Rust Bus Engine Example

Important Links

Read this page as a runnable CCP smoke test, not as the main evidence page. It shows the wrapper on a familiar problem after the estimator assumptions have already been checked elsewhere.

The Rust bus-engine replacement problem is the canonical dynamic discrete choice example. A bus operator observes mileage and chooses whether to keep the current engine or replace it.

By default, load_rust_bus() returns a synthetic dataset with Rust-style states and choices. It does not load the historical Rust sample.

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

df = load_rust_bus()

model = CCP(
    n_states=90,
    discount=0.9999,
    utility=rust_bus_reward_spec(90),
    num_policy_iterations=3,
)
model.fit(df, state="mileage_bin", action="replaced", id="bus_id")

for name in model.params_:
    print(f"{name}: estimate={model.params_[name]:.6f}, se={model.se_[name]:.6f}")
print(f"termination={model.termination_reason_}")
print(f"npl_converged={model.npl_converged_}")
print(f"parameter_residual={model.npl_parameter_residual_:.6e}")
print(f"policy_residual={model.npl_policy_residual_:.6e}")

Result

operating_cost: estimate=0.000995, se=0.000421
replacement_cost: estimate=3.072211, se=0.074237
termination=fixed_k_complete
npl_converged=False
parameter_residual=1.452655e-02
policy_residual=2.148578e-02

The model completed all three requested NPL stages. The parameter and policy residuals did not both meet the fixed-point tolerance within those stages.

Interpretation

The Rust bus reward specification estimates two parameters. The first is the operating cost slope over mileage states. The second is the replacement cost. The fitted policy gives the replacement probability by mileage state.

for state, row in zip(
    [0, 10, 50, 89],
    model.predict_proba([0, 10, 50, 89]),
    strict=True,
):
    print(f"{state}: keep={row[0]:.6f}, replace={row[1]:.6f}")

Result

keep=0.955732, replace=0.044268
keep=0.947600, replace=0.052400
keep=0.912779, replace=0.087221
keep=0.884699, replace=0.115301

Counterfactual Replacement Cost

cf = model.counterfactual(replacement_cost=4.0)
print(f"P(replace | state=50)={cf.policy[50, 1]:.6f}")

Result

P(replace | state=50)=0.054908

Raising replacement cost from the fitted value to 4.0 reduces the replacement probability at state 50 from 0.087221 to 0.054908.

Replication Boundary

This page is a package smoke example, not the full historical Rust replication. The printed robust standard errors are conditional pseudo-likelihood estimates. They hold the fitted transition model fixed. The reported simulation study is also synthetic and supplies the true transition tensor. The replication ledger separates two results. Its CCP/NPL section checks fixed-point equivalence to NFXP on bundled Group 4 data. On the official STORDAT panel, converged NPL reproduces the Rust Table IX estimates and joint full-likelihood standard errors. The JSON receipt records the data checksum, specification, residuals, and numerical comparisons.