Linear methods overstate fiscal stimulus in HA models
A new working paper by Javier Bianchi and Greg Kaplan finds that linear methods substantially overstate the effects of fiscal stimulus in heterogeneous-agent (HA) models. The study, published in June 2026, highlights that marginal propensities to consume (MPCs) are highly non-linear in wealth.
Non-linear MPCs skew stimulus
A working paper by Javier Bianchi and Greg Kaplan highlights significant non-linearities in marginal propensities to consume (MPCs) within heterogeneous-agent (HA) models.
MPCs fall sharply with wealth, particularly away from borrowing constraints, leading to a concave aggregate consumption response to fiscal transfers.
Larger transfers shift households out of high-MPC regions, dampening the overall consumption effect.
The authors find that linear methods consistently overstate the impact of fiscal stimulus at empirically relevant sizes.
For instance, a transfer of 5 percent of annual household GDP (around $8,300) results in linear methods overstating the consumption response by approximately 42 percent.
Even for a smaller transfer of 0.5 percent of annual GDP ($830), the overstatement is more than 17 percent.
This suggests that local methods, even higher-order approximations, may be unreliable for assessing moderate fiscal interventions where high MPCs are central to the transmission mechanism.
Beyond linear approximations
The research contributes to a growing body of work employing heterogeneous-agent (HA) models to study fiscal and monetary interventions.
These frameworks are crucial for capturing realistic household spending and saving behaviors, particularly matching the distribution of marginal propensities to consume (MPCs).
The failure of Ricardian equivalence, a key feature of HA models, underpins the aggregate and distributional effects of fiscal policies.
While much of the recent literature uses first-order methods like the "sequence space Jacobian," which approximate non-linear equilibria for small disturbances, this paper challenges the definition of 'small.'
It argues that for typical fiscal interventions, the necessary 'smallness' is far less than empirically relevant shock sizes, undermining the accuracy of linear approximations.
Linearity's fiscal blind spot
This paper delivers a critical blow to the widespread reliance on linear approximation methods for fiscal policy analysis in heterogeneous-agent models.
The systematic overestimation of stimulus effects suggests that past policy recommendations derived from such models might have been overly optimistic regarding their real-world impact.
Policymakers and researchers must therefore exercise greater caution, recognizing that the 'small shock' assumption often fails to hold for interventions of practical significance.