The Distributive Effects of Risk Prediction in Environmental Compliance: Algorithmic Design, Environmental Justice, and Public Policy
This repository contains the code necessary to reproduce the results, figures, and tables for the article "The Distributive Effects of Risk Prediction in Environmental Compliance: Algorithmic Design, Environmental Justice, and Public Policy" published in the Proceedings of 2021 ACM FAccT Conference in Toronto, Canada (link).
Please cite this article as follows, or use the BibTeX entry below.
Elinor Benami, Reid Whitaker, Vincent La, Hongjin Lin, Brandon R. Anderson, and Daniel E. Ho. 2021. The Distributive Effects of Risk Prediction in Environmental Compliance: Algorithmic Design, Environmental Justice, and Public Policy. In Toronto ’21: ACM FAccT, June 03–05, 2021, Toronto, Canada. ACM, New York, NY, USA, 16 pages. https://doi.org/10.1145/1122445.1122456
@article{Benami2021Distributive,
author = {Elinor Benami, Reid Whitaker, Vincent La, Hongjin Lin, Brandon R. Anderson, and Daniel E. Ho},
doi = {10.1145/1122445.1122456},
journal = {Toronto ’21: ACM FAccT},
month = {6},
title = {{The Distributive Effects of Risk Prediction in Environmental Compliance: Algorithmic Design, Environmental Justice, and Public Policy}},
year = {2021}
}This code was tested on a system with the following specifications:
- operating system: macOS Mojave Version 10.14.6
- processor: 4 GHz Intel Core i7
- memory (RAM): 16 GB 1867 MHz DDR3
The main software requirements are Python 3.7.3, and R version 3.6.1 (2019-07-05) with Rstudio version 1.2.5019 (Elderflower). For scripts in R, it is recommended that the user opens the snc-distributive.Rproj R project file. The complete list of packages for R scripts are listed in src/prediction/0_dependencies.R, which is sourced in the beginning of each R scripts. The list of packages for Python scripts are listed within the scripts and should be installed prior to running the scripts.
The src/prediction folder contains scripts necessary to generate prediction results and distributive impact results. Model outputs are saved in the output/models folder.
- Run ARIMA models:
1_run_arima.R - Upload ARIMA results to database:
2_upload_arima_to_db.R - Run Random Forest models for both classification and regression:
3_run_random_forest.R - Produce distributive impact results:
4_distributive_impact.R
The src/figures and src/tables folders contain scripts necessary to generate figures and tables in the article. Outputs are saved in the output/figures and output/tables folders.
Most scripts in this repo depends on access to RegLab's AWS database. Most EPA data contained in the database, including DMRs, are sourced from EPA's public data download page. We are not able to publish detailed prediction results due to sensitivity of the information. Interested persons should contact Elinor Benami ([email protected]) for data access issues.