Replication package for Catastrophic in Level, Progressive in Rate: The Distributional Incidence of Nigeria's 2023 Petrol Subsidy Removal
Description
This replication package reproduces every quantitative result in the paper - Catastrophic in Level, Progressive in Rate: The Distributional Incidence of Nigeria's 2023 Petrol-Subsidy Removal and the Failure of Compensation. It contains a self contained Python script (2023_Subsidy.py) and two household level analysis extracts derived from the Nigeria General Household Survey-Panel (GHS-Panel), Wave 4 (2018-2019, pre-reform baseline; N = 4,976) and Wave 5 (2023-2024, post-reform, excluding Borno State; N = 4,637), World Bank LSMS-ISA. All statistics are survey weighted, with the official 2018-2019 national poverty line (₦137,430 per person per year) and a survey based temporal food deflator of 2.469. Running the script regenerates the paper's core tables, FGT poverty measures (headcount, gap, severity) by wave (Table 4); the inequality suite = Gini, Theil-T (GE1), mean log deviation (GE0), and Atkinson indices at ε = 1 and 2 (Table 5); the progressivity of the petrol burden through the concentration index, the Kakwani index (C − Gini), and the Suits (1977) index (Table 6); and the weighted targeting logit for the probability of receiving compensating assistance (Table 8). Point estimates are deterministic; standard errors and confidence intervals come from a household bootstrap (SEED = 20240115, NBOOT = 1000), and the reference console output is included as Expected Output.txt. The raw GHS-Panel microdata are public use and available from the World Bank Microdata Library (https://microdata.worldbank.org) under the World Bank-NBS access terms; they are not redistributed here. The one identifier not included in the public extract is the household urban-rural and geopolitical zone variable used in the full Table 8 specification, which is available from the World Bank under its restricted geovariable terms; the released covariates reproduce the paper's targeting result. Code is released under the MIT License and the derived extracts under CC BY 4.0 (see LICENSE.txt).