63.7 million people use England's GP system. The government says 58.6 million live here. The gap is 5.1 million — and nobody will count them.
Last updated: March 2026
The latest official England population estimate is 58.6 million for mid-2024. The GP register still shows 63.7 million patients in England as at 1 March 2026, even after months of list-cleaning. Water companies' WRMP24 plans still point to roughly 62.3 million people served. Different systems, different methods, same broad direction.
This does not prove an exact hidden number to the last digit. But it does show a planning problem on a national scale. ONS has repeatedly revised migration and population inputs upward, is replacing the old method with an admin-based model, and outside institutions now treat this as a serious measurement failure rather than a fringe theory.
NHS England's official monthly publication shows 63,894,110 patients registered at GP practices in England on 1 January 2026 and 63,707,358 on 1 March 2026. That slight decline is consistent with ongoing list-cleaning, but the register remains above 63.7 million.
The key credibility point is not that GP lists are perfect. They are not. The key point is that even after known inflation and active removals, the register remains more than 5 million above the latest official England estimate. "List inflation" explains some of the gap. It does not, by itself, explain why the residual remains so large.
The chart below uses official annual snapshots. GP registrations are January counts from NHS England. ONS is the annual England mid-year estimate. The dates are not identical, but they are close enough to show the broad trend clearly: the GP line keeps pulling away from the official population line, even after upward ONS revisions.
Sources: NHS England Digital, Patients Registered at a GP Practice January releases 2020 to 2026; ONS England population mid-year estimate series `ENPOP`. ONS line ends at 2024 because that is the latest official annual England estimate currently published.
The evidence does not suggest that a 5 to 6 million discrepancy has always existed. A smaller gap appears to have been present for years because GP lists are imperfect and some groups are always hard to count. But the size of the mismatch has grown.
NHS Digital's own comparisons show the GP register was already above ONS population estimates by around 2.2 million in 2013. By 2020 that had grown to almost 3.9 million. By the mid-2020s, after the post-pandemic migration surge and repeated upward revisions to migration estimates, the visible gap had moved above 5 million.
The most likely reading is: a baseline discrepancy has existed for a long time, but the much larger modern divergence built through the 2010s and appears to have accelerated after 2021. That is exactly the period in which ONS began relying more heavily on admin-based methods and revising migration and population inputs upward.
Water companies publish statutory Water Resources Management Plans showing the populations they plan to serve. Across the English companies used here, the WRMP24 total is about 62.3 million. That is not a census, but it is also not a random forecast: it is the number used for infrastructure and supply planning.
The Environment Agency's latest annual performance review adds an important credibility check. It reports that actual household consumption in England was still 136.5 litres per person per day in 2024-25, and that companies supplied more water than would typically be expected. That matters because it means elevated demand is showing up in real system use, not just in long-range planning spreadsheets.
In 2014-15, the ONS said 257,000 EU migrants arrived. In the same period, 630,000 National Insurance numbers were issued to EU nationals. A gap of 373,000 in a single year. NI numbers are permanent — once issued, they're yours for life. There is no system to track whether those people left.
The 2021 Census Coverage Survey — designed to measure who the census missed — achieved only 61% of its target interview rate against a 90% target. ONS judged the results usable, but explicitly noted wider confidence intervals and a greater reliance on modelling and bias adjustment.
That context matters because ONS is now in the middle of a full transition away from the old population method toward admin-based population estimates. In other words, the statistical system itself accepts that measuring the population accurately has become harder and that the current method needs replacing.
We took eight government administrative systems — each counting real people — and back-solved the implied population from each. The results split into two groups.
Systems you actively sign up for (electoral register, driving licences, birth registrations, school enrolment) agree with the ONS at roughly 57 million. Systems that count you whether you signed up or not (water supply, GP registrations, council tax dwellings, state pensions) all suggest 60-62 million.
That pattern is exactly what you'd expect if several million people use public services but don't appear in official registers. They live in homes. They use water. They see a GP. But they don't register to vote or hold a driving licence.
The ONS's own replacement model no longer relies mainly on the census plus annual adjustments. Its admin-based population estimates combine a Statistical Population Dataset with additional "signs of activity" to infer who is likely to be a usual resident.
Those sources include the Personal Demographics Service, English School Census, Higher Education Statistics Agency, HMRC frameworks, Hospital Episode Statistics, the Emergency Care Dataset, HMRC Child Benefit, HMRC P14 pay records, and prison records. In other words: the statistical system itself is moving toward the same logic used on this site — track real presence and public-service activity across multiple administrative systems, then deduplicate and adjust.
This does not prove the exact size of the undercount. But it does make one thing harder to dismiss: the UK statistical system now accepts that population measurement increasingly depends on cross-checking administrative footprints, not just trusting the old headline estimate.
"GP lists are just inflated." Yes, partly. NHS England and Parliament have both acknowledged list inflation for years. But the current register is still above 63.7 million after active cleaning. The more serious question is how much of the gap remains after reasonable adjustments, not whether inflation exists at all.
"Water figures are forecasts, not counts." Also true. But WRMPs are statutory planning numbers, and the Environment Agency's latest performance review says actual household demand remained above 136 litres per person per day in 2024-25 and that companies supplied more water than expected. Forecasts and actual demand are pointing in the same direction.
"ONS already adjusts for this." It adjusts, and it revises. That is the issue. ONS has repeatedly updated migration and population estimates, published admin-based alternatives, and is replacing the current method because population measurement has become harder, especially in high-turnover places.
"So what is the true number?" This site does not claim to know an exact hidden total to the last digit. The narrower claim is that several independent administrative systems keep signalling a population higher than the headline estimate, and that gap is large enough to matter for planning.
"Could the real gap be 10 million?" It is possible in theory, but the current evidence on this page does not justify using 10 million as the headline claim. The strongest passive systems here cluster around a gap of roughly 5 to 6 million. Pushing beyond that without stronger corroboration would weaken the argument by making it easier to dismiss as exaggeration.
"What about SIM cards or mobile phone data?" Raw SIM totals are a weak population measure because many people have multiple subscriptions, work devices, tablets, and data-only plans. Ofcom reported 91.0 million active mobile subscriptions in the UK in Q3 2025, far above the UK population, so SIM counts cannot be read as resident headcounts. Proper telecom-presence analysis could still help at local level if it focuses on repeated overnight presence rather than total subscriptions.
Policy Exchange published The Case for a 2026 Emergency Census in England in December 2025, arguing that immigration-driven population growth since the 2021 Census makes a mid-cycle count necessary.
Centre for Cities called for a "corrections census" after the 2021 base was published, warning that unreliable estimates distort planning and local funding.
Parliament's own research service has documented the longstanding gap between GP registrations and ONS population estimates and the role of list inflation, which means this discrepancy is recognised inside the system rather than invented outside it.
ONS itself has moved toward admin-based population estimates and continues to revise migration and population figures as better administrative evidence arrives.
The government's response: No emergency census planned. The next scheduled census is 2031.
Water data: WRMP24 Supply-Demand Data, published by Defra via environment.data.gov.uk. "Total Resource Zone Population" per company for 2024-25 is used here as the planning-population signal. The table below also back-solves an implied population using household water delivered and a 137 l/p/d benchmark, close to the latest Environment Agency national figure of 136.5 l/p/d for 2024-25.
GP register data: NHS England Digital, Patients Registered at a GP Practice, latest monthly releases. Headline figures used here: 63,894,110 on 1 January 2026 and 63,707,358 on 1 March 2026.
NI numbers: DWP National Insurance Number Allocations to Adult Overseas Nationals, compared to ONS International Passenger Survey estimates.
Official population comparator: ONS population estimates for England and Wales, mid-2024. England total used here: 58.62 million.
Admin-based evidence: ONS mid-2024 admin-based population estimates and the linked data-quality overview. These set out the official shift toward combining coverage datasets with signs-of-activity sources such as school records, HMRC records, hospital activity and emergency care.
Census coverage: ONS Maximising the quality of Census 2021 population estimates. Target CCS rate 90%, achieved 61%.
Caveats: The WRMP population is a planning figure, not a resident headcount. It includes growth assumptions, company-boundary issues, and some cross-border complexity. GP totals include genuine ghost patients, delayed de-registrations, and duplicates. NI numbers are cumulative and include people who may have left. The argument here is not that every excess registration equals one hidden resident. It is that even after allowing for known inflation, the residual discrepancy across independent systems is too large to dismiss casually. On current evidence, the most defensible headline range is roughly 5 to 6 million, not 10 million.