Until very recently, it seemed like nearly everyone believed that monetary policy worked something like this:
The Federal Reserve buys government bonds in the open market and thereby swells bank reserves. The banks then put these reserves to work by lending them out.
Here's Paul Samuelson writing in 1948 (via Paul Krugman) and more or less adopting this way of thinking:
For in the middle of a deep depression just when we want Reserve policy to be most effective, the Member Banks are likely to be timid about buying new investments or making loans. If the Reserve authorities buy government bonds in the open market and thereby swell bank reserves, the banks will not put these funds to work but will simply hold reserves. Result: no 5 for 1, "no nothing," simply a substitution on the bank's balance sheet of idle cash for old government bonds. If banks and the public are quite indifferent between gilt-edged bonds—whose yields are already very low—and idle cash, then the Reserve authorities may not even succeed in bidding up the price of old government bonds; or what is the same thing, in bidding down the interest rate.
That was the basic model: monetary policy is loosened by increasing reserves, which the banks either do or do not "put to work" depending on whether they are feeling timid. Falling interest rates, in this way of thinking, were a by-product of rising reserves.
You can hear a lot of echoes of this line of thinking when people voice concerns that quantitative easing or zero-interest rate targets will rapidly lead to inflation once banks start lending out all of those reserves that they've accumulated.
But that's not how our system works anymore. Quite the opposite really.
Here's Peter Stellar explaining the new reality very clearly:
Bank reserves, or deposits at the central bank, have a singular role, to facilitate management of the payments system. They are used to settle transactions among banks and never leave the possession of the banking system. Indeed, in most systems, deposits at the central bank can only be held by banks, thus they are never lent "out" of the banking system. Consequently the notion that the quantity of bank reserves somehow constrains lending in a fiat money world is completely erroneous. Even in systems where legal reserve requirements are imposed on certain bank liabilities, all modern central banks allow the quantity of reserves to be demand driven in normal times. Consequently, monetary policy actually has little to do with money. It is more accurately thought of as the control of precisely the interest rate at which the central bank provides the reserves (over which it has a monopoly of creation) that are demanded by banks. A corollary of this fact is that the money "multiplier" is nothing of the sort. Bank reserves are not "multiplied" by the banking system nor is such a multiplication necessary for the expansion of bank credit and monetary liabilities. If there is a single fact that illustrates this point it is that total commercial bank deposits at the Federal Reserve in 1951 ($20 billion) were larger—in nominal terms—than at end 2006 ($19 billion) while total US credit market assets rose by over 10,000 percent in nominal terms during the same time period.
If this way of thinking about reserves were more widespread, there might be a little less panicked talk about the costs of current Fed policy.
—By CNBC's John Carney. Follow me on Twitter @Carney