Rethinking expectations in heterogeneous agent macroeconomics
Benjamin Moll in this paper presents a critical examination of the rational expectations assumption in heterogeneous agent macroeconomic models. Moll argues that “the assumption of rational expectations about equilibrium prices is unrealistic, unnecessarily complicates computations, and should be replaced.” He revisits early writings on rational expectations to highlight their original purpose and limitations.
Key findings and arguments:
- In heterogeneous agent models, the assumption of rational expectations implies that decision makers forecast equilibrium prices by forecasting entire cross-sectional distributions. This is both unrealistic and computationally challenging.
- The rational expectations assumption hinders rather than helps the “operability” of heterogeneous agent models, contrary to its original purpose in simpler models.
- Existing solution methods for heterogeneous agent models with aggregate risk are either limited in scope (e.g., linearization) or still too complex (e.g., solving the full “Master equation”).
- Alternative approaches should model decision makers as forecasting prices directly rather than indirectly via forecasting distributions.
- Moll proposes three criteria that alternative approaches to modeling expectations should satisfy:
– Simplification of computational solution
– Consistency with empirical evidence
– Endogeneity of beliefs to model reality (addressing the Lucas critique) - Promising directions for alternative approaches include:
– Temporary equilibrium and internal rationality concepts
– Incorporating survey expectations and hypothetical vignettes
– Least-squares learning and stochastic approximation methods
– Reinforcement learning
– Use of heuristics and simplified forecasting models
The Trouble with Rational Expectations in Heterogeneous Agent Models: A Challenge for Macroeconomics
Author: Benjamin Moll
From: LSE
Cognitive skills, overconfidence, and HANK
Pfäuti et al. in this paper develop a heterogeneous agent New Keynesian (HANK) model that incorporates persistent heterogeneity in cognitive skills and overconfidence about those skills. The authors analyze U.S. micro data from the American Life Panel to examine relationships between cognitive skills, overconfidence, and hand-to-mouth (HtM) status.
Departure from rational expectations
The core innovation of this model is its departure from the rational expectations assumption. Specifically:
- Heterogeneous beliefs: The model introduces persistent heterogeneity in how households form expectations about their future productivity and income. This contrasts with rational expectations models where all agents have the same (correct) expectations.
- Overconfidence: A key feature is that some households (38% in the baseline calibration) are persistently overconfident about their future productivity. This means they systematically overestimate their chances of reaching higher productivity states and underestimate the probability of lower states.
- Cognitive skills heterogeneity: The model links overconfidence to lower cognitive skills, creating a persistent relationship between cognitive ability, beliefs, and economic outcomes that is absent in rational expectations frameworks.
Partial retention of rational expectations
Despite these departures, the model retains some elements of rational expectations:
- Rational subset: In the baseline calibration, 62% of households still form rational expectations about their future productivity.
- Aggregate variables: The model does not introduce behavioral biases or information frictions regarding aggregate variables, maintaining rational expectations at the macro level.
- Model structure: Apart from the expectation formation process for some households, the rest of the model structure (e.g., budget constraints, market clearing conditions) follows standard rational expectations HANK models.
Implications for macroeconomic dynamics
The introduction of persistent overconfidence leads to several departures from rational expectations outcomes:
- Higher HtM share: Overconfident households underestimate their need for precautionary savings, leading to a higher share of hand-to-mouth consumers than in rational models.
- Realistic MPCs: The model generates average marginal propensities to consume in line with empirical estimates, which rational expectations HANK models struggle to match.
- Policy effectiveness: The model predicts different effects of fiscal policies compared to rational expectations models, particularly for policies aimed at providing insurance or liquidity. In particular, overconfident households undervalue self-insurance, leading to:
– Lower effectiveness of targeted fiscal transfers
– Reduced crowding out of private savings when public insurance is introduced
– A lower optimal level of government debt
Bad luck or bad decisions? Macroeconomic implications of persistent heterogeneity in cognitive skills and overconfidence
Authors: Oliver Pfäuti, Fabian Seyrich, Jonathan Zinman
From: University of Texas at Austin, Free University of Berlin, Dartmouth College