How Interpolation and Extrapolation Aid Data Analysis and Forecasting

Population data rarely arrives on a perfectly convenient schedule. A census happens once every ten years, surveys cover only some years, and disasters or administrative delays can leave entire stretches of time without reliable numbers. Yet planners, researchers, and policymakers need population figures for every year, for every district, and often for years that haven’t even happened yet. This is where interpolation and extrapolation step in as two of the most practical tools in demographic analysis, helping us estimate the unknown from what we already know.

Why these techniques matter in population studies 🔗

Interpolation estimates values that fall between two known data points, while extrapolation projects values beyond the known range. In demography, the distinction is more than mathematical. Intercensal estimates are referred to as interpolation and postcensal estimates as extrapolation, and both feed directly into how governments allocate resources, design welfare schemes, and plan infrastructure.

The Office of the Registrar General and Census Commissioner has been preparing these estimates for the Indian government since 1958, originally to support the Third Five Year Plan. With the last Census conducted in 2011 and Census 2027 now scheduled in two phases beginning April 2026, the gap between two official counts will stretch to sixteen years. Every estimate produced during this period – from district population figures to per capita income calculations – relies on interpolation and extrapolation done well.

Estimating intermediate values: filling the inter-census gap 🔗

A decennial census provides a snapshot, not a continuous record. If we know the population of a state in 2001 and 2011 but need a figure for 2007, interpolation gives us a reasoned estimate based on the growth pattern between those two known points. The technique assumes that growth follows a predictable mathematical form – linear, geometric, or exponential – and then locates the missing year along that curve.

How growth rate methods work in practice 🔗

The arithmetic method assumes the population changes by a constant absolute number each year. The geometric method assumes a constant percentage growth, which suits regions where the population multiplies in proportion to its existing size. The exponential method treats change as occurring continuously rather than in discrete jumps, and is often preferred for high-growth developing regions. The Technical Group on Population Projections, headed by the Registrar General, uses the mathematical method for seven north-eastern states – Arunachal Pradesh, Manipur, Meghalaya, Mizoram, Nagaland, Sikkim and Tripura – which studies past and present population data to map trends, while applying the cohort component method elsewhere.

Why intermediate estimates matter for policy 🔗

Government schemes do not pause between censuses. The National Food Security Act, the National Health Mission, school enrolment targets, and even Finance Commission devolution formulas all need annual population figures. If a district’s population in 2015 is undercounted, its share of foodgrains under the public distribution system may fall short of actual need. Interpolation closes this gap with a defensible estimate rather than a guess.

Handling missing data when records break down 🔗

Demographic records are not immune to disruption. Natural disasters, conflict, administrative collapse, or simple data mismanagement can leave gaps in registers that should have been continuous. Interpolation provides a way to reconstruct what is missing without abandoning the dataset entirely.

When disasters disturb the data 🔗

Sudden population shifts due to events like cyclones, earthquakes, or pandemics can render standard estimates unreliable. A study on small-area population estimation in the United States found that absolute percentage errors were especially high for estimates of maximal county population in 2006 (192.4%) and 2007 (91.8%), with discrepancies partly explained by sudden changes in population size due to natural disasters such as Hurricane Katrina. The lesson translates directly to India, where events like the 2004 Indian Ocean tsunami, the Kerala floods of 2018, and the COVID-19 pandemic created sharp deviations from expected demographic trajectories.

Reconstructing incomplete sub-population data 🔗

Sometimes data is missing not for entire years but for specific groups. Caste-disaggregated, tribal, or migrant-worker figures may be partially recorded, or recorded inconsistently across decades. Interpolation allows demographers to estimate these sub-populations by drawing on the trend of the available years and the behaviour of the total population. Care is needed, though – when sub-population estimates are interpolated independently, their sum may not match the recorded total, and adjustment becomes necessary.

Forecasting population trends through extrapolation 🔗

Extrapolation moves beyond the known range to estimate future or past values. For policy formulation, this forward-looking ability is indispensable. Schools need to be built years before children arrive in them, hospitals need to be planned for population sizes a decade ahead, and pension systems must be designed for retirees not yet retired.

India’s official projections to 2036 🔗

The most authoritative example in the Indian context is the Population Projections for India and States 2011-2036 report, prepared by the Technical Group on Population Projections under the National Commission on Population. Using Census 2011 figures as the base, the report projects that India’s population will rise to roughly 152.2 crore by 2036, with the urban share growing from about 31.8 per cent in 2011 to 38.2 per cent. These figures are not predictions in the certain sense – they are extrapolations based on assumptions about fertility, mortality, and migration, and they shape decisions in ministries ranging from health to housing.

The limits of looking forward 🔗

Extrapolation is less reliable than interpolation because the future can break from the past. Analysis of past Indian projections shows that the Registrar General’s 2006 projections underestimated the 2011 count for Tamil Nadu by 6.5 per cent and for Bihar by 6.1 per cent, while Kerala’s population has been systematically and substantially overestimated for several cycles. Errors of this size matter – a 6 per cent miss in a state like Bihar translates to several million people whose needs may be under-planned for.

Different methods, different horizons 🔗

Short-term extrapolation, say one to five years beyond the last census, tends to be reasonably accurate because demographic momentum carries forward. Long-term extrapolation, beyond fifteen or twenty years, becomes increasingly sensitive to assumptions about fertility decline, life expectancy, and migration. This is why the cohort-component method, which projects each age group separately based on its own mortality and fertility behaviour, is preferred over simple growth rate extrapolation for long horizons. Both the UN World Population Prospects and the Registrar General’s projections are based on the cohort component model, in which the components of population change – fertility, mortality, and net migration – are projected separately for each birth cohort or age group.

Ensuring data uniformity for accurate comparisons 🔗

A less appreciated use of interpolation is the standardisation of data intervals. Demographic data comes from many sources – the decennial Census, the Sample Registration System, the National Family Health Survey, the Periodic Labour Force Survey, and various state-level registers. Each operates on its own timetable.

Aligning irregular time series 🔗

If one dataset reports figures every five years and another every ten, direct comparison is unreliable. Interpolation can bring both onto a common annual or quinquennial grid, allowing apples-to-apples comparison. This is especially important when constructing long historical series – researchers studying the relationship between fertility decline and female literacy over forty years, for example, need both variables expressed at the same time points.

Adjusting for delayed or skipped enumerations 🔗

Census timing is not always perfectly synchronous. The 2021 Census of India was postponed and is now scheduled as Census 2027, creating a sixteen-year gap rather than the usual ten. When such delays occur, the assumption of regular intervals built into many growth rate formulas no longer holds neatly. Demographers must either lengthen the interpolation interval and accept greater uncertainty, or supplement with data from administrative sources like the Civil Registration System to anchor the estimate.

Cross-country comparisons 🔗

International agencies like the United Nations Population Division use interpolation to produce comparable estimates across countries that hold censuses in different years. Without this standardisation, a global ranking of, say, the under-five mortality rate would be meaningless because each country’s figure would correspond to a different reference year.

Where these tools fall short 🔗

Both techniques rest on the assumption that the patterns observed in the data will continue. That assumption is reasonable for most variables most of the time, but it fails precisely when failure matters most – during demographic transitions, economic shocks, or unexpected events. The COVID-19 pandemic, for instance, produced excess mortality that no extrapolation based on pre-2020 trends could have anticipated. Similarly, the rapid fertility decline in southern Indian states from the 1980s onwards outpaced projections made in the previous decade. Good demographic practice therefore combines mathematical interpolation and extrapolation with judgement, sensitivity analysis, and willingness to revise estimates as new data arrives.

What do you think? If Census 2027 reveals that current extrapolated projections have significantly overestimated or underestimated India’s population, which schemes or policy decisions made over the past decade do you think would have been most affected? And when official data is missing for your own district, would you trust an interpolated estimate enough to use it for local planning?

References

1. https://irispublishers.com/abba/fulltext/population-projection-and-adjustment-methodologies-for-household-sample-surveys-an-overview-of-methodology.ID.000503.php
2. https://www.pib.gov.in/PressReleasePage.aspx?PRID=2246847&reg=3&lang=1
3. https://nhm.gov.in/New_Updates_2018/Report_Population_Projection_2019.pdf
4. https://ijpds.org/article/view/1160
5. https://www.dataforindia.com/measuring-pop-projections/