CENSUS POPULATION ESTIMATES

The Comment is important enough to include. It would be good to have an observation from Tony Redpath of RBKC

Dear Dame,

I am having a distinct sense of deja-vue.  May I refer you to pages 10 and 11 of http://www.burisa.org.uk/Pdfs/154.pdf which deals with the 2001 underenumeration.

 

I have produced the same response rate graph for 2011, see attached, for comparison.

 

I see the Consultation and Partnerships Team say 'Despite an 18 per cent increase, from 64 per cent in 2001 to 82 per cent in 2011. . . . . ' I sense they are trying to claim responsibility for improving (by 'consultation and partnerships'?) K&C's response rate in 2011. If you compare the graphs you can see that actually *everyone* has improved their response rate - due to ONS's new procedure of post out / post back and targeting the defaulters from that.

 

You can also see that the two graphs have exactly the same form, with three peaks, a tail, and K&C an obvious outlier even from the worst by the same margin proportionally as in 2001. So the response rate may have improved, but the performance of K&C relative to everyone else is identical to 2001 - no improvement at all, and certainly no cause for self congratulation.

 

It was generally reckoned that Westminster spent near £1m challenging their 2001 census figure, and whilst they did eventually get ONS to increase the figure, which would have cascaded down the subsequent Mid Year Estimates, because the Standard Spending Assessment Formula is a zero-sum game it was by no means clear that it was financially worth it.  Just a word of warning for K&C.

The Council is entirely correct to contest the ONS estimate for the population decrease in the Borough. This well put together paper gives a good rationale for fighting the estimate. 
Getting the right result is very important for Labour wards and the poorer community areas in terms of education, social services, and the rest.

ROYAL BOROUGH OF KENSINGTON AND CHELSEA

COUNCIL RESPONSE TO THE 2011 CENSUS POPULATION ESTIMATES

1 Introduction

1. The first release of the 2011 Census estimates that the population of Kensington and Chelsea has decreased by 2.2 per cent in the intercensal period (2001-2011). The estimates suggest that the population of every other London borough has increased and that the overall population of London has grown by 11.6 per cent.

2. The first release of the census data also suggests that the number of households in Kensington and Chelsea has decreased by 0.6 per cent to 78,500.

3. This summary highlights the difficulty the Office for National Statistics (ONS) has had estimating the population of Kensington and Chelsea in the past, and examines the grounds on which the Council might seek to challenge the initial estimates.

2. ESTIMATing THE POPULATION OF KENSINGTON AND CHELSEA

1. ONS has traditionally found it difficult to produce reliable population estimates for inner London boroughs, especially Kensington and Chelsea. The reasons for this are well known. The Royal Borough exhibits all of the physical and demographic characteristics associated with poor census return rates and problems in producing reliable population estimates namely high population density, a large proportion of flatted properties (many of them with concierge entry or other forms of restricted access), an ethnically very diverse population and very high churn rates (over 70% a year in some parts of the borough).

2. These difficulties were illustrated at the previous (2001) census and are evident again with the 2011 census. The initial 2001 population estimate for Kensington and Chelsea (158,919) was far removed from the preceding mid-year estimates (190,300). This discrepancy led to a review and ultimately the rebasing of the 2001 mid-year estimate several years later to 162,600.

3. There is now a marked difference between the 2011 census population estimate (158,700) and the 2010 mid-year estimate (169,494) a difference of 10,794 residents or -6.4%

4. Despite an 18 per cent increase, from 64 per cent in 2001 to 82 per cent in 2011, Kensington and Chelsea still has the lowest response rate nationally. The confidence interval on the 2011 Census estimate is 3.79 per cent, the highest percentage nationally and equivalent to a margin of error of around 3,600 people. Because of the comparatively low response rate the census population estimate for Kensington and Chelsea is more heavily reliant on imputation, which accounts for some 29,100 of the total estimated population. Imputing the characteristics of non-respondents based upon the characteristics of respondents is notoriously difficult because the characteristics of these two groups may vary markedly (e.g. in household size, ethnicity, age profile).

3. GROUNDS FOR A REVIEW OF THE CENSUS POPULATION ESTIMATE

3.1. Arguments for a review of the initial census population estimate might be based on a number of grounds:

1. The intrinsic difficulty of estimating population in an area such as Kensington and Chelsea

Certain characteristics of Kensington and Chelsea make population enumeration and estimation difficult. The borough has the second highest population density nationally (131 people per hectare) and many residents live in flats that are accessible via buzzer-entry, security gates or a concierge style system, which inhibit efforts made by enumerators to ascertain who lives in a property but also whether a property is properly classified. Furthermore the borough is culturally and ethnically diverse with many non-English speaking residents. The ONS have also previously cited the large weekday populations that are drawn to the borough due to its proximity to large financial districts as another contributory factor.

2. Counter-veiling evidence to suggest that ONS estimates of population or household numbers are not correct

The Council holds a wealth of time-series service data about the properties and residents of Kensington and Chelsea. Analysis of this data may challenge some of the assertions from the first release, especially the assertion that the resident population has decreased.

3. Technical critiques of the imputation methodology and other elements of the census calculation

Some 29,100 residents were imputed for the Kensington and Chelsea population estimate. As mentioned previously (2.4) the weakness of a population estimate that contains a large number of imputed residents is that there may be a poor match between the characteristics of non-respondents and the respondents from whom data is imputed.

4. A critique of the quality assurance methodology used by ONS to verify their estimates.

The Quality Assurance process is undertaken by the ONS to ensure that census estimates are fit for purpose. The Main Quality Assurance Panel met 31 times to assess estimates for each of the 348 local authorities. This panel could request further work to be undertaken on the mid-year estimates and comparator data and ultimately consider adjustments to the census estimates. The ONS have stated that some Local Authority estimates were reviewed multiple times but no specific details have been published.

4. NEXT STEPS

Our best chance of a successful challenge to the ONS estimate is to mount a case based on a combination of the above grounds. Officers therefore intend to

1. meet with the National Statistician to query the reduction and discuss the impacts on future mid-year estimates.

2. request disclosure on whether Kensington and Chelsea was discussed by the Main Quality Assurance Panel. Specifically information on the nature of any discussions regarding Kensington and Chelsea, whether any recommendations were made and crucially if any adjustments were escalated and ultimately signed off by the National Statistician.

3. obtain and analyse a slice of service data (e.g. data from the electoral register, Council Tax information, the property gazetteer, and planning data) that relates to the number of households and residents in the borough to challenge the argument that the population has decreased since 2001.

4. understand the imputation methodology more fully. The Quality Assurance Pack that forms part of the first release of Census data gives details of the various types of imputation that combine to give the population estimate (e.g. overcount adjustments, bias adjustments, estimation and sample bias). Officers from Kensington and Chelsea should work closely with Westminster, who are pursuing a challenge along similar lines, to share expertise and results. This is likely to be the most technical aspect of a challenge and would require careful consideration as the ONS are the experts in this field.

Consultation and Partnerships Team

July 2012