Hybrid census to generate spatially-disaggregated population estimates

Housing census data and up-to-date population data at subnational levels are some of the basic data needs any country should possess to ensure efficient planning and delivery of services, effective development programming, election preparation and timely response to disasters. In countries like Afghanistan, where we face challenges on different fronts, including the collection of timely data, access to recent and reliable local-level demographic data is challenging. National population data is outdated and excludes key groups. Similarly, high rates of migration and urban growth make existing population data quickly outdated.

The last national population census in Afghanistan was conducted in 1979, and this census covered only 67% of the districts in the country due to security issues. Recent plans to conduct a full population census have also been hampered by ongoing insecurity and it is unlikely a population census will be conducted in the near future. Current national and subnational population estimates are based largely on projected population counts using a 2.03% growth rate from a 1979 baseline population, updated with information from pre-census household listing activities, where available. Current estimates using these approaches put Afghanistan's population at 31.5 million in 2018. However, analyses show that under scenarios of growth rates differing by only 1 percentage point, as recent data suggest is possible, the national population could be as low as 19.1 million or as high as 40 million.

Given the urgent need for reliable and accurate national population data, the government of the Islamic Republic of Afghanistan (GoIRA) requested assistance from the United Nations to generate quick population counts and structures that would feed into the planning and programming system of the government. Several organizations, including UNFPA; WorldPop at the University of Southampton, UK; Flowminder; Bill and Melinda Gates Foundation and National Statistics and Information Authority of Afghanistan (NSIA), formerly known as Central Statistics Organization (CSO), committed to working together and developed an innovative methodology to generate geospatially-disaggregated national and subnational population estimates -- a hybrid census. This joint effort facilitated the updating of current population distribution data through the use of updated satellite imagery and/or remote sensing and survey data, in lieu of a census.

The goal of a hybrid census is to produce population estimates for small areas of uniform, detailed grids in the absence of a traditional national census. Hybrid censuses rely on complete counts of population within small, defined areas, through ‘micro-census surveys', selected across an area of interest, and collected relatively rapidly and at a fraction of the cost of a full national census. Statistical models are then used to link these micro-census data to spatial data with full coverage over the regions of interest to predict population numbers in the unsampled locations. By aggregating these high-resolution predictions, population totals can also be produced for administrative units or for the national level, if required.

The ‘hybrid census project' to generate geospatially-disaggregated population estimates was planned 2015 and its pilot was initiated in March 2016. The initial pilot was conducted in three selected provinces using Socio-Demographic and Economic Survey (SDES) data, recent satellite imagery and a geospatial dataset. In October 2016, upon successful completion of the pilot phase, a statistical model was developed to produce population estimates for all 34 provinces using 9 provinces' SDES data and 19 provinces' micro-census survey data. The final population estimates were released in June/July 2017. By using the existing data, high-resolution satellite imagery and computer-vision based approaches, the statistical model was able to predict numbers in areas where no population data were previously available, and it was possible to generate high resolution maps of population estimates disaggregated by age and sex, together with uncertainty metrics.

(left) High resolution satellite imagery of a rural area of Afghanistan; (right) Settlement areas detected automatically by computer algorithms.

We understand the hybrid census approach can never replace the rich production of data on the individual, family, household or community that is generated by a traditional population and housing census. However, at present, we are unable to fully execute a traditional census at the national level due to insecurity and high costs. The hybrid census approach can be used to produce population estimates for small areas or uniform, detailed grids in the absence of traditional census data.

The data will be used to support decision-making and the measurement of SDG indicators. Of the 232 SDG indicators, up to 98 require population data for their calculation. The collaboration has also contributed to further strengthening capacities of the NSIA and other related government agencies to generate population estimates based on the integration of geospatial data and ground surveys.

In order to meet the future need of maintaining reliable and up-to-date national population data using hybrid census, NSIA aims to strengthen its institutional capacity and work with partners. NSIA plans an Afghan-owned and led effort with technical assistance from partners to conduct the next hybrid census and advocate for the public acceptance of hybrid generated population census data.

We appreciate the efforts of UNFPA, WorldPop at the University of Southampton, Flowminder, Bill and Melinda Gates Foundation in providing assistance to conduct the project, and we look forward to long-term technical support and partnerships to assist National Statistics and Information Authority (NSIA) in adopting this innovative methodology institutionally.

The Afghanistan National Statistics and Information Authority looks forward to presenting our work at the upcoming United Nations World Data Forum in the session entitled "SDG Data Challenges and Opportunities: Geospatial Reference, Infrastructure and Demographic Data for Development". We look forward to sharing our experiences and discussing how hybrid censuses can support other countries in enhancing their developments, while also learning from others at the Forum new and innovative strategies to boost national data availability.

Mr.Bahadur Hellali is the Deputy Director General, Information Systems Management and Development at the Afghanistan National Statistics and Information Authority (NSIA).