Landslides affect millions of people worldwide, but theoretical and empirical studies on the impact of landslides remain scarce, especially in Sub-Saharan Africa. This study proposes and applies a method to estimate the direct impact of landslides on household income and to investigate the presence of specific risk sharing and mitigation strategies towards landslides in a tropical and rural environment. An original cross-sectional household survey is used in combination with geographical data to acquire detailed information on livelihoods and on hazards in the Rwenzori mountains, Uganda. Ordinary least square regressions and probit estimations with village fixed effects are used to estimate the impact of landslides and the presence of mitigation strategies. Geographical information at household level allows to disentangle the direct impact from the indirect effects of landslides. We show that the income of affected households is substantially reduced during the first years after a landslide has occurred. We find that members of recently affected households participate more in wage-employment or in self-employed activities, presumably to address income losses following a landslide. Yet, we see that these jobs do not provide sufficient revenue to compensate for the loss of income from agriculture. Given that landslides cause localized shocks, finding a significant direct impact in our study indicates that no adequate risk sharing mechanisms are in place in the Rwenzori sub-region. These insights are used to derive policy recommendations for alleviating the impact of landslides in the region. By quantifying the direct impact of landslides on household income in an agricultural context in Africa this study draws the attention towards a problem that has been broadly underestimated so far and provides a sound scientific base for disaster risk reduction in the region. Both the methodology and the findings of this research are applicable to other tropical regions with high landslide densities.