Mapping Community Development Projects: Spatial Analysis in Myanmar

Background and Purpose

Aid policy has greater potential to combat global poverty when targeting areas of concentrated need. Literature in this line of study examines national development profiles as a determinant of aid volume. However, relatively few studies analyze subnational characteristics, and even in those studies the units of analysis are at the higher administrative level (e.g., state). This study intends to fill this knowledge gap by assessing the degree of pro-poor targeting at the lower administrative level that is consistent with the unit of intervention (e.g., township).

Linking poverty with interventions, this paper focuses on community-led development projects (CLD) in Myanmar. The deepest pockets of poverty in the world are concentrated in conflict-prone, fragile states where CLDs has become an increasingly favored aid modality. Two on-going, multi-million CLD projects in Myanmar, the SMU and NCDDP, offer a rare opportunity to assess the targeting of area-based interventions overall, and by different aid models.

Using a geo-spatial analysis, this paper first explores the question of the extent to which CLD projects take place in proximity to poor people. How much of the variance in the presence of, and the distance to, CLD projects is explained by the wealth of the project location? Then, it compares the targeting of the SMU and the NCDDP projects with their respective village selection criteria: accessibility and poverty.

Methods

To answer these questions, this paper examines the degree of association between the wealth index from the 2016 Demographic and Health Survey (DHS) and the presence/number of CDL projects. Given the inconsistency between the locations of wealth data (441 village clusters) and the locations of CLD project sites (162 villages), a gridded wealth map is created and overlaid onto the geo-referenced project data. The extent to which aid goes to poorer villages is captured by presence and distance analysis. The former estimates the odds of being selected as a project township given its wealth while the latter computes project event occurrence rate, using Kernel density estimation techniques.

Results

The results of presence analysis show that a one-unit increase in the 2016 DHS wealth index is  statistically significantly associated with an estimated 35% (e^0.304=1.35) increase in the odds of being assigned to the treatment township. Similarly, CLD aid flows disproportionately to areas with higher nightlights (composite cloud-free radiance values ß=0.00115), further distance to border/conflict areas, and low population density. Aid “hotspots” are clustered in the central region such as Mandalay. On average, the SMU site is associated with a 0.58 higher wealth index than that of the NCDDP project sites.

Conclusions and Implications

To conclude, CLD project villages are less poor than non-project villages, suggesting that aid is not responsive to measures of regional need. The findings also imply difference in CLD models as designed. The SMU focuses on building “model villages,” and scaling them up, while the NCDDP project targets poverty reduction in poorer regions. The results of this study, however, are sensitive to the spatial interpolation methods, and future studies that provide finer estimations of poverty would be beneficial.