Health insurance benefit package design by the poor: can we predict their choice?

ABSTRACT Only 3-5% of the Indians are insured under some kind of health insurance. A mechanism that is developed in order to bring the poor in a community health insurance scheme is CHAT (Choosing Healthplans All Togethers). In a CHAT experiment each participant has to select a health insurance package individually and as a group with a budget of Rs. 600 (±US$13). Till now two CHAT experiments have been conducted in India. This research is based on the second CHAT experiment, which was conducted in Maharastra, Karnataka and Rajasthan. The objective of this research is to see whether the decisions made could be explained and predicted. Therefore two analyses have been conducted. First, the individual choice has been analyzed. The results show that the choice for the three major benefits is highly influenced by the state the individual lives in. A probable explanation for this is that the health care services differ between the three states. The choice for the minor benefits is highly influenced by the selection of the major benefits. This result leads to the conclusion that the choice for the minor benefits highly depends on the remaining budget. Therefore three suggestion have been proposed in order to adjust the CHAT experiment in such a manner that this effect could be narrowed down. First, the participants could be asked to prioritize their selected benefits. Second, in the post-questionnaire of the CHAT experiment participants could be asked why a certain benefit has been selected and a third suggestion is that the budget of the participants and groups should not be fixed in advance, but instead the participants and groups are free to decide their own budget. The second analysis that has been conducted, was to see whether the background characteristics of the individuals correlate with a so-called matching-score (which shows the similarity between the individual and the group choice). Unfortunately the results show that only eight variables have a correlation coefficient that is statistically significant and the correlation is extremely weak (between -0.1 and 0.1). This leads to the conclusion that in fact non of the background characteristics could explain why some individuals have a higher matching-score than other individuals. Overall, the conclusion is that the choices could not be explained and predicted accurate. This is probably because the CHAT exercise is not suitable for this kind of analyses. The main goal of the CHAT exercise is to bring the poor into a community based health insurance scheme and not to predict and explain the choices made.