Thursday, October 14, 2010

More Confusion From Very Little Data

I wish economists and journalists and other leaders would read books like Understanding Variation by Donald Wheeler. Today the news is that unemployment claims went up, for the first time in 3 weeks. Analysts are now speculating that maybe the recovery is not as strong as it appears.

Excuse me. What are you saying? Are you saying that one upward blip can be extrapolated into a trend? Are you saying that because we flipped a coin 3 times and got heads (unemployment going down), and now we got a tail (unemployment going up) that our coin is defective? Are you saying that because temperatures were warmer than usual last week that we won't have winter?

One data point does not make a trend. We can't say the recovery is soft unless we know what the normal variation is on the out-of-work claims from week to week. Any result from week to week is going to be up or down (like heads or tails). Just because it goes up doesn't mean that a trend has reversed, when we didn't have a significant trend in the first place. 3 weeks of claims going down, analogous to coin flipping, could happen 1 in 8 times. The chance of it going down a fourth time is 1 in 16, unless there's a change in the system that creates unemployment claims (or better, creates jobs). Until we get to the odds of 1 in 64 (6 weeks in a row in the same direction--or some say, 1 in 256 is less risky for identifying a false trend) we can't say that we had an unemployment trend at all.

Please, Mr/Ms Economist and Mr/Ms Journalist, don't get confused and become Chicken Little thinking the economic sky is falling based on one piece of data that was going to show up normally.

No comments:

Post a Comment