Correlation ≠ Causation

Feb 28, 2012



In today’s broadcast of Marketplace on NPR, David Brancaccio led a great discussion about using statistical indicators to predict presidential elections. Which works best? GDP? Disposable income? American war casualties? Unemployment? Gas prices? Durable goods spending? (whatever that is) The list goes on and on. Sean Trende, Senior Elections Analyst for RealClearPolitics, drove the last nail into the coffin of this discussion, by observing that all this is nothing more than hindsight. Trende said, “If the president does win, people will say it is GDP and unemployment. If he doesn’t win, they’ll say it’s real disposable income.”

This is no different from lots of other analysis, whether it’s an election, a sports event, or which company will sell the most smartphones. There are a lot of “experts” out there and not many possible outcomes – in the case of elections and sports events, sometimes there are only two possible outcomes. So to impress people the “expert” not only has to pick the winner, but to explain WHY the winner will win, or has already won. Whether that explanation is valid or even rational is sort of beside the point.

Describing this phenomenon of pontificating experts, a finance professor I once had asked our class if we could imagine a stock market analyst getting on TV one night and saying, “Well, the stock market behaved randomly again today.”

The American electorate is going to face thousands and thousands of elections and other ballot issues this year. Well, is the electorate going to behave randomly? The odds as 64% that they will, but that probability will go to 71% if an American League team wins the World Series. I have spoken.

“Painting with Numbers” is my effort to get people talking about financial statements and other numbers in ways that we can all understand. I welcome your interest and your feedback.



Purchase your copy of painting with numbers today