Little is known about perceptions of medical expenditure risks despite their presumed relevance to the demand for health insurance. This is the first study to examine households’ beliefs about their future spending on health care. The study made a unique elicitation of subjective probabilities of medical expenditures from rural Ethiopians participating in a panel survey and offered the opportunity to enrol in a health insurance programme. The vast majority of respondents give logically consistent responses to the subjective probability questions. The data indicate that the cross-sectional variance of realized expenditures, which is often used to proxy risk exposure, greatly overestimate the risk faced by any single household. Consistent with the serial correlation observed in realized expenditures, expectations are positively correlated with past expenses. They are revised upward in response to an increase in realized expenditure and, to some extent, they predict expenditure incurred in the year ahead. Despite containing information on future medical expenditures, there is no evidence that expectations influence the decision to take out health insurance, although plans to insure are positively related to the perceived volatility of expenses.

These results suggest that adverse selection may not threaten the viability of voluntary health insurance. A caveat is that measurement error in the reported probabilities may weaken the test for adverse selection. Notwithstanding this limitation, measurement of household-specific distributions of future medical expenses is feasible and avoids relying on the cross-sectional variance, which provides an upwardly biased estimate of medical expenditure risk.

, , , , ,
, , ,,
International Institute of Social Studies of Erasmus University (ISS)