Aggregate Implications of Heterogeneous Inflation Expectations: The Role of Individual Experience
What this paper finds — and why it matters
Consumers’ inflation expectations are heterogeneous across birth cohorts and history-dependent: using panel data from the Survey of Consumer Expectations (SCE), the paper documents that each cohort’s inflation forecast is anchored to its cumulative inflation history, with the degree of anchoring estimated structurally. The authors model this via an experience-based Kalman filter in which each agent’s forecast combines a common Kalman-filtered signal (derived from food prices) with a cohort-specific reference term built from the cohort’s entire prior sequence of expected inflation. The estimated history-weight parameter θ is negative, confirming that agents positively weight their inflation history rather than overreacting to current news — a pattern that holds not only in US SCE and Michigan Survey of Consumers data but also across six European countries in the ECB Consumer Expectations Survey. Embedded in a Blanchard–Yaari perpetual-youth OLG New Keynesian model — where households hold experience-based expectations but firms set prices under rational Calvo frictions — the mechanism produces qualitatively different aggregate dynamics from full-information rational expectations (FIRE): after inflationary shocks, expectations initially underreact (agents anchor to the low-inflation steady state) and then persist well beyond the shock horizon as high inflation is gradually incorporated into cohort memory, generating hump-shaped expectation dynamics. For monetary policy, the optimal Taylor rule must be more aggressive after cost shocks than under FIRE: an energetic early response prevents the high-inflation episode from entering cohort memories, avoiding a self-reinforcing upward drift in inflation expectations. Applied to the 2021 high-inflation episode, the model predicts that the youngest cohorts — experiencing high inflation for the first time — will exhibit persistently elevated inflation expectations long after the supply shocks that caused the episode have dissipated.
Summary of a forthcoming paper, AI-assisted and human-reviewed. See the linked original for the authoritative claims and full conditions.
What are the four empirical patterns in the survey data, and do they hold outside the US?
Using the New York Fed’s Survey of Consumer Expectations (a monthly panel from 2013), the paper documents four patterns: (i) inflation expectations differ substantially across birth cohorts; (ii) cohort-specific inflation experience is age-clustered; (iii) individual inflation history is positively correlated with individual inflation expectations; (iv) cohorts do not differ in how they update to current information once their own inflation history is controlled for. These patterns hold in the Michigan Survey of Consumers and, with cohort fixed effects, across the ECB Consumer Expectations Survey covering six European countries — suggesting the mechanism is not US-specific.
How does the experience-based Kalman filter work, and what does estimation yield?
Each consumer’s forecast has two components: a standard Kalman filter signal common to all agents (extracted from food price data) and a cohort-specific reference term that is a weighted average of all past expectations formed by that cohort, governed by the parameter θ. Structurally estimated from SCE data using time fixed effects, θ is negative — meaning consumers positively anchor to their inflation history rather than over-extrapolating from current news. In a goodness-of-fit regression, the experience-based Kalman filter predicts observed cohort-level heterogeneity with a slope coefficient of 1.069, dominating lifetime average inflation and lagged inflation as predictors.
What is the general equilibrium model, and how do heterogeneous expectations enter the IS curve?
The model is a Blanchard–Yaari perpetual-youth OLG New Keynesian economy. Each surviving cohort solves a standard Euler equation using the experience-based expectations operator rather than rational expectations, yielding a history-dependent IS curve in which the effective real rate depends on the weighted average of each cohort’s reference inflation. Intermediate goods producers set prices under Calvo frictions with rational expectations, yielding a standard New Keynesian Phillips curve. The central bank follows a Taylor rule. The IS curve’s history-dependence means that past inflationary episodes — absorbed into cohort memory — affect present aggregate demand.
What do the impulse responses show under experience-based versus FIRE expectations?
Under a taste (demand) shock, experience-based expectations generate lower inflation on impact — agents anchor to the low-inflation steady state — but inflation remains elevated for longer as the shock is incorporated into cohort memory. Under a cost (supply) shock, two forces compete: anchoring to the steady state damps initial price pressure, but rational firms can raise prices by more because the IS curve becomes more inelastic; the net effect requires a stronger interest rate response than under FIRE. In both cases, household expectation dynamics are hump-shaped — initial underreaction followed by gradual build-up — consistent with evidence in Angeletos et al. (2021) and Pfajfar and Roberts (2018).
How does the optimal Taylor rule change under experience-based expectations?
After a cost shock the central bank should be more aggressive than under FIRE. The social cost of tolerating a transitory inflationary episode is much higher under experience-based expectations because it permanently shifts cohort memory upward, creating self-reinforcing dynamics in future periods. An aggressive early response prevents the episode from entering cohort references. After a taste shock the optimal response is similarly strong under both FIRE and experience-based expectations, so the memory channel adds little incremental urgency on the demand side.
What does the model predict about the 2021 high-inflation episode?
Feeding the model with actual monthly data through December 2021, average inflation expectations post-2021 are predicted to be both higher and more persistent under experience-based expectations than under FIRE or diagnostic expectations. Young cohorts, who experienced only low inflation in the 2010s, are updating their memory of inflation upward for the first time, creating a cohort-specific anchoring shift. The model implies that the 2021 episode could have long-lasting effects on consumer price expectations even if the supply shocks that caused it are fully transitory.