A SEASONAL QUANDARY
Tom McClellan – McClellan Market Report
When every other segment on financial TV is about the question over whether to “Sell In May And Go Away”, it is natural to wonder whether the belief in seasonality has become overdone. When everyone believes in something so thoroughly, especially about the stock market, can it still remain valid?
If you were to average together the market’s performance in one year chunks of time, it might look something like this week’s chart. It is made up by averaging together the performance of the DJIA each year from 1976-2007. I throw out 1987, just because the magnitude of the up and down moves that year tend to drown out the other voices to an extreme extent. I also prefer not to include the most recent years, when the Fed has had its foot on the neck of the financial markets. If you want to get an average picture of what “normal” looks like, don’t include the abnormal and steroid-enhanced.
The first point that jumps out at you about the Seasonal Pattern is that the very old saw about “Sell In May And Go Away” is off by a month. The average seasonal pattern really tops out in June, or at least that has been the case since 1976. This may be a change from earlier years when traders vacated New York City at the earliest opportunity to head to “the shore” in May and stay away for the entire unbearable muggy summer. Now, thanks to the insights and advancements of Willis Haviland Carrier, traders can stay in New York and toil away well into June.
So the old saying about “Sell In May…” ought to get updated to “Sell In June, and, um, (something witty that rhymes with June).” It is not quite as satisfying this way, which perhaps explains why the old version persists.
So, now we know that we should not sell in May, if we are to believe the historical pattern. But here’s the problem: the current market is off schedule by a week. If we were to align the current DJIA pattern with the average seasonal pattern, it does not come out quite right. I’ve found that I have to offset the pattern by a week just to get the structures to line up. And once I make that adjustment, they line up quite nicely:
The implication is that the current market is running a week behind schedule. You can insert your own joke here.
The further implication is that prices should keep on trending higher through June and into July. That’s if you believe the annual seasonal pattern.
The truth is more complicated, and it boils down to this: It’s different in an election year.
This chart of the Democratic Presidential Cycle Pattern substitutes the Seasonal Pattern’s 1-year chunk of time for a 4-year chunk, averaging it together in the same way, but with a twist. In this version, I am including 4-year chunks of time only from when a Democrat has been in the White House. When viewed in this way, the whole rule about “Sell In May…” goes completely out the window. May’s mission in election years seems to be to convince everyone to sell, and May seems to have recruited all of the business news reporters to tell that story on its behalf.
The real story during election years, especially when a Democrat is in the White House, is that May is a month for correcting, while June is a month for screaming higher. The timing of the turns is more important with this tool than calculating the magnitude of the move. Coming up just ahead is a bottom due May 14, which our subscribers have known about for the past month and so they have been anticipating its arrival. There is also a bottom due at the end of the May, which helps to illustrate the point that sometimes a bottom is what prices go down into, and sometimes it is what prices go up out of.
A big strong June and July is wholly contrary to the old saw about “Sell In May…”. Most of the time that rule does work, at least in part, but in election years a whole different rule goes into effect. If the correlation persists this year as well as it has been doing up until now, we can look forward to a big rally in June before the market finally enters a plateau in July, when the media’s attention is tuned to the campaign promises being slung by the presidential candidates.
So I suppose that the revised motto should really be, “Sell In May And Go Away, Except If It Is An Election Year, And Seasonality Is Running A Week Behind, In Which Case Expect A May 14 Bottom, And A Really Strong Month Of June”.
As far as mottos go, that one will never catch on.
Related Charts
Feb 03, 2012 Eurodollar COT Indication Calls For Big Stock Market Top Now |
Jan 13, 2012 RASI Above +500 Says Bull Market Not Done |
Nov 05, 2010 Entering the 3rd Presidential Year |
8 Comments
I definitely appreciate the article and like a lot of what Mr. McClellan does, but I must ask a question in regards to one of the foundational premises of the whole article:
“I also prefer not to include the most recent years, when the Fed has had its foot on the neck of the financial markets.”
In order to throw out that data for use in predictive modeling, wouldn’t one have to assume the Fed had taken its foot off the neck of the markets? Does anyone think that is currently the case?
I often think historical patterns going back 100 years are less useful than those going back 40. Even though you have less time and fewer samples, you have more that fit in to the current monetary system – and that is a big difference.
“Sell in June, don’t be a loon.”?
We are still very early into the month of May, but based on Sunday night’s future’ market showing SPX is likely to be off another 14+ points on Monday, those who followed the “sell in May” maxim on 5/1 have done very well so far.
Cullen, early this year in Jan & early Feb you warned your readers to lighten up on bonds, and you were right, they sold off into the middle of March. But no one on this site, even none of the commenters that I know of, advised us to go back into bonds. We are seeing a tremendous flight to quality Treasury bond rally right now – how much further does it have to go??
1987 happened.
As did the move up into it. In fact isn’t that what Paul Tudor Jones used to analog his call leading into 87-88.
The 1929 analog plotted almost with an ink jet done daily?
Why do intelligent people throw out data so easily. In fact how can u rationalize this? How do you intelligently choose what suites your bias?
Should Stanford fans throw out the “band is on the field” play. I mean that will never happen. Just a one off… Yep throw that out.
I remember at the CFA institute session in 2009 listening to one of the speakers throw out 2008-2009 as a sample as well as 1929 because it was such am extreme event.. Maybe… Ah forget more analogies..
It happened. Use it. If u dont use 1929- then throw out 21-29 also
If u don’t use 87then throw out 85-87 and 88-95 .
Better yet .. As Montier likes to point out.. Next time some expert starts erasing years to fit there thesis run don’t walk away from this jibberish
I am looking to initiate long positions around 1300 down to 1200 within the time frame of late may to late june.
i respect b ferro’s analysis but like i stated in other posts, i rather hide out in tsys right now and buy in around 1300-1200.
I think we’d all rather buy in at 1200 than at 1350.
The question is whether it gets back there and when.
P. Neophyte asked whether the Fed does not indeed still have its foot on the neck of the financial markets. That’s a reasonable point. But Opn. Twist is far smaller in magnitude than either QE1 or QE2 in terms of the net increase in money supply and the Fed’s balance sheet.
Better than applying pure logic about the Fed’s actions is looking at the actual data. The current market seems to be following the Democratic Presidential Cycle Pattern quite well, and so any ideas we may come up with about why it should not work have to be seen as secondary to noticing that it is working. “Is” trumps “should”.
As for VII’s point about setting aside 1987, I am in general agreement. I don’t like throwing out anything (you should see my garage!). But when 1987 is included, the result ends up looking like 1987 and not like anything else. The magnitude of what happened in 1987 is so huge that it drowns out all of the other voices. That’s what happens with averaging, and so we have to grapple with what to do about it. Setting aside the outliers is a recognized statistical technique, especially when there is a logical reason for doing so.
And again, the outcome is what matters most, and the ex-1987 version just works better. You all are free to do it yourself and include 1987 if you think it works better for you.