How to lie with numbers, from the WSJ

Tip to Paul Krugman for this funny and yet not funny illustration of how you can lie with numbers.

The WSJ chart on defense spending (note the title):

And here's Krugman's correction:

Unleaded gasoline lowered violent crime by 56%

According to a recently released study, examining the effects of removing lead from gasoline, unleaded gas substantially reduced lead poisoning in children. Because lead poisoning is linked to more aggressive behavior is linked to violent crime, the study finds a substantial reduction in violent crime associated with the removal of lead from gasoline.

This is in addition to the substantial reduction in violent crime from the legalization of abortion.

It's crazy how things are linked together...

Liquor advertising revisited

Statistician friend Rick criticized my earlier post on liquor advertising, noting that there's more to the problem than just the portion of minors reading a particular magazine.

I agree, but that's because we both have tunnel vision. I argued that the liquor marketing rule - of no ads in publications where >30% of readers are under 21 - was pathetic because 30% of people are under 21.

Rick argued that you have to take into account the portion of magazine readers who are minors:

This is true, but relevant only if the under-21 crowd reads magazines at the same rate as the over-21s.

If, say, 85% of all magazine readers are over 21, then requiring 70% of-age readership is a weak criteria and should be criticized.

If only 50% of magazine readers are over 21, then 70% is excellent (because a magazine with 70% of-age readership is much higher than average)

We're both wrong. Consider this: Publication A might have 65% readership by over-21s but total circulation of 10,000 and publication B might have 90% readership by over-21s and 1 million circulation. The decision to not advertise in A protects 3500 kids whereas the decision to advertise in B gets 100,000 of them.

In that case, it doesn't matter what portion of the magazine reading population are minors (or whether the ad threshold is 70% or 85%). The point is that the threshold isn't designed to minimize the number of minors exposed to liquor advertising. They'd have more luck banning liquor advertising based on circulation than on the demographics of readers.

There's probably 1000 other ways to crunch the numbers, but I stick with my first point: liquor advertisers aren't doing very much to ensure that a disproportionate share of their advertising hits the over 21 market.