QE is just a huge open market operation. The Fed buys Treasury securities and issues bank reserves instead. Why does this do anything? Why isn't this like trading some red M&Ms for some green M&Ms and expecting it to affect your weight? (M&M of course stands for "Modigliani Miller" if you didn't get the joke.)
The usual thinking is that bank reserves are "special." They are connected to GDP in a way that Treasuries are not. In the conventional monetary view, MV = PY. Bank reserves, through a multiplier, control M. The bank or credit channel view says that bank reserves control lending and lending affects PY. The red M&Ms, though superficially identical, have more calories.
In Andy's view (my interpretation), that is turned around now. Now, Treasuries supply more "liquidity" needs than bank reserves, and (more importantly) the supply of treasuries is more connected to nominal GDP than is the supply of bank reserves.
Part of this inversion of roles is supply. In place of the usual $50 billion, we have $3 trillion or so bank reserves. Bank reserves can only be used by banks, so they don't do much good for the rest of us. Now, they just sit as bank assets in place of mortgages or treasuries and don't make a difference to anything. More treasuries, according to Andy, we can do something with.
More deeply, constraints only go one way. Normally, the banking system is up against a constraint. Reserves pay less interest than other assets, so banks use as little as possible. Now, they are awash in liquidity. You can't push on a string, as the saying goes. Much "constraint" economics forgets that once the constraint is off, the relationship doesn't hold any more.
Andy describes the repo market and the sense in which Treasuries are "special" in providing low-haircut collateral. Lots of academic research is now viewing Treasuries as special or liquidity-providing in the shadow banking system.
So, this is at least a gorgeous possibility: In a frictionless world, open-market operations, buying one kind of government debt (Treasuries) and issuing another (reserves) have zero effect on anything, by the M&M theorem. Monetary economics thinks the M&M theorem is violated, because one kind of government debt (M) is connected to nominal GDP and the other is not.
But financial systems change. When the textbooks were written, banks mattered a lot, so bank reserves, leveraged to loans and checking accounts, were the "special" asset. In today's market, and given today's glut of reserves, Treasuries, leveraged to mortgage backed securities and money market funds through the repo market and "shadow banking system," might be the "special" asset connected to nominal GDP. In that case, the effects of open market operations might have the opposite sign. As Andy says,
... the Federal Reserve's policy—to stimulate lending and the economy by buying Treasurys..—is creating a shortage of safe collateral, the very thing needed to create credit in the shadow banking system for the private economy. The quantitative easing policy appears self-defeating, perversely keeping economic growth slower and jobs scarcer.I'm not totally convinced, though this story and the alleged enormous demand for Treasuries is being bandied around as established fact. I'm also not convinced that this is all a good idea. Maybe the Fed should starve the shadow banking system.
You repo a security so that you can borrow against it. For example, you might buy a mortgage-backed security, then leave (repo, really) that security as collateral for a loan, which you used to buy the security in the first place. But what sense does it make to repo-finance a Treasury? You can't borrow at lower interest rate to make money on a Treasury! You could, possibly, if it's a long term Treasury and you're borrowing short, betting that interest rates don't rise. But I would think an interest rate swap or future would be a cheaper way to make that bet, and anyway betting on the slope of Treasury yield curve doesn't add up to the necessary GDP-linked lending that Andy has in mind.
In short, if you have money to buy a Treasury, why do you need to borrow? For any of this to get off the ground, you have to have some other, not totally rational, reason for buying the Treasury, and then you want to borrow against the Treasury so you can buy the risky asset that you really wanted all along. Who is that? Why is this such a necessary part of our financial system? Can't we fix things so they just buy the MBS with their initial cash?
Andy points out that repos are re-hypothecated. You use your Treasury as collateral against a loan, then the guy you gave it to uses it again as collateral to get the money to give to you. So one Treasury is used as collateral against two or three loans. Hmm. As the money multiplier creates run-prone structures, so using the same thing as collateral two or three times is a lot of what makes banks "too big to fail." If we all go down, who has the collateral?
A system awash in all kinds of liquidity, following the Freidman optimal quantity of money, seems a lot safer to me. I'd rather we expand the "bank reserve" concept -- fixed-value, floating-rate, electronically-transferable Treasury debt, and lots of it, washing the shadow banking system in liquidity and putting the run-prone structures out of business. Of course, open market operations would then have no effect in my world either, as I have removed the liquidity constraint in the shadow banking system just as Mr. Bernanke has removed it in the conventional banking system. But violations of M&M always mean the system can be made better.
If you want to comment and explain shadow banking, please use little words that the rest of us can understand.