Bob Murphy has a good post up on my conversation with him on Bastiat. You can read him for context, but what it boils down to is that he and I started parsing out the costs and benefits and we got to a point where he thinks I'm going back on an earlier assumption. In one way I was - and I didn't mean to because I still maintain my earlier assumptions about crowding out. But in a more accurate sense, saying "Daniel's going back on an earlier assumption" is too generous for me because my conversation with him in the comments was simply wrong. So let me clean this up a little bit.
We have a Keynesian economy where 0 < MPC < 1, Y=E, there is some autonomous spending, and there is some investment expenditure unrelated to Y or E, which we'll call "S" (for Bastiat's "shoe spending"). This should bring to mind some familiar dynamics, and it's presented in the first panel of the figure below. Vertical lines mark C and Y=E.
Then, disaster strikes. Underneath that first panel we have a new autonomous component of expenditures, R (for "repairs"). R is done by the same people that do S, and so S declines in response to an increased spending on R.
Now, the $787 billion question is "how does S respond to the introduction of R?". Bastiat's position is that it is completely displaced. Bob Murphy showed us where Bastiat wrote that in this post, which was one of the few enlightening Bastiat posts I've read recently (Do you see a pattern here? Read Bob Murphy!). Keynesians disagree with this, and I disagree with this - under certain conditions. This is the point I botched in the conversation with Bob. On the one hand I assumed a multiplier effect, but on the other hand I said it was completely crowded out (Bob could have pointed out that this was contradictory which means I was dumb, but he gracefully assumed I am not dumb and just agree with him).
Right now we think that a lot of money is being kept liquid, so that new spending doesn't necessarily crowd out old spending. Seems reasonable (to me). So spending on S doesn't completely disappear in the second panel, but it is curtailed. The addition of R2 to S2 and C brings the new Y=E up above the old Y=E: higher GDP due to spending on R at a time when there isn't perfect crowding out. Voila.
Now, I still am maintaining that a net benefit for the economy seems very unlikely, although I have just demonstrated I think there is a gross benefit in the form of an increase in GDP and employment. The reason for this is that there's still a lot of crowding out, so only a small portion of the increase in Y=E (see the horizontal axis) associated with an increase in R (see the vertical axis) is an actual increase - the rest is just a displacement of Y=E attributable to spending on S1.
Hold on a second - what was that? Why does Daniel think so much crowding out is happening? Don't he and Krugman and DeLong and all those other kooks think there are all sorts of free lunches all over the place?
Well you have to think about exactly what is being proposed here. The presumption is homeowners and shopowners are rebuilding, right? They don't have much of a choice but to divert a lot of their spending towards repairs. Doesn't this contradict what I say about the stimulus? No, it doesn't. Why?
2. The Stimulus
The key here is to remember what Keynesians actually propose, which is why the other day I wrote "The boilerplate Keynesian position is to increase spending and lower taxes during a downturn. So there is no proposal of taking money from anybody. The point is to create money or other safe, liquid assets (like, say, Treasury debt) for which there is an excess demand". One commenter said it was "stupid". Another called me a "loon". But the point of this is crucial.
If you really want to analogize cleaning up after a disaster to economic policy, it would have to be analogized to a situation where the government pays for a $787 billion stimulus by levying a $787 billion tax on investors. That would be a real broken window situation. You might have something like I demonstrated in panel 2 of the figure above: a modest increase in output due to the fact that perhaps some of those taxes would be paid out of hoards rather than investments. After all, the government's liquidity preference is essentially zero and investors' liquidity preference is some positive figure. So you might get some growth from that. But not really much of any. That's why Obama didn't demand a $787 billion tax increase with the ARRA bill. That's why Krugman wasn't clamoring for a $787 billion tax increase. That's why I - unlike Krugman or Obama - didn't want to repeal the Bush tax cuts.
So for the reasons Bastiat laid out, paying for window repair with money budgeted for buying shoes is not going to increase growth. A Keynesian might quibble that if the shopkeeper had a non-zero level of liquidity preference you might get a little growth (because not all of his shoe spending would be displaced), but I'd maintain that once you take the loss of wealth into account, you're still likely to end up with a net loss in economic welfare, despite the increase in GDP. Same goes if you tax the shopkeeper and then pass a stimulus package with that tax money.
Of course, tax money and money budgeted for buying shoes isn't the only way to fund these things. Enter, the bond market.
This is what really sets the stimulus apart from any of this discussion about Bastiat. This is why I can say that with a few quibbles I agree with Bastiat, but that I also agree with Krugman and Keynes (and that Bastiat doesn't really contradict Keynes in a lot of the ways people seem to think that he does).
Let's think of the loanable funds market, which I find to be a much better way to think about the origins of the IS curve in the IS-LM model than the modern tendency to derive it from the national income equation.
A lot of Keynesians (maybe not all?) think that the loanable funds market is not clearing because of the zero lower bound on interest rates. There's a lot of bitching and moaning on this point, but let's take that as given for the sake of illustrating why regardless of what you think of the zero-lower bound it's still very different from Bastiat's shopkeeper. In the figure below, Id is investment demand, Ss is savings supply (loanable funds supply), and the market clearing equilibrium point is below the zero lower bound. Since that's inaccessible, the actual equilibrium point is of course the point where the Id curve intersects the x-axis. What happens when government borrows in this market? It depends on how much they borrow, of course.
If they borrow at G1 and hit the market clearing equilibrium exactly, there is no crowding out, because at this price you still have the same amount of investment demanded that you did before. If the government borrows more than that (say, at G2), then you start to crowd out investment. Notice this is very different from supporting stimulus with taxes (which still have a very large opportunity cost) or confiscation, or anything else like that.
This is why the Bastiat talk sounds like such a non-sequitor to Keynesians. Bastiat doesn't sound wrong to us - he sounds irrelevant. When you quote Bastiat to talk about the stimulus we hear "you shouldn't tax people to spend on stimulus", to which our response is "duh". When you quote Bastiat to talk about the GDP response to disasters, we say "well actually what's kind of cool is that you could get something of a bump in GDP because people are hoarding more than usual right now, but disasters still suck".
I feel like these two conversations collide a lot which leads to a lot of the confusion. I think I've clarified my/our position a little bit here.
Multipliers depend crucially on the extent of crowding out. That depends on (1.) how spending is financed, and (2.) the macroeconomic facts on the ground. Bob talks about a 1.84 multiplier. Fine. That's a 1.84 amplification of whatever spending you increase, but it's also a 1.84 amplification of whatever spending you reduce (paradox of thrift, anyone?). So the empirically measured multiplier from a debt-funded stimulus, a tax-funded stimulus, and a spending-replacement stimulus (i.e. - cut the Dept. of Education and the Pentagon's budget to build bridges) are all going to be different because there's different levels of crowding out associated with each. What I should have said to Bob Murphy was that we need to take:
1. Minus the $1 billion in damages
2. Plus the $1 billion in repairs
3. Plus the $840 million in the multiplier from the repairs
4. Minus the $X million in reduced shoe purchases
5. Minus the $(X)*(0.84) million in reduced shoe purchase multiplier
Bastiat thinks X = 1,000. I think X < 1,000 but still pretty big. And ultimately, you still have the $1 billion loss of wealth, so I don't see how any of this is likely to result in a net benefit (although it will result in a gross benefit to output.
Now - if instead the government funded $1 billion in repairs through deficit spending, then we'd be looking at some net benefits. But that would be a completely different story from what Bastiat tells - and we don't have to wait for a broken window to fund $1 billion in spending through deficits.