This repository contains code & data for "The Hidden Effects of Judge Ideology in Federal District Courts" by Ryan Hübert and Ryan Copus. A draft of the paper is available at ryanhubert.com/abstracts/revisiting-ideology/
Ryan Hübert, Assistant Professor, UC Davis, rhubert %at% ucdavis %dot% edu
Ryan Copus, Climenko Fellow, Harvard Law School, rwcopus %at% gmail %dot% com
Data Collection & Preprocessing
Note: the code for the data collection and preprocessing is not available in this repository.
We collected every docket sheet on file with PACER in the U.S. District Courts in Washington, Oregon and California from 1995 to 2016.
We then used an automated process to parse the text of the docket sheets to generate a quantitative dataset.
Finally, we merged in additional data from the FJC's Integrated Database available at https://www.fjc.gov/research/idb.
Files for Current Project
01a.Clean.py: imports and cleans dataset for use in prediction and estimation
01b.Clean.R: imports and cleans token data from docket sheets' first entries (this is separated due to its reliance on
quanteda R package)
02.Identification.py: performs predictions for propensity scores using random forest
03.Train.py: trains stacked ensemble models to generate predicted probabilities for each outcome (predicted potential outcomes)
04.Predictions.py: loads models generated by
03.Train.py and generates predicted probabilities
05.Effects.py: estimates average treatment effects for main analyses as well as effects for bias analysis
06.Visualize.R: generates tables and plots for paper (not available in repository)
We will post anonymized data for replication upon publication of the paper.