Verändert der Zugang zu formalen Finanzdienstleistungen soziale Beziehungen und Normen? Experimentelle Evidenz aus dem ländlichen Indien
Zusammenfassung der Projektergebnisse
This project exploited the randomized branch rollout by a large Indian financial institution (henceforth LFI) in rural Tamil Nadu in order to examine how formal financial access alters social networks as well as other-regarding preferences and generalized trust. Each LFI offers a broad range of financial products, including loan, savings, and insurance products. DFG funding partially covered endline surveying activities. These included full social network mappings in a subset of villages (205 in total) as well as incentivized games, which were added to the representative household survey covering over 700 villages. Endline surveys occurred approximately 18 to 24 months after branch openings in treatment areas, a point at which approximately 40% percent of households in treatment areas had taken up at least one financial product. In Paul-Delvaux et al. (2019) we provide rigorous evidence on the effects of LFI access on village-level network structure, risk-sharing capacity and public goods provision. We collect novel network data on the carrying capacity of each network link through hypothetical survey questions on potential insurance and borrowing amounts available from each link. In addition, we collect data on public goods contributions from every network member. The scale of our intervention allows us to capture the general equilibrium effects of formal financial intermediation on network structure. Two years after the randomized rollout, we observe significantly lower informal borrowing among community members and significantly lower levels of social interaction, similar to what has been observed in quasi-experimental settings. We extend this literature by showing that this substitution away from informal financial intermediation is also associated with a large reduction in the insurance capacity of village networks, as measured by the maximum amount that households state they could borrow in times of need from each link. As predicted by theoretical models of insurance provision on networks, the lower quantity of insurance is associated with specific changes in village risk-sharing network structure: average closeness falls, suggesting that individuals are less connected to farther away households, and the network exhibits a higher clustering coefficient and lower node degree, suggesting that households are embedded in more tightly knit groups. Finally, we also see evidence of reduced cooperative capacity in the form of lower levels of public goods provision on networks in banked villages, in terms of both time and money spent on village-level public goods. In light of the network structure results, this suggests that the contagion benefits of closer networks outweigh the punitive benefits of clustering for facilitating public goods provision. In Pande et al. (2019), we draw on data from several incentivized games to examine how increased formal financial access alters other-regarding preferences and generalized trust. Overall, we find little observable changes in prosocial behavior in response to the intervention. In particular, the results from the dictator games show no increase in altruism in treatment relative to control service areas approximately two years after bank branch openings, while we do find some evidence for an increase in recipients' willingness to punish unfairness in the ultimatum game in treatment relative to control service areas. Results from the binary trust game suggest that respondents' willingness to choose trust and cooperate fell in treatment relative to control service areas, though results are not significant at conventional levels. Self-reported survey responses indicate no change in generalized trust, but a strong decrease in trust in community members other than family. These results suggest that increased access to formal financial instruments erodes local trust, consistent with insurance and cooperation outcomes from the previous paper.
