Essays on formal insurance and informal solidarity in developing countries

In developing countries, shocks such as illness, death of family members, natural catastrophes, price fluctuations, unemployment and many more constitute a real economic threat. Especially poor and vulnerable households are often forced to sell productive assets, reduce consumption, take children out of school and send them to work, all of which may have long-term negative consequences on human capital accumulation and income. Hence, low-income households face the danger of being trapped in a vicious circle of poverty by economic shocks. This dissertation deals with the potential to insure against such risks in developing countries. I do not consider insurance in the narrow sense. Instead I focus on two aspects: First, I assess informal risk-sharing arrangement in social networks that act as an informal insurance. While many arrangements are based on explicit reciprocity (reward and punishment), I focus on pro-social solidarity transfers that require trust and altruism. Second, I consider the potential of formal insurance products for the poor – so-called microinsurance. In different chapters I explore conditions under which solidarity transfers occur and how they can be affected by the availability of formal insurance. The latter is also the reason for focusing on solidarity instead of reciprocity: While reciprocity based on rational self-interest should be easily restored after a change in economic incentives, the same is probably more difficult if pro-social motivations were crowded out. In the last part of the dissertation I focus on a positive side effect of reduced vulnerability and evaluate the impact of formal health insurance on child labor. The data is collected using different techniques. Standard household surveys are used for items that can be quantified in a relatively objective way, while behavioral experiments are employed to gather pro-social solidarity that is otherwise hard to measure. All causal analyses rely on experiments, i.e. truly random variation. I also use theoretical models in two cases: to further advance identification of effects from the data analysis and to rationalize empirical results. I believe that this mix of methods is well suited for the variety of analytical challenges. The three chapters of the dissertation are described in more detail in the following. In the context of economic uncertainty and when formal insurance is weak, informal risk-sharing mechanisms play a large role. Hence, informal transfers are especially important in developing countries. One important component of such transfers is solidarity within the social network. Chapter 2 (joint work with Björn Vollan) examines two potential drivers of such solidarity transfers: network strength and collective action. While the former relates to trust in a specific relationship, the latter is often associated with more generalized trust and conditional cooperation type in the literature. We build a simple model to illustrate how solidarity might be influenced by specific trust, generalized trust and cooperator type. We then exploit the unique setup of a behavioral game we conducted in the Philippines to test the causal effect of network strength on solidarity and whether collective action relates to solidarity through generalized trust or cooperator type. We find that network strength has a positive effect on solidarity transfers. We also find that – depending on organization type – collective action might relate to both generalized trust and cooperator type. Chapter 3 (joint work with Markus Frölich and Björn Vollan) tests the impact of insurance on solidarity in a behavioral experiment with rural Filipinos when people have the possibility to hide part of their income. We find that offering insurance crowds out solidarity if income shocks are observable such that benefits of insurance are completely offset. Reduced solidarity might persist even if insurance is removed in a later round. However, when hiding income is possible, solidarity is already low without insurance and there is no additional negative effect of insurance. Hence, formal insurance is not effective when the monetary situation can be closely monitored, but it might be an important complement when financial resources cannot be observed. Chapter 4 (joint work with Markus Frölich) assesses another aspect of insurance for the poor. Microinsurance is an actively advocated financial instrument at the moment, and thought to reduce vulnerability of poor households to economic shocks. Many evaluations of microinsurance explore effects on financial protection, access to health care and risk taking in other domains (e.g. agricultural investment). Yet, there is an especially undesirable consequence of shocks that might be mitigated by microinsurance and has received surprisingly little attention until now: child labor. Taking children out of school and sending them to work may have long-term consequences on children’s human capital or health. We are therefore interested in the impact of insurance on school attendance and child labor outcomes. We exploit the recent extension of a health and accident insurance scheme by a Pakistani microfinance institution (MFI) that was set up as a randomized control trial and accompanied by a household panel (baseline plus four follow-up surveys). Together with increased coverage the MFI offered assistance with claim procedures in the treatment area. Difference-in-difference techniques are employed to control for accidental selection across the limited number of randomization clusters. We find evidence for lower incidence of child labor and strong evidence for households to rely less on child labor earnings after the innovation in the treatment branches. Separating the two parts of the innovation package, the effects of claim assistance are mostly insignificant, while increased insurance coverage has large and highly significant effects on child labor outcomes and days missed at school. Consistent with a theoretical model we develop in this paper, the effect is largely due to an ex-ante feeling of protection as opposed to a shock-mitigation effect