The Death Cross
On Friday (07/02) the S&P 500 made a “Death Cross”.
This ominous-sounding event occurs when the 50-day moving average crosses under the 200-day, and to some technicians it signals the start of a long-term bearish bias.
In this post I’ll put the Death Cross to the test and compare it to its bullish cousin, the Golden Cross.
[logarithmically-scaled, growth of $10,000]
The graph above shows the results of going long the S&P 500 following a Death Cross (red, 50-day SMA < 200-day SMA) versus the Golden Cross (green, 50-day > 200-day) from 1930. I’m assuming the investor traded at the close on the day of the cross.
Geek note: see end of post for assumptions about return on cash and trade frictions.
The graph appears to show that the market has been far less bullish following a Death Cross. But to be fair, the graph isn’t really comparing apples to apples because the market has spent so much more time in a Golden than a Death Cross (64% vs 36% of all days), so let’s look at the numbers…
The numbers confirm that while the market has eked out some small gains following the Death Cross, it has also exhibited far more downside risk (see drawdown) and been far less bullish on days invested (see daily return vs volatility).
The moral of the story is twofold:
1. The market has clearly exhibited weakness following a Death Cross, and investors of the longer-term variety should take heed. However…
2. That weakness has not (in and of itself) justified shorting the market. The market has still on average gained following a Death Cross, and without other confirming signals, long-term investors are better off simply moving to cash.
[Click for a summary of our posts related to the Golden Cross]
Happy Trading,
ms
Side note: some interpretations require that the 50 and/or 200-day MA be falling to call a Death Cross. This additional rule increases the bearishness of the Death Cross, but also greatly reduces the number of days that qualify as a Death Cross.
Test assumptions: (a) I’ve calculated the moving averages using the S&P 500 cash index (which investors generally use), but calculated returns based on daily dividend-adjusted data (dividends interpolated from quarterly data), (b) results frictionless (i.e. do not account for transaction costs/slippage) but could be closely reproduced in today’s market using mutual funds less part of an annual expense ratio, and (c) I have not accounted for return on cash.
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Trading Strategies, Trend-Following | 21 Comments
using vfinx, a dividend reinvested surragate for the sp 500, I got the cross the day the market soared, wednesday 7/7. ironic, huh?
Hi Michael. Thanks for another great post. Would you consider putting EMA50/EMA200 crossovers to the test also, please?
Thanks,
Rasmus
RE to Rasmus: already on the to-do list for tomorrow. Great minds think alike. michael
Wonderful! Thanks :-)
You will find it here http://prudenttrader.com/2010/07/golden-death-crosses/
S&P from 1963 Dow Jones from 1916 – Just price no dividends are considered. Enjoy!
Hi Michael,
what tools do you use to backtest?
Thanks,
PHIL
RE to Philipp: guten tag…see this post:
http://marketsci.wordpress.com/2009/01/06/faq-software-and-data/
michael
It would be interesting to see the test run on detrended data as there was clearly a bullish bias in the timespan of the test (depends if you assume whether the bias will prevail or not…)
RE to Jez: hello sir…long time no talk. My initial thought is to show a rolling X-year risk-adjusted return of the GC vs DC vs B&H to show that in all market types the relationship in the post holds. Thoughts on how you’d show results detrended?
P.S. the comment made me giggle just a bit. Usually you’d see this type of comment on a very short test during one particular market regime. Here we’ve covered 80 YEARS of market history. Having said that, you still made a good point.
michael
hehe… Ok, it is a long backtest but still with a fairly clear upside bias, and on that basis a bullish strategy should naturally perform better than a bearish one. Pushing the argument we could also say that since the industrial revolution we are on the same macro-macro trading regime…
But to be honest I am not 100% sure where I stand on detrending, especially for trend following strategies… I just feel that it is easy to adopt/embrace the bullish bias and draw conclusions that a long strategy is better than a short one (as in the Death v. Golden Cross study). If you believe the bias might turn long-term to bearish (over-population, oil disappearing, etc.), then the Death cross would probably over-perform the Golden Cross. But then, we are talking more about very long-term macro views…
With regards to detrending, you could apply detrending to every trade and compute the stats for death and golden crosses:
Step 1: The underlying trend is defined by the difference in price from the start to the end of the backtest:
Trend = LOG(Price_end / Price_start)
Step 2: Simply divide the underlying trend by the number of days in the backtest data:
Daily Trend Drift (DTD) =
LOG(Price_end / Price_start) / NumDays
Step3: For each trade, apply the detrending adjustment calculated above:
LOG(TradeReturn) =
LOG(SellPrice / BuyPrice) – DTD x NumDaysTrade
The B&H approach would obviously return a 0% performance
This is an interesting graph, Michael. The performance of the death cross shows you exactly what you are missing if you employ the golden cross. In other words, you can see that during some long time periods (decades), golden cross traders would consistently miss out on some returns and underperform the market, e.g. 1955-1970, and the 80′s and 90′s. (The underperformance is actually sizable but doesn’t look like much on a small graph.)
This is an issue for people to keep in mind if using the golden cross (and similar) moving average strategies for risk reduction. Assuming the future is like the past, you may have to put up with decades of underperformance (possibly painful tracking error) but not give up, or you’ll miss the few times when the strategy plays catch-up during clean bear market drops like 2000-2002 and 2008-2009.
RE to Scott: I agree that this strategy (like any strategy) is going to go through long periods of underperformance. Good point.
Side note: I wouldn’t characterize it as decades. Remember that in this test I didn’t account for return on cash, so some of that time spent out of the market is more productive in cash than equities even though the market might have a gained a bit in that time.
michael
Long-term moving average strategies have underperformed for several decades at a time, even taking into account a T-bill return on cash… Mebane Faber’s study of the 10-month SMA strategy had the same behavior. That timing strategy outperformed in the 1900′s, 1910′s, 1920′s, and 1930′s, underperformed in the 40′s, 50′s, 60′s, 80′s, 90′s, almost tied in the 70′s, and outperformed in the 2000′s. The end result was slight outperformance over the full century. See pages 39-40:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=962461
Looking at studies of the 50/200 SMA system, it seems to have very similar behavior. (I’m actually interested in which system might perform better and am leaning toward the 10-month SMA, or actually 15-month SMA which has done even better due to fewer round-trip trades and higher annualized return, while still offering very good risk reduction).
RE to Scott: I have volatility (risk) adjusted returns on the brain. I reran the numbers based on your email and I agree, in terms of return there have been consecutive decades of underperformance. Not by much, but a slow drip, drip, drip. In terms of vol-adjusted returns, LT crossovers (with the exception of the late 1990′s) have consistently outperformed the market over 10-year periods. michael
Yeah but what if you factior in the situatin where you hold the sticks and they pay dividends and the dividends are reinvested (which during a downturn would result in picking up new shartes at a lower price). Does that not help mitigate the damage done by holding rather than selling following a death cross?