How Indirect Estimation Complements Population Analysis

Counting people sounds simple – until you try to do it accurately across a country of 1.4 billion. Births go unregistered in remote villages, deaths are reported months late, and entire categories of demographic events slip through the cracks of official records. So how do demographers build reliable population statistics when the raw data itself is full of holes? The answer lies in a clever family of techniques called indirect estimation, which use the information that is available to mathematically reconstruct what is missing.

Why direct data so often falls short 🔗

In an ideal world, every birth, death, and migration would be promptly recorded by a national vital registration system. In reality, especially across South Asia, Africa, and Latin America, civil registration systems have historically struggled with under-reporting and delays. India’s Civil Registration System (CRS) has faced longstanding challenges of incomplete coverage and delayed reporting, which is precisely why a parallel demographic surveillance mechanism – the Sample Registration System (SRS), covering roughly 0.5% of the national population through continuous enumeration – was created in the 1960s to fill the gap.

Even with the SRS, gaps remain at the district level, among migrant populations, and for cause-specific deaths. When a demographer wants to know the under-five mortality rate in a small region, or fertility trends from a decade before a census, direct measurement is often impossible. Indirect estimation steps in here. Rather than asking “how many children died last year?”, it asks easier, more reliable questions – like “how many children have you ever borne, and how many are still alive?” – and uses demographic models to extract the underlying mortality or fertility signal.

What indirect estimation actually does 🔗

The core principle is elegant. Indirect estimation seeks to estimate a demographic parameter that is difficult to measure directly from some indicator that can be accurately recorded and is largely determined by the parameter of interest. The effects of confounding variables on that indicator are then mathematically stripped away, leaving an estimate of the quantity researchers actually want.

Think of it like inferring the temperature of a room from how quickly an ice cube melts. You cannot measure the temperature directly with the instrument you have, but you can measure something related, account for known confounders (ice size, humidity), and back-calculate the answer. Demographers do this with population data using model life tables, fertility schedules, and standardised mathematical relationships derived from populations where good data does exist.

The kind of questions indirect methods can answer 🔗

A short list of what these techniques routinely estimate includes infant and child mortality, adult mortality from sibling or orphanhood survival reports, total fertility rates from census data, and even maternal mortality. Each method works by taking a simple, robustly-collected indicator and transforming it through demographic theory into a refined estimate. The questions on a census form remain short and easy to answer truthfully, but the analytical output rivals what an expensive longitudinal survey could produce.

The Brass method: a turning point in demography 🔗

The single most influential indirect technique was developed by British demographer William Brass in the 1960s. Indirect methods pioneered by Brass and Coale in 1968 estimate child mortality from aggregate numbers of children ever born and children still alive, as reported by women classified by age group. This kind of data – called a summary birth history – is cheap to collect, can be included in any census, and does not depend on accurate date reporting.

The intuition behind the Brass method is that the proportion of a woman’s children who have died gives a strong clue about child mortality conditions, but the clue is tangled up with the ages of those children. The children of a 20-year-old mother are mostly very young, so few will have died yet. The children of a 40-year-old mother span infancy through their twenties, so a larger share will have died. Brass’s contribution was to design simple fertility and child mortality models that simulated proportions dead for different fertility patterns, producing conversion factors to adjust observed proportions for the age patterns of childbearing.

How the calculation works in practice 🔗

In a census or survey, women are grouped into five-year age bands – 15-19, 20-24, 25-29, and so on. For each band, two numbers are tabulated: the average number of children ever born (CEB), and the proportion of those children who have died. Multiplier coefficients, derived from model life tables, are then applied to convert these proportions into standard mortality indicators such as q(2) (the probability of dying before age 2), q(3), q(5), and so on.

A particularly useful refinement came later. Feeney’s 1980 insight allowed approximate reference dates to be attached to estimates from different age groups of mothers – the older the mothers, the further in the past the child mortality reflected by the survival of their children. This meant a single census could reconstruct not just current mortality but a rough trend going back 10 to 15 years, all from one set of questions.

The assumptions you cannot ignore 🔗

No method is magic, and the Brass approach rests on assumptions that can break down. The classical version assumes fertility has been roughly constant in the recent past, and that child mortality does not vary systematically by mother’s age. Neither is fully true. Children of young mothers appear to have systematically higher mortality than children born to women after age 25, which can distort estimates from the youngest age group. To address this, demographers developed alternative classifications – by duration of marriage (Sullivan, 1972) or time since first birth (Hill and Figueroa, 2001) – that reduce these biases.

Beyond child mortality: the wider Brass family 🔗

Brass’s logic extended naturally to fertility estimation. The P/F ratio method, also developed by Brass in 1964, compares current and lifetime measures of fertility to correct for misreporting of recent births. The “P” represents average parity (children ever born to women in each age group), while “F” represents cumulative fertility implied by current age-specific birth rates. When current birth reporting is incomplete – a common problem in census data – the P/F ratio for younger women, whose lifetime fertility is reasonably well-recalled, is used to scale up the directly observed fertility schedule.

This same family of techniques has been extended to estimate adult mortality (using questions about whether respondents’ parents or siblings are still alive), migration, and even maternal mortality. The entire toolkit is consolidated in the United Nations’ classic reference, Manual X: Indirect Techniques for Demographic Estimation, published in 1983, which remains a standard handbook in demography programmes worldwide.

Why this matters for countries with weak data systems 🔗

For most of the twentieth century, very little was known about mortality and fertility levels across large parts of Africa, Asia, and Latin America simply because civil registration was rudimentary. Indirect estimation changed that. The Brass method and its developments have been very widely applied and revolutionized knowledge of mortality conditions, particularly in sub-Saharan Africa but also in Latin America, providing the empirical foundation for international development policy, UN population projections, and the early work of UNICEF on child survival.

In the Indian context, indirect methods continue to complement the SRS and CRS. They are used to validate the SRS results, to produce estimates for smaller administrative units where the SRS sample is too thin, and to reconstruct historical trends from older census rounds. Even today, accurate estimations of demographic trends remain vital for planning, and India relies on both direct and indirect systems to count births and deaths. Without these techniques, planning for schools, vaccination drives, maternal health programmes, and pension schemes would be working with far blurrier numbers.

A note on modern refinements 🔗

Demographic science has not stood still. Research has shown that the Institute for Health Metrics and Evaluation (IHME) methods, particularly the cohort-derived variant, outperform the classical Brass method and can produce robust past child mortality trends across a variety of demographic regimes. Yet even these newer methods are direct descendants of Brass’s original insight – they take the same simple census questions, classify women by time since first birth instead of age, and apply more sophisticated statistical models. The data-collection side has stayed reassuringly simple, while the analytical side has grown more powerful.

Applications in population health and social planning 🔗

The practical value of indirect estimation shows up in unexpected places. State health departments use indirect child mortality estimates to identify districts with the highest under-five death rates, then channel resources – auxiliary nurse midwives, immunisation outreach, ICDS strengthening – to those areas. Researchers studying long-term fertility transition rely on P/F ratio reconstructions from successive censuses to test theories about education, contraception, and women’s autonomy. Insurance companies and pension planners use indirect adult mortality estimates to price products for populations where life tables would otherwise be unavailable.

Even global health goals depend on these techniques. Tracking progress towards the Sustainable Development Goals on child and maternal mortality would be impossible in many countries without indirect methods quietly working in the background, converting simple survey responses into the numbers that appear on UN dashboards.

What do you think? If a census question as simple as “how many children have you had, and how many are still living?” can unlock a country’s mortality history, what other neglected data sources might be hiding rich demographic information? And as digital birth and death registration improves, do you think indirect methods will fade into history – or remain essential for the next several decades?

References

1. https://www.dataforindia.com/crs-srs-explainer/
2. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7430426/
3. https://www.encyclopedia.com/social-sciences/encyclopedias-almanacs-transcripts-and-maps/estimation-methods-demographic
4. https://demographicestimation.iussp.org/content/indirect-estimation-child-mortality
5. https://www.sciencedirect.com/topics/computer-science/indirect-estimation
6. https://demographicestimation.iussp.org/content/overview-fertility-estimation-methods-based-pf-ratio
7. https://www.un.org/en/development/desa/population/publications/pdf/manuals/estimate/childmort/intro.pdf
8. https://www.demographic-research.org/articles/volume/34/39/