Nervous Markets, from Jeff Watson

December 29, 2007 |

Nervous Back in my apprentice days in the wheat pit, a wizened old guy took me under his wing. Since he had been trading successfully since the late 1930s, I was all ears. One of the anecdotal statements he made to me was that "Nervous markets always close lower." I've remembered that sage bit of wisdom for all of these years, and have followed that advice with reasonable success. Lately, I've been wondering if there was a way to quantify that statement, or if anyone in this group can lead me in the right direction. I'd like to see what the numbers and percentages of nervous markets really are.

One might ask, "What is a nervous market?" The only answer is that I can't define a nervous market, but that I know one when I see it.

Bill Rafter replies:

There has been some work done relating volatility to subsequent price behavior. Volatility (depending on how it is defined) may agree with your description of nervousness. Generally the premise is that volatility is bearish, which would be in agreement with what the wizened old guy told you.

There are many definitions of volatility. My suggestion is to look at measures other than 1-day rates of change. Earlier this year I asked around for other measures of volatility, and got approximately 20 variations. Additionally, there are “handmaidens” of volatility, such as institutional holdings.

Jim Sogi adds:

Vic and Laurel recently hypothesized that afternoons closer to the low of the day in S&P could be thought of as you say "nervous." I've also been playing around with NYSE declining volume. Today 857k. Fairly nervous. Over 1m down before the end of the day is very nervous.

Phil McDonnell remarks:

"Volatility is bearish" requires considerable qualification. In my ruminations the results have generally shown that rising volatility is bearish on a contemporaneous basis and over the short run. However high levels of volatility can be quite bullish. It is important to define what volatility we are talking about.

Victor Niederhoffer investigates:

Victor Most definitions of nervousness refer to trembling, quivering, and agitation. I thought I would look at some qualitites of markets that seem to be of that nature. I started with markets that were up at the open, down at 11 am, up at 1pm and down at 3 pm. I found 17 cases since 1999 in the S&P futures, and nine of them went down from 3 pm to the close, with an expected move of three points. Such moves occur about twice a year. I found that a similar gyration, but down at 2 pm rather than 3 pm, which happened eight times since 1999, to be visited with four up from 2 pm as of the close and four down. I found eight cases of the first pattern in bonds, and the next day was 50-50 up with an expectation of four ticks up. In general, I would say that there is not much evidence that a quivering market with numerous crossovers to the down side is bearish. In another study, I found 22 cases since 1999 where the market was up at open, down at 10 am, up at 11 am, and down at noon. Eight of these occured in 2007. Thus, the market appears to be gyrating more frequently in a way most people would get nervous about if it were their limbs or sinews. Seven of the 12 cases showed a rise from noon to the close. Thus, nervous stock markets, defined in this way, did not show any non-random predictive tendencies, although the jury is out as to whether the market shows an inordinate tendency to tremble between hours.

Vincent Andres adds:

One classical way to proceed would be to have someone provide : 100 examples of nervous/non-nervous/"don't know", status markets. (Maybe more than those three classes). Each example being the price series and the graph. Of course, it's possible to ask several people to do the classification.

Some counting would probably teach things, hopefully reveal possible clusters in the appropriate measure space. Well used, tools like neural nets (many NNs are nothing else than stat learners) may be of use here. However, there is a preprocessing to be done on the price series before feeding the counter … and this will often be the clue. (Some ideas have been provided in this thread).

It is however quite a painful job, and requires a rigorous methodology. And even if classification succeeds, it's only a piece of the job. As P.McDonnell said, nervous => bearish is not a 100% sure implication. And markets are dynamic, not well-defined, etc.

Comments

2 Comments so far

Ken Millhone on December 28, 2007 9:54 pm
Interesting thought from your mentor. A mentor of mine always said when "Grandma" buys in, it's time to get out. I know there were a lot of "Grandmas" buying in in 2000-2001. I don't see anything like that going on now.
Zyxo on January 1, 2008 3:28 pm
Interesting discussion. Especially since I am a dataminer and datamining (with neural nets or whatever other means) is what is needed. I am sure, you will all agree that a lot of parameters can have an effect on the subsequent stock prices, and volatility is only one of the candidates. A lot of people did some dataminings on stock prices. What you see in the reports is that it does not work (or at least not worthwhile). Which is the same as saying that we should forget about technical analysis. Obviously if someone creates a datamining model that works, he will keep it quiet and use it …

Recent Posts

List of Authors

Dec

29