Weekly Unemployment Claims

Today's headline: Jobless claims rise more than expected. Last week's headline: Jobless claims unexpectedly fall.

There are thing about which we cannot be certain with absolute precision. Then there are things like the weekly jobless claims, where the noise in the weekly statistics exceeds the trend by a factor of ten. In truth, there is hardly any trend.

But every week, there's the "expected" value. It is derived by asking a lot of economists what they think the number will be and averaging the answers. It reminds me of an old, politically incorrect joke. Nobody has ever seen the emperor of China. How do you calculate the length of the emperor of China's beard? Answer, you ask 100 Chinamen how long they think the beard is and take the average.

So every week, the news services calculate the average of a few dozen opinions from people who don't know, and compare it with a government report that contains no statistically significant information. They then always have a headline, because the two are never the same. OK, it's possible that due to some fluke they would match, so there would be a headline, "Economists finally get one right."

It's much the same thing when the Weather Channel broadcasts some ninny breathless announcing that the high temperature in Des Moines today will be 12 degrees "higher than it should be."

Public education would be well served by taking an hour a week for two weeks for each graduating high school senior and explaining the uses and misuses of statistics and probability. Not the formulas, just the rough principles. It would be really useful.

Here's a pop quiz. If you have a room with 35 randomly selected people, what are the odds that you'll have two of them with the same birthday? Day and month, not year. email me with your guess and I'll explain it to you. You're almost certain to be far wrong. Statistically speaking.