GO TO THE SOURCE

July 21, 2008
Dan Hebert here, Putman Editor. Here's an excerpt from a recent column that I wrote for our sister pub Control Design. Many journalists don’t dig for data and don’t know how to analyze numbers using basic statistics. Instead, they present sensationalist headlines that generate visceral fears and inaccurately influence opinions. For example, a common headline over the past few decades suggests that higher oil prices drive stocks...
Dan Hebert here, Putman Editor. Here's an excerpt from a recent column that I wrote for our sister pub Control Design. Many journalists don’t dig for data and don’t know how to analyze numbers using basic statistics. Instead, they present sensationalist headlines that generate visceral fears and inaccurately influence opinions. For example, a common headline over the past few decades suggests that higher oil prices drive stocks down. You know that a declining stock market is not good for sales of your machines, for your career, and for your investment portfolio; so the recent rise in oil prices may be causing you great concern. Should it? Using the Internet and Excel, you can run a correlation analysis between virtually any published data sets from oil prices to stock prices to machine tool sales. The Excel function “=CORREL” checks correlation between two sets of data. Correlation values are always between -1 and plus 1. A correlation of 1 means the two data sets move in perfect lock step, a correlation of -1 means the data sets move in a perfectly opposite fashion. A correlation of 0 means the data sets are random and have no relationship to each other. Put another way, correlation numbers near 1 or -1 mean that changes in one variable cause the second variable to change positively or negatively respectively. Correlation numbers near 0 mean that changes in one variable don’t cause changes in the second variable. To check correlation of oil prices and stock prices, we found stock prices at yahoo.com and oil prices at inflationdata.com. We imported monthly data from January 1974 to November 2007 into an Excel spreadsheet. The Jan 1974-Nov 2007 time period was selected because it had wide fluctuations in stock and oil prices, it is relatively recent, and it contains enough data points to generate a valid correlation. The S&P 500 stock index is a widely used index representing almost half of worldwide stock market capitalization. When comparing virtually any economic data, one must first strip out inflation effects or else the correlation will always be strongly positive because prices tend to increase over time. This is done by comparing the percent change from month to month instead of the raw price data. Comparing percentage change in monthly stock and oil prices over the specified time period yields a negative correlation of -.15. Per wikepedia.com, this is a very weak tending toward insignificant negative correlation. That means that stock prices go down a bit when oil prices go up, but not enough to matter or to worry about. You can perform similar correlation analysis on any data sets of interest to you and your company and reveal the truth behind the headlines.