Sunday, April 18, 2010

Confidence Intervals for for Changes In the Number Unemployed

Carl Bialik’s column in The Wall Street Journal this past week pointed out that what is normally announced about government data is a point estimate.

Of course, the probability that any point estimate is exactly correct is generally zero.

It would be better to use a confidence interval instead.

If you do though, you find that it isn’t clear if the number of unemployed has gone up or down over the course of the last year or so.

That ought to be the point though. If you’re worried about the unemployed, we’re not seeing evidence that the situation is getting better. Alternatively, if you think activist policy is a problem, we’re not seeing any evidence that it is working.

No comments:

Post a Comment