[keithpeter]

"Something about it seems like a just-so story. The inherent noisiness of economic evidence is lost, all the graphs are cherry picked to support his view. And yet you'd have to dig quite a lot to find specific things to complain about."

Perhaps your friend could do a really popular debunking book called 'Stop trying to manage noise' and make a fortune? I'm actually a bit serious, so many people do not understand the idea of sampling error and associated noise in data.

[scribu]

You mean like Huff's "How To Lie With Statistics"?

https://en.wikipedia.org/wiki/How_to_Lie_with_Statistics

[lordnacho]

Yeah that's a classic. But it takes a lot of effort to pick apart those kinds of things. Some of the logic is very subtle and gets lost if you're also in a partisan debate.