Negative income shocks can either be the consequence of risky choices or random events. A growing literature analyzes the role of responsibility for neediness for informal financial support of individuals facing negative income shocks based on randomized experiments. In this paper, we show that studying this question involves a number of challenges that existing studies either have not been aware of, or have been unable to address satisfactorily. We show that the average effect of free choice of risk on sharing, i.e. the comparison of mean sharing across randomized treatments, is not informative about the behavioural effects and that it is not possible to ensure by the experimental design that the average treatment effect equals the behavioural effect. Instead, isolating the behavioural effect requires conditioning on risk exposure. We show that a design that measures subjects preferred level of risk in all treatments allows isolating this effect without additional assumptions. Another advantage of our design is that it allows disentangling changes in giving behaviour due to attributions of responsibility for neediness from other explanations. We implement our design in a lab experiment we conducted with slum dwellers in Nairobi that measures subjects' transfers to a worse-off partner both in a setting where participants could either deliberately choose or were randomly assigned to a safe or a risky project. We find that free choice matters for giving and that the effects depend on donors' risk preferences but that attributions of responsibility play a negligible role in this context.