The Last Day of the Month Blahs
We’ve covered “turn of the month” seasonality before. The market exhibits consistent strength the first and last handful of days of the month (read more & more). This observation is included in the free daily State of the Market report.
But there is one particular day in that bullish period that has been pretty blah over the last 20 years: the very last day of the month.
[logarithmically-scaled, growth of $10,000, monthly-interval]
In red, the graph shows the results of a trader only going long the S&P 500 index on the final day of each month (from the previous day’s close) from 1950. Geek note: results are frictionless.
As a benchmark, I’ve included a fictional portfolio in grey that returned each month’s average daily return (excluding the last day of the month). If there was nothing special about the last day of the month we’d expect the red line to be much more volatile than the grey (because it only represents a single day instead of the average of many days), but returns to be more or less inline.
The graph shows that from 1950 to 1990, the last day of the month was consistently much more bullish than the broader market, with an average daily return of 0.17% (vs 0.02% for all other days).
But notice how since then the edge has apparently disappeared. Below is the same chart starting from 1991.
[logarithmically-scaled, growth of $10,000, monthly-interval]
Where did that last day of the month strength go?
I would speculate that it shifted forward a day to the first day of the month.
Below is the same test, but this time looking at the first rather than the last day of the month, from 1950.
[logarithmically-scaled, growth of $10,000, monthly-interval]
Right around the time that the last day of the month broke down, the first day of the month took off, and at this particular moment in history, is the strongest of the turn of the month plays.
Summary
Turn of the month seasonality appears to be alive and well in today’s market, but the very last day of the month specifically has been blah over the last 20 years. It appears that that strength shifted forward to the first day of the month, which is now the strongest of the turn of the month days.
Happy Trading,
ms
P.S. this observation will be included in the soon-to-be-launched monthly Super Duper Seasonality map.
. . . . .
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: Time-based | 7 Comments
nice,
keep up the good work
funny how the breakdown in the 90′s is similar to the breakdown of the famous Turtle System performance (starts at 216% CAGR for the first 15 years and ends up completely flat since the 90′s).
On the other hand by changing the system parameters (in hindsight: making the system more long term), the performance appears much better (ie robust/smooth).
I am wondering what the best systematic methods are to switch from one system to the next/alter system parameters: walk-forward testing with periodic robust re-optimisation, equity curve trading of a pool of systems (ie start trading each system when equity > MA), trading using regimes…
Have you found any good approach for this sort of problematic?
PS: for reference, the post where I discussed the Turtle system: http://www.automated-trading-system.com/turtles-just-lucky/
Mike: some time ago you published a study where you showed the results of a system that used an average of the signals for the three time frames of the State of the Market Report. Can’t find it. Could you please be kind a enough to post a link to that ?
Thanks in advance.
eber
RE to eber: here you go sir…
http://marketsci.wordpress.com/2010/01/15/testing-the-sotm-the-aggressive-swing-trader/
…and also was accompanied by this report:
http://marketsci.wordpress.com/2010/01/14/testing-the-sotm-the-sleepy-trend-follower/
michael
One important thing that happened back in 1995 was the controversy over an alleged “mark up” of Fidelity Magellan’s stocks the last three days of every month in 1995 – Bill Fleckenstein calculated that while Magellan was ahead of the S&P 500 for most of the year, more than all of the outperformance came in the last three trading days of every month, so that PM Jeff Vinik was marking his stocks up.
Here’s the money quote from the NY Times:
“Consider the research promoted last December by a Seattle money manager, William Fleckenstein, who contended that a group of technology stocks in which Magellan had big positions mysteriously ran up in price in the final three trading days of each month from August through November, implying that Magellan was somehow forcing prices up to make its record look better.”
http://www.nytimes.com/1996/05/05/business/mutual-funds-who-s-out-to-topple-jeff-vinik.html?n=Top/Reference/Times%20Topics/Subjects/F/Futures%20and%20Options%20Trading&pagewanted=4
EVEN IF you think that Fleckenstein’s allegation was totally false (but I think you have to admit the statistic happening by chance might be unlikely), the allegation by itself would likely have led PM’s to trade not quite so aggressively on the buy side the last couple of days of the month (and quarter and year end) for a while.
Or not.