Correlation is not causality

Contributed by: filbert Monday, August 14 2006 @ 09:39 AM CST

Much of what passes for news reporting, in all areas, is of the nature of "X happened, causing Y."  The cause-and-effect relationship is simply assumed, asserted, taken for granted.  A typical example is this story:

Mideast cease-fire boosts Wall Street[*1]
NEW YORK - Wall Street welcomed a cease-fire between Israel and Lebanon Monday, sending stocks higher as oil prices also dropped sharply.
With no economic data and little corporate news of note, investors saw the cease-fire as a buying opportunity after last week's losses. Crude futures fell as traders saw less risk of a supply disruption in the Middle East after the United Nations-mandated cease-fire took effect. A barrel of light crude was quoted at $72.80, down $1.55, on the New York Mercantile Exchange.
. . .
"We're up now on the cease-fire and oil prices, but it's hard to be an optimist right now, at least in the short term, because of the uncertainty over the economy and rates and the Fed," said Jay Suskind, head trader at Ryan Beck & Co. "As the week wears on, everybody's going to be focusing on the economic numbers and the debate over inflation will come back again."
Early on in any academic course in statistics, you learn the phrase "correlation is not causality[*2] ."  This means that just because we notice that two things seem to happen at the same time, it is not necessarily true that one causes the other.   There may be another, unidentified factor at work.  Or, it might just be a coincidence.  Or, it could just be that someone's trying to yank your chain.

There is a natural human desire to assign easily-understood causes to observed effects.  Many of these asserted causes have little or no actual basis in reality.  This effect is often seen in business reports which purport to explain why the stock market is up one day, and down the next.  When reading these reports, it's very useful to remember another common phrase:

"It ain't necessarily so."


Comment (0)

Medary.com - Correlation is not causality
http://www.medary.com/article.php/20060814103944253

[*1] http://news.yahoo.com/s/ap/20060814/ap_on_bi_st_ma_re/wall_street
[*2] http://web.augsburg.edu/%7Eschield/MiloPapers/95ASA.pdf