26/12/2010

Operation Blank Slate, 2010 Edition

A year ago I wrote:
In my bookmarks, there is a folder called "to blog", and in there is a subfolder called "Kommentar" (comment) and you can find links in there that are very old. So [...] I'll give you the links to stuff from around the web that at one point in the past I thought I had something interesting to say about, in the order in which I bookmarked them
And here's this year's batch:

1. Will Wilkinson: "Pangloss and the Rock of Keynesian Providence"

2. Philip Brickman et al.: "Lottery winners and accident victims: Is happiness relative?" (abstract)

3. Whiskey: "Feminized TV: How PC Kills Revenue"

4. Eric Crampton: "Intrafamily Effects and Externalities"

5. Steven Pinker vs. Elizabeth Spelke: "The Science of Gender and Science"

6. Felix Salmon: "Helen Thomas, Christopher Hitchens, and Being Wrong"

7. Tom English: "The Level of Punditry Is Patronising and Insulting"

8. marcozanni: "30 engrossadas do Brasil na Copa de 70" (video)

9. Tom Rees: "Why Some Countries Are More Religious Than Others"

10. Steve Sailer: "What Would Google's Stock Price Be If They Ever Smarten up about Advertising?"

11. Robin Hanson: "Control Variables Avoid Bias"

4 comments:

Eric Crampton said...

I'd still be keen for comments on intrafamily effects!

LemmusLemmus said...

As usual, I don't remember what I was going to say; here are a few thoughts off the top of my head after rereading the post:

1. I know very little about economic cost-benefit analysis, but the more you allow for imperfection of information, including asymmetric information, the weaker your case seems to get.

2. This is especially true if you allow for a variant of The Winner's Curse: Marriage (as a standin term for all such relationships)is more likely to result when its utility is overestimated than when it is underestimated.

3. Enter emotions. Choosing a spouse is not the same as choosing an employer in this respect.

4. The expected utility of exiting a relationship the utility of which was initially overestimated can be prohibitive. Martin Daly and Margo Wilson concluded that (a) women in abusive relationships often say they don't leave it because they fear their partners' revenge; (b) this fear is often justified. (I don't remember the exact reference.)

5. The old libertarian standby that "of course X should be illegal" is not helpful. This would only be the case if the existence of a law against X lead to X not happening at all; we know this isn't so. What we really want to know is the prevalence of X in a world with vs. without (for example) alcohol consumption, all other things, including all laws about assault etc., equal. This is but an application of what is known as the "standard interpretation" of the counterfactual concept of causality in philosophy, which, in effect, is used by everyone who performs a randomized experiment.

6. It appears that alcohol consumption increases aggression. For a summary of the older experimental literature, see

Bushman, Brad J. and Harris M. Cooper, 1990: “Effects of Alcohol on Human Aggression: An Integrative Research Review”, Psychological Bulletin 107 (3): 341-54

The only paper that I know of that even tries to establish a causal effect of alcohol consumption on violence is

Fals-Steward, William, 2003: “The Occurence of Partner Physical Violence on Days of Alcohol Consumption: A Longitudinal Analysis”, Journal of Consulting and Clinical Psychology 71 (1): 41-52

7. So, in summary, I think that both an estimate that counts all of the negative effects of a person's alcohol consumption on his/her spouse as externalities and an estimate that counts no such costs would be naive; the correct estimate would be somewhere in between. In practice, I suspect that you can't make a convincing case for one point estimate being the right one, but will have to make do with presenting upper- and lower-bound estimates (and perhaps a few in between), carefully discussing the assumptions behind them.

Eric Crampton said...

Excellent. Will mull over. Some initial thoughts:

1. Asymmetric info only matters here if one partner has private knowledge, before the contract, about likelihood of doing something nasty like becoming a raging drunk, putting on a lot of weight, ceasing to like particular marital activities, or whatever. It doesn't count as asymmetric info if things change later in a way unanticipated by both parties, or if one party is willfully blind. If we allow intrafamily effects based on asymmetric information, the implications are pretty broad.

2. Doesn't winner's curse hit both parties expectationally equally in this kind of case? So the woman who most underestimates the man's likelihood of being a horrible drunk is most likely to wind up with him, but the man who most underestimates the woman's other problems winds up with her. What happens when both parties are hit by winner's curse AND we then start correcting only one set of failures?

3. Meh. Emotions enter into all kinds of purchasing decisions. Cars, houses...

4&5. This is probably your best argument, and I need to think more on it. Initial thoughts: Higher exit costs make people less likely to enter into relationships in the first place, right? To spend longer in the evaluation period where exit costs are lower? To spend more on screening before entering into the evaluation period? I think that the argument then requires ex ante undetectable propensity for a bad type draw from the urn. Otherwise ex ante investment in screening devises I'd think would have to make ex post results optimal - spent X to reduce the probability of a bad draw down to Y, additional dollar spent on screening must be worth expected gain in partner quality, etc. We'd then need to ask what proportion of abuse cases (where exit is extremely costly) are of the ex ante undetectable sort and what proportion are partner self-deception about likelihood of "fixing the guy" or voluntary tradeoffs against other desirable characteristics in an ex ante bundle? The larger proportion that is the latter, the fewer costs we can consider external.

6. I can buy that there's some causal relationship. It would be damned hard to sort out the magnitudes without some really good instrument. The paper you cite doesn't cut it: suppose the husband drinks a lot when he's had a bad day and he's more likely to be abusive on bad days, regardless of alcohol consumption. You really need some other instrument to sort things out. Maybe alcohol excise tax rates in a panel, but even that fails if alcohol excise tax rates jump when there's a local maximum in domestic abuse figures due to folks thinking there's a relationship; then it's just mean reversion. Causality is hard.

7. The baseline for intra-contract cases is going to have to be no external cost if we stick with a neoclassical kind of set up. Adding in an ex ante undetectable chance of an irreversible bad choice would then give an upper bound social cost of intra family behaviour, but we'd still need some measure of the strength of the causal link as it won't be the costs of all abuse in those cases but only that portion of it that's higher than it would have been absent alcohol.

But I need to think more on it. Thanks!

Eric Crampton said...
This comment has been removed by the author.