A total eclipse of the brain

Give an economist a computer, reams of data, and a mandate to publish something (anything!) in an academic journal and you never know what you might get. Pile on top of that our modern day obsession with trying to "beat the market" andit seems that no data is too obscure to mine for stock market gold.

The latest example is highlighted in an article in The Economist magazine. The magazine reports on the work of Garbrielle Lepori of the Copenhagen Business School who explores the link between stock market returns and solar and lunar eclipses. According to Mr. Lepori, the market tends to do slightly worse in the days leading up to an eclipse and slightly better in the days following the eclipse.

The slight difference in performance adds up over time. Using Mr. Lepori's data, The Economist reports that a "buy and hold" investor in the Dow Industrial Average would have seen a $1 investment increase to $37 between 1928 and 2008. An investor selling before eclipses and buying back right after would have seen than same investment grow to $55. Give me a break.

With all due respect, Mr. Lepori's paper highlights a common problem with the way many people think about the market. Just because a pattern appears in the historical data does not make it the basis of a sound investment strategy in the future. Even random events such as a series of independent coin flips exhibit what appear to be trends and patterns and may even correlate with other completely unrelated data.

Let me illustrate this with a simple experiment. If I flip a coin 1,000 times and record the outcome of each flip, I would expect to end up with roughly 500 heads and 500 tails. However, within the data there can be persistent winning and losing streaks. To confirm this, let's play a game. Let's assume you and I flip a coin 1,000 times. For every heads, you pay me a dollar for every tails I pay you a dollar.  Like any good financial advisor, I keep an accounting of my winnings in a spreadsheet and plot them in a nifty chart. After 100 of these games, my results look like this:

Each line plots the value of my winnings or losses, flip by flip, over the course of one game. All 100 games are plotted so the graph gets kind of messy. But the message is clear. What we see are a bunch of games where my gains or losses are very small and a few outliers where my gains or losses are very big. Think about it. Despite the fact that these coin flips are completely random and independent, there are times when I win big (my account balance grows to almost $100) or I lose big (my balance goes negative by almost $100.) If we didn't know better, we might be tempted to dig through this historical data looking for clues to improve my "performance" in future games.  But that would be futile. Why? Because the data is a completely random series. There is absolutely no information in the data despite the presence of what might appear to be "patterns and trends."

The absurdity of looking for information in the historical data of our coin flipping game is readily apparent to most people. Unfortunately, that lesson is often lost on people who sift through stock market data. Again and again on TV, radio and in magazines we find promoters touting "sure fire investment strategies" based on backtested data. These trading systems are more than a joke. They are dangerous waste of time and money.  A wise investor will stay away from them.

Now please don't get me wrong. I don't mean to imply that there is no information in the historical market data. And I know several people who are profitable traders. My point is simply that we have to be very careful when we look at trading or investing based on historical price patterns. While there is clearly some information in historical prices, there is also a lot of random noise. It is very difficult to distinguish one from the other.

A better approach is to focus on fundamental principles that govern investor returns. These principles include reward for risk and relative value. We will talk more about these principles in future posts.