Risks, insurance, and the loss of work due to illness among low-income households

For low-income households in developing countries, formal financial markets usually do not function well due to asymmetric information and lack of enforcement. Informal risk-coping strategies often arise, compensating for market failures. Using a large household survey data set from southern India, the ICRISAT data, this dissertation examines the effectiveness of village-level mutual insurance, and the interactions between consumption, health, and labor supply.; The dissertation starts with a review of the literature on informal risk-coping strategies, and offers some insights on the methodologies and empirical findings in this area. The socioeconomic conditions of the ICRISAT households are discussed afterwards.; The theoretical analyses compare the predictions on consumption and sickness leave under full insurance for three preference specifications—additive CARA, multiplicative CRRA, and nonseparable preferences. It shows that the first two specifications offer very similar predictions, i.e., individual consumption and sickness leave can be fully explained by the village aggregates and a set of individual factors. Sickness leave does not affect consumption and vice versa. Under nonseparability between consumption and health, however, individual consumption and sickness leave interact after controlling for village aggregates and individual factors, which is still compatible with full insurance.; The empirical analyses of this dissertation consist of three chapters. The first chapter addresses the problem of missing observations in the labor data. The results suggest serious selection biases towards low-income households. A selection factor is constructed and brought to empirical estimations later. The second chapter revisits the famous work of Townsend and tests full insurance with more years of data, village-specific prices, and sample selection adjustment. The results strongly reject full insurance and also suggest serious biases in Townsend's regressions. The last chapter tests full insurance and nonseparability structurally. The results reject full insurance and separability between consumption and health on the grounds that socioeconomic factors such as household income, landholding, and demographic characteristics which may restrict the scope and coverage of mutual insurance affect individual consumption and sickness leave. The results also suggest that the ICRISAT households under-invest in health by taking less sickness leave and more consumption, other things equal. Policy implications are discussed afterwards.