This paper investigates the effects of rainfall variability on agricultural input demand while controlling for risk preference and other covariates. For the empirical analysis, rural household survey data, which was matched with rainfall variability data and experimentally generated measures of risk preference, was used. The results show that increased rainfall variability prompts households to reduce the application of productivity-enhancing inputs, such as fertiliser, but bolsters the application of low-risk inputs such as manure. These results are robust to alternative specifications and support the theoretical predictions developed. The findings suggest the following policy implications for chemical fertiliser use among risk-averse smallholder farmers in areas characterized by rainfall variability. First, developing more weather-resilient crop varieties and irrigation could stimulate higher use of chemical fertiliser by producing more stable yields. Secondly, weather index insurance (WII) could incentivize higher chemical fertiliser use by reducing income risk and easing liquidity constraints. Thirdly, social protection such as cash transfer programmes could lead to a higher use of chemical fertiliser by serving as insurance against income risks (i.e., through providing regular and predictable financial resources).