Tuesday, February 27, 2007

Big Down Days: What Happens Next?

That seems to be the question everyone is asking following the meltdown on Tuesday. What we observed is a scenario in which everyone simultaneously runs for the exits. China's market was down over 9% in Shanghai prior to our open. The Yen soared to multiweek highs, a risk factor to the carry trade, as mentioned here a while back. As selling in the U.S. markets begat selling, we saw weeks' worth of profits eroded in a single session. Among money managers, who fight tooth and claw for each quarter's performance, such a drop was a major threat. Volume picked up dramatically, reflecting the exit of the institutional traders, and--with that--the decline accelerated, now wiping out months' worth of performance.

Such an occurrence is rare. I went back to 1990 (N = 4304 trading days) and could only find 25 occasions of a single day drop of 3% or more in the cash S&P 500 Index. What's even more rare is to have such a decline in the context of a bull market. How many times in the last 17+ years have we seen a 3+% decline in a market that had been up over the prior 20 and 60 days? It's only occurred three times out of all those trading days.

But shifts happen. And that, my friends, is why money management and risk control are all important. If you double your money, double it again, and then double it again, only to lose 90% in the next debacle, you wind up down 20% on your initial capital, needing a 25% gain just to return to where you were at the start.

So what happens after big down days? Let's start with the 25 occasions in which we've dropped 3% or more in a single day. At just about every time frame from one to twenty days out, returns following such a large single-day drop are quite bullish. One day later, the S&P averaged a gain of .47% (17 up, 8 down), much stronger than the average single day gain for the rest of the sample of .03% (2256 up, 2023 down). Twenty days later, the S&P was up by an average of a whopping 4.47% (20 up, 5 down), again much stronger than the average 20-day gain of .73% for the remainder of the sample (2641 up, 1638 down). In all, large down days have tended to represent buying opportunities since 1990.

Those findings may be a bit deceptive, however, because--of those 25 large down days--15 occurred in the context of a market that had *already fallen* 3% or more over the past 20 sessions. In other words, large down days have tended to occur toward the end of market downmoves--as a kind of washout. We don't typically see large down days following intermediate term strength, as noted above.

I relaxed my criteria a bit and found six occasions in which the market had not been down more than 2% on a 20 and 60 day basis prior to the large down day. All six occasions were up the next day, and all were up over the following 20 sessions. Indeed, the average gain over the next 20 trading days was an impressive 3.89%.

In short, big down days have tended to represent panic and, even when they've occurred in relatively strong markets, they have tended to occur nearer to market bottoms than tops. On average, traders have made money by buying into such panicky markets. While the current volatility may be more than many traders wish to tolerate, at the very least the pattern of superior returns following panicky declines should offer caution to those tempted to arrive late to the bear's party.


Lord Tedders said...


I'm sure I'm not the first person to say this, but I'm very impressed with your analysis. This site is becoming a go-to site for me.



John Friedrich said...


(I can't resist jumping in with my humble musings.)

A few bullet points that augment your astute observations:

We have not had a 2% drop in a single session in nearly 1000 days - the longest since 1929 - so I have read recently.

Notice the breakaway gap.

Notice the single-bar action. The day opened at the top and closed near the bottom with little "wick." This engulfing action, as you point out, engulfs nearly 3 months of gains. This also illustrates the "shift" in volatility. Market action cycles between periods of contracting ranges and expanding ranges.

3% down days are rare but in the context of expanding ranges as we had in the late nineties, not necessarily indicative of a shift the expansion/contraction cycle.

I don't know what the ATR of the SPX has been lately, but it looks to be quite small compared to todays action. On the order of 40x or so if the ATR of the SPX is about 10pts which it looks like at a glance of your chart

So, could you tell me how many days have this charactistic?

1) > 3% down day

2) One day range expansion > 40x the previous 20days ATR (you get the idea)

3) The drop happens after multiple multi-year highs.

4) The drop wipe out 3 months of gains.

5) Daily volume is 70% greater than the average daily volume. (I know the average NYSE volume is about 2.3B shares. Todays was 4.2B

To sum up, I would not characterize today as a "blow off" because todays action does not occur after a prolonged down turn. As such, I am interested to know whether there is any evidence that would support the hypothesis that this kind of action is a precursor to an extended down turn.

This question is not in conflict with your findings since I am interested in a longer time frame.

I also am a bif fan. Keep up the good work

cpptrader said...

The jump in volatility was nice for long vix call options, but I think it was discounted based on some of the possible bounce ideas I was looking at. There seemed to be an overly aggressive sell of in QQQQ calls, and the delta there seems pretty attractive as a particular way to take advantage of a short term bump (and the next ones, which I have since gone long in, expire in 18 days - very close to your 20 day study). Just an idea on how to play this.

Thanks as always for the great post.


Brett Steenbarger, Ph.D. said...

Thanks, Tedders; I really appreciate that. A very different analysis will be posted before the open on Wednesday--


Brett Steenbarger, Ph.D. said...

Hi John,

Great idea for an analysis; that would take a very, very long-term historical look. I'll see if I can get to it sometime tomorrow. Thanks much for your comment--


Brett Steenbarger, Ph.D. said...

Very interesting ideas; thanks Matt--


Brandon Wilhite said...

Just a few observations from the currency side of things:

1) Yesterday was also a bloodbath over there, but only if you look at the right crosses. If you're just looking at eur/usd or gbp/usd you won't see it!

2) The right crosses are the carry-trade crosses, specifically at the least usd/jpy aud/jpy, gbp/jpy, and gbp/chf.

3) The move in currencies began at about 1:00am EST or so, a savvy trader might have picked up on this and been able to profit in the equity markets also.

Both the BOJ and the Swiss National Bank have been warning carry-traders. I haven't seen it mentioned anywhere yet (this being the first place I've been :) ), but is it possible we are seeing the beginnings of an intervention there? Personally I have no clue and I'm just speculating...I've never been through a day like that before so it was a good learning experience.

In summary I think we could say it was a great day for daytraders but a lousy day for position traders. I'm mostly position myself, but I don't trade just EOD, so I escaped mostly intact. I'm not convinced it's over there.


Trading Nerd said...

Hey Dr. Brett,

I have also done similar analysis, only my criteria are a 3% drop and a 21-day (one month) high for the S&P 500 the previous day.

Going back to 1930 (yeah, it's a long time, but there haven't been many times that the S&P 500 has declined by 3% after making a 21-day high the previous day) ...

The average 1-day return of the entire sample is .03% (x = .03%), with a standard deviation of 1.14% (s = 1.14%). When the above-mentioned criteria are met (n = 27), the average 1-day return is 2.40%, which is statistically significant at the 99% level (or put another way, p < .01).

The average 5-day return since 1930 is .13% (x = .13%) and s = 2.54%. When the above-mentioned criteria are met (n = 27), the average 5-day return is 1.49% --statistically significant at the 99% level.

The average 10-day return since 1930 is .27% (x = .27%) and s = 3.56%. When the above-mentioned criteria are met (n = 27), the average 10-day return is 1.31% --not quite statistically significant at the 95% level (but very close -- p = .065).

The average 21-day return since 1930 is .57% (x = .57%) and s = 5.31%. When the above-mentioned criteria are met (n = 27), the average 21-day return is 2.91% --not quite statistically significant at the 99% level (but very close -- p = .011).

It is worth noting that the 1-day returns are greater than the 5-day and 10-day returns, which means that typically, returns are negative (on the net) for days 2 through 10.

jessechan said...

Very good analysis. My own opinion is that yesterday's drop was similar to the decline last May or June when "Carry Traders" unwound their positions when Yen appraised in value, but this time the triggering event came from the debacle in China's stock markets plus appreciation in Yen.

By the way, I'm re-reading "Enhancing Trader Performance" the second time, need to digest the ideas slowly. Thanks for your contributions to the traders' world.

Brett Steenbarger, Ph.D. said...

Thanks for the forex insights, Brandon. Great example of how you can gain perspective about one market by watching other, connected ones.


Brett Steenbarger, Ph.D. said...

Great research, Trading Nerd. Many thanks for passing along--


Eric said...

Brett and anyone else that could provide assistance,

Where do you get your historical data from for use in your research? Are there any free sites that provide good data that you could recommend? Thanks.


Brett Steenbarger, Ph.D. said...

Hi Eric,

I've never used free data. My sources are listed in the Trader Development links on my personal site:
www.brettsteenbarger.com. Thanks--


googleaddict said...

Hi Dr. Brett,

Do you have an idea about the Saudi Arabia stock market? It had crashed earlier on the 25th of February, 2006. It has turned to a very acute bear market since then. People have been trying to spot the bottom, only to find they see transient bottoms. Can you give us a guide on how to make sure we have reached the real market bottom. Thanks.

Brett Steenbarger, Ph.D. said...

Hi Google Addict,

You're asking a great question, but unfortunately I don't have enough data on that market to provide an informed opinion. My flow studies are limited to this point to U.S. equities. I'd love to expand that, but I'm early in the research project. Thanks for the interest--