Replication Data for: Renewable Capacity, Coal Spillovers, and Grid Reliability: Spatial Durbin and LeSage-Pace Panel Evidence from Indian States Under SDG
Description
This dataset contains the state-year panel and replication scripts for the paper "Renewable Capacity, Coal Spillovers, and Grid Reliability: Spatial Durbin and LeSage-Pace Panel Evidence from Indian States Under SDG 7," submitted to Environment, Development and Sustainability (SDG 7: Affordable and Clean Energy collection). The panel covers 35 Indian states and union territories from 2020 to 2023 (N = 140 state-year observations) and is constructed from sixteen publicly available datasets sourced from the Central Electricity Authority of India (cea.nic.in) and India's open data portal (data.gov.in). All R scripts required to replicate the empirical analyses reported in the paper are included. See README.txt for execution order and package requirements.
Files
Steps to reproduce
Software requirement: R version 4.4 or higher. Required R packages: Install all dependencies before running any script: install.packages(c("plm", "lmtest", "sandwich", "spdep", "splm", "spatialreg", "quantreg", "relaimpo", "MASS", "dplyr", "readr", "lubridate", "tidyr", "stringr")) Data source: The sixteen source datasets used to construct the panel are publicly available from the Central Electricity Authority of India (cea.nic.in) and India's open data portal (data.gov.in). Download and store them in a local folder. Update the folder_path variable at the top of each R script to match your local path before running. Execution order: Step 1 — Run 01_panel_construction.R. This merges the sixteen source datasets into the balanced state-year panel (panel_main_2020_2023.csv). The compiled panel is also provided directly in this deposit; skip this step if using the pre-built panel. Step 2 — Run 02_primary_FE_model.R. This produces the two-way fixed effects estimates under clustered and Driscoll-Kraay standard errors (Table 1), the Pesaran CD cross-sectional dependence test, the Hausman test, and the Mundlak-Chamberlain correlated random effects model (Table 2). Step 3 — Run 03_hansen_threshold.R. This produces the Hansen panel threshold test result and the Goodman-Bacon decomposition reported in Sections 4.3 and 4.5. Step 4 — Run 04_spatial_weight_matrices.R. This constructs the grid-region and geographic contiguity spatial weight matrices (W_region.rds and W_contig.rds) used in the Spatial Durbin Model. Step 5 — Run 05_spatial_durbin_model.R. This produces the Spatial Durbin Model estimates (Table 3), Moran's I pre-tests, and the LeSage-Pace direct, indirect, and total impacts decomposition (Table 4). Step 6 — Run 06_robustness_checks.R. This produces all five robustness checks reported in Table 5: the outage-rate model, the coal consumption subsample, panel quantile regression, LMG relative importance decomposition, and Mundlak-CRE coefficient stability. Expected output: All tables reported in the paper (Tables 1 through 6 and Table A1) are reproducible from these scripts. Running time for the full sequence is approximately 15 to 20 minutes on a standard laptop, with Steps 3 and 5 taking the longest due to bootstrap replications (300 draws each). Operating system: Scripts are platform-independent and have been tested on macOS (aarch64, R 4.4).
Institutions
- Gokhale Institute of Politics and EconomicsMaharashtra, Pune