Data and code for: Digital strategy, ESG integration, and green supply chain efficiency in the Thai agri-food industry
Description
This dataset contains the compiled panel data and the complete Python code that reproduce all results, tables, and figures in the article "Digital Strategy, ESG Integration, and Green Supply Chain Efficiency in the Thai Agri-Food Industry: Evidence from an Undesirable-Output SBM-DEA and System GMM Analysis" (Cleaner Logistics and Supply Chain). The data cover all 51 firms listed in the agriculture and food sectors of the Stock Exchange of Thailand (SET) over 2021–2025, or 255 firm-years. They were compiled from publicly available sources: 56-1 One Report filings, the SET sustainability assessments (the Thailand Sustainability Investment list for 2021–2022 and SET ESG Ratings for 2023–2025), and corporate sustainability reports. Variables include total assets, operating expenses, revenue, net income, Scope 1 and 2 greenhouse gas emissions, disclosed R&D and IT expenditure, export revenue, ESG ratings, leverage, size, and firm age. Monetary values are in million baht and emissions in tCO2e. The code estimates: (1) green supply chain eco-efficiency with a non-oriented, variable-returns-to-scale modified slacks-based measure (Sharp et al., 2007) that treats emissions as an undesirable output, together with an m-out-of-n subsampling bootstrap; (2) radial Malmquist and Malmquist-Luenberger productivity indices; (3) a two-step system GMM model with forward orthogonal deviations, collapsed instruments, the Windmeijer correction, and Hansen, difference-in-Hansen, and Arellano-Bond diagnostics; (4) robustness analyses, including pooled OLS, two-way fixed effects, bootstrap bias-corrected dynamic fixed effects, fractional probit, a Heckman selection correction, and alternative variable definitions. Files: - data/panel_51firms_2021_2025.csv: compiled panel (255 firm-years); a variable dictionary is given in README.md - data/Supplementary_Table_S1_firm_coverage.xlsx: firms and year-by-year data coverage (Supplementary Table S1 of the article) - code/: analysis modules and numbered pipeline scripts (01–08) - run_all.py: runs the full pipeline - README.md: data sources, variable definitions, construction rules, and a map from each script to the tables and figures it reproduces - requirements.txt: Python dependencies
Files
Steps to reproduce
DATA COLLECTION 1. Sample: all 51 firms in the SET agriculture and food sectors listed before 2021, observed over fiscal years 2021-2025 (255 firm-years). 2. Financial data (assets, SG&A expenses, revenue, net income, liabilities, equity; million baht) from 56-1 One Report filings and financial databases, using one source per firm across all years. 3. R&D and disclosed IT capital expenditure from the notes, One Report, or MD&A; export and foreign revenue from geographic segment notes. 4. Scope 1 and 2 GHG emissions (tCO2e) from sustainability reports, for Thai operations and the group; undisclosed years are left missing, not imputed. 5. ESG: THSI membership (2021-2022) and SET ESG Ratings BBB-AAA (4-7) for 2023-2025; unrated = 0. Listing year from SET factsheets. VARIABLE CONSTRUCTION (code/prep.py, code/s2lib.py) 6. DIGI = (R&D + IT capex) / sales x 100; SIZE = ln(assets); LEV = liabilities / equity; EXP = export revenue / sales; AGE = year - listing year. 7. GHG uses the Thai-operations inventory where reported, otherwise the group inventory. 8. THSI firm-years take the firm's first SET ESG Rating in 2023-2025, or 4 if none. REPRODUCING THE RESULTS 9. Keep the folder structure (README.md, run_all.py, requirements.txt, code/, data/). 10. Install Python 3.10+ and run: pip install -r requirements.txt 11. Run: python run_all.py (about 3 minutes). Scripts 01-07 compute efficiency scores, bootstrap, Malmquist indices, system GMM and static models, robustness checks, and figures. 12. Outputs go to outputs/. Seeds are fixed (2026 and 11), so all results are reproduced exactly. README.md maps each script to the tables and figures.