Agroecological knowledge sources among smallholder farmers in Busia and Siaya counties, western Kenya

Agroecological knowledge sources among smallholder farmers in Busia and Siaya counties, western Kenya, 2024-public use file

This dataset records where smallholder farmers in western Kenya actually obtain agroecological knowledge and how those sources combine into individual information repertoires. It supports the article Supplements, not entry points: the stratification of agroecological knowledge access among smallholder farmers in western Kenya.

Study design. Structured household interviews were conducted with 386 smallholder farmers in Busia (n = 176) and Siaya (n = 210) counties, western Kenya, between September and October 2024. Three sub-counties were covered in each county and two wards in each sub-county: Nambale, Butula and Teso South in Busia; Bondo, Rarieda and Ugunja in Siaya. Farmers were purposively recruited from county-supported agroecology projects. Interviews were administered on KoboToolbox tablets by trained enumerators. Both counties fall within the Lake Victoria basin and are dominated by small-scale mixed crop–livestock farming.

Contents. One Excel workbook with three sheets: READ_ME (provenance and full disclosure-control log), Data (386 records × 365 variables), and Data_Dictionary (variable type, coverage, and notes for every column). Multiple-response questions appear as a combined text column plus one binary indicator per option.

Variable groups cover sources of agroecological knowledge across eleven channels; topics sought; media used; preferred learning modalities; formal training received; perceived quality, relevance, and comprehensiveness of information; barriers to access; agroecological practices implemented; preferences for a digital advisory tool; farm characteristics, land tenure, cropping and tillage systems, soil and moisture management, livestock management, crop–livestock synergies, residue recycling, and produce marketing; and household demographics.

De-identification. This is a public use file derived from an internal research file that is not published. Direct and indirect identifiers were removed: the field sample code, interview date and duration, GPS coordinates, altitude, and exact land size in four variables. Ward is replaced by an arbitrary cluster identifier so that cluster-robust standard errors remain reproducible; the cluster-to-ward mapping is withheld. Age was collapsed to two bands and farm size to three top-coded bands. Local suppression was then applied so that every record lies in an equivalence class of at least five records on the key of county, sex, age band, education, and farm size band. Suppressed values are marked "Not reported" and are missing by design, not non-response. Free-text fields were screened for names, telephone numbers, and e-mail addresses; none were present. Full details are in the READ_ME sheet.

Reproducibility. Channel prevalence and county comparisons, the repertoire structure and sole-source analysis, the latent class model and the preference–provision analysis all reproduce exactly from this file. Because exact land size is released only in bands and education is suppressed for 8% of records, the regression coefficients reproduce approximately the following: effect sizes are essentially unchanged, but two coefficients reported as marginal in the article fall just below conventional significance here. Analysis code is deposited alongside the data.

Ethics. Approved by the University of Eldoret; research authorization from NACOSTI, Licence No. NACOSTI/P/24/39785; administrative permission from the County Governments of Busia and Siaya. Consent was verbal, an approved procedure for this minimal-risk paperless survey, and was recorded electronically in a mandatory field before any interview data could be collected.

Conditions of use. Users must not attempt to re-identify any respondent or household and must not link this file to other datasets for that purpose. Some residual risk is inherent to open survey microdata with this many variables and is not eliminated by the treatment described above.

Funding. Flemish Interuniversity Council - University Development Cooperation (VLIR-UOS), Project No. KE2022TEA516A101.

Related article. Bisase L, Ariya G, Maina F, Otinga A. Supplements, not entry points: the stratification of agroecological knowledge access among smallholder farmers in western Kenya. Discover Agriculture.