It’s in the bag: Stock prices will be up this year based on the Super Bowl indicator! Stock market legend has it that years in which a team from the original NFL wins the Super Bowl, stock prices will go up. If a team from the original AFL wins, stock prices are likely to fall. Since the Saints are the NFC team this year and the Colts trace their history back to the original NFL, stock prices should rise regardless of the outcome!
But before you call up your broker on Monday and buy stocks on heavy margin, consider the following: stock prices are also positively correlated with butter production in Bangladesh. MarketWatch.com has published an article with more details regarding the silliness behind the Super Bowl indicator and various derivative indicators that have been developed over the years. While it is good for a few laughs, we can also draw some important lessons about the nature of statistics.
Lies, Damned Lies, and Statistics
One of the fun games that students often play in Econometrics classes involves finding correlations between seemingly unrelated data. Computers today provide the tools required to quickly obtain large sets of data on nearly any activity. It is very easy to develop correlations between such data and to come up with “insights” to prove a point. The correlation between the Super Bowl winner and stock prices, or Bangladeshi butter production and stock prices are obviously silly. Few would trade on the outcome. However, it is also easy to come up with correlations that are more persuasive.
Those who understand statistics are never fooled by such tactics. However, others may easily mistake the presence of a correlation with causation. It is one thing to say that Bangladeshi butter production is correlated with United States stock prices. It is quite another to claim causation.
The next time you read an article that uses statistics to prove a point, whether the topic is related to investing, politics, or a pitch for a consumer product, consider whether the presenter is making an attempt to imply causation through the mere presence of a correlation. If additional evidence is not produced to show causation, it is likely that you are being manipulated in some way. Legitimate studies on causation will never merely present a correlation and expect readers to accept the presence of causation without additional evidence.