econ-ark/HARK

What it solves

HARK provides a standardized, open-source toolkit for structural modeling of economic choices. It removes the need for researchers to write custom, idiosyncratic software for every new model, lowering the barrier to entry for economists using heterogeneous-agent models (where microeconomic agents like consumers or firms solve structural problems to derive aggregate outcomes).

How it works

Built in Python and leveraging libraries like NumPy and SciPy, HARK uses object-oriented programming to provide a modular system. It allows users to define agent types (e.g., PerfForesightConsumerType) and solve their optimization problems—often expressed in Bellman form—using a .solve() method. The toolkit is organized into Tools (mathematical functions), Models (economic models and solvers), and Applications (simulations of economic phenomena).

Who it’s for

• Students: Those learning structural modeling and how to move beyond simple pattern recognition in big data.
• Economists: Researchers in macroeconomics and structural microeconomics who need to simulate the equilibrium behavior of populations of heterogeneous agents.
• Computational Economics Developers: Software developers building extensible platforms for partial- and general-equilibrium structural models.

Highlights

• Modular Architecture: Separates mathematical tools from specific economic models and applications.
• Ready-to-use Models: Includes pre-built models like the Perfect Foresight Consumer.
• Extensible: Designed for developers to create their own models and tools.
• Educational Resources: Provides a comprehensive set of example notebooks for a gentle introduction to structural modeling.