Friday, September 15, 2006

The Value of Market Blogs

In my previous post, I made the distinction between descriptive and inferential statistics. When we describe a sample of the world around us, we take the first steps toward formulating hypotheses about that world. When we employ inferential statistics, we report tests of those hypotheses. Generating hypotheses, testing them, refining them based on tests: this is much of what science is all about.

But there is one crucial, missing ingredient: theories.

Theories are our explanations of the world we see. When we observe and describe our sample of the world, we build a model based on our perceptions and say to ourselves, "This is what I think the world is like." That model--our theory--provides us with the hypotheses that we test. When we check out and revise our hypotheses based on those tests, we're really refining our internal models of the world: our theories.

So where do market blogs come in?

It is rare indeed to find market blogs reporting statistical significance tests. Blogs are online journals: they describe; they do not infer. The really good blogs show you how an experienced investor/trader thinks. They not only provide valuable observations, but show readers how the writer moves from description to explanation: from data to model of the world.

Barry Ritholtz recently mentioned that he has three objectives in market analysis: 1) determining objective reality; 2) determining consensus on market issues; and 3) identifying where consensus varies from reality. In other words, he builds the most accurate model of the world that he can and then tries to find where market prices are not factoring in that model. This is the essence of opportunity: points at which the majority of market participants have not adequately updated their own models of the financial landscape to account for new realities.

The financial blog attempts such updating--hence the popularity of links--but also illustrates how the blogger uses this information to generate his or her own map of the world. Specific trade ideas generated from this theory can be thought of as partial tests of the blogger's map; over time, it is objective reality itself that provides the tests of the trader's ideas. In that context, every trade is a hypothesis: the trader's P/L over time reflects the degree to which the trader is skilled at generating and updating models of the world that are superior to the models of the consensus.

How does one learn to become a scientist? In graduate school, you join a research lab and apprentice yourself to an experienced investigator and the senior students of that investigator. Such an apprenticeship teaches skills, but also produces a modeling of the scientific process. For many traders, reading the ideas of experienced market participants in the blogosphere is as close to an apprenticeship as they're going to get.

That is the value of the best blogs. They don't just produce thoughts about markets; they model how to think about markets.

Subscribe to TraderFeed

6 comments:

Paul Stiles said...

I am a postdoc at Northwestern (chemistry/materials department) and am relatively new to the financial blog world. I often ask myself the same question raised in your post and took particular interest in your comparisons to the science world.

Here is my question to you. Is is ever really possible to empiracally test hypotheses about moves in the stock market? I realize that it is possible to make meaningfull predictions about economic directions based upon established theories, but, from what I've learned, there is no theory that can do the same for trading markets. I think this is relevant because most people read financial blogs to gain insight into stocks and not for economic trends. Also, it is my impression that the economy can affect the direction of markets, but it can also be anti-correlated. Isn't it true that focussing on macroeconomic conditions can just cause someone to completely miss large moves in the market?

I have many more questions for you, but I don't want to ramble.

Brett Steenbarger, Ph.D. said...

Hi Paul,

You're asking great questions. I do think it is possible to observe regularities in the markets, construct explanatory models, and then test these models against the data. If you're familiar with the data mining tools from Salford Systems (www.salford-systems.com), you can see how non-linear patterns in the market might be identified, explained, and tested for future occurrence.

As for the role of the economy in predicting stock movements, this depends entirely upon the time frame of the stock movements being studied. Markets are moved in the short run by factors that are very different from those macroeconomic variables that account for the broad trends. Those other factors, such as sentiment, volatility, etc., can be included in multivariate models such as MARS, however, or can be used as categorical predictors in a program such as CART (both programs from Salford).

Brett

Paul Stiles said...

Do you find that these data mining techniques are more predictive for stocks, sectors or for the larger market, in general?

Of course these methods all rely on the assumption that market patterns repeat themselves,which brings me to my last question (or two). Has anyone published their research that establishes a connection between market moves in the past with moves at some later time? I realize that academics have long been proponents of the "efficient market", but are their minds changing?

Thanks

Brett Steenbarger, Ph.D. said...

Hi,

Yes, I think if you survey the behavioral finance literature, you'll see examples of market inefficiencies documented through research. I do not trade individual stocks or sectors, so cannot speak to the issue of whether their patterns are more reliable than those found in the equity indices. It would not surprise me if patterns were most robust among small cap and microcap issues that are not typically included in index arbitrage trades.

Brett

Anonymous said...

Thanks Brett. You have one of the more informative and helpful blogs that I have seen.

Brett Steenbarger, Ph.D. said...

Thanks, Paul; I appreciate the feedback and support--

Brett