Comments on: UC Davis Econ in the News

By: pushmedia1

pushmedia1 — Fri, 14 Mar 2008 10:39:45 +0000

There doesn’t seem to be rhyme or reason as to how the profession chooses nulls (except maybe conservativism).

In any case, I pulled up OP and see that sigma (substitution between ferners and merikns) is between 6 and 10. Those numbers are, last time I checked, less than infinity and greater than one so it seems you’re right.

I notice, looking at the “preferred” regressions in the borjas paper (table 4), that if you do a one-tailed test, the p-value is 6%. They’re rejecting the “proper” null but not by much. (And does anyone think they’re doing IV because they’re worried about endogeneity? “Oops, we can’t find a good instrument… and look at those standard errors! I guess we can’t reject the null now!”)

BTW, if Borjas ever complains about teenage wages being compressed by immigrants, well, I just don’t what I’ll do.

By: notsneaky

notsneaky — Fri, 14 Mar 2008 06:30:37 +0000

Well yes. BUT. The OP finding does not rely on complementarity either. In fact, unless I’m messing something up, all it needs is that workers are simply less than perfectly substitutable. Even less comp. than Cobb Douglas. When you think about it that way than, yes, the null of “perfect substitutes” doesn’t seem to be very realistic (I’m flipping the nulls here btw, to be consistent with the ‘fishing for substitution’ thing). And if that null doesn’t pass the common sense test then why should we be so adamant about wanting to reject it?

By: pushmedia1

pushmedia1 — Thu, 13 Mar 2008 17:18:32 +0000

I suppose the retort is something like “well, no one’s a perfect substitute for anybody.” The point being if chop the data up enough, we’d never reject that null.

By: notsneaky

notsneaky — Thu, 13 Mar 2008 16:46:30 +0000

“It seems the proper null hypothesis (the one that should be harder to reject) would be complementarity.”

I was thinking that too. Why is the post titled “Fishing for Complementarity” where it seems like “Fishing for Substitution” (by dropping unconformable data) is more appropriate?