Households, organisations and governments commonly engage in risk sharing. However, residual risk often remains considerable, especially in low-income countries. In response, many policy makers have considered the introduction of insurance. But this raises the question of how demand for insurance depends on the extent of pre-existing risk sharing. We contribute, first, by showing in a simple model that risk sharing is a substitute for indemnity insurance but a complement to index insurance. Second, in an artefactual field experiment with Ethiopian farmers, we are the first to vary the extent of risk sharing exogenously. The predictions from theory are confirmed.