econirl

Structural dynamic discrete choice and inverse reinforcement learning in Python.

EconIRL helps estimate forward-looking choice models, recover reward functions, and evaluate counterfactual policies from panel data.

Documentation: https://econirl.readthedocs.io/

Install

pip install econirl

Quick Start

from econirl.datasets import load_rust_bus, rust_bus_reward_spec
from econirl import NFXP

df = load_rust_bus()

model = NFXP(n_states=90, discount=0.9999, utility=rust_bus_reward_spec(90))
model.fit(df, state="mileage_bin", action="replaced", id="bus_id")

print(model.params_)

cf = model.counterfactual(replacement_cost=4.0)
print(cf.policy[50, 1])

Example output:

{'operating_cost': 0.001002924937407198, 'replacement_cost': 3.072263682263484}
0.055196266692073837

Public Estimator Guides

The public docs split the estimators into a core roster and the rest. NFXP is the reference within the core, the exact estimator we replicate to Rust (1987) Table IX. See Choosing an Estimator for how the methods relate.

Core: NFXP, CCP, MCE-IRL, Neural MCE-IRL, AIRL, NeuralAIRL, and GLADIUS. GLADIUS is the package's neural estimator, the GLADIUS class is the NeuralGLADIUS implementation.

Other: TD-CCP, AIRL2, and every other implemented estimator outside Core. See the exhaustive Other Estimators list.

Package Surface

The recommended API is sklearn-style:

from econirl import NFXP, CCP, NNES, TDCCP, MCEIRL, RHIP, AIRL, GLADIUS

Additional estimators and lower-level configuration objects are available under econirl.estimation, econirl.estimators, and econirl.contrib for advanced workflows.

Repository Layout

• src/econirl/: package source.
• tests/: unit, integration, and validation-evidence tests.
• docs/: public Read the Docs source.
• validation/: reproducible validation runners and machine-readable results.
• examples/: public examples and notebooks.

Manuscripts, PDFs, local research workspaces, and assistant-specific notes are not tracked in this public package repository.

License

MIT