We're rather unimpressed with media coverage of daily or monthly stock market moves.

Typically, you'll see a breakdown of stock moves in terms of industry groups.  Occasionally, you'll hear that small-caps outperformed large-caps.  But do you ever hear "stocks with increasing volume gained today"?  How about decreasing volume?  Losses yesterday?  Gains two days ago?  High volatility?  Low volatility?  A large differential between yesterday's high and close?  Large gains over the course of the year?  These groupings are ignored, yet our data shows the big gainers and losers are often found in groupings that are unrelated to industry!

October of 2002 was an excellent month for bottom-fishing. The financial media picked up on this, albeit well after the trend was already apparent, but attributed it to some sort of anomalous short-selling. Never mind that a glimpse at data gathered from the last few Octobers showed exactly this sort of bottom-fishing environment…certainly no one-time anomaly.

The January effect is well-known and receives a bit of media coverage. Other monthly or quarterly seasonal effects, however, receive little or no coverage.

The other "effect" that gets some coverage would be that of the general market performing poorly from May to November.  We've identified any number of groups, however, that have gained during the period.  Looked at from the point of view of stock groupings, as opposed to the broadest market, "sell in May and stay away" looks a bit cliched..

Regarding the January effect, some folks tell you to hold on to your early-January bottom dwellers until mid-February. Looking at the recent data, that's a truly horrid strategy. Chances are, though, some market guru will be offering that same horrid advice late next December.  

These gurus will also tell you that you should buy stocks of low market-cap, December losers, or yearlong losers to maximize your January gains.  We'll tell you right now...these aren't nearly the most significant predictors of January gains.

Perhaps you hear that "beaten-up" stocks performed well last month. True enough, but in some cases it would have been far more accurate (and profitable) to say, "cheap stocks performed well". Of course, there's a good deal of overlap between the two groups, but how can investors pinpoint trends with that sort of skewed, half-true reporting?

One month, you hear that stocks fell because of the Federal Reserve's failure to lower interest rates. Next month, the Federal Reserve fails to lower interest rates, but stocks don't fall because the market had already "discounted" the possibility of lower rates. The same sort of worthless market thinking can be applied to political events…regardless of whether the market rises or falls, today's move can be related to the situation in Iraq.  Of course, there are times when the market spikes sharply immediately after a political or economic announcement…we don't mean to demean every cause/effect attribution that emanates from the gurus' mouths.

When analyzing, say, the historical price of gold stocks, some folks will haul out charts that go back to 1850. Are we supposed to believe that trends that existed 100 years ago are relevant in today's world, where information flows at the speed of the internet and the gold standard has been largely abandoned around the world?

In fact, we question the wisdom of looking beyond, say, 10 years of compounded historical data.  In some of our backtests, better results have been gotten by mirroring only the previous year's patterns than by relying on data compounded over longer periods.

Another approach towards seasonality is this:  if the stock in question fared well in a particular period historically, then buy it in the same period this year.  The assumption, normally left untested (or "tested" without the slightest sense of statistical rigour), is that this behavior is likely to repeat.  Depending on the period looked at, however, we've found that such a strategy could well backfire.

Seasonal analysts have a wonderful capacity to make statements along the lines of "the market has gone up in 15 of the last 19 presidential years".  The human brain, though, has an extraordinary pattern-seeking talent...given a spreadsheet of entirely random numbers, there are still those who will be quite adamant in pointing out trends they've extracted from the jumble of numbers.  The key question is whether these trends exhibit any statistical significance.  In the above (contrived) example, it's essentially impossible to gather the significance, since we have no way of knowing all the permutations of data that were sorted through by our mythical analyst before arriving at one of interest. 

In retrospect, it seems possible to find some quasi-logical explanation for just about any move the market makes. We also take a retrospective view, but we stick with the same set of indicators:  saying "Tuesday's most significant predictor of gains was the differential between Monday's high and close values" is quite a bit different than saying that "fears of terrorism dampened the spirit on Wall Street". Another important difference is that it might be possible to discern an evolving trend (e.g. the market is favoring a large high/close differential OR yesterday's emphasis on high/close differential is likely to reverse) with our indicators and historical data, but the terrorism angle is essentially worthless for investment purposes.

Rant Part II (!!!)

Seasonal investments home page

Copyright © 2010 MarketSynopsis.com. All rights reserved.