Read on Twitter
I'm kind of surprised at all the criticism that that PNAS paper on income and happiness has been getting. Here are some responses to the criticisms I've seen. After an admittedly quick read, the paper, to me, seems fine /1
Criticism #1: Not big enough contribution for PNAS: I thought we all agreed that PNAS was bad cuz they publish flashy findings with questionable methods. I'd love it if they published more studies like this: Simple question with marginally better data than what's come before /2
Moreover, the paper to which this one responds was written by two Nobel prize winners. Given how much it takes to get people to question ANY published effect, I can see why a prestige journal pub is need to counteract the original. I don't like that that's needed, but I get it /3
To be clear, I still think that PNAS's "choose-your-own-editor" policy for NAS members is problematic, so I won't support them by reviewing (yes, I did in the past). But this paper seems like an improvement over much of what gets published there. /4
Criticism #2: Paper focuses on log(income): Economists and psychologists have argued about this for as long as income/happiness effect has been studied. I think using log totally makes sense here, psychologically. And the author is extremely clear that this is what he's doing /5
Criticism #3: Effect size is tiny! This one's tricky. Yes, r is small (.17 for LS). But this is one of those cases where different effect sizes don't match up. Comparing rich & poor, d looks to be about .65. That's about that size of the hit people take when their spouse dies /6
Also, there aren't many things that do predict happiness with effect sizes higher than this (other than *self-reports* of other things). So it's not obvious that this effect is too small to be important (though we can certainly debate that) /7
I do agree that the effect sizes for the experiential measures are getting pretty small. That's really consistent with other evidence: There isn't much that correlates very strongly with experiential measures other than things measured at the same time using the same method /8
So maybe I'm missing something, but given how popular (and problematic) the original $75K paper was, I think that this is a nice counterpoint /end
