India’s last census was conducted in 2011, and it is now 2026. Every policy decision, programme allocation and baseline survey that refers to “population” is using data from a country whose shape has since changed.
Most development programmes land at the district. A District Collector sits at a desk, receives a budget, and tries to allocate it across blocks and sub-blocks. The data she has, from the census, NFHS and HMIS, is almost always at the district level. She knows the district average. She does not know which blocks are doing well and which are falling behind without anyone noticing.
The estimation method was built to solve this problem.
What the method does
The method uses small-area estimation. It takes district-level survey data of the kind that already exists in NFHS rounds, HMIS reports and various programme MIS systems, and uses statistical modelling to produce block-level estimates. The estimates state what they are interpolating and what they are guessing, and each comes with a confidence interval.
The technique comes from survey statistics (the Fay-Herriot model and its extensions), adapted to Indian administrative geography. Its key assumption is that blocks within a district share structural characteristics, such as infrastructure, health system staffing and road connectivity, that can be modelled, while still differing enough that the district average hides real variation.
Why this matters now
The 2011 census is fifteen years old. The 2021 census was postponed indefinitely. NFHS-6 is in the field but will not produce block-level estimates. HMIS data is reported by facility and is not representative of the population.
A programme manager in Khunti district (Jharkhand) or Dhubri (Assam) needs to know which blocks to prioritise for a maternal health intervention. The district average does not help her, because it smooths over the blocks where the problem is worst.
The method fills that gap. It does not replace a census and does not claim to. It gives a programme manager a defensible number below the district level, with a clear statement of how far the estimate can be trusted.
The politics of granularity
Choosing what to measure is a political decision. Measuring at the district level instead of the block level decides who gets seen. A district-level indicator can show that things are improving while specific blocks get worse, because the average hides the variation.
The method makes that variation visible, and this is its main contribution. The statistics are standard: multilevel models, area-level random effects, and auxiliary data from administrative sources. The politics is harder. Once you show that Block A is doing twice as badly as Block B, someone has to explain why the budget was split evenly.
Where this sits
The method was developed for a climate and health fund working across four districts in India, and the methodology is documented in a technical paper. The approach generalises. Any country with district-level surveys and no recent census faces the same problem, and the same estimation approach applies.
The field kit that accompanies this work, The Measurement Checklist, asks practitioners twelve questions before they commission a study. Question 11 is: “What do you refuse to count?” The method answers the reverse question, which is what you do when the census has stopped and you still need to count.