Intertemporal Substitution
One agent saves more.
Low-Wealth
High MPC; respond to income
Middle-Wealth
Precautionary motive stifles response
High-Wealth
Low MPC; respond to prices
Click on any year in the timeline to explore the key developments in heterogeneous agent modeling that led to today's powerful HANK frameworks.
Optimize
consume & save
Aggregate
wages & prices
Invest & hire
The goal of Econ-ARK is to provide cutting-edge, modular, open-source implementations of each component in this circular flow, from household optimization algorithms to aggregation and simulating general equilibrium solutions.
This dashboard explores fiscal multipliers in a Heterogeneous Agent New Keynesian (HANK) model with Search and Matching frictions. To capture the distributional effects of fiscal policy, the model features households with heterogeneous preferences facing idiosyncratic income risk, unemployment dynamics, and endogenous job creation.
Key insight: UI extensions typically generate the highest multipliers due to targeting unemployed households with high marginal propensities to consume.
Captures distributional effects across wealth and employment status
Endogenous job creation and unemployment dynamics
Compare effectiveness of UI, transfers, and government spending
The notebooks below are ordered in a "bottom up" fashion to show you what HARK can do as quickly as possible. It begins with an introduction to the absolute basics of HARK's core framework for representing populations of heterogeneous agents, the AgentType superclass, including our workhorse example IndShockConsumerType; the second notebook on the path provides details of that model and its implementation in HARK.
Next, we explain how information about the agent's problem (parameters, etc) is structured conditional on which AgentType subclass is used, with a focus on how HARK's flexible time structure allows for both finite and infinite horizon problems. The fourth notebook on this path offers a tutorial on how to generate simulated data after solving an AgentType instance's problem.
The last three notebooks give instruction on the "sequence-space Jacobian method" for computing impulse responses to so-called MIT shocks in heterogeneous agents macroeconomic models. The first of these gives and overview of the method itself, including the computational motivation and the mathematics that the method uses. We then provide a tutorial on HARK's syntax for constructing HA-SSJs for (almost) all AgentType subclasses. Finally, we put it all together and show how those HA-SSJs can be used with the sequence_jacobian toolkit to specify a macroeconomic framework in which the microeconomic agents live, and compute impulse responses to unexpected macroeconomic shocks.
If you are an ambitious (or impatient!) reader, feel free to jump ahead to the final notebook and "start at the end" to get to the most macroeconomically substantive material first. You can then work backward through the notebooks to learn more about what was happening in each layer.
The entry point for the HARK package, providing syntax and basic examples of its fundamental AgentType representation.
Documentation for the workhorse IndShockConsumerType model of consumption-saving under permanent and transitory income risk, which is used by the notebooks to follow.
An explanation for how problem data is structured for AgentType instances, and how to use HARK's flexible timing structure.
Detailed documentation and examples for how simulation works for our AgentType subclasses, including both the legacy system and a new model-file based system.
An overview of the mathematics of the sequence-space Jacobian method.
A tutorial for the syntax and scope of HARK's HA-SSJ constructor tool.
An example application that combines HA-SSJs constructed from a HARK AgentType instance with the sequence_jacobian toolkit to compute impulse response functions to an unexpected macroeconomic shock. Includes a discussion of how the inclusion of heterogeneous agents in a new Keynesian model (HANK) matters substantively for the predictions of a macroeconomic model, relative to a traditional representative agent (RANK) or two-agent framework (TANK).
Walk‑through of computing the stationary equilibrium in an
incomplete‑markets Aiyagari model
(ConsAggShockModel/AiyagariSteadyState.ipynb).
Shows how to iterate on the capital‑labor ratio and verify market
clearing with a heterogeneous household block solved in HARK.
Useful as a micro foundation for RBC–style general‑equilibrium
explorations or as a warm‑up before adding nominal rigidities.
Full reproduction of Carroll, Slacalek, Tokuoka & White (2017)
using HARK (econ-ark/DistributionOfWealth).
Matches the cross‑section of U.S. wealth and delivers the high MPCs
that are essential for evaluating temporary fiscal transfers.
Classic aggregate‑shock benchmark recreated in
econ-ark/KrusellSmith.
Demonstrates law‑of‑motion iteration for the mean of the wealth
distribution and provides a baseline for Elyesian policy experiments
under real rigidities.
Replication of Cocco, Gomes & Maenhout (2005) in
econ-ark/CGMPortfolio.
Useful for stress‑testing calibrations of risk aversion and
portfolio‑rebalancing costs before embedding a life‑cycle block
inside a larger HANK framework.