Understanding population estimates in the World Development Indicators

This blog is based on the Population Estimates and Projections for the World Bank: Technical Note and describes the sources of the World Bank’s total population estimates, as well as the criteria used to determine the source.
 

Population in the World Development Indicators

The World Bank publishes annual mid-year total population estimates and projections (SP.POP.TOTL) for over 200 economies from 1960 in the World Development Indicators (WDI) database (estimates only) and the Population Estimates and Projections database (both estimates and projections up to the year 2050) biannually.

The World Bank’s population estimates serve as the denominator for many per capita measures calculated and published by the World Bank, including per capita Gross National Income estimates that are used to determine the lending rates and credit terms for World Bank operations. The population estimates and projections are also used to calculate the World Bank’s poverty estimates, and they are used to produce population weighted averages for regions and income groups.

Source of data and selection criteria

The main sources of the total population estimates are (1) United Nations Department of Economic and Social Affairs, Population Division’s (UNPD) World Population Prospects (UN WPP), and (2) national statistical offices (NSOs) where the World Bank collects data directly from NSOs’ publications/databases, or through Eurostat’s database. A set of criteria was developed to determine whether estimates from NSOs or UN WPP are used. The purpose of developing these criteria is to be guided by data quality when deciding the population source, while being transparent and objective.

NSO estimates (including Eurostat) are used when they all the following criteria are met:

• an economy’s births and deaths registration are estimated to be at least 90% complete, or utilize another source estimated to be reliable by the United Nations Statistics Division (UNSD)’s Population and Vital Statistics Report;
  
  
• statistics of births, deaths, and migration are applied when time adjusting population estimates, or population estimates are sourced annually from a population register as reported in UNSD’s Demographic Yearbook Metadata;
  
  
• a population census was conducted within the past 15 years, or population figures are compiled from administrative registers;
  
  
• the NSO’s population estimates (not projections)¹ are available in annual time-series; 
  
  
• and definitions of population/territory are in line with the World Bank’s definition.

_{(1). Estimates after a census (postcensal estimates) are calculated by using the population data from the census as a base, and then adding births, subtracting deaths, and adding net migration using empirical data. Projections are calculated by using future assumptions of births, deaths, and migration subsequent to the latest year of estimates.}

These criteria are met for roughly one third of all economies in the WDI database.

When national statistical offices produce population estimates, population censuses often play a critical role. A population census is the official counting of a population in the territory of an economy. Most economies conduct a population census at least once every ten years. Typically, population estimates after a census (postcensal estimates) are calculated by using the population data from the census as a base, and then adding births, subtracting deaths, and adding net migration using empirical data.

For economies that have a recent census, one might think that there is little to disagree about in terms of the population of the economies. Yet, virtually all censuses have issues of under- (or over) counts. Ideally, after a census, an economy should conduct a post enumeration survey (PES) to measure the magnitude of under- (or over) counts. Low- and middle-income economies often publish census figures without PES and/or without making such adjustments. Under-count is much more common than over-count because missing people during censuses is much more likely to happen than double-counting people (United Nations 2017). Therefore, such census figures are likely to be under-estimated. The UNPD carries out an analysis that takes into account those issues and seeks to establish past population trends by resolving the inconsistencies (Pelletier 2020, United Nations 2024). This can lead to notable differences between estimates from national statistical offices and the UNPD.

In addition, the UN WPP takes into account other data from the NSOs, such as official estimates, demographic surveys, vital event registries, and population registries. The UN WPP population estimates are constructed using the cohort-component method, which takes into account the three components of population change -- births, deaths and migration -- to project the population forward in time (Leslie, 1945; Preston et al., 2001; United Nations 2024). The estimates for each component are taken from relevant data sources and adjusted for under- (or over-) reporting and under- (or over-) enumeration as needed.

As a result of these adjustments and its methodology, the UN WPP estimates differ from population estimates and census estimates of national statistical offices. We consider the UN WPP estimates to be necessary for reducing possible biases and errors especially for many low- and middle-income economies, and thus relies on the UN WPP as the primary source for population estimates for many low- and middle-income economies.
 

Population Estimates and Projections Technical Note

In the Population Estimates and Projections for the World Bank: Technical Note, we describe the selection criteria more in detail. We also cover more topics such as adjustments made to reference time, missing data, geographical areas, as well as calculations of population by age, by sex, urban/rural, and population projections. Please take a look.