Over the next year, ILC-UK, supported by the specialist insurance company, Partnership Assurance Group plc, plans to undertake a series of events to explore the relationship between our changing demography and public policy.
We started the series by exploring how proposals to change the way we undertake our Census may impact on our ability to understand our future society.
The Census was first carried out in 1801 - when the official population of Great Britain was revealed for the first time at 9 million. But current plans may mean significant changes to the future collection of data. In September 2013, the ONS initiated a three month consultation on the future of the national Census.
The ONS has proposed two options for reform. Either continuing with a Census each decade, but conducted primarily online; or using annual but smaller surveys in conjunction with existing government administrative data. The motivation is partly cost. However, the ONS has also stressed that any decision needs to be based not on cost, but on how to get the best and most timely information given technological advances.
Census findings are a tool to help governments allocate spending and plan ahead. The smaller annual survey would identify demographic and social trends more quickly but would be less detailed and comprehensive.
The Census has uncovered social phenomena that would otherwise have remained hidden – slum housing, fertility rates and transport among them. For example, the 1971 Census revealed how many people were living without hot running water. These findings can have a marked impact on policy. Danny Dorling, Professor of Human Geography at Oxford University, said “If you want to highlight the inequalities in a society there is no better way than to ask everybody how many bedrooms they have and how many people live in their house.”
The case for replacing the traditional Census with an annual alternative is based on a number of tenets, one of which is cost. The 2011 Census cost £480m; in 2021, the cost is expected to be £800m if the same, paper-based system were used. Replacing the Census would also allow for more timely data for planners and decision makers and could potentially avoid statistical surprises such as the unexpectedly big population growth uncovered by the 2011 Census.
The debate will feed now into an ILC-UK response to the Census consultation.
Throughout this debate and resulting policy brief we explored:
■How important is the Census to policy makers and industry (including the financial services industry)?
■Might the loss of some very local data make identifying exclusion more difficult?
■Might other datasets prove to be better than the Census in helping us understand our population and how it is changing?
■Are there any unintended consequences of scrapping the Census in its current form?
■How can we ensure that reforms to the Census do not risk our understanding of demographic change and ageing?
The End of the Census? Population Patterns Seminar Series supported by Partnership
1. The End of the Census?
Population Patterns Seminar
Series supported by Partnership
Thursday 28th November 2013
This event is kindly supported by Partnership
#populationpatterns
3. Andy Teague
Head of Statistical Development for Beyond 2011
ONS
This event is kindly supported by Partnership
#populationpatterns
4. International Longevity Centre – UK, Population Patterns
Seminar Series, 28 November 2013
The Census and future provision of
population statistics in England and Wales
– Public consultation
Andy Teague
Beyond 2011
5. The Beyond 2011 Programme
• Census – every 10 years for over 200 years
• Review a normal part of the census cycle but
the need greater than ever
Rapidly changing society
Evolving user requirements
Technological advances
Improved data sources
DRIVERS : Cost, efficiency, opportunity, burden
CRITERIA : Cost V social and economic benefit
of outputs, privacy, public acceptability, risk etc
• Government proposes to Parliament (with NS advice)
• Beyond 2011 findings will be published in 2014
7. Beyond 2011 : Two potential approaches
A census once a decade
-like that conducted in 2011,
but primarily online
A census based on
administrative data
and large annual surveys
8. Beyond 2011 : An online census – what it is
A census once a decade
-like that conducted in 2011,
but primarily online
A compulsory questionnaire for every household (and communal)
Majority of responses online (and mix of ways to complete)
1% survey to adjust for those who don‟t respond
Administrative data to check the quality
Population estimates produced annually using births, deaths, etc
Questions and topics similar to 2011 Census – but will consult later
9. Beyond 2011 : Administrative data and surveys – what it is
NHS Patient Register
DWP/HMRC Customer Information System
Electoral roll (> 17 yrs)
School Census (5-15 yrs)
Higher Education Statistics Agency data (Students)
Birth and Death registrations
NO PERSONAL DATA HELD – ALL NAMES &
ADDRESSES AND DATES OF BIRTH
ANONYMISED
A census based on
administrative data
and large annual surveys
Re-use of admin sources to produce annual population estimates
Anonymous data from eg NHS, DWP, HMRC, DfE, HESA
Annual compulsory 1% survey to adjust for error in the admin sources
Annual compulsory 4% survey to collect characteristics information
Majority of responses online (and mix of ways to complete)
Questions and topics similar to 2011 Census – but will consult later
10. Beyond 2011 : An online census – what you get
A census once a decade
-like that conducted in 2011,
but primarily online
Huge richness of data
Data for very small areas and very small populations
Detailed cross tabulations – nearly 6 billion cells
Continuity – tradition
A benchmark – a definitive snapshot of the nation - certainty
Data that is (or might be) out of date most of the time
An illusion of knowledge (some of the time) but the best we‟ve got
11. Beyond 2011 : The admin data approach – what you get
A census based on
administrative data
and large annual surveys
Administrative data and survey approach
population estimates
12. Administrative data and survey option
NHS patient
register
DWP/HMRC
Customer
information
system
1%
coverage
survey
HESA data
(students)
population
estimates
14. Population Pyramids using admin data
Administrative data method population pyramid with Census
comparison: England & Wales
Females
Males
2.5
2.0
1.5
1.0
Admininstrative
data method
Census
90+
85-89
80-84
75-79
70-74
65-69
60-64
55-59
50-54
45-49
40-44
35-39
30-34
25-29
20-24
15-19
10-14
5-9
0-4
Admininstrative
data method
Census
0.5
0
0
population (millions)
SPD 5
0.5
1.0
1.5
2.0
2.5
15. greater than 15% higher
between 14% and 15%
between 13% and 14%
between 12% and 13%
between 11% and 12%
between 10% and 11%
between 9% and 10%
between 8% and 9%
between 7% and 8%
10
between 6% and 7%
12
between 5% and 6%
14
between 4% and 5%
between 3% and 4%
between 2% and 3%
between 1% and 2%
16
within 1%
between -2% and -1%
between -3% and -2%
between -4% and -3%
between -5% and -4%
between -6% and -5%
between -7% and -6%
between -8% and -7%
between -9% and -8%
between -10% and -9%
between -11% and -10%
between -12% and -11%
between -13% and -12%
between -14% and -13%
between -15% and -14%
%
greater than -15% lower
Percentage difference between administrative data approach
and Census estimates – Output Area level
0%
88% of OAs
within +/-10%
Ave 30
people
8
6
4
2
0
Note that central bar covers double the range of the other bars
16. Administrative data and survey approach
What you get – population estimates
• Annual population estimates – for all
geographic levels – down to output areas
• Annual estimates age & sex – for all levels
down to LSOA
• OA level currently unproven but all the signs are
that this will be possible
• Research is ongoing
17. Administrative data and survey option
NHS patient
register
DWP/HMRC
Customer
information
system
1%
coverage
survey
4%
characteristics
survey
HESA data
(students)
population
estimates
characteristics
18. How survey works
• Most characteristics not currently covered by
an administrative source, although lots of
unproven potential
• Need compulsory 4% survey
• Reliable statistics could be produced for
characteristics representing:
800 or more people using 1 year‟s data (4%)
230 or more people using 3 years‟ data (12%)
130 or more people using 5 years‟ data (20%)
• Statistics for smaller populations would be
produced but lower accuracy (CI > 40%)
21. Administrative data and survey approach
spotting change over time
45,000
HARINGEY
Country of Birth:
Other EU
40,000
?
35,000
Estimated annual
population
30,000
Non-overlapping significant change
25,000
20,000
Assuming constant
rate of change by 2003 a
significant change
in the „Other EU
Country of Birth
Category would be
identified
Estimate
15,000
Lower Bound
10,000
Upper Bound
5,000
0
2000
2001
2002
2003
2004
2005
2006
Year
2007
2008
2009
2010
2011
2012
22. Administrative data and survey approach
spotting change over time
5,000
WELLINGBOROUGH
Tenure: Private
rented from landlord
or letting agency
4,500
4,000
Estimated annual household
population
3,500
3,000
?
Non-overlapping significant change
Assuming
constant rate of
change by 2006 a
significant change
in the „Private
rented from
landlord or letting
agency‟ Tenure
Category would be
identified
2,500
2,000
Estimate
1,500
Lower Bound
Upper Bound
1,000
500
0
2000
2001
2002
2003
2004
2005
Year
2006
2007
2008
2009
2010
2011
2012
23. Administrative data and survey approach
What you get - characteristics
Box F: Statistics possible using survey data
1 year’s data
(800 threshold)
3 years’ data
(230 threshold)
5 years’ data
(130 threshold)
LA
Average
number of
residents
160,000
Detailed crosstabulations
(c 200 cells)
Detailed crosstabulations
(c 500 cells)
Very detailed
cross-tabulations
(c 1000 cells)
MSOA
7,800
Some single
Very simple
Simple crossvariable statistics cross-tabulations tabulations
(c 10 cells)
(c 30 cells)
(c 50 cells)
LSOA
1,600
Not available
Some single
Some single
variable statistics variable statistics
(c 5 cells)
(c 10 cells)
OA
300
Not available
Not available
Area type
Not available
24. Admin data for characteristics - the
potential ??
Key advantage – broad coverage sources allow
statistics at lowest geographies ? – every year ?
Household composition
Income
Economic status
Health status (index?)
Qualifications
Industry of employer
Carers
Ethnicity
– various
– HMRC / DWP
– HMRC / DWP
– HSCIC / NHS Wales
– Census / DfE / BIS
– HMRC?
– HMRC / DWP (limited)
– NHS (quality?)
– HMRC / DWP (limited)
(Full list in paper M12)
27. Beyond 2011 : Advantages and disadvantages
A census once a decade
-like that conducted in 2011,
but primarily online
RISKS
• Increasingly difficult to
get high response
• Other methods of
completion required for
some households
• Considered an invasion of
privacy by some
STRENGTHS
• A rich set of statistics for a range of
geographies, a wide range of topics, small
populations, detailed cross tabulations
• Proven ability to deliver - proven and tested
• High degree of continuity
• A single, high quality snapshot of the nation
WEAKNESSES
• Only every 10 years (except for LA
population estimates) – reduces usefulness
• Costs more than the admin data option £625m per decade - £1.10 per person per
year
• Build-up and run-down challenging
• A burden on all households
OPPORTUNITIES
• Online completion will be cheaper and
more efficient
28. Beyond 2011 : Advantages and disadvantages
STRENGTHS
• Continually updated – statistics on an
annual basis
• Changes and trends identified more quickly
• Less expensive - £460m a decade – 80p
per person per year
• Reduced burden on households
WEAKNESSES
• Will never produce the detail provided by
the census
• Data combined for several years – makes
date to which it refers more complex
• Loss of a single historical record (options to
store more – but not yet developed)
• Requires new legislation
OPPORTUNITIES
• Use of admin data can be extended over
time
• Potential to be more flexible in questions
• New opportunities for historic research (in 2121)
A census based on
administrative data
and large annual surveys
RISKS
• New and untested methods –
other countries have taken
decades
• Some discontinuities
• Requires access to admin data
• Survey response will be
challenging here too
• Requires public acceptance of
use of admin data
30. Richard Willetts
Director of Longevity
Partnership
This event is kindly supported by Partnership
#populationpatterns
31. The End of the Census?
• A commercial perspective
Richard Willets
•
International Longevity Centre – 28 November 2013
32. Uses of census information
32
• Some use of the detailed information provided
– (including data published at Output Area level)
• Very widespread use of individual age mid-year population estimates
for England & Wales
Life expectancy
projections used to
value pension
scheme liabilities
and assess
insurance company
solvency
Mortality rate (for each age) =
death count / mid-year population
estimate
2-d ‘surface’ of mortality rates
November 13
33. Impact of the 2011 Census on population
estimates
33
Revision to England & Wales population estimates for mid-year 2011 following 2011 Census,
by age
5%
Females
Males
0%
40-44
45-49
50-54
-5%
-10%
-15%
-20%
•
Source: own calculations using ONS data
November 13
55-59
60-64
65-69
70-74
75-79
80-84
85-89
90-94
95-99
100+
34. Impact on apparent mortality improvement rates
Annual rates of mortality improvement, males in England & Wales, 2001-2011, by
age,
before and after publication of 2011 Census results
3.0%
2010-based
2.5%
2.0%
1.5%
1.0%
0.5%
0.0%
80-84
•
Source: own calculations using ONS data
November 13
85-89
90-94
95-99
34
35. Impact on apparent mortality improvement rates
Annual rates of mortality improvement, males in England & Wales, 2001-2011, by
age,
before and after publication of 2011 Census results
3.0%
2010-based
Revised
2.5%
2.0%
1.5%
1.0%
0.5%
0.0%
80-84
•
Source: own calculations using ONS data
November 13
85-89
90-94
95-99
35
36. Example consequences
36
• Small but material reductions in the projected lifespans of
pensioners/annuitants (typically circa 1%)
• Likely to see more significant revisions in the projected number of
elderly individuals
The ONS publication “What are the chances of reaching age 100?” published in Spring
2012 projected that 12% (95,000) of individuals aged 65 would reach age 100*
A more realistic projection – following the publication of Census 2011 results – might be
approx 7%-8% (about 60,000 individuals)**
•
*Source: ONS
November 13
•
**Source: own calculation
37. More significant revisions in the US
• In 2004 the US Census Bureau projected there would be:•
•
114,000 Americans aged 100 plus in 2010
1.1 million centenarians in 2050
• Following the 2010 Census, figures revised to:•
•
•
53,364 Americans aged 100 plus in 2010
0.59 million centenarians in 2050
Source: Wall Street Journal
November 13
37
38. Options for Beyond 2011
• „Once a decade‟ approach
•
•
Population estimates at high ages can „drift‟ significantly away from reality
between censuses
But we can be reasonably confident that figures are correct on the census
date
• „Administrative data‟ approach
•
•
•
May provide estimates with more consistent accuracy
Could be difficult to monitor degree of accuracy over time
Particular challenges in making high age estimates
If the latter approach is chosen:need to test accuracy of higher age estimates
(ideally with a transition period of „dual running‟)
November 13
38
39. Thank you
Partnership is a trading style of the Partnership group of Companies, which includes; Partnership Life Assurance Company Limited
(registered in England and Wales No. 05465261), and Partnership Home Loans Limited (registered in England and Wales No.
05108846).
Partnership Life Assurance Company Limited is authorised by the Prudential Regulation Authority and regulated by the Financial
Conduct Authority and the Prudential Regulation Authority. Partnership Home Loans Limited is authorised and regulated by the
Financial Conduct Authority. The registered office for both companies is Sackville House, 143-149 Fenchurch Street, London EC3M
6BN.
November 13
39
40. Professor Peter Goldblatt
Deputy Director, Department of Epidemiology
& Public Health
UCL Institute of Health Equity
This event is kindly supported by Partnership
#populationpatterns
41. Professor Heather Joshi
Professor of Economic and Developmental
Demography, Emeritus Professor, Centre for
Longitudinal Studies
Institute of Education University of London
This event is kindly supported by Partnership
#populationpatterns
42. ‘The End of the
Census’
ILC-UK, 28th November 2013
Remarks by Heather Joshi, IOE.
cls.ioe.ac.uk
43. Census Continues
FOR
Comprehensive Details
Geographical granularity
locality of disadvantage
Relationships- variables
-people
Households and families
– Eg Complex households
Carers – co-resident/ others
Comparability over time
Continue the ONS LS
AGAINST
Only once in ten years
Not timely
Cost
Irregular work-flow
Uncertain success on-line
Unpopular?
44. Admin census + surveys
FOR
Up to date information
Timely
Cheaper
Could include new variables
AGAINST
Limited spatial resolution
Quality not proven
Limited scope for linkage of
individual records across
people or data sets
May not provide continuity or
historical record
45. The best of both
Keep census for at least another round
Develop admin sources for current updates
Enrich admin data sources, developing address and
e-mail data bases.
Pilot new approaches for more gradual introduction
of new modes.
46. Professor Ludi Simpson
Beyond 2011 Independent Working group
University of Manchester
This event is kindly supported by Partnership
#populationpatterns
47. Beyond 2011 Independent Working
Group
http://popgeog.org/beyond-2011-independent-working-group/
International Longevity Centre
28 November 2013
Ludi Simpson, University of Manchester
President British Society for Population Studies 2011-13
(Drawing on comments by many others)
48. Option: Admin records for population
estimates, 4% survey for attributes
• Very positive potential for population
estimates
• Attractive promise of more annual statistics
• But lost, by design
– Flow data for migration and commuting
– Longitudinal data from linking census records
– Nationally comparable data for small areas
• Risks of new legislation and methods
49. Option: online census
• Maintains a known structure and continuity
• Challenges of response, that also affect a
survey
• Option not developed sufficiently to prove
opportunities for efficiency
50. How will we interpret area attribute data
pooled over different numbers of years?
• Helpful survey accuracy tool from ONS
– 40% CI is very wide for adequate estimates?
• Smaller populations (areas, or social groups)
– Large
Each year
– Medium 3 year average
– Small
5 years or more average
• Key need currently satisfied once each decade:
Compare social change in area X with area Y
– Often impossible even for averages of less than 10 years
– When possible, best data may be from different time
periods
51. Examples of information loss without a
census
• Causes of death linked to life circumstances
– Recent and lifetime: Longitudinal Study
• Age-sex-structures national and sub-national
– Unproven validation of estimates from
administration records (RSS comments;
legislation)
– Survey insufficient for detailed age structure of
social groups within local authorities, even once in
10 years
52. Examples of information loss without a
census
• Life chances linked to place
– Unable to identify from a survey even over ten
years:
• Health related to private renting within cities
• Overcrowding in housing neighbourhoods
• Cultural demand for burial services in towns
• Response to emergencies (flooding, industrial, ...)
– Affected areas not predictable in advance, local
information now from census will not be available
53. Costs and benefits
2013 prices across one decade
• Costs in 2021
– Census option (£625m) more than Survey (£460m):
+£165m
(ONS consultation paper C1)
• Benefits to users
– Census option (£797m) more than Survey (£557m):
+£240m
(preliminary, ONS consultation paper C3)
• “Census saves £1bn”
– (Head of Census Glen Watson to Commons S&T Cttee 2011)
54. Biggest decision on UK social statistics
for many decades
• Neither option alone is optimal
– A disaster to rely on administrative data, but
stupid not to take advantage of them
– Knowledge of small populations is essential to
stable government, but a mainly paper census is
inefficient
55. An organised transition to...
• A replacement for the Census
– Validated attributes from administrative records
together with survey modelling, offering mainly
equivalent and better products than
Census, tested with a mainly online 2021 Census
• The Census as a basis for updates
– A mainly online census, with legislation to allow
samples to be drawn from it
– A mainly online census, with legislation to allow a
personally updateable record
56. Consultation to December 13th 2013
• The seriousness with which government will treat
their decision
• The examples which will be taken into the
political and media arenas
• Individuals and organisations
We are interested in a copy of your comments:
AreaStatistics@gmail.com
http://popgeog.org/beyond-2011-independent-working-group/
has link to presentations and examples from October 21st event
57. Phil Rossall
Research Manager
Age UK
Vivienne Avery
Research Adviser
Age UK
This event is kindly supported by Partnership
#populationpatterns
58. Voluntary Sector use of Census data –
supporting the daily work of Age UK
Phil Rossall and Vivienne Avery
Age UK
28 November 2013
59. Introduction to Age UK
•
•
•
•
•
•
Mission – to improve the lives of older people
Social enterprise and charity
7 areas of activity
170 local Age UKs
3000 employees
50,000 volunteers
There are over 200,000 charities in the UK
and some much bigger than us!
60. Who uses research and statistics
in Age UK?
• Campaigns Team: basis of Care in Crisis and Spread the Warmth
campaigns
• Research Team: in-house research, Economic Monitor, research
publications
• Media Team: stats and survey-led press releases (e.g. loneliness,
winter alerts)
• Policy Team: Agenda for Later Life report, policy briefs, submissions
to parliamentary committees, consultation responses
• Services: stats support for local Age UKs, Information and Advice,
special projects (e.g. digital inclusion programme)
• Local partners: independent charities, e.g. bids for commissioned
services
• Enterprises and shops: support services, product choice e.g.
mobile phones, customer intelligence (market failure)
• Training: nurses, social workers awareness training, empathy
61.
62. Examples of Age UK Research Impact
Development of the exercise programmes in the NSF falls
services (2001)
Development of the DWP ageing strategy in „Opportunity
Age‟ (2005)
Development of the Social Exclusion Units strategy in „A
Sure Start to Later Life‟ (2006)
Development of treatment for reducing post-event
inflammation in thrombotic stroke (2009)
Changes to the diagnosis and treatment of urinary
incontinence (2010)
Removal of the default retirement age (2011)
63. Examples of using LSOA data (1)
• Strategic planning
• Targeting services for older people - growing focus on
loneliness and vulnerability need for data with multiple
characteristics
64. Examples of using LSOA data (2)
Combining Census, benefits,
transportation & advice service data to
help locate a new local office to reach
those with the greatest needs
Data also used for
• Fundraising
• Tendering for commissioned
services
• Impact and evaluation
65. Local Age UK views on Census proposals
“It is already hard to find detailed information using the current available data.
The changes would make this even trickier to look at need in localised areas,
making it hard to apply for funding and produce evidence or track changes
which may be positive in terms of work relating to the prevention agenda.
Lack of data may cause further inequalities to develop for populations which
are already disadvantaged. This is especially relevant for Stockport as it is very
varied - we have some of the wealthiest pockets and some in the top ten IMD.
These pockets are small so would cease to be significant.” Age UK Stockport
“It seems shocking to me that HMG are so poor at capturing and manipulating
reliable current data. Tesco knows my age, my politics, my favourite television
programme and in all likelihood the hour of my death.” Age UK Northampton
66. Ready for Ageing?
Future Census and population statistics must address
• Population growth
• Life expectancy gap
• Changing services
Health & social care – aiding independent living
Urban planning / Housing
Retirement income / planning financially for
older age
Addressing inequality – interplay between age,
gender, generation & wealth
67. Population Growth and Life Expectancy Gap
Living with long-term / multiple
conditions (dementia)
Local Authority level data masks
the detailed picture
Medical records provide data on
the condition rather than it‟s impact
on daily life
68. Our thoughts on ONS proposals
Internet Census
•
•
•
We know the methodology works down to low levels
The data are rich and valuable
Should be easy to get >16% response overall but will need to
take care with 80+ age group – numbers who have never used
the internet are increasing
Admin records/Survey Census
• Timeliness benefits – by 2010 much unreliable data – poor
resource allocation and decision making
• Survey potential for more detail at local authority level but
compulsory? Can merged survey years really produce reliable
LSOA data?
• Merged admin data should produce better neighbourhood data
over time but concerned there will be a data gap in short term
69. Our conclusions
• Small State v Big Society We need the best information possible
–little difference between an annual cost of 80p & £1.10 per person given
costs of market failure due to lack of information. NHS annual budget per
person is c£2,000 + education, transport etc
• Before taking a final decision we need a better impact assessment
beyond population numbers for local authorities – more evidence of how
combined survey and admin data will measure up at smaller areas
• We need a better development programme and more dual running to
enable us to decide when not if a traditional Census can be replaced.
That programme should include identifying broader future information
needs not just for ageing but more widely.
72. The End of the Census?
Population Patterns Seminar
Series supported by Partnership
Thursday 28th November 2013
This event is kindly supported by Partnership
#populationpatterns
Notas do Editor
England & Wales total for total population = -
Quniary age/sex totals as compared to the census estimate.Main feature: overcoverage of men between 35 and 50Females track the census estimates pretty well.
Both options involve challenges, but the admin+survey simply cannot offer some of the key data output types that we need.
Difficult to publish even for one year simple results.40% CI a low threshold – 10%?
Each over one decade. Not sure if comparable. Benefits certainly incomplete on both sides of equation.
Pop Stats strategy in 1990s – didn’t go far, but now an opening. Importance of small populations – area data but not only – needs to proven in this consultation.
Arising from consultation already, which you may comment on or add to.