The master file for preparation of both the pairwise randomized treatment assignment, and for construction of the synthetic control, is 0_master.R. All output (tables and figures) produced by these scripts is available in the Data subfolder.
The master file first calls 1a_matching_data_prep.R to prepare the data of the pilot participants for matching. Data are prepared based on Leistungsbezug_TeilnehmerInnen.dsv (for benefit_receipt as a proxy of prior income), Arbeitsmarktstatus_mon_uni_status_int_TeilnehmerInnen.dsv (for construction of total days unemployed in the last 10 years), and Arbeitslose_TeilnehmerInnen.dsv (for a series of further demographics).
In 1b_pairwise_matching.R, the package nbpMatching is then used for pairwise matching and random assignment of treatment within pairs. This script exports TeilnehmerInnen_Welle_1.csv and TeilnehmerInnen_Welle_2.csv to the Data/ subfolder, which are used by the implementing partner organization for assigning participants to waves.
Lastly, 1c_matching_quality_checks.R is invoked to run a series of descriptive checks and produces some visual summaries of the quality of matches which are also stored in the Data subfolder.
In the second step, the master file calls 2a_synth_data_prep.R to prepare the data for the construction of the municipality-level synthetic control. The script prepares (1) the cross-section data on the unemployed, (2) the longitudinal data on the unemployed, (3) the cross-section data on the working age population, (4) the longitudinal data on the working age population, (5) the mean wage level, and (6) the communal tax.
(1) and (2) are from the AMS internal registry (Arbeitsmarktdatenbank), (3), (4) and (5) are from the social security registry (via AMS BMAFJ Erwerbskarrierenmonitoring or AMDB), and (6) is from the national statistical agency (STATcube - Statistische Datenbank of Statistik Austria). In step (7), the script merges all variables constructed and exports "municipalities_merged.csv" to the "Data/" subfolder, which is used in the next step to construct the synthetic control.
In 2b_synthetic_controls.R, a subset of 26 municipalities is selected which are closest to Gramatneusiedl in terms of Mahalonobis distance. The package Synth is then used to construct the synthetic control based on these 26 municipalites. The package furrr is used for parallelizing permutation inference.
Lastly, 2c_synth_figures.R prepares tables and visual summaries for evaluation of the synthetic control. All output is stored in the Data subfolder.