We identified the poor by comparing their levels of consumption against the reference measure, which was the poverty line as established by the government authorities. This approach tends to look at poverty essentially from an economic standpoint in a rural context where the concept of poverty is relative and complex to comprehend . This community-based experiment was conducted exclusively in a rural setting, and therefore the results must be considered in this particular context. We cannot say anything about the feasibility and effects of this type of community-based experiment in an urban setting, which remains to be tested. A second limitation is related to our small sample sizes, which do not allow for detailed analyses such as stratified analyses in population subgroups. Fortunately, the different indicators provide convergent results that strengthen the internal validity of the study.
A community-based process that minimizes inclusion biases
In a previous article , we described how this participative process for selecting indigents was considered appropriate and was much appreciated by the social actors. The present study shows that this community-based process is, additionally, potentially effective with respect to its first criterion which is the ability to minimize inclusion biases in the identification of people who should benefit from the exemption. In fact, the results suggest that the COGESs appropriately controlled inclusion biases when they prepared their final list from the names of indigents proposed by the VSCs.
They were able to retain people from households living in extreme poverty and from very significantly vulnerable situations. This bodes well, since many studies have demonstrated the ineffectiveness of administrative processes for selecting indigents in Africa  and in Burkina Faso [16, 17]. Equally numerous are the authors who promote such community-based approaches [8, 10] without necessarily having sufficient evidence. In another context, in Cambodia, three experiences were also found to be effective in reducing inclusion errors. One of these demonstrated that a community-based process was definitely more effective in selecting the extremely poor , which we also observed to some extent in this study.
In this rural experience, it would also be particularly useful to understand better, by means of qualitative studies, the process that led to the selection of a majority of elderly persons. In a social context where intergenerational solidarity is disintegrating , the care of the elderly by means of social solidarity has become a new issue. For example, Senegal decided, in 2007, to eliminate user fees in public healthcare facilities for persons over the age of 60. It could be useful to understand better the social, cultural and religious values underlying the selection of elderly indigents, in order to adapt targeting policies to context.
A method that does not cover all the indigent
The study revealed a major limitation of the experience: it was not able to ensure acceptable coverage of indigents. The VSCs and, even more, the COGESs were very conservative in their indigent selection processes. However, we do not believe their decisions were biased in favour of important or influential people. Indeed, instructions had been given to ensure a certain neutrality in the VSCs' composition and to avoid having influential leaders (dignitaries, religious leaders, or persons with official status) sitting on them. These recommendations were, in fact, respected , although this obviously did not prevent a few rare attempts at influence. Still, these remained the exception and their impact on the committees's choices was marginal. Even though the indigent are predominantly much poorer than the rest of the population, they represent only a very small segment of the population. We expected a larger selection, based on some of the literature on the subject [6, 26] as well as on government directives which indicated that the indigent make up between 10% and 20% of the country's population . Also, the United Nations Development Programme considers that indigence corresponds to extreme poverty , that is, 9% of the population of Burkina Faso . One might think that, in a context of generalized poverty, this restraint reflects an emic conception of poverty among the populations of that region. However, we tend to believe that the restrictive nature of the selection owes more to the decision to use local and endogenous funding to pay for the exemption. Thus, the limited coverage might reflect not so much the emic perspective of indigence, but rather the COGESs' perception of their limited capacity for endogenous funding. Our study of selection processes showed that some COGESs had influenced the choices of the VSCs because no outside resources had been provided to fund the fees exemption. Other African experiences using a similar strategy reached the same conclusions . One such experience in Mauritania was able to retain only 0.67% of the general population , while some in Asia, funded by outside funding agencies, selected more than 20% of the population . In another region of Burkina Faso, the directive given to groups of three people per village to select 20% of the population that they considered the poorest to receive an outside subsidy for the health mutual premium was well respected . Thus, the conflict of interest between identifying the poor and ensuring the financial viability of the services provided is definitely a core factor in explaining the very small coverage of this selection .
Continuation of the experience and the public will
It appears that this community-based process should be retained, because it is efficient and appreciated by the people, but it covers only a very small segment of the population. Thus, the funding approach could be modified in other experiences in order to see how the communities would react to increase the number of people who would be eligible for fees exemption.
To move the COGESs forward in their thinking, one option would be to show them that their cost recovery system generates enough profit to support a certain number of indigents. We have shown that, in the district involved in this study, on average the COGESs had the financial capacity to take on six times more indigents than were retained . This solution would have the advantage of being endogenous and sustainable, but to make it possible, the system's decision-makers would have to demonstrate a much stronger political will. For example, a recent (2007) policy to combat maternal mortality in Burkina Faso instituted a full exemption from delivery user fees, funded by the national budget, for 20% of pregnant women who were considered indigent . Nevertheless, almost none of the health workers are aware of this possibility  and the government has done nothing as of yet to determine the best process to select the indigent, even though it had committed to doing so as far back as 1992 . Beginning in 2010, the Ministry of Health has advised the CSPSs to use 200,000 F CFA (considered as a spending ceiling) per year of their own funds to exempt the indigent from user fees . Hopefully, that planning directive will be an incentive to replicate the process we have described in this article.