Economic growth does not reduce unemployment equally across Indonesia: it helps a lot in some districts and barely at all in others. Because growth in one district also affects jobs in its neighbors, job policies work best when they fit local conditions and are coordinated across nearby regions.
Python companion to the R and Stata Double LASSO tutorials — same data, same five estimators, plus a hands-on introduction to the DoubleML library (DoubleMLPLR, DoubleMLIRM, and learner-robustness across LASSO, RandomForest, XGBoost).
Stata companion to the R Double LASSO tutorial — same data, same five estimators, replicating the Belloni-Chernozhukov-Hansen 284-control extension of Donohue and Levitt's abortion-and-crime panel with pdslasso, rlasso, and cvlasso.
Did legal abortion lower crime in the United States, as a famous 2001 study claimed? Instead of hand-picking a few control variables, we let a data-driven method choose among 284 candidates and check whether the original finding holds up. This R tutorial comes with an interactive app that runs in your web browser.
Does mining wealth help or hurt local development, and do good local institutions make the difference? This Python tutorial uses causal machine learning on simulated district data with known true effects to measure how the impact of mining and mineral prices varies from place to place. It comes with an interactive app that runs in your web browser.
Does mining make places richer or poorer, and does the answer depend on the quality of government? Using simulated data that mirrors a published study, we use machine learning in Stata to estimate how the effects of mining and mineral prices on development differ from place to place. The tutorial comes with an interactive app that runs in your web browser.
Does access to a workplace retirement savings plan make American households save more, or do the households that get it simply earn more? This Python tutorial uses double machine learning, which lets flexible prediction models account for income and other differences, on 1991 survey data. It comes with an interactive app that runs in your web browser.
Does access to a workplace retirement savings plan help some households build more wealth than others? Instead of reporting one average effect, we use machine learning in Stata to estimate how the effect differs across households by income and other traits. The tutorial comes with an interactive app that runs in your web browser.
Does a cash bonus help unemployed workers find jobs faster? This Python tutorial uses double machine learning, which lets flexible prediction models strip out the influence of background characteristics, on data from a real experiment in Pennsylvania. It comes with an interactive app that runs in your web browser.
Can satellite images tell us how well each municipality in Bolivia is developing? This beginner-friendly Python tutorial trains a random forest, a machine learning model that averages many decision trees, tests it on places it never saw, and finds that the images hold real but limited information. It comes with an interactive app that runs in your web browser.