Blog@MPIDR

MPIDR on Tour – Counting the Uncounted

© istockphoto.com / OlegAlbinsky

Every other year, the field of demography receives special attention on the 17^th of July, when the United Nations releases their updated population estimates and projections to honor World Population Day. The researchers who generate these data belong to the United Nations (UN) Population Division, which is the mecca for demographic research. I was fortunate to visit them for two months in the spring of 2025. The UN Population Division’s substantial contributions to the development of demographic methods and demographic statistics have historically shaped the field of demography, provided gold-standards for population data, and also educated the wider public about demography. However, this group is operating in an era of rapidly changing times. Population issues have marked the ground zero of polarizing global debate and policy gridlock since at least the 1968 publication of the controversial “Population Bomb” by Paul Ehrlich. While new demographic concerns have emerged since then (think declining rather than expanding fertility), international organizations are under growing financial strain as UN member states appear to be increasingly veering towards isolationist policies. Therefore, in this blogpost, I will present my impressions from working at the United Nations Population Division. My name is Henrik Schubert and I am a research scientist in the Laboratory for Social Demography at the Max Planck Institute for Demographic Research (MPIDR) in Rostock.

United Nations Headquartes in New York City. © iStockphoto.com / GarethLowndes

A Sneak Peek Into Demography Outside Academia

In summer 2024, during the final stretch of my PhD journey, I found myself in an existential crisis, questioning academia generally and my future in academia. To combat this crisis, I looked for demographic positions outside of academia. Around that time, the new update of the World Population Prospects (WPP) — the official United Nations population estimates and projections for 237 countries or areas — was published and reminded me of my love for classic demographic investigations and methods. I discussed my idea of visiting the UN Population Division with my supervisors Mikko Myrskylä and Christian Dudel and received full support to undertake this adventure. I took the initiative to apply for a research stay and landed in NYC few months later.

The United Nations Population Division (UNPD) was founded in 1964 and sits in New York City in the US (although is technically not part of the US) adjacent to the United Nations headquarters. UNPD is part of the Department of Economic and Social Affairs, receives its directives from the UN Social and Economic Council; and pursues three major goals:

• providing statistical reports on demographic pattern and trends at the global level, the primary outputs being the World Population Prospects (WPP) that are published biannually;
   
• assisting intergovernmental processes at the United Nations in the area of population and development, which culminates in the UN Commission on Population and Development (CPD);
   
• building the capacity to produce and analyze population data in UN member states. As an example of this, the supervisor I worked with at UNDP, Patrick Gerland, traveled to Uganda in July 2025 to work with the Ugandan government to present the models behind the World Population Prospects (WWP).
   

I contributed to the first two objectives by working on components of the WPP and observing and providing assistance during the fifty-eight session of the UN Commission on Population and Development (CPD).

Patrick Gerland (left) and Henrik Schubert in a conference room at the UN. © Henrik Schubert

A brief look inside the UN Security Council Chamber. © Henrik Schubert

Sneak peeking into the UN general assembly. © Henrik Schubert

Using Estimations and Data to Ensure Individual’s Rights Globally

During my demographic adventure, I worked with Patrick Gerland, the head of the Population Estimates and Projections Section at UN Population Division, on measuring the completeness of birth registration in vital statistics data between 1950 and 2025 around the globe. Knowing the completeness of birth registration is important for several reasons:

• it indicates the country’s statistical and administrative capacity (Sustainable Development Goal 17),
   
• being registered at birth is an individual’s right and guarantees access to social support measures (Sustainable Development Goal 16), and
   
• it may serve for correcting birth statistics and other demographic indicators, e.g. infant mortality rate. 

Birth registration completeness refers to the share of births in a country that are observed, meaning that the completeness is one if all births are counted. When no birth is being registered the number drops to zero. The challenge is to understand that we compare the number of observed births to the number of true births, which is unknown.

I cracked my head about the question how to estimate the completeness despite not knowing the true number of births. Five major approaches exist.

1. The  aggregate approach suggests to simply assume that a different source, for instance the WPP number of births, which is an estimate rather than a count is true and to enter this number for the “true” value.
    
2. Validation studies, which are considered to be the gold-standard, use a second source of birth records, for example hospital records or school enrollment records, to compare the registration completeness. We then would match these birth records to all the existing individual data we have, and then estimate the overlap and the non-overlap between the two samples we have. This idea can be illustrated by counting two samples of the number of fishes in a pond. First you catch a sample of fish, mark them with color, throw them back in, and re-catch another sample of fish. In this second sample of fish, the number of fish that are marked in the indicate the size of the fish population in the overall pond. If the second sample consists only of marked fish, it is likely that “true” population is close to the population of marked fish, but if you draw only unmarked fish in the second sample, this provides an indication there are so many fish in the sample that the chance of drawing a marked fish twice becomes so small.
    
3. The survey-based approach consists of making use of surveys in which households are asked about whether their children (usually those under the age of five) have been registered and then estimating the share of registered children out of all children.
    
4. With the help of demographic relationships the true number of births is indirectly estimated. Several methods belong to that approach, but one idea is using reverse-survival techniques to school enrollment data, which simply means that the number of births is assumed to be the number of children entering school five years later multiplied by the inverse of the probability to survive to age 5. The idea is simply, in a country with compulsory schooling, every child will enter school at age five except for those who die before, so that the number of school children just needs to be corrected for mortality.
    
5. Country experts assess the completeness of birth registration.

Drawing on these approaches and the existence of data for the necessary information, my supervisor and I created a data base consisting of completeness estimates for all countries, however gaps and inconsistencies persist as different countries have different availability and completeness of data and registration systems.

Methodological Challenges

Having created the data base, my next goal was to understand how to deal with a large set of data with gaps and inconsistencies. Given that the world has more than 187 countries and the UN reports indicators over a range of 75 years, the number of observations becomes high. I needed to develop an automatic procedure that produced reliable results in a variety of settings. Coming up with a one-size-fits-all approach to very different settings is a big challenge. Therefore, I quickly figured out that I needed a procedure that could correct biases in data, extrapolate to settings where data is scarce or even non-existent, and that could provide uncertainty of estimates. I collected validated indicators from a variety of globally validated studies to have a set of variables that might predict completeness in data scarce settings.

Using this data, I trained a machine-learning model to predict the best values for completeness and implemented a secondary spline-regression model to smooth the trend over time. This model allowed me to fill the data gaps and to resolve inconsistencies in the results.

XGBoost (machine learning) prediction of birth registration completeness in South Africa between 1950 and 2024. © MPIDR

Learning From Real-World Applications

The time at the United Nations gave me valuable insights into a different demographic world, which centers around social policy and application. The challenges look different, but they are neither fewer nor simpler. My work with the UN taught me that the approach of the UN Population Division is inherently different to academic research: In academic research you come up with a relevant research question and look for the ideal setting to answer it (the question is flexible, the context is fixed), while the UN has a mandate to report a list of indicators for all places in the world, irrespective of the existence and the quality of the data (the question AND context are fixed, so the approach needs to be flexible).  

My research showed that it is possible to estimate completeness of birth registers using a range of predictor variables in a machine-learning model and to smooth the predictions in a hierarchical Bayesian model. The finding is illustrated in the figure below, which shows the completeness of birth registration across the world, which might be included in a future WPP publication.

During my time working with the UNDP team, I was most impressed by Patrick Gerland’s drive to improve the World Population Prospects during every single biannual iteration, despite working on this project for years. As they say, “what gets measured, matters”, and this department’s commitment to measuring important indicators of human development is clear. I got the impression that he has big plans for the future, so expect some fascinating innovations in the 2027 World Population Prospects update. The research of his team remains cutting-edge despite the growing challenges posed by the financial, technical equipment, shrinking manpower and political support.

Part 2 – Life in New York

Estimate of completeness of birth registration around the world in 2022. © MPIDR

How might it feel to chair a UN session? © Henrik Schubert