Sunday, January 07, 2007

DIA, The Diamonds ETF: What Happens After a Dow Reversal Week?

Past TraderFeed posts have focused on daily reversals in the S&P 500 (SPY) and the NASDAQ 100 (QQQQ) markets. During this past week shortened by the New Year's holiday and the day of closing to commemorate President Ford's passing, we had a downside reversal week in the Dow Jones 30 Industrial Average (DIA). By that, I mean that the week traded above the high of the previous week, but closed down for the week overall. Do we see any historical patterns with respect to such downside reversal weeks? I decided to go back to February, 1998 (N = 459 weekly periods) and take a look.

First off, let's take a look at what happens after a week in which we make a price high compared with the week previous (N = 257). One week later, the Dow averages a gain of only .03% (136 up, 121 down). Conversely, when the Dow fails to better its previous week's high (N = 202), the next week in the Dow averages a gain of .21% (114 up, 88 down). We thus see a mild tendency for stronger weeks to be followed by subnormal returns the following week and weaker weeks to be followed by better returns. This tendency has been somewhat more pronounced since 2004. When we've had a week that's made a higher high (N = 89), the next week in the Dow has averaged a loss of -.07% (41 up, 48 down).

Now let's break down those weeks in which we made a higher high into those occasions in which the Dow finished the week higher (N = 178) vs. finished the week lower (N = 79). Note that this latter group (traded above the previous week's high, but closed lower on the week) represents our downside reversal group. When the Dow has made a new high and closed higher, the next week in the Dow averaged a loss of -.04% (97 up, 81 down). When the Dow made a new high but reversed and closed lower on the week, the next week in the Dow averaged a gain of .21% (39 up, 40 down). In other words, the Dow was just as likely to close lower as higher one week after a reversal, but the average size of the gains was larger than the average size of the losses.

Interestingly, however, if we look out to *five* weeks following the Dow downside reversal, we see a more distinct pattern emerge. When the Dow has made a week-over-week price high and also has closed higher on the week, the next five weeks in the Dow have averaged a gain of only .04% (96 up, 82 down). When, however, the Dow has made a week-over-week high and reversed to close lower, the next five weeks in the Dow have averaged a gain of .94% (48 up, 31 down). By way of comparison, the average five-week change in the Dow over the entire sample has been a gain of .52% (268 up, 191 down). It thus appears that, if there is any tendency at all in the data, it is for downside reversal weeks to reverse over the coming weeks and display relative strength. We don't see any bearish edge to such a reversal, contrary to trader lore.

Ah, but wait. If we just focus on the time period from 2004 to the present, we find 31 occasions of downside weekly reversal in the Dow. One week later, the Dow averaged a *loss* of -.25% (12 up, 19 down). Interestingly, however, there was no downside edge on a five-week basis. Thus it seems that during the past three years, downside weekly reversals have tended to continue in the short run before themselves reversing. Indeed, 24 of the 31 reversal weeks traded below the lows of their reversal moves during the following week. Based on that view, a short-term or daytrader might consider starting the week out with a bearish leaning.

Still, there's one element missing from the above analyses. It's an important element, and it offers an important trading lesson that we need to consider with respect to historical market patterns. That missing element is context. Is the downside reversal occurring within the context of a larger market upmove or in a flat or downtrending market?

Since 2004, when we've had a downside reversal week that also made a five-week price high (N = 19), the next week averaged a loss of -.54% (6 up, 13 down). The following week traded below the low of the reversal week on 16 of the 19 occasions. Five weeks out, the Dow averaged a loss of -.44% (10 up, 9 down)--no evidence of reversal at all. By contrast, when the downside reversal week did not occur at a five-week high (N = 12), the next week in the Dow averaged a gain of .18% (6 up, 6 down) and a sizable average gain of 2.1% five weeks out (9 up, 3 down). We need to be careful because these are small samples. Still, they suggest that downside reversals that follow weeks of strength have not tended to bounce back and produce superior returns over the next several weeks.

What we find is that regimes shift over a period of years: the patterns of 2004-present are different from those of, say, the late 1990s. Moreover, the same exact price pattern may have different expectations depending upon the larger context in which it exists. The recent pattern of weakness following a weekly Dow downside reversal in a strong market is one that I will be keeping in mind for next week. Since 2004, weekly Dow reversals to the downside have indeed lived up to the trader lore and produced subnormal returns the following week.


Anonymous said...

On some random day in November I set up a portfolio to track indices relative to that day. There was nothing magic about the day, it was simply because I wanted a sense of what the various indicies were doing relative to each other.

Well, the dow is still up over 100 pts since that day, the nasdaq is down about 10 and the s&p is up about 10.

Given the stellar performance of the nasdaq from summer to that point, well, a little claw back isn't unreasonable. The nasdaq was up as much as 21% over its summer low, which is utterly enormous. It is still up 19.6% since its summer low.

If you could do that every year, in 10 years you'd have turned $1000 into $6000. Well, there has to be some correction in there.

You do like the macro economics.

I like the micro economics, but it sure makes sense to use the macro to help decide when to buy.


Brett Steenbarger, Ph.D. said...

Hi Deborah,

Thanks for the comment and the blog link. I do think you have an important point: with longer-term holding periods, the macro variables become increasingly important. For an intraday trader (my own holding time averages 20 minutes), the big macro factors are less crucial, although there are short-term intermarket relationships that can be quite important. (I'll be posting something on that tonight on the Trader Performance page of my personal site). Thanks for the perspective--


Glen said...

I find it useful, as a short term trader, to have a sense of the macro trends. For example, as oil prices have trended downwards, my bent has been to look for opportunities to get short oil/gas stocks. When oil prices were rising, I only considered long trades. Same applies to the Dow and the Russell. They have started trending downwards, and these incipient trends have been semi-confirmed by deteriorating technical indicators. So I will start out looking for shorting opportunities in DIA, IWM, SPY, etc. this coming trading week, and modify my expectations as information arrives.

As for the statistical distribution of historical patterns, I'm from Missouri regarding their usefulness. I was once fooled into choosing Door #2 on the basis of what appeared to be meaningful statistics, but Door #1 was the prize and Door #2 was the tiger. I try to trade more on the basis of what I see now, not what I think, but do take the historical probabilities into account to build an expectation of duration of trend. Seems that's what you are also trying to determine, and the verdict appears to be that chances are relatively remote for a major decline or rally presently. I'm still going to keep an eye out for the tiger.

Brett Steenbarger, Ph.D. said...

Hi Glen,

You've stated that nicely; I think it makes a lot of sense to take a Missouri approach to historical market patterns. They provide a heads up for what might come, but the market has to "show me" that the pattern is playing itself out before I'll get on board. Thanks--


Science_of_Speculation said...

Dr. Brett,

It seems to me that the weakening in the Dow might be related to the current strengthening of the Dollar.
Perhaps large American value stocks are seeing profit-taking on speculation that a stronger dollar will hurt overseas revenue?

It might be that the current asset reallocation we are seeing is in response to the overwhelming recent economic data suggesting that a soft landing is increasingly likely. If so, it would seem that funds and international market participants are betting on increased odds of rate hikes later this year. Your recent analysis of the 10-year interest rates rally in December confirms. This also explains why commodities like gold are being sold.

As you said a few weeks ago, value has become the new growth. But what about real growth? Looking at large cap growth like SMH, it appears that volume dried up near the bottom of a range in December. Actually, SMH had the lowest volume this past December in the last 2 years. QQQQ and SMH are showing some strength at the start of the year. Is it time for growth and technology to lead again?

Let the interest rate hike speculation begin!

By the way, great site. Your intermarket analysis has inspired me to begin my own studies!


Brett Steenbarger, Ph.D. said...

Excellent, excellent observations Frank. Thanks for passing along. I appreciate the comments and by all means send me a heads up if you decide to post any of your own studies. I'll be happy to link--