Limited fiscal foresight
Martin Eichenbaum, Joao Guerreiro, and Jana Obradović conduct an online survey experiment with 6,000 U.S. adults aged 22–65, recruited through Prolific. Most responses were collected between December 2024 and January 2025, using demographic targets to approximate the U.S. population. Respondents were randomly assigned to hypothetical scenarios involving a $1,400 transfer: an individual rebate, a universal transfer to all households, or a universal transfer accompanied by an explicit $1,400 tax increase the following year. They reported how much they would spend, save, or use to repay debt and revised their expectations for taxes, income, inflation, and interest rates at different horizons.
The authors then incorporate the survey estimates into a heterogeneous-agent New Keynesian model with cognitive discounting. This allows households to place insufficient weight on future taxes and other general-equilibrium consequences when deciding how much to spend today. They compare this “inattentive HANK” model with a conventional full-information rational-expectations HANK model.
Findings
- Households planned to spend about 31% of the individual rebate and 33% of the universal transfer, a statistically insignificant difference. They also did not significantly revise their expected future taxes after receiving the universal-transfer scenario. Together, these results indicate that households do not fully connect current deficit-financed transfers with their future tax liabilities, contradicting Ricardian equivalence.
- When respondents were explicitly told that their own taxes would rise by $1,400 the following year, their planned spending declined to approximately 27%, but remained substantially above zero. Thus, information about future taxes affects behavior, but does not restore full Ricardian equivalence.
- In the calibrated model, tax inattention raises the first-year transfer multiplier from 0.24 to 0.30, an increase of 25%. Under an alternative calibration that matches households’ near-zero expected tax response, and therefore assumes greater inattention, the multiplier rises to 0.36, 50% above the full-information benchmark.
Ricardian Non-Equivalence
Authors: Martin Eichenbaum, Joao Guerreiro, Jana Obradović
From: Northwestern University, UCLA
