Dataset on HYV Adoption, Climate Stress, and Agricultural Income Across 33 Districts of Maharashtra, India, 2014-15 to 2023-24
Description
This dataset is a balanced district-year panel covering 33 districts of Maharashtra, India, over ten agricultural years from 2014-15 to 2023-24 (330 observations, 30 variables), compiled to study the relationship between high-yielding variety (HYV) seed adoption, climate stress, and agricultural income. Maharashtra was selected because its agro-climatic heterogeneity, spanning the humid Konkan coast, the semi-arid Deccan plateau, the drought-prone Marathwada belt, and the cotton-soybean Vidarbha region, produces the climate stress variation necessary for threshold identification, while its persistently high incidence of farmer distress gives the question high policy stakes. The dataset integrates variables from seven public secondary sources: the Directorate of Economics and Statistics (DES) Maharashtra, the India Meteorological Department (IMD), the Indian Council of Agricultural Research (ICAR), the National Bank for Agriculture and Rural Development (NABARD), Maharashtra Agricultural Produce Market Committees (APMC), the Pradhan Mantri Gram Sadak Yojana (PMGSY) road database, and three rounds of the Agriculture Census of India (2010-11, 2015-16, and 2021-22). Variables cover six domains: (1) agricultural income, measured as Gross District Value Added from Agriculture and Allied Sectors at constant 2011-12 prices; (2) seed system quality, proxied by HYV cultivation area in thousand hectares; (3) climate stress, including Maximum Consecutive Dry Spell in days (the primary threshold variable, computed from IMD daily records), standardised rainfall anomaly, and standardised temperature anomaly; (4) input intensification, including pesticide and NPK fertiliser consumption, irrigation coverage, and variable input cost per hectare; (5) institutional access, including KVK training intensity, extension worker density, PACS societies, APMC market yards, and rural road density; and (6) farm structure, including average holding size, small farmer share, cropping intensity, the Herfindahl crop concentration index, and rainfed area share. Key construction decisions: Thane and Palghar districts are merged into a single unit reflecting the pre-2014 administrative boundary; Mumbai City and Mumbai Suburban are excluded due to negligible agricultural activity; Agriculture Census variables are linearly interpolated to annual frequency between census rounds; and MCDS is computed as the longest consecutive run of days with rainfall below 2.5 mm within each agricultural year. The dataset supports the companion research article published in Land Use Policy (Manuscript No. LUP-D-26-02373) and accompanies four R replication scripts covering Hansen Panel Threshold Regression, the Spatial Durbin Model, and Canay panel quantile regression. A full variable dictionary and README with step-by-step replication instructions are included in the deposit.
Files
Steps to reproduce
Step 1: Software requirements R version 4.3 or higher. The following packages are required and can be installed by running the install.packages() command at the top of Complete_R_Code_Submission.R: plm, splm, spdep, quantreg, ggplot2, sf, viridis, dplyr, tidyr, sandwich, lmtest, stargazer Step 2: Download the deposit files Download all eight files from this deposit into a single folder on your local machine: master_panel_maharashtra.csv, Master_Panel_Maharashtra.xlsx, Complete_R_Code_Submission.R, Layer1_Hansen_PTR.R, Layer2_Spatial_Durbin.R, Layer3_Panel_Quantile.R, README.docx, and Variable_Dictionary.docx. Step 3: Set the working directory Open R or RStudio and set the working directory to the folder containing the downloaded files: setwd("your/folder/path/here") Step 4: Run the replication code To reproduce all results from the companion article in one run, source the complete script: source("Complete_R_Code_Submission.R") To reproduce individual estimation layers separately, run the layer-specific scripts in order: Layer1_Hansen_PTR.R reproduces Table 5 and Table 5A (Hansen Panel Threshold Regression results). Layer2_Spatial_Durbin.R reproduces Table 6 (Spatial Durbin Model impact decomposition). Layer3_Panel_Quantile.R reproduces Table 7 and Table 8 (Canay panel quantile regression results and slope equality tests). Step 5: Spatial map (Figure 1) The spatial mapping section downloads Maharashtra district boundaries from GADM (gadm.org) at runtime. An active internet connection is required for this section only. The Thane-Palghar merge and Mumbai exclusion are applied automatically within the mapping code. Step 6: Expected runtime Approximately 8 to 12 minutes on a modern laptop. The Hansen bootstrap procedure (300 replications) and the Canay quantile bootstrap (500 replications) account for most of the runtime. Progress messages are printed to the console at each stage. Step 7: Output All tables are printed to the R console in formatted form and saved as .csv files to the working directory. Figure 1 (six-panel spatial map) is saved as a .png file at 300 dpi. Console output matches the numerical results reported in Tables 4 through 9 of the companion article (Land Use Policy, Manuscript No. LUP-D-26-02373).
Institutions
- Gokhale Institute of Politics and EconomicsMaharashtra, Pune