FlowsBook a callSign inCreate free accountSign up
Population and demographics

Gridded population vs census tables: why the two numbers rarely match

A hexagon estimate and a census table can give different counts for the same district. The five usual reasons, how to reconcile them, and which to use when.

Part of the guide Location intelligence automation

By the xMap team · · 6 min read

When you put gridded population vs census figures side by side for the same district, they almost never agree to the person, and that is expected. A census table counts people at one moment and reports them inside official boundaries. A grid takes official counts and spreads them over small cells, so you can add them up inside any shape you like. The difference comes from timing, boundaries, rounding and the way people are placed on the ground, not from one of them being wrong.

This piece is for planners, analysts and expansion teams who have a census table on one screen and a hexagon map on the other and need to explain the gap to a manager. It walks through where each number comes from, the reasons they part ways, a made-up reconciliation, and a way to keep both in one report.

Two ways of answering how many people live here

A census is a count. Every household is asked who usually lives there on a reference night, and the answers are published as tables for areas the statistics office defines: wards, tracts, communes, districts. The table is exact for those areas on that date, then grows older every year until the next count.

A population grid answers a slightly different question: roughly how many residents live in each small cell of the map. Grids of this kind are generally modelled by taking official totals and placing them where people are likely to live, using clues such as buildings and land use. The result is a figure for every cell, including cells that cross a district line, which a census table cannot give.

xMap's population grid uses hexagons of about 0.1 km2, the cells of the H3 system at resolution 9, and covers about 100 countries. The Demographics block adds them up within a circle around a place, inside a city or area you set, or returns every hexagon for a heat map, each with men, women and nine age groups.

Five reasons the counts drift apart

  1. Different dates. A census table describes the reference year. A grid may be built from that same census or from a later estimate. Fast-growing suburbs and new estates show the gap first.
  2. Edges that do not line up. Hexagons do not follow a district boundary. Adding up the cells whose centre falls inside a district takes in a sliver of the neighbour and leaves out a sliver of its own.
  3. Spreading, not counting. A single cell holds a share of a larger official count. Over one or two cells that share can be well off; over hundreds the errors tend to cancel.
  4. Who counts as resident. Census rules decide where students, prisoners, care home residents and people with two homes are counted. A grid inherits those choices from its source, which may not be the table you are holding.
  5. Rounding and privacy. Small census areas are often rounded or slightly altered so no household can be identified. Small grid cells are estimates. Neither should be read to the last digit.

A made-up reconciliation in Lyon

The figures here are invented to show the method; they are not measurements. A health insurer has a census table for four districts of Lyon and pulls the same districts from the grid by drawing each one as a shape in an Area of interest.

DistrictCensus tableGrid totalGap
District A (old centre)48,20047,100-2.3%
District B (riverside)31,50032,900+4.4%
District C (new quarter)22,80026,400+15.8%
District D (hillside)29,60028,700-3.0%

Invented figures for illustration only.

Three districts sit within a few per cent, which is about what different dates and ragged edges explain. District C stands out. A look at the heat map shows a cluster of busy hexagons along its eastern edge where new apartment blocks have gone up since the census year. The grid is likely closer to today there, and the census closer to the official record. The insurer keeps the census for anything that must match published statistics and uses the grid for where to put an agent.

Which number to use, and when

  • Use the census table when your figure must match official publications: grant applications, regulatory filings, anything an auditor will check against a public source.
  • Use the grid when the shape is your own: the 1 km around a shop, a delivery zone, a 15-minute drive, a territory that crosses three wards.
  • Use both when you present to a board: lead with the grid for your own areas, and show the census for the districts it covers so nobody is surprised by the gap.

Mistakes to avoid

  • Treating one hexagon as a fact. A cell of 0.1 km2 is a modelled estimate. Judge areas of many cells, not a single one.
  • Comparing a grid circle with a census district. Same name, different shape, different number.
  • Calling the gap an error. A gap of a few per cent is normal. Investigate only the outliers, and look at the map before deciding which side moved.
  • Expecting every country. The grid has no data for some countries, among them the United Kingdom and the United States, where you will need the census alone. France, Canada, India, Thailand, Egypt and Australia are covered.
  • Mixing residents with visitors. Both sources count people who live somewhere. Neither says how many pass through; that is footfall, a separate measure.

Keeping both side by side with an AI employee

If you report on the same districts every quarter, an AI employee in xMap Flows can rebuild the comparison for you and flag any district where the gap moves. It is a small piece of location intelligence automation that saves an analyst a morning each time.

  1. Upload your census table to My data with one row per district and its official count.
  2. Schedule: every quarter.
  3. Area of interest: each district drawn as a shape, named as in your table.
  4. Demographics: Totals for each area, so each district comes back as one item with its residents and age groups.
  5. Match with my table: match on the district name and add the census count to each item.
  6. Calculated field: gap_pct = (population - census_count) / census_count * 100.
  7. Filter or Sort and limit: largest gaps first, or only gaps above 10%.
  8. Email to the planning team.

What lands in the inbox: one row per district with both counts, the gap and the age split from the grid, the biggest gaps at the top. Add Show on a map with a second Demographics block set to Every hexagon if you want the heat map that explains each outlier. The demographic profile of each city template is a quick starting point.

Questions

Is gridded population more accurate than the census?

Not in general. The census is the official count for its areas and date; a grid is a modelled spread of official counts. The grid is more useful for shapes the census does not publish.

Why does adding up hexagons inside a district not give the census total?

The hexagon edges do not follow the district line, the dates may differ, and each cell holds an estimated share. Gaps of a few per cent are normal.

How small an area can I trust from a grid?

The more cells you add up, the steadier the figure. A single 0.1 km2 cell is a rough guide; a 1 km circle or a district of many cells is far more reliable.

Can I get grid figures for the United Kingdom or the United States?

Not from xMap's population grid today. Use the official census tables there.

Templates that do this

Read next

More on population and demographics · Every article · Templates

Put it to work

An AI employee in xMap Flows does this on live data, on its own timetable. Start from a template or describe it in plain words.