Adaptive Expectations

Adaptive expectations is a hypothesis about how economic agents form forecasts of future variables — most famously inflation, but also output, exchange rates, and asset prices. Under the simplest version, the expected value of a variable next period equals last period's expectation plus a fraction of the most recent forecast error. Formally, if πᵉₜ is expected inflation and πₜ₋₁ is realised inflation, then πᵉₜ = πᵉₜ₋₁ + λ(πₜ₋₁ − πᵉₜ₋₁), where λ is between 0 and 1. Expectations therefore "adapt" gradually to new information, and the forecast is a geometrically weighted average of past observed values.

The idea was developed in the 1950s, notably by Phillip Cagan in his 1956 study of hyperinflations and by Milton Friedman in his work on the consumption function and, later, the natural-rate hypothesis. Friedman's 1968 American Economic Association presidential address used adaptive expectations to argue that the Phillips curve trade-off between inflation and unemployment is only temporary: workers eventually revise their inflation expectations upward, shifting the short-run Phillips curve and returning unemployment to its natural rate.

Adaptive expectations were largely displaced in academic macroeconomics after the rational expectations revolution led by John Muth (1961) and Robert Lucas in the 1970s. Critics pointed out that adaptive agents make systematic errors — they are always behind the curve in a rising-inflation environment — which is implausible if agents learn. The Lucas critique (1976) specifically attacked policy models built on backward-looking expectations.

Nonetheless, adaptive expectations remain useful. They are tractable, fit some survey data on household inflation expectations reasonably well, and underpin many "learning" models and central-bank forecasting tools. Behavioural and heterogeneous-agent macroeconomics has revived interest in them as a realistic depiction of bounded rationality.