After reading the title of this post, my hope is that a look of disbelief is cast on your face. Of course butter production in Bangladesh has nothing to do with prediction of US stock market returns. However, through data mining all sorts of 'relationships' can be demonstrated. Many mutual funds, ETFs, and trading strategies are built upon these data mining strategies--most often to the harm of investors that utilize them.
As Jason Zweig describes in a recent article in the Wall Street Journal, entitled Data Mining Isn't a Good Bet For Stock-Market Predictions, "The stock market generates such vast quantities of information that, if you plow through enough of it for long enough, you can always find some relationship that appears to generate spectacular returns -- by coincidence alone. This sham is known as data mining."
I recall in from my business statistics course in grad school, how I was able to show that ice cream consumption was correlated to the murder rate--the more ice cream that was consumed the more people were murdered. Should we therefore ban ice cream? On a completely object note, my stomach says, "Heck no!"
The thing I learned from the ice cream case is that correlation does not equal causation. Simply put, your idea should make sense on top of any evidence found in data. And then if the idea makes sense and data seems to support it, you must test and re-test to make sure you're still not data mining. This is the process of prudent research.
What History Can Tell Us
There are some things history can and cannot tell us, and it is often very difficult even for the trained eye to know the difference. However, it is critical to build a sound investing strategy and at least in part, utilize what history can tell us. So what can we learn from history?
As Jason Zweig describes in a recent article in the Wall Street Journal, entitled Data Mining Isn't a Good Bet For Stock-Market Predictions, "The stock market generates such vast quantities of information that, if you plow through enough of it for long enough, you can always find some relationship that appears to generate spectacular returns -- by coincidence alone. This sham is known as data mining."
I recall in from my business statistics course in grad school, how I was able to show that ice cream consumption was correlated to the murder rate--the more ice cream that was consumed the more people were murdered. Should we therefore ban ice cream? On a completely object note, my stomach says, "Heck no!"
The thing I learned from the ice cream case is that correlation does not equal causation. Simply put, your idea should make sense on top of any evidence found in data. And then if the idea makes sense and data seems to support it, you must test and re-test to make sure you're still not data mining. This is the process of prudent research.
What History Can Tell Us
There are some things history can and cannot tell us, and it is often very difficult even for the trained eye to know the difference. However, it is critical to build a sound investing strategy and at least in part, utilize what history can tell us. So what can we learn from history?
- Stocks generally will return more than bonds over time. This is called the 'equity premium.'
- Value companies generally will return more than growth companies over time. This is called the 'value premium.'
- Small companies generally will return more than large companies over time. This is called the 'small premium.'
All of these premiums have been well documented in the academic literature. Studies show these premiums persist over different time periods and in different countries, so they are not data-mined farces. These premiums can add higher rates of return to an investor's portfolio while also adding diversification benefits. An investor not building a portfolio utilizing these premiums is ignoring useful history and not the data-mined farces that harm many investors.