If anyone knows the Neumark, Salas, and Wascher critique of the contiguous counties approach as well as the Dube et al. response well and wants to write up a guest post, I'd be interested in hearing from you. I haven't gotten to these papers in my recent push to familiarize myself with the literature and am afraid I might not. But it's clearly important. As my last post suggests, I think the spatial heterogeneity point looms over all the other issues in this debate, so if it's poorly dealt with that matters.
I suspect it's not - I kind of got a sense of Dube et al.'s response to Neumark, Salas, and Wascher in one of their responses to Meer and West. But I'd be interested in hearing from someone that's done a closer reading.
Without dealing with that I really don't see where the model identification is. Without the county match you've really just got a fixed effects model that you're trying to pretend has the identification properties of a DID. Right?
Friday, January 24, 2014
Subscribe to:
Post Comments (Atom)
Posts
My admittedly naive understanding of the two papers (high-level technical econometrics makes my eyes go all googly but I can understand the text pretty well):
ReplyDeleteNSW: A higher minimum wage represents a trade-off between higher wages for some and disemployment for others. The two effects are related in their impact, when you see a large increase in wages following a minimum wage increase, you also see a large disemployment effect. When you see a negligible change in wages, you see only a negligible change in employment. A lot of the studies (Dube et al, for example they name specifically) show the latter.
Dube et al: There is no reasonable explanation for why there would not be significant discontinuity created by a state border when the minimum wage is increased, therefore NSW must have made a mistake in the continuity estimates, but we can't seem to find it. Nonetheless, if you assume that our methodology captures a wage discontinuity they way it's supposed to, there is no significant disemployment effect.
I've read the Dube et al response multiple times trying to understand why it's not terrible, and I cannot figure that part out.