Thread by @joseazar

@Econ_Marshall @MartinSGaynor @martincschmalz @judy_chevalier @ProfFionasm @crampell @Noahpinion @steventberry Note: when I say we use concentration in other cities, I mean average of log (1/Number of firms), not HHI so it's not subject to endogeneity of market shares.

@Econ_Marshall @MartinSGaynor @martincschmalz @judy_chevalier @ProfFionasm @crampell @Noahpinion @steventberry Also note that this identification strategy was not possible in the old SCP work because the unit of observation was a national level industry, not geographic/occupational markets. So what we're doing is in a way closer to the NEIO approach (but still reduced form).

@Econ_Marshall @MartinSGaynor @martincschmalz @judy_chevalier @ProfFionasm @crampell @Noahpinion @steventberry Some thoughts on the "own-time" criticisms of SCP from @steventberry 's slides:

1. "Chicago" theoretical endogeneity of market shares. Low cost -> high concentration even with low prices. True, but this biases the effect towards zero, so implies SCP regs underestimate effect.

@Econ_Marshall @MartinSGaynor @martincschmalz @judy_chevalier @ProfFionasm @crampell @Noahpinion @steventberry 2. "Everyone: accounting data are terrible." Agreed! So, we don't use accounting data.
3. "Many: there is no single cross-industry theory of markets to guide cross-industry studies." True, but this critique doesn't apply to airlines market, and probably not totally to labor mkts.

@Econ_Marshall @MartinSGaynor @martincschmalz @judy_chevalier @ProfFionasm @crampell @Noahpinion @steventberry 4. "L. Weiss (and others): econometric endogeneity of shares and Herfindahls: what are possible IVs / what is excluded?" We can instrument HHIs by using number of firms instead of shares? Why is this worse than using, say, number of products to instrument for prices?