Recent projections on the impact of climate change argue that eastern and southern Africa will be two regions around the globe that will experience dramatic reductions in maize yields by mid‐century. Absent from these projections is any consideration for farmer adaptation of cropping practices or land reallocation. This research quantifies risk, loss and ambiguity aversion for a sample of smallholder Kenyan farmers using framed field experiments. This behavioral information, directly elicited, is used to condition the selection of maize varieties differentiated by drought tolerance, pest resistance, maturity, and seed price. Overall, the willingness to pay for drought tolerance and other attributes is highly heterogeneous as determined through a Latent Class modeling approach. Failing to account for farmer heterogeneity biases the potential welfare gains from this technology. Secondly, willingness to pay estimates identify segments of farmers that are seed‐price sensitive and this elastic demand may limit technology purchase and the eventual impact of this adaptation strategy without seed market intervention.