Poor households in developing countries typically rely on the help of family or friends in times of economic hardship. However, these informal exchanges of gifts, loans and labour, which serve de facto as risk pooling and which are an important insurance source of households, generally fail to provide full protection. While literature highlights limited commitment and information asymmetries as the two main reasons for incomplete informal insurance, our study focuses on risk taking and hence the responsibility for experiencing a negative shock as another possible explanation. In a computer-supported laboratory experiment with slum dwellers in Nairobi we investigate whether solidarity towards a needy person depends on whether he had just bad luck or whether he self-inflicted his situation by taking high risk. In a between-subject design with two different treatments, each participant could either choose or was randomly assigned to a safe or a risky asset, whereby the latter involved a one-half probability to end up with a zero payoff. After being randomly matched with another person, subjects could make voluntary transfers to their partners. Comparing transfer behaviour across treatments allows investigating whether solidarity is affected by the possibility of choosing the individual level of risk exposure. Moreover, exploiting the two randomization stages of the experiment (random assignment to treatment and project) we develop a novel econometric approach which is able to identify whether the treatment effect stems solely from an indirect effect via actual project choice or whether there is a direct effect on conditional transfers.