It's a choice between being surprised by immediate small changes, and being surprised by warnings of much bigger future changes. A simple and maybe obvious point. I don't know if it's been made before; if not, it should have been.
The Bank of Canada has eight Fixed Announcements Dates per year. At each FAD it announces its target for the overnight rate of interest until the next FAD.
Suppose the Bank of Canada did it differently. At each FAD it announces the overnight rate target for the following FAD. This would give financial market participants 6.5 weeks advanced warning of any change in the overnight rate.
I think I am right in saying that financial market participants in general, and the commercial banks in particular, would prefer that second way of implementing monetary policy. I think they would say that it is a better communications strategy, with greater transparency. I think what they mean is that they don't like being surprised, and prefer to have advanced warning of any changes in the overnight rate, which is perfectly understandable.
But why stop there? Wouldn't it be better still if the Bank of Canada announced at each FAD what it would do at the FAD after the next, so everyone would get 13 weeks advanced warning of any change in the overnight rate?
But why stop there? And so on.
In the limit, when the Bank of Canada first introduced the 2% inflation target, it should have announced the whole time-path for the overnight rate target from then until eternity. Which is a contradiction, because it wouldn't be able to hit its inflation target if it couldn't ever adjust its monetary policy instrument.
It's tempting to say that it's like asking the driver of a car to give the passengers advanced warning before moving the steering wheel, which makes the driver's job of keeping the car on the road much harder. But that analogy is not quite accurate. Because the economy is like a car whose direction depends not just on the current position of the steering wheel but also on the car's expectation of the future position of the steering wheel.
In the simplest case (the pure expectations hypothesis) the current two year interest rate is equal to the (geometric) average of the current one-year interest rate and the expected one-year interest rate for next year. So by announcing changes in the interest rate one year in advance, the Bank of Canada would be able to influence all current interest rates for terms greater than one year, but not less than one year. The longer the advanced warning, the smaller the subset of the term-structure of interest rates the Bank of Canada can influence. And if Aggregate Demand depends on the whole term-structure of interest rates, from shortest to longest, the smaller the subset it can influence, the bigger the change in the overnight rate that will be needed to have the same effect on Aggregate Demand.
Take a simple example to illustrate my point. Suppose Aggregate Demand depends equally on the one-period interest rate and the two-period interest rate, and the two-period interest rate is the simple average of the current one-period and expected future one-period rates. And suppose the economy is hit with random shocks that are permanent (the natural rate r* is a random walk). If the Bank of Canada changes the one-period interest rate immediately it observes a 1% shock to r*, it changes the one-period rate by the same 1%, and the two-period rate changes by the same 1% (because everyone knows that shocks to r* are permanent). But if the Bank of Canada has to give one period's advanced warning, it needs to change the current two-period rate by 2% (since it cannot change the one-period rate), and it needs to announce a change of 4% in the future one-period rate (since you need to change the expected future rate by twice as much as you want to change the two-period rate). And in the following period, if there are no additional changes to r*, it announces a future cut of 3%.
It gets even hairier if changes to r* are transitory. I will leave that as an exercise for the reader.
Your call, financial market participants who don't like surprises.
[We could talk about conditional commitments, where the Bank of Canada announces in advance what it will do, conditional on future data. But that inevitably leads to the Bank of Canada committing itself to an instrument rule (like the Taylor Rule for example). Which opens up another can of worms. So I will leave that aside.]
I marvel at your ability to shift perspective on obvious things in useful and interesting ways. It's one of the reasons I read this blog.
One response to this issue might be to poll financial market participants about how long in advance to set the rate. But then you run into the same infinite regress of how long in advance to poll them.
It seems like there should be some kind of information theoretic way to calculate the timing that would provide the most information to the market over an infinite period of time.
Posted by: Brett Reynolds | September 16, 2017 at 08:03 PM
Brett: Thanks!
The logic of my post leads to the conclusion that the Bank of Canada should eliminate Fixed Announcement Dates, and simply announce an immediate change in the overnight rate on any day as soon as it thinks the data warrant it. Which is what the Bank of Canada used to do, until a few years back. It changed to FADs for 2 reasons: market participants didn't like the uncertainty of never knowing in advance when the Bank was going to make an announcement; doing all the preliminary stuff for deciding what the overnight rate should be is quite a major job for the Bank, that it wouldn't want to do daily (and it felt this would lead to inertia "let's wait a bit"). Plus some data comes in quarterly, though 4 FADs per year seemed a bit too few, so it compromised on 8 (though maybe 12 i.e. monthly, might be better, because some data come in monthly).
Posted by: Nick Rowe | September 16, 2017 at 08:27 PM
When it becomes serious, you have to lie.
Posted by: Tel | September 18, 2017 at 07:18 AM
Tel: yep, that's another problem with communication, that never arises if you just do it.
Posted by: Nick Rowe | September 18, 2017 at 09:07 AM
Is there a practical difference in your "advance warning by one period" case and the current situation? As soon as the BoC announced that it "would" be changing rates as of 6.5 weeks from now, that information would be instantly priced into the capital markets. Just like it is today when it is priced in immediately.
The practical benefits for the markets in having 8 fixed days per year seem real as compared to random announcements, but I just don't see practical differences between making an announcement today with immediate effect, and an announcement today with "delayed" effect. It would be "delayed" in name only.
Posted by: PelinoC | September 18, 2017 at 12:03 PM
PelinoC: I tend to agree. But if you borrow very short, and lend long (like banks do), and if your long assets are illiquid and take time to sell, I wonder if you could make a case (ignoring the case against in my post)?
Posted by: Nick Rowe | September 18, 2017 at 12:14 PM
The purpose of the central bank is to strategically destroy information; this is not a "problem" with communication, it is a key feature.
If they always lied about everything, of course no one would listen. If they never tell lies they have no leverage.
Thus, there's a balance in terms of just the right amount of dishonesty to introduce into any given situation. This is the reason we can never replace the central bank with a simple algorithm... but it is in the central bank's interest to give the appearance that they do operate by a simple algorithm, except when they don't.
Posted by: Tel | September 22, 2017 at 07:56 PM