Studying the Equitability of Policy Interventions Using a Heterogenous Policy Effect Approach: An Application to the Dutch Decentralization of the Social Domain

├── src
## Setup datasets
│   ├── 00a_gen_wmo_demog_data.R
│   ├── 00b_gen_municipality_crosswalk.R
│   ├── 00c_gen_continuous_income_data.R
│   ├── 00_fetch_educ.do
│   ├── 00_fetch_num_income.do
│   ├── 00_functions.R
## Merge data into analysis dataset
│   ├── 01_functions.R
│   ├── 01_merge_data.R
│   ├── 01_visualize_output.R
## Generate basic descriptives
│   ├── 02_basic_descriptives.R
│   ├── 02_functions.R
## Establish basic determinants of social assistance use
│   ├── 03_functions.R
│   ├── 03_model_descriptives.R
## Estimate main models (causal tree, causal forest, ates)
│   ├── 04a_gen_ates.R
│   ├── 04b_grf_heterogeneity.R
│   ├── 04c_ct_heterogeneity.R
│   ├── 04_functions.R
## Analyze main models (causal tree, causal forest)
│   ├── 05a_analyze_grf.R
│   ├── 05a_prep_sets.R
│   ├── 05b_analyze_ct.R
│   ├── 05c_analyze_groups.R
│   ├── 05_functions.R
## Generate output for paper
│   ├── 06a_make_tables.R
│   ├── 06b_make_figures.R
│   ├── 06c_gem_analysis_plots.R
│   ├── 06d_group_plots.r
│   ├── 06_functions.R
## Robustness analyses
│   ├── r_2015_2016_treat.R
│   ├── r_education.R
│   ├── r_functions.R
│   ├── r_pre_policy_trends.R
│   └── styles.R
└── tables

00.* scripts

This set of scripts is used to merge and clean data at the registry level for analysis. It relies on the standard demographic sets based on the stapelingsmonitorYYYY generated yearly by Statistics Netherlands.

These are located in the H:/data/ folder within CBS' secure Remote Access Environment which is a folder that is not reflected in this repository.

00a_gen_wmo_demog_data.R : Merge demographic data with social assistance use data

00b_gen_municipality_crosswalk.R : Account for mergers amongst municipalities

00c_gen_continuous_income_data.R : Change income data to a numeric

00_fetch_educ.do : Get education data from .dta files

00_fetch_num_income.do : Get income data from .dta files

01.* scripts

This script stacks population registry data for the years 2016-2019. It includes income data and ensures only individuals living in a municipality that conistently reported data in 2016 and 2019 are included.

01_merge_data.R : Stacks 2016 and 2019 and includes income and municipality data.

02.* scripts

This script generates a number of descriptive tables, including:

• Overall descriptives of the population in 2016 and 2019 as well as the analysis set.
• Descriptives by social assistance use in 2016 and 2019.
• Number of municipalities submitting data to the registry.
• Use of WMO in 2016 and 2019 by univariate demographic groups
• Differences between use per municipality

02_basic_descriptives.R : Generates five descriptive tables.

03.* scripts

This script generates basic model insights when relating social assistance use to demographics, including:

• LPM on 2016 and 2019 data
• LPM on 2016 data
• LPM on 2016 data including education
• LPM on 2016 data including random intercepts at municipality level
• LPM on 2016 data including random intercepts at municipality level and education

03_model_descriptives.R : Generates 5 model objects and 7 tables describing the model coefficients.

04.* scripts

This script estimates the main models for the results section including ATEs using conventional methods and the Causal Tree and GRF approaches.

04a_gen_ates.R : Script to generate ATEs using covariate balancing, entropy balancing and propensity score matching.

04b_grf_heterogeneity.R : Script to generate causal forests for a sample of the entire data set (including municipality codes) and excluding municipality codes

04c_ct_heterogeneity.R : Script to generate causal trees for the alternate municipalities and the set of consistent municipalities