This Washington

A Failed Experiment

By Camden Swita April 7, 2010

In the interest of furthering man's knowledge of a tax hike on beer, which the Senate proposed in their latest revenue package, Sightline Institute super wonk Eric de Place put some elbow grease into the spreadsheet and chart I made yesterday comparing state tax rates on beer and the rate of consumption.

He wanted to find out if there was a correlation between the tax rate and beer consumption by running a "regression," which is an erudite term for math we at PubliCola do not understand, on the data I'd compiled. Or as he explained in an email shortly after I posted: "I want to run a regression on it. To see if there’s any meaningful statistical correlation."

His conclusion? This:

The fact that you can't really draw a trend line through those dots, said de Place, means there's only negligible correlation between tax rate and consumption.

"Darn it, it’s not very correlated," he wrote in his follow-up email, clearly distraught. "The r2 is 0.0017, which is sort of a fancy way of saying that the tax rate 'explains' about 2/10ths of a percent of the variation in consumption. In other words, they’re not correlated."

Once again, the ol' "r2" lets me down.

A couple of caveats that de Place pointed out after his discovery: "1) Real statisticians hate doing this kind of thing because doing a regression on 50 items is kind of silly. 2) Plus, the tax itself is only a small portion of the total price, so it’s not clear if looking for a correlation here is just a waste of time. Anyway, it’s perfectly possible that price and consumption are correlated, but tax rates and consumption are definitely not."
Show Comments