As a practical matter, I think Brad is right. As a technical matter, I still think Cochrane has a point - why am I wrong?

Cause I'm sure I'm wrong somewhere.

So reading some thoughts from Brad DeLong and Andrew Bossie made me think more about this... I've moved somewhat closer to them, but I still think Cochrane has reason to expect more than this.

Brad is making the point that the advantage of these state-level studies is that they hold monetary and financial policy constant. This is very true.

But because they use exogenous federal expenditures, paid for with federal taxes, monetary and financial policy isn't the only thing that's differenced out in the cross section. Taxes are too, and taxes are a drag on demand. This is Cochrane's point. Of course a gift-wrapped present from Massachusetts to California will not only make California look better, but it will make Massachusetts look worse (they had to pay for the gift).

Where I've moved closer to Brad and Andrew is that since we're really concerned about deficit financed stimulus, differencing out the taxes shouldn't really matter all that much. We're looking for stimulus not paid for with taxes. Taxing is austerity.

But I really can't blame skeptics for saying "Wait a minute - so you're differencing out the impact of taxes, which means you're differencing out the cost and measuring the impact of a free lunch. In the real world there are no free lunches - you have to got to the loanablefunds market".

Brad, Andrew, and I would then all pull our hair out and start talking about liquidity traps.

There is a free lunch.

But arguing that these state-level studies (which difference out the cost of spending and don't difference out the benefit) are applicable to today precisely because we are faced with a free lunch is assuming our own conclusions. I am convinced. Andrew is convinced. Brad is convinced. But there is nothing in the state-level analyses that makes us convinced of this. We came to the table convinced.

To convince Cochrane that we can generalize these state-level studies to today's situtation because we have a free lunch today, we need something more.

That "something more" inevitably is a lot of good theory, work on the Depression and WWII, and well specified national-level analyses (and maybe better specified state-level analyses that don't difference out the costs in this way).