A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Agents must form expectations about future inflation, interest rates, output, and other macroeconomic variables using available information and forecasting models. Alternative mechanisms—rational expectations, adaptive expectations, rule-of-thumb heuristics, and sticky information—imply vastly different inflation dynamics and policy transmission. The choice of expectation mechanism critically affects predictions about inflation acceleration, policy credibility, and the efficacy of stabilization policies.
From rational expectations, you know the benchmark assumption: agents use all available information and understand the true model of the economy, so their forecasts are correct on average. This is elegant but demanding — it requires that households and firms solve the same model economists do and update instantly when new data arrives. Expectation formation mechanisms are the broader menu of assumptions about how agents actually forecast the future, and each assumption leads to dramatically different macroeconomic predictions.
Adaptive expectations are the simplest alternative: agents forecast future inflation by extrapolating from recent past inflation, typically as a weighted average of observed values. If inflation was 3% last year and 4% this year, an adaptive forecaster might expect roughly 4% or slightly higher next year. The key property is backward-looking behavior — agents ignore structural changes in policy and learn only from experience. Under adaptive expectations, a central bank that tightens monetary policy to reduce inflation faces a painful adjustment period because expectations lag behind reality. Inflation persistence is built into the model because expectations are anchored to the past, not the future. The Phillips curve becomes accelerationist: any attempt to hold unemployment below its natural rate causes continuously rising inflation, as expectations ratchet upward period after period.
Sticky information models (associated with Mankiw and Reis) offer a middle ground. Agents *would* form rational expectations if they could, but they update their information sets infrequently — perhaps because acquiring and processing macroeconomic data is costly. In any given period, only a fraction of agents have the latest information; the rest operate on stale forecasts. This produces inflation dynamics that look partly forward-looking and partly backward-looking, matching empirical patterns better than either pure extreme. A credible disinflation still takes time to work because many agents haven't yet absorbed the policy change, but it works faster than under pure adaptive expectations because the agents who *have* updated immediately adjust their behavior.
The choice of mechanism has enormous policy implications. Under rational expectations, a credibly announced disinflation can be costless — if everyone believes the central bank will reduce money growth, expected inflation drops immediately, and actual inflation follows without any need for a recession. Under adaptive expectations, the same disinflation requires a prolonged period of high unemployment to force observed inflation down, which then slowly drags expectations lower. Under sticky information, the cost is intermediate. This is why the expectation formation assumption is not a technical detail buried in model appendices — it is the single most consequential modeling choice for evaluating whether monetary policy can painlessly reduce inflation, whether fiscal stimulus will be offset by forward-looking consumers, and whether central bank communication and credibility matter for real economic outcomes.