Effective Methods for Interpolation and Extrapolation

Census data arrives once every ten years, but planning for schools, hospitals, housing and welfare schemes cannot wait that long. Demographers therefore rely on two essential tools – interpolation to fill gaps between census years, and extrapolation to project numbers forward into the future. Both rest on a small family of mathematical techniques: linear and binomial expansion methods, and arithmetic and geometric growth methods. Understanding how each works, and where each fits, is the foundation of credible population estimation.

Why interpolation and extrapolation matter in population studies 🔗

Interpolation estimates an unknown value that lies between two known data points, while extrapolation estimates a value that lies beyond the available range. In a country where the Census of India is conducted only once a decade, and where the last full enumeration was carried out in 2011, almost every population figure cited in policy documents today is the product of one of these two techniques.

The need is immediate and continuous. Healthcare planners need to know how many children will need vaccination next year. Education departments must anticipate enrolment three years ahead. The Technical Group on Population Projections headed by the Registrar General publishes intercensal and postcensal estimates precisely because waiting for a fresh census is not an option. As one analysis notes, district and block-level estimates in the absence of fresh enumeration are routinely derived using intercensal or postcensal projection methods.

Linear interpolation and extrapolation 🔗

The linear method is the simplest of all techniques. It assumes that the variable being estimated – usually population – changes by a uniform amount over equal intervals of time. If the population in year t₀ is P₀ and in year tₙ is Pₙ, the population at any intermediate year t can be estimated as:

P_t = P₀ + (t/n) × (Pₙ − P₀)

Suppose a town had 1,00,000 residents in 2011 and 1,20,000 in 2021. The estimated population for 2016 would be 1,00,000 + (5/10) × (1,20,000 − 1,00,000) = 1,10,000. The same logic, extended past 2021, gives a linear extrapolation.

Strengths and limits of the linear method 🔗

Linear interpolation works well for short gaps and for populations whose growth is slow and steady – small towns, mature urban centres, or regions where fertility and mortality are stable. It is intuitive, requires only two data points, and is computationally trivial. However, it cannot capture accelerating or decelerating growth, cyclical fluctuations, or sudden shocks like migration waves or pandemics. Used over long horizons, it tends to underestimate growth in expanding populations because real demographic change usually compounds rather than adds.

The binomial expansion method 🔗

When more than two data points are available and they are evenly spaced, the binomial expansion method offers a more refined alternative. It is built on the idea that if a finite difference series is taken far enough, the higher-order differences eventually vanish. For n known values of the dependent variable, the relation (y − 1)ⁿ = 0 is expanded using Newton’s binomial theorem and solved for the missing term.

For four known values, the expansion is:

y₄ − 4y₃ + 6y₂ − 4y₁ + y₀ = 0

For five known values:

y₅ − 5y₄ + 10y₃ − 10y₂ + 5y₁ − y₀ = 0

If one of these y-values is missing, the equation can be rearranged to solve for it. The technique works under three conditions: the values of the independent variable must be in arithmetic progression, the common difference must be uniform, and the value at which estimation is required must align with one of the original x-values.

Where binomial expansion fits in demographic work 🔗

The binomial expansion method is particularly useful when handling decennial census series – say, populations for 1981, 1991, 2001, 2011 and 2021 – and one of those figures is suspect, missing, or needs verification. Because it uses the structure of the entire series rather than just two endpoints, it captures curvature in the data that the linear method misses.

Arithmetic progression method 🔗

The arithmetic method extends the linear idea to growth-rate language. It assumes that population increases by a constant absolute number in every unit of time – typically a year or a decade. The formula is:

P_n = P₀ + n × r̄

where r̄ is the average increment per period. The average is calculated from past census decades. For instance, if a city grew by 10,000 in the 1991-2001 decade and by 12,000 in 2001-2011, the average decadal increment of 11,000 is added to the 2011 figure to estimate the 2021 population.

This method is best suited to small towns, rural settlements, or areas where population growth is linear or slow. It is also the preferred technique when a city has reached its saturation stage and net additions have stabilised.

Geometric progression method 🔗

The geometric method assumes that population grows by a constant percentage rather than a constant number – much like compound interest in finance. Each year’s increase builds on the previous year’s population, so growth accelerates over time. The standard formula is:

P_n = P₀ × (1 + r)ⁿ

where r is the average growth rate per period, expressed as a decimal. The growth rate itself is usually derived from past census data by taking the geometric mean of decade-wise percentage increases.

Consider a city of 50,000 people in 2015 growing to 55,000 by 2020. The annual geometric growth rate works out to (55,000/50,000)^1/5 − 1 ≈ 0.0183, or about 1.83% per year. Projecting forward, the 2025 population would be 55,000 × (1.0183)⁵ ≈ 60,219 – slightly higher than the arithmetic projection because the compounding effect builds up.

When geometric is the right choice 🔗

The geometric method is appropriate for young, rapidly growing cities and for regions in the early or middle phases of demographic transition. Urban centres such as Bengaluru, Hyderabad, Gurugram and Surat – where in-migration and natural increase together produce compounding effects – are best modelled this way. The compound annual growth rate (CAGR) has been used to project intercensal and postcensal populations of India from 2011 to 2030, capturing age-wise and gender-wise distributions.

The catch is that geometric extrapolation can overshoot when applied for too long. Real fertility rates eventually decline, mortality stabilises, and migration changes direction. Pushing a constant 1.8% growth rate three or four decades into the future for a country whose total fertility rate has already fallen below replacement would produce wildly inflated numbers.

Applications in population analysis 🔗

Each method has a clear role in the demographer’s toolkit, and credible work usually involves more than one.

Estimating intercensal populations 🔗

Between two census years, linear or binomial methods are typically used to estimate annual populations for indicators that need yearly denominators – birth rates, death rates, school enrolment ratios, immunisation coverage. The Sample Registration System uses such mid-year estimates as the base for computing vital rates.

Filling data gaps 🔗

When a particular year’s data is missing from a series – for example, if a survey was disrupted by an emergency – binomial expansion can reconstruct the value using the surrounding observations. The same logic is applied to smooth out anomalies caused by undercount or miscount in specific census rounds.

Forecasting future trends 🔗

For projections beyond the last census, the arithmetic and geometric methods serve as the foundational mathematical tools. The Registrar General’s Population Projections for India and States, 2011-2036 applied the mathematical method (essentially extrapolation using past trends) to seven small north-eastern states, while reserving the more sophisticated cohort component method – which separately projects fertility, mortality and migration – for the 21 larger states.

Policy and resource planning 🔗

Different methods produce different scenarios – low, medium and high growth – and planners often present all three to capture uncertainty. Housing schemes, water supply infrastructure, electricity demand forecasts and budget allocations under programmes such as the National Health Mission rely on such projections to anticipate the size of the population they will serve.

Choosing the right method 🔗

No single technique is universally correct. The choice depends on three factors: the growth pattern of the population in question, the length of the projection horizon, and the quality of available data. Linear and arithmetic methods are reliable for short horizons and stable populations. Binomial expansion adds precision when multiple evenly-spaced data points exist. Geometric methods are essential for rapidly growing regions but must be tempered for long-term forecasts. As an evaluation of India’s projection track record shows, even sophisticated official projections can miss state-level counts by several percentage points, which is why demographers usually apply multiple methods and triangulate the results.

What do you think? If you had to estimate the population of your own city or town for 2026 using only the Census 2011 figure and the 2001 figure, which method would you choose – arithmetic or geometric – and why? And how would your answer change if you knew that migration into your city had slowed sharply after 2020?

References

1. https://censusindia.gov.in/census.website/
2. https://nhm.gov.in/New_Updates_2018/Report_Population_Projection_2019.pdf
3. https://telanganatoday.com/opinion-census-2027-at-a-crossroads
4. https://en.wikipedia.org/wiki/Binomial_theorem
5. https://track2training.com/2025/08/07/arithmetic-geometrical-and-decadal-population-growth-methods/
6. https://www.ijhsr.org/IJHSR_Vol.16_Issue.2_February2026/IJHSR-Abstract41.html
7. https://censusindia.gov.in/census.website/data/SRSREP
8. https://ruralindiaonline.org/en/library/resource/population-projections-for-india-and-states-2011-2036/
9. https://www.dataforindia.com/measuring-pop-projections/