Central banks run monetary policy not so much by doing things, but by threatening to do things. If their threats are credible, we never observe them carrying out those threats, and we often observe them doing the exact opposite
A credible central bank is a bit like Chuck Norris. (Apologies to Lars Christensen for stealing his metaphor.) Chuck Norris simply looks at the target variable, and it moves to wherever he wants it to go. It looks like magic. But it works because nobody wants Chuck Norris to carry out his implicit threat. So he doesn't need to.
You might, if you were imprudent, say that Chuck Norris is like the confidence fairy, who also seems to move things by magic.
(Just in case you are not familiar with Chuck Norris jokes, they go like this: Chuck Norris doesn't read books, he stares at them until they tell him what he needs to know.)
Thesis. We teach the monetary policy transmission mechanism like this: the central bank pulls a lever, and that lever pulls other levers, which eventually move the target variable in the direction the central bank wants to move.
Antithesis. That's wrong. A credible central bank is exactly like Chuck Norris. It looks at the thing it wants to move, and the thing moves, and all the other levers fall into place where they should be. Causation runs backwards from the target variable. Credible central banks don't actually do anything. They just threaten to do things. But a credible central bank never needs to carry out its threats.
Synthesis. That's not quite right either.
1. Even Chuck Norris can't make the impossible happen. A credible central bank can move the economy just by saying that it wants the economy to move. But it must be a new equilibrium that it moves to. And maybe that new equilibrium won't be an equilibrium unless the central bank moves its lever. Chuck Norris can't clear the room if he is standing in the only doorway. He has to step aside to let people exit, even if he doesn't need to throw anyone out.
2. Chuck Norris wasn't always Chuck Norris. He had to earn his reputation. In the early days, or in unfamiliar territory, he actually had to carry out his threats. Till people learned the new regime.
Here's an example.
Every 6 weeks, the Bank of Canada announces its overnight rate target for the next 6 weeks. And the actual overnight rate instantly moves to where the Bank of Canada wants it to be. It's pure Chuck Norris. The Bank doesn't actually do anything. It just lets the market know where it wants the market to move, and the market moves there. The market moves because of the Bank of Canada's threat. If the overnight rate is above the Bank's target, the Bank will add however much settlement balances as are needed for as long as is needed until the overnight rate falls to the target. And if the overnight rate is below the Bank's target, the Bank will subtract however much settlement balances as are needed for as long as is needed until the overnight rate rises to the target. And because this threat is credible, the Bank doesn't need to carry it out.
But the Bank's overnight rate target is normally at the middle of two other interest rates: that interest rate it charges on loans to the commercial banks, and the interest rate it pays on reserve balances. The Bank has to adjust those two interest rates, so they don't create a disequilibrium with what the Bank wants to happen. It can't get the overnight rate below the interest rate it pays on reserves, for example. Chuck Norris can't block the exit if he wants to clear the room.
Here's a second example.
The Bank of Canada wants to keep inflation at the 2% target. How does it do this?
One way it could keep inflation at the 2% target would be by following some sort of Taylor Rule. Set the overnight rate equal to: some assumed constant real natural rate, plus 2%, plus 1.5 times (inflation minus 2%), plus 0.5 times (output minus potential output).
That's not really what the Bank of Canada does. It doesn't know the natural rate, which is not a constant. It doesn't know what potential output is either; even though it tries to estimate potential output its real time estimates are very different from its final revised estimates. So it can't follow a Taylor Rule, even if it wanted to. But most importantly, if you estimate the Bank of Canada's reaction function, it is very hard to find a coefficient on inflation anywhere near as big as the Taylor Rule requires. This is initially puzzling. But it's not puzzling at all when you think of Chuck Norris.
The Howitt/Taylor Principle is well-understood. If inflation moves 1% above target, the central bank must increase the nominal interest rate by more than 1% (by 1.5% in the Taylor Rule above), in order to increase the real interest rate, reduce Aggregate Demand, and bring inflation back down to target. But the Bank of Canada doesn't seem to react this strongly to inflation, and yet it has succeeded in keeping inflation at target. Which is puzzling.
The puzzle is easily resolved. As long as the Bank of Canada is credible, expected inflation stays at the 2% target. People believe that any fluctuations in inflation will be temporary, and don't change their expectations of future inflation. So if actual inflation rises 1% above target, but expected inflation stays at target, the Bank of Canada need only raise nominal interest rates by (say) 0.5% to raise real interest rates, and bring inflation back down.
What the Howitt/Taylor Principle really says is that the central bank must raise the nominal interest rate by over 1% for every 1% rise in expected inflation. The central bank must threaten to increase the nominal interest rate by whatever it takes to bring expected inflation back to target. But if that threat is credible, we would never observe it being carried out. Because expected inflation would never deviate from target. It wouldn't dare risk the wrath of Chuck Norris.
If an econometrician ever observed the Howitt/Taylor Principle in action, he would know that he was not observing a rational expectations equilibrium, and that the public was in the process of learning and testing the central bank's credibility. He would be seeing Chuck Norris carry out his threat.
How does the Bank of Canada keep inflation on target when it doesn't observe the natural rate of interest, and doesn't observe potential output?
The Bank of Canada keeps inflation on target, and expected inflation on target, by making threats. Those threats do not have a finite magnitude or duration. It threatens to raise or lower interest rates by however much it takes for however long it takes to bring actual and expected inflation back to target. That threat is what keeps expected inflation well-anchored at the 2% target. And keeping expected inflation on target is well more than half the battle to keeping actual inflation on target.
Here's a third example.
If a central bank wants to keep expected inflation on target, it must threaten to reduce nominal and real interest rates if expected inflation falls below target, in order to make expected inflation rise. But suppose a central bank wants to increase its target rate of inflation. It announces that the new target is 3% inflation. If the central bank is credible, and expected inflation increases to 3%, the central bank must increase the nominal interest rate to prevent the real rate from falling.
Chuck Norris is standing in the doorway. He threatens to go forward into the room to throw people out. But his threat is credible, so he actually steps back to let people exit. He does the exact opposite of what he threatened to do, because he doesn't need to carry out his threat.
A central bank that wants to increase expected inflation must threaten to cut nominal interest rates if expected inflation does not increase to where it wants it to go. But if its threat is credible, it does the exact opposite of what it threatened to do. It raises nominal interest rates.
At the zero lower bound on nominal interest rates, the standard threat strategy cannot work. It is no longer credible for the central bank to threaten to cut nominal interest rates if expectations don't move to where it wants them to move. It has to threaten to do something else. Like start buying stuff. How much stuff? For how long? That answer is exactly the same as when we are talking about interest rates. As much as is needed for as long as is needed.
But if the threat is credible, it need never be carried out. In fact, if the threat is credible, and expected inflation increases, the demand for money will fall, and the central bank will need to do the exact opposite of what it threatened to do. It will need to start selling stuff, to reduce the supply of money in line with the reduced demand for money.
How much QE will the Fed need to do, to get expected inflation to increase?
That is the wrong question to ask. The Fed needs to communicate its target clearly. And it needs to threaten to do unlimited amounts of QE for an unlimited amount of time until its target is hit. If that threat is communicated clearly, and believed, the actual amount of QE needed will be negative. The Fed's balance sheet is much bigger than in normal times. But in order to shrink its balance sheet back to normal, the Fed must threaten to expand its balance sheet by an unlimited amount. And be seen to be ready to carry out that threat.