The empirical analysis of precautionary savings based on household level has long suffered from the assumption of no risk-sharing behavior and measurement errors. Using a unique dataset from China Health and Nutrition Survey (CHNS), we provide an explicit estimate of the relative prudence coefficient of a representative consumer in the Chinese community. As the main innovation of this paper, focusing our estimation on the community, rather than individual household level, is motivated by two salient features of Chinese communities: the virtual isolation between the communities and a significant amount of risk-sharing within each community. Averaging consumption growth across individual households in the same community in China’s context can reduce measurement errors. We show that the relative prudence coefficient for a community representative consumer is dramatically greater than that of an individual household, which is consistent with the widely agreed belief about the magnitude of risk-aversion in the life-cycle model.