Bandwagon Effects in the Stock Market

A little while back I posted about the increasing number of one-sided days in the stock market and how those are affecting traders. Yesterday I noticed another kind of one-sidedness: Seven of the last nine trading sessions in the S&P 500 Index (SPX) have closed either in the top or bottom quarter of their day's range.

This led me to examine more broadly the issue of how often stock indexes close near their day's highs or lows. If we divide each day's range into quartiles, then we would assume, over time, that a chance distribution would put 50% of market closes in the top or bottom quarter of the day's range and 50% in the middle two quarters.

That is not what we see in the data, however. Since July, 2007, two-thirds of all trading days closed either in the day's top or bottom quarter: 132 out of 200 days, to be precise. From January 2000 through June 2007, that ratio was 61% (385 out of 627 days).

If we tighten the criteria and divide each day's range into deciles, we can observe how often the market closes in the top or bottom 10% of its daily range. One would expect, by chance, that this would occur 20% of the time. Since July 2007, however, it has occurred 77 out of 200 days: 38% of all occasions. From 2000 through June 2006, we closed in either the top or bottom 10% of the daily range in SPX: 209 out of 627 days, or 33% of the time.

What we are seeing is that markets are closing near their day's highs or lows more frequently than we would expect by chance. This may reflect a bandwagon effect, in which traders and investors observe market movements during the day and don't want to miss out on them. This would lead them to buy rising markets and sell falling ones, creating late-day strength or weakness. Regret and the fear of missing out on a market move would lead to increased trending behavior as the day progresses, creating a kind of one-sidedness to the trade.

This bandwagon effect, exaggerating market movements late in the day, tends to be unwound the next day. Going back to 2000 (N = 2082 trading days), we find the following average next day changes in SPX as a function of the location of the prior day's close:

Prior Day Closes in Top Quarter of Range (N = 727): -.06% (343 up, 384 down)
Prior Day Closes in Middle Two Quarters of Range (N = 743): -.02% (381 up, 362 down)
Prior Day Closes in Bottom Quarter of Range (N = 612): .11% (357 up, 255 down)

What this suggests is that, once a bandwagon starts during the day, it tends to persist into the close. Fading one-sided days, particularly of late, has not been a fruitful endeavor for traders. Expecting bandwagons to persist into the next day's trade, however, has also not been profitable. It appears that traders segment their performance day by day, perhaps jumping aboard trends in an effort to finish their days on winning notes. By the close of the next day, however, any such bandwagon effect has been erased.

Comments

[Moby Waller:]

"What this suggests is that, once a bandwagon starts during the day, it tends to persist into the close. Fading one-sided days, particularly of late, has not been a fruitful endeavor for traders. Expecting bandwagons to persist into the next day's trade, however, has also not been profitable."

-- good info and research for day traders of the indexes

2008 Apr 18 03:47 PM

[Zeon:]

Good work, thanks.

2008 Apr 18 08:37 PM

[User 164492:]

hmm..its a known fact .in a bull market you buy early when the market opens.

2008 Apr 18 10:04 PM

[EtherDoc:]

I'm not sure your assumption (half of market closes in outside quartiles) is correct. One would need to run a simulation of volatility and average daily market moves to decide. Especially with large moves, I suspect it is likely to spend most of its time between the open and closing values rather than outside those values. The market ends up moving in a continuous line, not a random scatter.

2008 Apr 19 09:45 AM

[WACG:]

There is another explanation that fits the phenomena. With the advent of "online" or automated retail trading, "investors" could be using more trailing stops and targeted profit points. Trades would thus tend cluster in the upper and lower quartiles.

2008 Apr 19 10:28 AM

[cat & mouse:]

I tend to agree, except that the volumes of the stocks bought or sold during the first one third of the trading hours are huge, compared to the rest of the trading hours. In the first one third, the variations are huge, while in the rest of the trading hours the variations are small. You can book huge profits during the first one third period in a small interval of time, while you can book small profit in huge intervals of time in the rest of the trading period. In the second interval of time, there is tendencies to lose, if you wait too long to book profits.

2008 Apr 19 10:51 AM

[Brian in Mo...:]

I agree with Cat and Mouse, I've been out of the market by as early as 11:30am even after waiting until 10 or 10:30, and booking profits that don't appear again until the end of the day, if at all. Friday was a classic example of very little after morning volatility (thing dropped after the open then cameback up). Interesting enough, I've been shorting Gold stocks (and I'm a gold bug too) in the morning, even on Good Gold days these stocks have tended to comeback down after a good morning jump. Now that's where you find some really volatility to trade, like the dollar only in reverse (MOST of the time).

2008 Apr 19 12:44 PM

[LR European:]

Is this really a change to 'normal' market behaviour? It doesn't seem untoward that the market is moving in a direction (medium as well as short term) and closes further along its road at the end of the day than where it started.... It would be fascinating to see the data over a more extended period and how these proportions fluctuate in varying market conditions.

2008 Apr 19 01:52 PM

[phdinsuntan...:]

cinema is in fire and people is running in all directions...

2008 Apr 19 03:53 PM

[drposhmoney:]

Ermmm, could it be that there are usually 1 or 2, good or bad, major items of news per day? Given that the flow of good or bad news, or good or bad reaction, is essentially random from day-to-day, I don't see how this really helps anyone to make money? For example, are we going to have an ^, v, /, \, - kind of day? All of which end in the top or bottom 10%! Looking for day-to-day patterns in any *broad* based index is pretty pointless because it's based on a random flow of news InMyHumbleOpinion. Try doing an FFT (Fast Fourier Transform) on the last 2048 days of trading on the SPX, and you won't find *any* frequencies/cycles that have *any* significance (proof enough that it's random, and that 'technical analysis' on indices is simply voodoo). The only thing you can say about indices is that they tend to have alpha-stable distributions, but if you can't calculate the parameters, that doesn't really help either. Individual companies or sectors are the only things you can trade where you've got more than half a chance of predicting the direction IMHO :-).

2008 Apr 19 08:25 PM