RDP 9803: Forward-Looking Behaviour and Credibility: Some Evidence and Implications for Policy 4. Conclusions

The evidence on inflation expectations is mixed, with the empirical evidence in this paper suggesting that inflation expectations are a combination of simple extrapolation from the past and explicit consideration of what is likely to happen in the future, including the expected behaviour of the central bank. There are also differences between groups of people in the economy, with, for example, financial market economists' inflation expectations being much more consistent with rational expectations than households' expectations. The tests in this paper are not exhaustive, and so this assessment is only tentative.

We also examine some policy implications that arise from the different ways in which expectations can be formed. Since we present and analyse these implications using a model, the assessments are, to some extent at least, model dependent. We have three main conclusions.

First, inflation and output are substantially less variable when policy-makers are forward looking, setting interest rates using model-consistent expectations of inflation and the output gap. For the particular model used, forecasts of inflation and the output gap four quarters out are more efficient than forecasts over other horizons if interest rates are set by a Taylor-type rule. Given that the output gap contains information about future inflation, this implies that policy is reacting to forecasts of inflation four to seven quarters ahead.

Second, the effect of shocks depends on how private-sector expectations are formed, with forward-looking behaviour being more stabilising than simple extrapolations from the recent past. This additional stability is accounted for by people recognising that the central bank will respond to ‘shocks’ in order to stabilise inflation and growth. For example, if people think that the central bank will accommodate inflation shocks, but the central bank in fact does not do so (which is the case of an inflation target with no credibility), then inflation is still tied to the target in the medium term – because there is a policy response – but it is considerably more variable, as is output.

Also, in the simple model we use, it is straightforward to generate heterogeneous inflation expectations in the private sector, such that the ex ante real interest rates that affect activity and the exchange rate differ from each other. This can induce either oscillations or overshooting in the exchange rate (depending on the particular mix of expectations), with implications for the variability of inflation and output.

Third, optimal policy – by which policy-makers set a path for current and future interest rates which minimises variations in inflation around target and output around potential – does better than a simple rule. But expectations about the future which are only simple extrapolations from the past produce more variability in inflation and output than more explicitly forward-looking behaviour, and optimal policy cannot fully compensate for this difference in expectations (given that the shocks are the same in each case). In this sense, policy-makers have to do what they can, given the constraints imposed on them by the structure of the economy.