What to expect from the first Census results for London
On June 28th, ONS will publish the first results from the 2021 Census. This blog offers an overview of what we expect to learn from the new data, as well as some of the challenges in interpreting the results.
- The first release will contain only limited data about the size and age structure of the population as well as numbers of occupied household spaces.
- These results will have an important role in informing both future estimates of the population and revisions to our view of how the population changed over the last decade.
- The results of the 2011 Census exposed a number of major issues in the accuracy of official population estimates and helped prompt a change in expectations about London’s future growth.
- We should anticipate the results of the 2021 Census will also reveal issues in current estimates for London, including inflated numbers of children and older adults.
- Naïve comparisons of the results of the census with recent population estimates are likely to conflate the impacts of the pandemic and errors in the estimates data.
- The Census figures will not be an accurate reflection of London’s population in 2022, as the Census preceded the return of many young people to London that took place over the following months as the economy reopened.
What’s in the first release?
This first release will provide us with census-day estimates of the population, broken down by local authority of residence, five-year age group, and sex, as well as estimates of the number of occupied household spaces.
While this won’t give us anything like the rich and detailed picture of the population that subsequent releases will provide, these results will be keenly followed by anyone whose work relies on an accurate understanding of the size of the population.
Significance of new census data
The significance of even these limited results largely hinges on the critical role they play in the production of annual population estimates and projections. In the UK, as in many other countries, annual population estimates are anchored to the results of the most recent census. Estimates for subsequent years ‘roll-forward’ from this starting point, accounting for estimated births, deaths, and migration flows.
Because of the difficulties in accurately measuring migration, errors tend to build up in population estimates over time. For parts of the country – particularly within London – that experience high levels of migration, there is potential for very large errors in estimates to accumulate over the course of a decade.
The results of the 2021 Census will effectively supersede existing estimates of the size and age structure of the population. The data released on June 28th therefore could have important implications across a wide range of areas, including: local government finance, anticipated demand for health and education services, our understanding of the impacts of the pandemic – and how we prepare for potential future waves of COVID-19.
What did the 2011 Census results show us?
To get an idea of the potential scale of readjustment that new census data might bring, we can look back to 2011 and compare official estimates before and after they were revised to account for the results of the census.
Estimates for London as a whole were revised up from 7.94 million to 8.20 million, an increase of 3.2 percent on the previous 2011 population estimates. With most of the difference being accounted for by a previous underestimate of the population between the ages of 10 and 35.
A 3.2 percent difference, though much larger than for any other region, may seem like a minor adjustment with limited consequences for planners and policy makers. However, this innocuous headline figure hid some more significant (if not completely unexpected) findings.
Firstly – the 2011 population might only have differed by three percent, but the difference in growth since the last census that this implied was both much greater and much more relevant for the city’s planners.
The census data revealed that, over the decade, London had been growing 43% faster than previously estimated. The implications of this difference is perhaps most starkly illustrated by comparing population projections produced before and after the census figures were incorporated into official estimates:
- The 2008-based projections, published by ONS in 2010, gave a figure of 9.19 million for London’s total population in 2033
- The 2012-based projections, published in 2014, gave a figure of 10.35 million for the same year
While this increase in the projected population wasn’t entirely due to the upward revision of estimates following the census results (birth rates were rising, mortality was falling, and domestic outflows dropped sharply in the wake of the financial crisis), it was an important component of it.
Secondly – looking at London as whole hides the much greater revisions made to estimates for individual local authorities. Those in London saw both the largest upward revisions and the largest downward revisions in the country.
The populations of central London authorities were far higher in the original estimates than in those based on the 2011 Census, with estimates for City of London (+62%), Westminster (+17%), Camden (+9%), and Kensington and Chelsea (+7%) all within the ten most inflated in England. At the other end of the scale, estimates for Newham (-22%), Brent (-17%), Haringey (-12%), and Waltham Forest (-11%) saw the country’s biggest upward revisions.
Such issues had important financial implications for the local authorities involved as a portion of the funding they received from central government was directly linked to their resident populations as captured in the official estimates.
The impacts of revisions were also apparent in any statistics where population estimates were used as denominators to calculate rates. Perhaps the most salient example of this is how estimates of life expectancy changed. The arrival of the 2011 census data revealed that a significant portion of the apparent wide (and much commented upon) variation in life expectancy within London had been an artifact of the underlying population data used to create them.
What can we expect from the 2021 results?
Impacts of the pandemic
The 2021 Census was conducted in March 2021 during the pandemic. At this time the population was coming out of lockdown, but recovery was only just underway. In our analysis of population change during the pandemic we concluded that many young adults left London during and following the first lockdown, and that many returned over the course of Spring and Summer 2021. With the census held just before this turnaround took place, it seems likely that the snapshot of London’s population it captured will be missing many of those who were temporarily displaced from the city.
Indexed change in London payrolled employees by age group (Indexed: March 2020 = 100)
Source: ONS, Earnings and Employment PAYE Real Time Information
This and other issues, such as how higher education students attending courses remotely will be reflected, mean that many of the results of the census are going to require careful interpretation. Such issues will make ONS’s task of creating the Mid-Year Population Estimates – which involves accounting for changes in population that occurred between the date of the census and the end of June – especially challenging this time around.
Issues with existing population estimates
Since 2011 ONS has made various changes to the methodologies used in their mid-year population estimates. A key improvement has been to the methods used to allocate inflows of international migrants between individual local authorities, which should help address the single most significant issue that affected the accuracy of the older series. Despite this improvement, however, the 2021 census results can still be expected to reveal problems in the annual estimates.
Fewer children in London than official estimates suggest
The one we’re most confident in predicting is that the census will show official estimates of the number of children in London to be significantly inflated. This is because a methodological change made after the last census exaggerates the contribution of international migration to the population of children. The nature of the issue means that it has the biggest impact in areas with high levels of international migration, and that it tends to compound over the course of the decade.
The most extreme impacts of this can be seen in areas like Westminster, where we think that the official estimates for some age groups are now fifty percent higher than reality. In our own projections, where we try to compensate for this effect as best we can, we put the number of resident children in London for mid-2020 at 1.21 million, over one hundred thousand below the official estimate of 1.32 million
Inflated estimates of older adults in some parts of London
The roll out of COVID-19 vaccinations in 2021 was key to reducing hospitalisations and mortality in subsequent waves of the pandemic. As coverage increased among the older population, there was intense interest in identifying areas of low take-up.
Using official population estimates as denominators revealed a very large variation between local authorities in the apparent proportion of the older population that had not been vaccinated. After further analysis considering alternative sources of data such as numbers of pension claimants, we concluded that real differences in take-up rates were likely being masked by inflated population denominators in many areas. The resulting uncertainty about the true level of vaccination coverage across London’s communities will hopefully be greatly reduced by the initial release of results from the census.
Issues with capturing returning international students
Measuring emigrant outflows is generally accepted to be the most challenging element of estimating population. While ONS was able to make real improvements to how they measure flows into the country, progress on the measurement of outflows has been more modest.
Potential issues with estimates resulting from a failure to accurately capture flows of returning international students were highlighted by the Office for Statistics Regulation last year in their review of ONS’s population estimates and projections.
The OSR review was triggered by concerns raised about the accuracy of estimates for Coventry (home to many University of Warwick students), but it’s possible that similar issues may also impact estimates for areas in London with large populations of overseas students.
Challenges in interpreting the results
Disentangling impacts of the pandemic from issues with the annual estimates
A temptation that I’d strongly urge resisting will be to compare the census results with the 2020 population estimates and conclude that the differences between the two figures represents the impact of the pandemic on the population. Such a naïve approach is liable to conflate the two effects and arrive at misleading conclusions.
It will require careful analysis to ascribe differences between the census results and annual estimates to specific causes. Even then it will likely prove impossible to remove all ambiguity – certainly before more detailed data from the census becomes available later in the year.
Limitations of what the census results will tell us about the current population
The 2021 Census has captured an extraordinary moment in the history of the country. All our analysis over the last year has suggested that the number of people in London at the point the census was conducted was likely much lower than was the case either a year earlier or a year later.
If the census has captured this moment accurately, then it has fulfilled its purpose but won’t necessarily allow us to say that it represents our best understanding of London’s current population (at least not without some caveats). Much challenging work lies ahead, building on the foundations of the census results to develop an understanding of how things have changed in the months and years since March 2021.
ONS actually published some tables of data from the Census earlier in the year to help inform the response to crisis in Ukraine. However, these tables were only based on raw counts and will be superseded by subsequent data that includes adjustments for variations in coverage.
To clarify – these comparisons are between ‘rolled forward’ estimates for 2011 using the ‘original’ method that was consistent with the official estimates series at the time, rather than the ‘indicative’ or ‘new method’ estimates that were also available and which included improvements to international migration estimates.