I'm still not sure I fully understand this, but I'm going to post it anyway. That's what blogs are for.
[Updated, see halfway down the post.]
The basic idea is that one of the ways quantitative easing may work (there are others) is that it allows the central bank to buy stuff directly from the non-bank public, and directly increase the quantity of money in public hands. "Public" here can be a household, or a firm, or a pension plan, or mutual fund, or any financial institution, except a bank. A "bank" is a financial institution whose liabilities are media of exchange.
In normal monetary policy the central bank buys stuff from the commercial banks. If those banks are capital constrained, they do not respond by increasing loans to the non-bank public, so the quantity of money in public hands does not increase. Quantitative easing is a way for central banks to bypass the blockage caused by banks being capital constrained, and allows the central bank to increase the quantity of money, M1, the medium of exchange. And the proximate cause of a recession, with excess supplies of goods, is always an excess demand for the medium of exchange.
Here's how monetary policy is supposed to work in normal times, when banks are not capital constrained. And how capital constraints will stop it working the way it is supposed to. There are two ways to tell the story, and I am going to tell them both, because it doesn't matter for my current purposes which of those two stories you like best.
Horizontalist Neo-Wicksellian interest rate version. The central bank lowers the rate of interest on commercial bank reserves. Commercial banks in turn lower their rate of interest to the public, and increase loans to the public. Monetary aggregates like M1 will expand as a by-product of increasing bank loans to the public.
Verticalist banking multiplier version. The central bank increases the quantity of bank reserves by buying something (e.g. a bond) from the commercial banks. Commercial banks now have excess reserves (more than they desire), and so lend out some of those extra reserves to the public. The banking multiplier tells us that monetary aggregates like M1 will expand by some multiple of the increase in reserves.
Again, it doesn't matter which of those two versions you believe. Or even if you are a post-modernist who "believes" both stories. Each version requires commercial banks to expand loans to the public as part of the monetary policy transmission mechanism. But if banks are capital-constrained, so are unwilling to expand risky loans to the public, the monetary transmission mechanism hits a brick wall right there. The only interest rate that falls is the interest rate on bank reserves. Demand deposits do not increase, currency in public hands does not increase, M1 does not increase; the new base money sits uselessly as additional reserves at the central bank. The banking multiplier is zero. The traditional reserve ratio is not the constraint; it's the capital ratio. The monetary transmission mechanism never gets to first base.
How does "Quantitative Easing" help? Now QE can mean a lot of things to a lot of people. In this context I want it to mean that the central bank buys something directly from the non-bank public. It doesn't matter what it buys: a government bond, commercial bond, shares, toxic waste, antique furniture, whatever. What matters is that it pays for it with $100 base money: either a cheque drawn on the central bank, or new currency. The seller of that something deposits the cheque or currency in his chequing account at the commercial bank. The commercial bank now has $100 extra reserves, but also has $100 extra demand deposits. M1 expands by $100. And if the seller prefers using currency to cheques, and so doesn't deposit the $100 in his chequing account, currency in public hands increases by $100, and M1 expands by the same $100. Under QE, the banking multiplier is now one, not zero.
Why the difference between injecting new base money into the commercial banks, and injecting it directly into public hands, when commercial banks are capital constrained?
It's easy to understand in the case where the public wants to hold the $100 as currency rather than demand deposits. It's a simple transaction between the central bank and the non-bank public. Commercial banks' balance sheets are unchanged, so their capital constraint is irrelevant. It's harder to understand in the case where the public wants to hold the $100 as demand deposits, because the commercial banks' balance sheets must expand by $100. Why doesn't the capital constraint bite? I think the intuition is that, on the margin, the commercial bank is holding 100% reserves against the extra deposit. A bank with 100% cash reserves (currency or deposits at the central bank) does not need capital. Nothing can go wrong (apart from bank robbers). It's just keeping its depositors' cash in a giant safe. (But I wish my intuition were stronger on why it can't also work when the central bank injects base money into the commercial banks.)
[UPDATE. Thanks to comments, especially from JKH and RebelEconomist, my intuition is now a bit clearer.
If the central bank buys something from the non-bank public for $100, and the public then deposits that $100 in a capital-constrained bank, why does M1 increase by $100? JKH reminds us that reserves are zero risk-weighted, so an expansion of the commercial bank's balance sheet by adding $100 reserves and $100 deposits does not worsen the capital constraint. That accords exactly with my intuition about 100% reserve banking, so I'm now much more confident with that part of the story.
On the other part of the story, on why M1 does not expand, when the banks are capital constrained, here's another way of thinking about it:
We can model a capital constraint in either of two ways: as a "hard" constraint, in which a bank that has insufficient capital must immediately call in loans or sell assets; or as a "semi-hard" constraint, in which a bank with insufficient capital may is not forced to contract, but may not expand by making new loans.
We also need to ask, if the central bank buys an asset from the bank, is it a risky asset or a safe asset? If it's a risky asset, that may relax the commercial bank's capital constraint, and allow M1 to expand.
So if there's a "hard" capital constraint, which is just binding, and the central bank buys a risky asset, then the capital constraint may go slack, and M1 mat expand.
There's also a very interesting exchange between JKH and Jon in the comments below, that I wanted to get my head around and summarise, but have failed miserably. It's better you read it yourself. They know way more than I do about actual operating procedures.
Back to original post.....]
OK, so QE, interpreted as the central bank buying something directly from the non-bank public, and paying for it with an expanded monetary base, can circumvent banks' capital constraints, and get M1 to expand. We have got to first base. What next?
From now on, the monetary policy transmission mechanism is whatever story you normally tell. Here's my story.
First off, just because someone sold the central bank something in exchange for $100 doesn't mean he wants to hold an extra $100 in money. If I sell my car for $4,000, because someone offers me what I think is a good price, that doesn't mean I want to hold that $4,000 as currency or sitting in my chequing account. Money is the medium of exchange, remember. We always sell stuff for money, but most of the time we don't want to hold that money; we want to use it to buy something else. So that extra $100 will get spent. And when it gets spent it doesn't disappear; it just goes into someone else's pocket.
Regardless of the original cause of the recession, the proximate or immediate cause of a generalised excess supply is always an excess demand for the medium of exchange. We live in a monetary exchange economy, not a barter economy. Money is traded in every market; nothing else is. Everything else has a market of its very own; money doesn't.
As i explained in a previous post, an excess demand for antique furniture means that people cannot buy as much antique furniture as they like, because nobody is willing to sell as much as they would like to buy. Unable to buy antiques, they have to buy something else instead. End of story (unless they decide to hold money instead). But an excess demand for money is very different. An individual can always get more money, simply by buying less of something else. They can't all get more money, of course, unless the supply of money expands, but their individual attempts to get more money by buying less of other things causes a generalised excess supply, and a drop in monetary receipts, so that each individual will need to cut his spending even more. And it doesn't end until prices have fallen, or income and employment have fallen, by enough to make people stop trying to get more money. So the excess demand for money is eliminated in the new (possibly quite nasty) equilibrium.
Since the proximate cause of a generalised excess supply is an excess demand for the medium of exchange in the original equilibrium, the cure must be to remove that cause. M1 approximately corresponds to money as a medium of exchange. QE can increase the supply of M1, and eliminate the excess demand for the medium of exchange, even when banks are capital constrained, and normal monetary policy can't.