Abstract
Public investment in agriculture is widely regarded as a lever for growth, yet its effectiveness depends on how spending is allocated across categories such as research, extension, and infrastructure. This study examines the relative effectiveness of five disaggregated categories of agricultural public expenditure in Burkina Faso over 2006–2023, using an ARDL bounds-testing approach suited to a short annual time series. Only agricultural research expenditure yields a statistically robust long-run relationship with agricultural performance, with an elasticity of 0.339 and a bounds F-statistic of 8.24, exceeding the Pesaran-Shin-Smith upper bound at the 5% level. Agricultural infrastructure, rural infrastructure, extension services and private investment show positive but statistically non-significant associations with agricultural output. CUSUM stability tests and Chow breakpoint tests find no evidence of parameter instability, and Variance Inflation Factors indicate no severe multicollinearity. We recommend prioritising agricultural research as the most statistically defensible policy lever, and offer elasticity-informed guidance for reallocating scarce public resources across the four remaining expenditure categories, while also flagging the small-sample constraints that should shape future panel-data or instrumental-variable extensions of this strand of analysis.