NEW COD-PS Standards and Process - DRAFT

The Process

COD-PS are based on the ‘best-available data principle’ and as such have no unique and standardized source. The preferred source is recent population estimates and projections produced by National Statistical Offices published after the census. However, these are not always available or do not provide the sex and age disaggregation required by humanitarian organizations. Further, as the date of the last census passes by (and changes to the administrative structure occur), more survey data becomes available to update estimates and projections of recent, postcensal population dynamics. In these cases, the ‘best-available data principle’ mandates to provide sex and age disaggregated projections that reflect the more up-to-date data environment. In this section, we describe the application of the “best available data” humanitarian standard to the construction of COD-PS datasets – noting the diverse array of population data and humanitarian landscapes that need to be considered.

Best Available Data Standard to humanitarian preparedness and operational response

There is no “one size fits all” approach when applying the Best Available Data Standard to humanitarian preparedness and operational response countries. Regional and country context matters, as well as, the population data landscape and humanitarian context which shapes the approach to constructing a COD-PS. Specifically, the following are key factors in determining how to apply the best available data standard to a given population data landscape:

1. The census round in which the last population and housing census in the country was undertaken;
2. The availability and usability of published sex- and age-disaggregated subnational population projections for the current year;
3. The availability of population census data and recent data on population dynamics and their usability to construct a new set of sex- and age-disaggregated subnational population projections up to and including the current year; and 
4. The necessity and feasibility of constructing a new set of subnational population projections.

When Government Data is available

When available, the most logical candidate dataset for a COD-PS is the latest set of subnational population projections compiled by the National Statistics Office or related competent authority. Such projections estimate the population at a baseline year (usually at the time of the last census) and project the population forward to the current year based on information about recent population dynamics (namely fertility, mortality, and migration patterns).  

When Government Data is available but outdated

When recent subnational population projections are not readily available, then to construct a COD-PS it is necessary to acquire the most recent census data and indicators on population dynamics to construct sex- and age-disaggregated population projections at either ADM-1 or ADM-2. In cases where recent population and demographic data can be utilized, the Bayesian population projection framework can be used - this will ensure consistency in the methodological approach of the UN’s official national population projections (known as the World Population Prospects) and COD-PS datasets constructed at lower administrative units (i.e. ADM-1 and below).

When Government Data is unavailable (or too outdated)

When neither up-to-date subnational population projections (that are sex- and age-disaggregated) nor source data needed to compile such projections are available, then either a formal data request to the National Statistics Office is needed, or the use of model-based estimates needs to be considered (by either using the available WorldPop gridded population estimates or undertaking a hybrid census modeling exercise). If modeling is needed,  UNFPA follows a set of principles and methods of demography to estimate and project the population by age and sex.

BayesPop framework: Given the data availability challenges and data quality issues in many humanitarian contexts, UNFPA preferred tool is the BayesPop framework of Bayesian population, cohort-component projection to construct COD-PS datasets from the best available population and demographic data. The BayesPop framework is used to produce the World Population Prospects (WPP), United Nations bi-annual population estimates and projections. All code libraries are publicly available for download and can be run in R which in turn is open source.  The use of these tools facilitates reproduction and transparency of UNFPAs estimates and projections, as well as, future updates as new population or demographic data become available later.  The BayesPop framework facilitates the flexible projection of population data and usage of a wide array of data on population dynamics (e.g. fertility, mortality, and migration). It also provides a probabilistic framework for combining multiple disparate types of population data, of varying quality, and communicating the uncertainty associated with resulting projections in the form of credibility intervals. 

Standards

The minimum standards identified in the current evaluation checklist enable data interoperability and harmonization. 

COD-PS Required Characteristics (minimum requirements in red)