OK, this post is a little on the whimsical side. The argument is a lot less strong and clear than I would like it to be. But sometimes you just have to say what's in your head, and hope that will help it get clearer. Read at your own risk (as if that needed saying).
Suppose, just suppose, that the Governor of the Bank of Canada became an existentialist. And everyone knew that.
He's a free man. Each morning he wakes up to a brand new day. It doesn't matter what he did yesterday. Bygones are forever bygones. What will he set the overnight rate at today? He decides to set it at 11%. Yesterday he set it at 1%. But so what? That was yesterday.
In the face of an unexpected one thousand basis point increase in the overnight rate, the bond market rolls its eyes, yawns, and goes back to sleep. Because the bond market knows that tomorrow will be yet another brand new day for the existentialist Governor, and what he does today tells us next to nothing about what he will do tomorrow. He will be a totally different person tomorrow. One day at 11% rather than 1% is nothing in the life of a 30 year bond.
An existentialist Governor is not really a Governor. He is a daily sequence of different governors. And because his daily actions would have so little effect, he would need to make truly massive surprise changes to have any effect at all. And since he knows this, he does make massive changes. But those massive changes have zero weight on his future selves.
That, surprisingly, is exactly how monetary policy works in economic theory. At least, that's how it works in theories where it is assumed that the central bank re-optimises every period, and cannot pre-commit its future actions. The central bank is assumed to be run by an existentialist. The only reason those theories work at all is because we imagine they use quarterly data. So the central bank only has an existentialist moment four times a year.
That is not at all how monetary policy works in practice. In practice, if the Governor made a surprise thousand basis point hike in the overnight rate, the bond market would have a nervous breakdown trying to figure out why the Governor did that, and what his actions mean for the future. "Does he think that demand and inflationary pressures have suddenly gotten much stronger, so he needs that massive thousand basis point hike to offset that pressure and keep inflation on target? What does he know that we don't know? Or maybe he wants to reduce inflation to a new lower target? Or maybe, just maybe, he's trying to tell us that he wants a higher inflation target, and if we figure that out, we will be expecting higher inflation from now on, so he needs to raise the nominal interest rate to keep the real rate from falling? Or was it just a fat finger on the press release, so it doesn't mean anything, will be corrected tomorrow, and we can all go back to sleep?"
The bond market, and the stock market, and the forex market, and the supermarket, and the labour market, will all want to know what the Governor means by raising the overnight rate one thousand basis points. "What is he trying to tell us about his future actions?"
Today's action by the Governor has an almost negligible effect on anything. How his action today affects expectations of his future actions is what matters. And the Governor knows this. So he will be choosing an action that changes expectations in the way he wants them to be changed. And the market knows this. And the Governor knows this too. And so on.
We can imagine a world in which conventions get established. Suppose the market has a superstitious belief that if the overnight rate target ends in a 5, it is likely to be a temporary change; and if it ends in a 0 it is likely to be a permanent change. And the Governor knows the market has this superstition, and the market knows the Governor knows this, and so on. It's common knowledge. That superstition is enormously helpful to the Governor. If he wants to tell the market he thinks this will be a temporary change, he makes sure the new target ends in a 5. If he wants to tell the market he thinks it will be a permanent change, he makes sure the new target ends in a 0. And as long as the Governor is honest, the market will believe what he "says", even if they subsequently forget the superstition.
In an economy with that convention, the IS curve would be very spiky. Or, there would be one IS curve joining the dots which end in 5, and another IS curve joining the dots that end in 0. Two very different IS curves, with different slopes, because the effect of a change in the overnight rate depends on whether it is expected to be temporary or more permanent.
Most English-speaking people have a similar superstitious belief that the word "cat" means cat. That superstition works pretty well. It lets us all talk about cats, and understand each other. It's just a convention, of course. It doesn't matter what word we use to mean cat. Just as long as we are not all existentialists, and keep changing what words mean every new day. True existentialists can't speak English. Or even Parisian French. Why should words mean what they meant yesterday?
The word "grue" means "green until 2013, and blue thereafter". The word "bleen" means "blue until 2013, and green thereafter". In the other half of the dictionary, the word "green" means "grue until 2013, and bleen thereafter". And "blue" means "bleen until 2013, and grue thereafter".
All emeralds so far observed have been both green and grue. If we believe in inductive inference, should we expect emeralds to continue being green, or continue being grue, after 2013? Will emeralds change colour or do nothing in 2013? What do you mean by "doing nothing"? Do you mean staying green, or staying grue?
Suppose you see monetary policy change today. Do people expect monetary policy to change again or do nothing in 2013? What do you mean by "doing nothing"? Because if you think that you can project the present into the future, so that the null hypothesis is doing nothing, your actions today will depend on what you mean by "doing nothing".
Take a simple duopoly model. Two firms in an industry, each produces a good that is an imperfect substitute for the other firm's good. Write down the demand functions and cost functions for the two firms. Assume each firm maximises profits. Assume all that is common knowledge between the two firms. Assume Nash equilibrium in a one-shot game with both firms moving simultaneously.
As economists have known for centuries(?), all that is not enough to predict the outcome. There's the Cournot-Nash equilibrium, where each firm chooses a quantity to maximise profits, given the other firm's quantity. And there's the Bertrand-Nash equilibrium, where each firm chooses a price to maximise profits, given the other firm's price. And the prices and quantities chosen in Cournot-Nash are different from the prices and quantities chosen in Bertrand-Nash. Any upper year economics undergraduate could do the math to prove this. I did it once.
It matters whether the two firms speak the quantity language or speak the price language. A change in the language in which we talk about and think about the concrete reality changes that concrete reality. We get lower prices and higher output if the duopolists speak P-language than if they speak Q-language.
Sure, P-language can be translated into Q-language, and vice versa. But an unconditional statement about what price the firm will set is equivalent to a conditional statement about what quantity the firm would set, as a function of the other firms quantity, to ensure that same price. (And vice versa). And then we have re-opened the game, because there is no auctioneer to solve those two conditional reaction functions for the equilibrium.
If language matters for nuts and bolts in something as concrete as industrial organisation theory, where the practical no-nonsense feet-on-the-ground applied economists hang out, how can it not matter in monetary theory? The central bank can say things in one language which it cannot say in another language. The interest rate language they currently use does not contain negative numbers. And it is an ambiguous language when it comes to whether an increase in nominal interest rates means a tightening or a loosening of monetary policy.
The people of the concrete steppes won't like this post. Because I'm saying their concrete steps will go up or go down, depending how you talk about them. They aren't made of concrete. But it's incredibly hard, when you have gotten used to one conceptual scheme for thinking about the world, to realise that your conceptual scheme is not concrete. There is nothing essentially catty that all cats have in common. "Cat" is just a more or less useful name, a bit fuzzy round the edges, that we use to divide off some animals from others. But in economics, unlike biology, how cats behave depends on whether we call them "cats".
Von Mises said, in Human Action IIRC, that economics is reasoning about reasoning beings. That's true and important. But economics is also talking about talking beings. It might be pushing things only a little too far if I said that economics is a sub-field within philosophy of language.
[I won't be near a computer much this weekend, so won't be responding quickly to comments.]