Sunday, 15 November 2009

Can we build a theory of expectations?

If Scott Sumner and Paul Krugman agree on something, it must be true. I think most of us would accept that as an axiom of economic sociology.

Accordingly (see here and here), we can take as read the idea that expectations of inflation and aggregate demand strongly influence actual inflation and aggregate demand.

There's nothing really controversial about this: the idea of multiple equilibria is well established. If people expect deflation and recession, they will try to save more, spend little and invest less - and deflation and recession will result. If people expect inflation or growth, they will spend more quickly and invest in the hope of protecting their assets and capturing a share of that growth; and as a result, the expectation will be fulfilled.

Or as it's caricatured by a self-help motto you've probably seen: if you think you can, or if you think you can't, you're right.

Presumably we want to have more growth rather than less, and (mild) inflation rather than deflation. So we'd want people to expect growth and inflation respectively. Can we, then, influence people to change their expectations? Certainly.

There are two separate expectations here: inflation, and growth. They are linked, but I'll start with inflation as it's a little simpler.

One way to change expectations, on which Scott and Paul agree, is for the central bank to announce a policy of higher inflation. Paul's view, which is more mainstream, is that the Federal Reserve should simply have an explicit inflation target - perhaps 4%. If people believe that the Fed has the power to achieve this, then they will act accordingly and the target will become self-fulfilling.

Imagine a bus driver announces that he'll drive on the left-hand side of the road. It's quite credible that he can implement this policy, because his bus is big and heavy. Therefore it becomes self-fulfilling, because everyone coming the other way drives on their left, and the bus driver doesn't have to work very hard to achieve his goal.

However, many people might not believe the Fed has the power to achieve this - perhaps they think it can't get money into circulation fast enough, or maybe that its tools are not precise enough, that it is at risk of overshooting and therefore may undershoot to avoid this risk.

Thus, the Fed announcing a target of 4% inflation might not be enough to get people to expect 4% inflation. Could we build a model of how individuals form their expectations? If so, we might get some idea of how to communicate, what actions to take, and what incentives to put in place to affect expectations just right.

Scott has a solution. His policy proposal cleverly bypasses the need to understand the mechanisms of expectation forming by using the price mechanism instead. It says that the Fed should not target inflation but should target the market's expectations of inflation. The Fed would measure market expectations by buying and selling futures contracts on next year's price level. By looking at the prices of those contracts they would know exactly what the average expectation is, and therefore could take exactly enough action to raise or lower expectations (i.e. the prices of those futures) by the right amount.

An ingenious scheme indeed. However, it still relies on the Fed having enough power (through monetary policy) to be able to adjust the prices of those futures contracts. Does it have this power? As a counterfactual, imagine the Fed decided to target the market price of a 12-month future contract on barrels of oil - it is not at all obvious that it would be able to do so.

And when we consider expectations of real growth instead of inflation (or nominal growth, which is Scott's preferred target) everything gets a bit more complex. Monetary policy can probably help us to get out of recessions (because negative growth is a perverse state for the economy to be in); but it surely can't increase the productivity of firms across the economy - which is what is needed to sustain real growth.

Expectations of growth, on the other hand, probably do influence productivity: my experience of firm behaviour is that they invest more in, and care more about, their output when they have an external benchmark. I would love to see some data on this if there is any. Perhaps monetary policy is still the best solution, because nominal growth may be just as relevant as real growth for setting expectations.

Let me touch briefly on the theory of rational expectations. This is the idea that any time expectations are out of line with market prices, market prices will immediately change to reflect expectations. For example if oil is currently $70 a barrel and everyone expects it to be $100 by Christmas, it will jump up to $100 right now (with a small discount for interest rates). Thus, expectations will always be the same as market prices, and market expectations by definition will be fulfilled. That is, whatever people expect is true, and whatever happens was already expected.

Another clever idea, but it begs two questions:
  1. Markets are composed of many people with different expectations - does it matter that the price level reflects a weighted average of a range, instead of a single opinion held by the whole market?
  2. Do expectations and prices really move instantly, or do illiquidity, imperfect information and bounded rationality mean that they take some time to change? If so, what if expectations change again before the price catches up? Will the arbitrage opportunity implied by the rational expectations ideal ever really exist?
Neither of these questions are well-understood; the first is glossed over by the idea of a representative agent, the second by doubtful assumptions about efficient markets. Attractive as Scott's idea is, I suspect it too is at risk from these same questions.

So let's get busy understanding - and learning how to change - some psychology. What factors determine the amount of growth, and inflation, people expect?

This knowledge will help us build an economy that achieves its potential, keeps growing and suffers from fewer and milder recessions.

Update: Scott has an excellent posting here summarising the relevant bits of macroeconomic theory and some interesting cultural points; Nick Rowe has another good (though technical) article here. Meanwhile, I realised after posting that I spent so much time summarising the background to the expectations debate that I didn't make any actual proposals on how to model them. I'll come back to that tomorrow.

No comments: