Refugees are, for the most part, excluded from global poverty statistics, due to the lack of micro survey data on displaced populations. This paper presents the first application of recent advances in cross-survey imputations to estimate poverty among Syrian refugees living in Jordan. The authors exploit data on about 40,000 households captured in both:
- Administrative data from UNHCR’s profile Global Registration System (proGres)—while proGres does not capture data on income, consumption or expenditure, it contains socio-economic data that are potential predictors of consumption; and
- Survey data from UNHCR’s Jordan Home Visits (HV) Round 2 (November 2013 to September 2014), covering a third of registered refugees in Jordan—HV includes data on income and expenditure, as well as a large set of individual and household socio-economic characteristics.
The authors test the accuracy of the poverty estimates imputed from proGres data, by comparing them with ‘true’ poverty rates produced directly from the HV survey. The main model specification uses only variables available in proGres: household size, and the characteristics of the Principal Applicant (age, gender, educational attainment, occupation group, marital status, religion, governorate/city of origin). The authors demonstrate that:
- Imputation-based poverty estimates are not statistically different from non-predicted consumption-based poverty rates. This result is robust to various validation tests, including alternative poverty lines and disaggregation by case size.
- Imputation-based poverty estimates are found to perform better or have smaller standard errors than other poverty measures based on asset indexes or proxy means testing.
- Imputation models require relatively few predictor variables that are already available in UNHCR’s proGres database.
- Relatively small survey samples may be combined with a census-type registration system to provide cost-effective and updated estimates of poverty. However, the methodology may not apply to other country contexts, sources of data or welfare measures.