ABSTRACT
This study employs unique household data collected in cyclone-affected communities in Bangladesh to investigate whether religious fractionalisation is associated with crime victimisation after disasters. The identification strategy relies on two characteristics of the study area: 1) its religious composition is stable; and 2) its households’ pre-disaster socio-economic status is uncorrelated with religious fractionalisation and disaster damage after controlling for the observed characteristics. The findings suggest that households in disaster-affected and religiously fractionalised communities are more likely to be victims after a natural disaster than are households in non-fractionalised communities. This study also finds empirical support for the idea that the result is driven by the misallocation of disaster relief in fractionalised communities.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1. In this paper, a community is considered to be religiously fractionalised when it consists of multiple religious groups.
2. Studies on risk sharing show that the arrangement is likely to be inefficient if the potential sanction against deviation from the arrangement is lower (Ligon, Thomas, & Worrall, Citation2002) and if individuals are self-interested (Foster & Rosenzweig, Citation2001).
3. EM-DAT, accessed on 13 May 2017; http://www.emdat.be/disaster_list/index.html.
4. A union is an administrative unit in Bangladesh. Each union contains multiple villages.
5. Montalvo and Reynal-Querol (Citation2005) suggest the use of polarisation rather than fractionalisation indices. However, since the samples include only four religious groups, the polarisation index is strongly correlated with the ELF index. Hence, although the use of the polarisation index does not qualitatively change the empirical result, it is not reported in the paper.
6. The data on risk-sharing network are obtained from the following question: How many households in the village could you call for help if you are in need? General trust is based on the subjective information elicited by the following question: Generally speaking, would you say that (1) most people can be trusted; (2) you can’t be too careful; or (3) no idea. The indicator of general trust takes unity if the answer is (1) or (3), and zero otherwise. Trust in the local government is also elicited by a similar question.
7. The survey households are considered to be connected to the village leader if they self-report that their relationship with the leader is that of a relative, friend, or neighbour. Investment in crime-prevention technology is captured by three indicators: locking the door of the residence (ρ = 0.02, p-value = 0.679) or livestock hut (ρ = −0.07, p-value = 0.307) and watching livestock when feeding them (ρ = 0.033, p-value = 0.635).
8. The results are not reported in the paper but are available from the corresponding author upon request.
9. The statistics are not reported in the paper but are available from the corresponding author upon request.