Selective Data: The State Unemployment Case

People who do selective data analysis should be beaten with sticks. Case in point: I got into a discussion with someone today who argued that the U.S. is far from monolithic in its economic downturn, with some states doing worse than others.

It was hard not to agree with that first point. It is, or should be, self evident. Some states were hit first and harder. But that doesn’t mean everyone won’t feel the pain eventually.

He disagreed, and here is the data he used. Not only are some states doing better than others, but more than half of the U.S. saw unemployment decrease in September, despite the gloom and doom of people (like yours truly) who say that unemployment is on a speedy ride to higher levels. Half! he said.

That caught me by surprise. Half went down in September? Really? So I went and had a look at the BLS data. And he was right: Unemployment rose in 21 states month-over-month in September, but fell in 23 states.

Q.E.D., right? Not really, or at least not in any meaningful sense. While more than half of U.S. saw unemployment decrease in September, unemployment is up in 47 U.S. states year-to-date, and national unemployment has risen from 4.7% to 6.1% over the last twelve months.

The upshot? Classic selective data analysis. Unemployment is up across the U.S. and it’s going higher: Anyone who says otherwise is selling something -– probably a line of patter.