blprs

Fast Berry–Levinsohn–Pakes (BLP) estimation for random coefficients logit models, written in Rust. The goal of blprs is to provide a feature-complete and performant alternative to pyBLP, accompanied by rich inline documentation and an ergonomic Rust-first API.

⚠️ This is an early preview. The demand-side estimator is implemented, but advanced functionality (supply, optimal instruments, bootstrapping, simulation aides, and the full suite of pyBLP conveniences) is still under construction.

Highlights

• idiomatic Rust translation of the core BLP demand estimator: product data validation, simulation draws, contraction mapping, and two-step GMM.
• ready-to-read inline documentation and unit tests that mirror pyBLP notebooks.
• modular architecture prepared for supply-side extensions, optimal instruments, and equilibrium pricing models.
• no unsafe code; builds on nalgebra, rayon, and serde.

Getting started

Add the crate once it is published:

cargo add blprs

For now, use a Git dependency while the crate incubates:

[dependencies]
blprs = { git = "https://github.com/eam398/blprs" }

Example

use blprs::{ContractionOptions, Problem, ProblemOptions, WeightingMatrix};
use blprs::data::ProductDataBuilder;
use blprs::integration::SimulationDraws;
use nalgebra::{DMatrix, DVector};

let market_ids = vec!["m1".to_string(), "m1".to_string(), "m2".to_string()];
let shares = DVector::from_vec(vec![0.3, 0.2, 0.4]);
let x1 = DMatrix::from_row_slice(3, 2, &[1.0, 10.0, 1.0, 15.0, 1.0, 12.0]);
let x2 = DMatrix::from_row_slice(3, 1, &[10.0, 15.0, 12.0]);

let products = ProductDataBuilder::new(market_ids, shares)
    .x1(x1)
    .x2(x2)
    .build()
    .expect("validated product data");

let draws = SimulationDraws::standard_normal(200, 1, 1234);
let problem = Problem::builder()
    .products(products)
    .draws(draws)
    .options(ProblemOptions::default())
    .build()
    .unwrap();

let sigma = DMatrix::from_row_slice(1, 1, &[2.0]);
let options = ProblemOptions::default()
    .with_contraction(ContractionOptions { tolerance: 1e-10, ..Default::default() })
    .with_weighting(WeightingMatrix::InverseZTZ);

let result = problem.solve_with_options(&sigma, &options).unwrap();
println!("beta = {:?}", result.beta);
println!("GMM objective = {}", result.gmm_value);

API surface

• data: builders and validation for product-level matrices (X1, X2, instruments).
• integration: Monte Carlo draws with shape checking and seeded reproducibility.
• demand: share prediction and the BLP contraction mapping.
• solving: solver configuration and diagnostics.
• estimation: high-level BlpProblem wrapper that mirrors pyblp.Problem for the demand side.

Roadmap

1. Implement supply-side marginal cost recovery and equilibrium pricing.
2. Add optimal instrument routines and support for demographic heterogeneity.
3. Provide full pyBLP-style convenience APIs (formulations, problems, counterfactuals).
4. Publish crate releases, CI, benchmarks, and docs.rs integrations.

Contributions and feature requests are welcome via pull requests or GitHub discussions.

Development

cargo fmt
cargo clippy
cargo test

License

Licensed under the MIT License. See LICENSE for details.