Data for PLS-SEM Analysis
Description
The study is guided by the hypothesis that the multidimensional adoption of rice innovations is not driven by digital access alone, but by the extent to which digital capabilities, agricultural information, training, relational capital and productive resources are converted into actionable support for producers. More specifically, the study hypothesises that access to agricultural information, digital capabilities, agricultural training, relational and institutional capital, and productive constraints are significantly associated with the adoption of rice innovations. The data come from the October 2024 baseline survey of the RIZAO programme conducted in Côte d’Ivoire, Senegal and Togo. The analytical sample includes 2,244 rice producers selected from major rice-growing areas across the agroecological formations of the three programme countries. Data were collected using tablets and covered producers’ socioeconomic characteristics, digital access, agricultural training, access to information, institutional contacts, production constraints and adoption of rice-related innovations. Adoption was measured as a multidimensional construct combining the number of rice varieties used, equipment items used and rice-farming activities reported by producers. The empirical analysis used Partial Least Squares Structural Equation Modelling with formative composite constructs and 5,000 bootstrap replications. Complementary robustness checks were conducted on 2,003 observations using robust OLS, ordered logit, Poisson and negative binomial models. The results show that digital equipment is widespread, with 92.9% of producers owning a mobile phone and 73.6% owning a smartphone, while awareness of agricultural applications remains low at 9.2%. Agricultural training and relational capital improve access to agricultural information. Digital capabilities, relational capital and productive constraints directly support adoption. Agricultural information does not have an automatically positive direct effect in the PLS-SEM model, although it becomes positive in the count models. The findings indicate that agricultural information is a conditional resource rather than an automatic adoption driver. Digitalisation should therefore be embedded in advisory systems that combine actionable information, training, producer networks, credit, water, climate services and technical support.
Files
Steps to reproduce
To reproduce the analysis, download the replication package and place the de-identified dataset, named 00_base_originale_chargee.dta, in the working directory specified in the Stata do-file. Open IMA_GJAE_UPDATED_ANALYSIS_PATH_FIXED.do in StataNow 19.5, revise the ROOT and DATA paths if necessary, and run the file from the first line. The program creates the output folders, reconstructs all variables and indices, defines the analytical samples, estimates the main village-clustered models, performs schooling, productive-resource, information-demand, exposure, interaction, nonlinear, and country-specific sensitivity analyses, and exports tables and logs. Verify that the principal engagement model contains approximately 2,003 producers in 269 villages and that the schooling sensitivity contains about 934 observations. The R script 33_IMA_cSEM_SUPPLEMENTARY_ONLY_UPDATED.R is optional and reproduces the exploratory PLS-SEM appendix after its root path is updated. Compare generated coefficients, statistics, sample sizes, and tables with the manuscript and archived outputs. Avoid manual edits after import, because they may prevent replication across software environments and machines.