FinMark Trust’s (FMT) partner Insight2Impact (i2i) sought to assess if short message service (SMS) survey data can be used to collect financial inclusion data in developing countries. Mathematica developed a predictive model to estimate financial inclusion (access to banking systems) in eight countries in Africa and Asia.
These countries are: Tanzania, Uganda, Nigeria, Kenya, Pakistan, Bangladesh, India, and Indonesia. The focus of the study was to test if we could accurately estimate financial inclusion using data from mobile SMS surveys, which are much less expensive than more traditional face-to-face surveys but also systematically under-represent certain target groups, including the rural poor, women, and the elderly. In order to adjust our estimates to account for this non-representative sampling, we used a technique called multilevel regression with post-stratification (MRP, or “Mr. P”). MRP is widely used across the social sciences to adjust non-representative data, and has shown to provide more efficient estimates than traditional (frequentist) post-stratification in cases with many small post-stratification cells. We found MRP to perform well with the SMS data. We also explored novel approaches of combining data from multiple survey modalities, including computer-assisted telephone interview (CATI) data and face-to-face data, with the larger SMS survey data.