This paper uses a unique household data set collected in Vietnam to empirically test the
necessary conditions for an extended version of the consumption risk-sharing
hypothesis. The test explicitly incorporates self-production and uses natural disasters
such as avian influenza, droughts, and floods to identify the effectiveness of market and
non-market risk-sharing mechanisms. With these additional treatments, full risk sharing
cannot be rejected, which suggests the presence of omitted variable bias in existing
studies that reject full risk sharing. We also find that credit constraints have a significant
impact, although limited commitment is not necessarily serious.