Methods: This study uses data from AssetsAfrica, an asset-building program piloted in the Masindi District of Uganda. The data, which was collected through face-to-face surveys, is a two level nested data-set. Level-1 consists of 374 household heads, and level-2 is made up of 57 communities. In the analysis, a multilevel modeling with the Full Maximum Likelihood estimation method is used to model the pattern of household bank savings because it is expected that the level of bank savings does not only vary by household characteristics but also by community level characteristics. The outcome variable is households' accumulated bank savings. One of the two level-2 predictors is proximity of banks, a dichotomous variable that measures whether the bank in the study area is perceived as far or near. The other level-2 predictor is also a dichotomous variable that measures whether there are at least three informal savings mechanisms in a community or not. The level-1 predictors include participation in a financial training session; perceptions about obstacles posed by bank rules, documentation, and procedures; household size, and self-perception about ability to make ends meet.
Results: The Intra-Class Correlation was found to be about 31% indicating that the community level characteristics account for a considerable amount of variability in bank savings. The fixed effects coefficients for household size, perception of livelihood, financial education, and informal savings are all statistically significant. There was no significant association between household's bank savings and the rules/procedures at the banks. Only the random effects coefficient for informal savings mechanism at the community level was statistically significant.
Conclusions and Implications: When youth in rural communities have to decide whether to save in a bank or not, the proximity of banks may not matter as much as availability of informal savings mechanisms. Also, although rural communities often have a lot in common, they are unique in many ways. It cannot always be assumed that these communities are the same. Asset building research and interventions in rural communities ought to be contextualized in order to account for differences between communities.