<p>The fragmentation of local government units has often been used as a device to enhance government’s territorial reach in many unitary developing countries. Yet, it is still unclear what its developmental effects are and, in particular, if it relieves poverty. We contribute to this debate by examining whether an increase in the number of local units has contributed to Bangladesh’s remarkable progress on poverty reduction. Using a new panel dataset on the socio-economic conditions of Bangladesh’s districts covering the period from 2001 to 2010, we find that spatial fragmentation (the increase in the number of units per land area) has reduced both moderate poverty and, to a lesser extent, extreme poverty. However, the effect is nonlinear, as the magnitude of the marginal effect on poverty is smaller in highly fragmented districts than in relatively consolidated ones. We speculate that its effect may be related to three channels: via greater outreach of health and education programmes; via improving infrastructures; by enhancing the ability of social safety net programmes to reach the poor. Finally, not all types of fragmentation reduce poverty. When using indicators of political fragmentation focussing on the number of units per capita or density, we find no significant impact. The results are robust to introducing a number of controls, to using a range of panel methods accounting for unobserved heterogeneity, and to using Instrumental Variables estimation.</p>