, I did not include other factors which would certainly affect actual portfolio performance, such as transaction costs, reinvestment of dividends, interest earned while sitting in cash, taxes, etc. Those factors do matter a lot, but they also complicate and sometimes contaminate the analysis when one is really trying to evaluate a simple phenomenon like trading a MA crossover event.
During some periods in history, dividend yields and money market rates are comparable, so throwing them both out does not really matter. At other times in history, they are widely apart, and so including them can skew the answers one gets and perhaps lead one to mistaken conclusions about the system performance.
We should also remember that money market funds are a relatively recent innovation, as is the ability to “trade” the SP500 via products like SPY, mutual funds, or futures. Those did not exist 50 years ago like they do today, so it would not be right to pretend that we could have traded this way back then. But we can see what the system’s performance would have been if we could, just as a way of evaluating the signals themselves.
“On The Money” is correct that there are mechanical trading systems that do show a much better long term track record than buy and hold. But not all of them have jazzy sounding names that catch the media’s attention like something called the “Death Cross”.
As for the comments made by James and Conventional Wisdumb, it is true that the hypothetical equity curve does indeed show dampened swings compared to the SP500 Index. So one could do some sort of risk adjustment process on the data to evaluate returns that way, but that too would be adding another ingredient into the mix.
The Y-axis denotes hypothetical dollar values of a portfolio with a starting value of $10,000. This is a standard convention that a lot of analysts use for trading system analysis. I left such explanations off the chart to reduce the clutter. Scaling the chart logarithmically allows us to better see the slope of the growth rate, whereas an arithmetically scaled chart would appear to the eye to be flat for 70 years and then a big swoop up at the right end of the chart.
From January 1928 to August 19, 2011, the buy and hold return on the SP500 Index (again, ignoring dividends, etc.) takes this $10,000 portfolio to $535,302. That is an annualized growth rate of 4.91% over the 83 years in this study.
For the same period, this MA crossover system grows to $906,068. That seems a whole lot bigger than buy and hold, but its annualized growth rate is just a little bit higher at 5.58%. This helps show the power of compounding just a slightly higher return over a long period.
Interestingly, if I had concluded this study in 1999 or 2000, the buy and hold record would have been better than the MA crossover system’s record.
One final point I did not address about this topic is whether or not a person could ever bring himself to continue trading such a system. Take yourself back to 1987, and imagine that you were trading this system. You just rode down the biggest crash in decades, and then your system gets you out on Nov. 5, 1987. And it does not allow you to get back in until the middle of 1988, such that you miss a big chunk of the rebound.
This process gets repeated multiple times over the years, getting out late, and missing the rebound. That kind of behavior tends to get frustrating, and a lot of people could not stick with such a system, regardless of what the long term track record looks like.
Point on!
Risk adjusted, returns would be far greater than shown in the chart. Never mind being better able to sleep at night during protracted downturns.
Besides, in an era of low single digit returns, being able to improve only a few percent is very significant.
It’s interesting how often these simple mechanical timing systems are decried by analysts and commentators, many of whom are in the business of selling investment advice with a far poorer long-term track record.
Mebane Faber’s paper ‘A quantitative approach to tactical asset allocation’ (downloadable for free) demonstrated another simple method for timing the S&P500 which continues to outperform the index as it has done for more than 100 years. Combined with proper diversification, these methods can generate consistent and highly impressive long-term gains.
It’s true there are disadvantages, in terms of missing occasional sharp turns at tops and bottoms. But over time, they are far outweighed by the advantages of exiting before those occasional huge bear markets which can decimate your long-term returns.
You have to use a combination of Fundamental,Technical, & Behavioral Analysis to get a full picture. We all know EMH isnt true, anymore at least. So its safe to say trying to trade and judge price direction off of group dynamics and the herds movement would be a logical strategy.
They shouldn’t be using a log scale on this data given the size of the chart it creates the impression of a small distance between the two at the current terminal point. I am not even sure what the scale on they y axis means.
I suspect if it used percentages on a shorter timeframe the differences would be more easily seen. Even in the above view, it looks to me like this technical trade would blow away buy and hold over the last 30 years.
It gets back to “have a system and stick to it” knowing full well no system is perfect. But what discipline does give is a good night sleep. That alone is more valuable than beating the averages by a couple of percent. So if death crosses and the like are your thing, stick to it and be happy. You will only lose if the entire system ends. Then guns and ammo are the trade of the day.
You guys all made really good points. You can see the depth of the losses with buy and hold while the death cross is relatively smooth. However with the y-axis all weird, you can’t really decipher how much they are different.
See http://en.wikipedia.org/wiki/Logarithmic_scale for a primer to help understand this style of chart scaling. In brief:
“A logarithmic scale is a scale of measurement using the logarithm of a physical quantity instead of the quantity itself.
“A simple example is a chart whose vertical axis increments are labeled 1, 10, 100, 1000, instead of 1, 2, 3, 4. Each unit increase on the logarithmic scale thus represents an exponential increase in the underlying quantity for the given base (10, in this case).
“Presentation of data on a logarithmic scale can be helpful when the data cover a large range of values. The use of the logarithms of the values rather than the actual values reduces a wide range to a more manageable size.”
So it is not so much a case of the Y-axis being “all weird” as it is a case of cramming 83 years of data into one small chart, and of the two sets of data staying relatively close to each other over that time span.
Without QE to save the day the “Death Cross” should be taken seriously.
bottom line is moving average crossovers alone are “lagging” indicators. not the best way to time a trade.
This only goes to show that technical indicators are not always reliable, as other circumstances influence charts outside the norm. These factors have no knowledge of how the chart is supposed to go, just as the Universe doesn’t know it is August the 20th, 2011. Heck, the Universe doesn’t know it’s 2011.
Nice piece, Mr McClellan! I just read another article which was suggesting that this could be used as a market timer…
Hear is one supporting your thoughts…One thing of interest, however, is the 12 month returns after the event…
http://www.schaeffersresearch.com/commentary/content/is+there+life+after+the+djia+death+cross+by+the+numbers+investigates/observations.aspx?ID=101226
stocktiming points out THIS time C-RSI is negative….last time was positive.
http://www.stocktiming.com/Monday-DailyMarketUpdate.htm