by Mary Benton Hummel
benton.hummel@state.mn.us
August 2018
Those of us who work in the Labor Market Information Office of the Department of Employment and Economic Development genuinely like numbers, at least one of us to four decimal places. We also respect them for both the good and the bad they can do. We thought this might be a good time to offer a few caveats about Abuse of Statistics.
You may need to start by realizing that while Minnesota has a Board of Examiners for Barbers (and thank heavens for that), it has none for scientists or, for that matter, economists. A barber you can have confidence has a demonstrated skill set; scientists and economists, not so much.
The first issue is always money. Scientists and economists would have to be independently wealthy to be able to follow something that interests them for years on end. That or work for a university with deep pockets and endless patience. The rest of us need regular paychecks. Are you amused at the assurances passed out recently by scientists that all of us either definitely should or definitely should not have a glass of wine every single day? Could it be that a group of vintners paid for the “should” study? Secondary warning – if the study mentions preventing Alzheimer’s Disease, it’s probably bogus. My personal favorite was a recent study that proved definitively that a wide circle of friends could prevent mortality in later life. Actually, the only thing that prevents mortality in later life is dying young. But I digress.
There are also bar statistics, the sort of thing that starts, “Well, everybody knows that . . .” We may all agree that the sun rises in the east and sets in the west, but you would be amazed how few people realize that “squirrels are the Devil’s oven mitts.”
Then we get to technical issues, meat and drink to economists. You might wonder, for example, about the sample size for An Important Study. Can a sample of 200 represent a Minnesota population of 5.58 million (2017)? Suppose they didn’t all reply? Sampling can be tricky, and response rate can damage An Important Study further. You might also consider that Regression Analysis can actually be Latin for “my sample was inadequate, so I had to fiddle with the data.” And did what you’re reading offer a source other than “scientists”? This publication is quite fierce about source lines under charts and graphs. We’re also big on footnotes which are occasionally where the really good stuff is hiding.
There is at least limited good news here. Statistical Stuff you get from government agencies is limited in scope but tends to get a goodly amount of peer review, admittedly some of it hysterical laughter. We may also have universe data, but again, it’s limited in scope. Our agency has a universe file of everyone working in covered employment. This is not everyone working in Minnesota, but it comes close. Public Safety has a universe file of licensed drivers in Minnesota. This isn’t everyone who drives here, but again it comes close. Note: both of these files are kept extremely confidential. We don't tell.
Much of the foregoing can be summarized by a saying I heard in a post office years ago (anecdotal evidence): figures don’t lie, but liars figure. Let the reader beware.
