Population Trends 107

About the Office for National Statistics The Office for National Statistics (ONS) is the Government Agency responsible for compiling, analysing and disseminating many of the United Kingdom’s economic, social and demographic statistics, including the retail prices index, trade figures and labour market data, as well as the periodic census of the population and health statistics.The Director of ONS is also the National Statistician and the Registrar General for England and Wales, and the agency administors the statutory registration of births, marriages and deaths there.

About Health Statistics Quarterly and Population Trends Health Statistics Quarterly and Population Trends are journals of the Office for National Statistics. Each is published four times a year in February, May, August and November and March, June, September and December, respectively. In addition to bringing together articles on a wide range of population and health topics, Health Statistics Quarterly and Population Trends contain regular series of tables on a wide range of subjects for which ONS is responsible, including the most recently available statistics. Subscription Annual subscription, including postage, is £75; single issues are £20. Annual subscription for both Health Statistics Quarterly and Population Trends, including postage, is £135. Online Health Statistics Quarterly and Population Trends can be viewed or downloaded as Adobe Acrobat PDF files from the National Statistics website www.statistics.gov.uk/products/p6725.asp (Health Statistics Quarterly) or www.statistics.gov.uk/products/p6303.asp (Population Trends).

A National Statistics publication National Statistics are produced to high professional standards set out in the National Statistics Code of Practice. They undergo regular quality assurance reviews to ensure that they meet customer needs. They are produced free from any political interference.

Contact points at ONS

Editorial board

People with enquiries about the statistics published regularly in Health Statistics Quarterly and Population Trends can contact the following enquiry points.

Peter Goldblatt (editor) David Pearce (editor) Angela Dale Paul Hyatt Graham C Jones Azeem Majeed Jil Matheson Ian R Scott Judith Walton

Topic enquiries Abortions: 01329 813618 Births: 01329 813758 Conceptions: 020 7533 5113 Expectation of life: 020 7211 2622 (Government Actuary’s Department) Marriages and divorces: 01329 813379 Migration: 01329 813889/813897 Mortality: 01329 813758 Population estimates: 01329 813318 Population projections: National – 020 7211 2622 (Government Actuary’s Department) Subnational – 01329 813474/813865

Contributions Articles: 5,000 words max.

W

in

te

r

n um ut A

Title

General enquiries Su m m er

Issue

Sp ri ng

Dates for submissions

Health Statistics Quarterly by 11 Sept by 11 Dec by 22 Mar by 21 June Population Trends

by 23 Oct

by 2 Feb

by 4 May

National Statistics Information and Library Service 1 Drummond Gate London SW1V 2QQ Tel: 0845 601 3034 E-mail: [email protected] Website: www.statistics.gov.uk

by 26 July

© Crown copyright 2002. Published with the permission of the Controller of Her Majesty’s Stationery Office (HMSO).

Please send to: Clare Parrish, executive secretary Health Statistics Quarterly/Population Trends Office for National Statistics B7/06 1 Drummond Gate London SW1V 2QQ Tel: 020 7533 5264 Fax: 020 7533 5103 E-mail: [email protected]

Applications for reproduction should be submitted to HMSO under HMSO‘s Class Licence: www.clickanduse.hmso.gov.uk Alternatively applications can be made in writing to: HMSO Licensing Division, St Clement’s House, 2-16 Colegate, Norwich NR3 1BQ ISBN 0 11 621535 6 ISSN 0307-4463

Spring 2002

N O 107

Population Trends In this issue

Page

In Brief

2

Demographic indicators

4

2000-based national population projections for the United Kingdom and its constituent countries The latest national population projections for the next twenty-five years, compiled by the Government Actuary’s Department Chris Shaw

5

Improving ONS spatial referencing – the impact on 2000 births and deaths data Description of improvements in the way that the Office for National Statistics allocates births and deaths data to administrative and health areas in England and Wales Elspeth McVey and Allan Baker

14

Implications of changes in the UK social and occupational classifications in 2001 for vital statistics Summary of recent changes in the method of reporting vital statistics by socio-economic status Angela Donkin, Yuan Huang Lee and Barbara Toson

23

The Millennium Cohort Study The latest birth cohort study, charting various aspects of individuals’ lives into adulthood, within the United Kingdom Kate Smith and Heather Joshi

30

Marriages in approved premises and register offices in England and Wales: the proportions of couples who marry away from home Evaluation of the provisions of the 1994 Marriage Act enabling marriages in approved premises and outside district of residence John Haskey

35

Tables List of Tables Tables 1.1 - 9.3 Notes to Tables

53 54 82

Report: Marriages in England and Wales during 2000

83

London: The Stationery Office

Population Trends 107

Spring 2002

in brief Civil Registration Review: White Paper Launched The long awaited Government White Paper on civil registration reform in England and Wales was published in January 2002. Civil Registration: Vital Change1 sets out the Government’s plans for modernisation and improvement of the civil registration service in England and Wales. The White Paper is the culmination of the review of civil registration that started with the publication of the consultation document Registration; Modernising a Vital Service.2 The current system for Civil Registration dates back to the 19th Century. The needs of society, families and individuals have changed dramatically since then. It is important for civil registration to adapt to reflect these changes. The framework described in the White Paper provides for improved and new registration services and enables the full use of modern technology. It will not compromise on the integrity and accuracy of registration records but, through innovation and effective application, will maintain standards and be responsive to future needs. The proposals will give people more choice and provide a better service at important moments in their lives. The main proposals include: • individuals will be able to register births and deaths3 on-line, in person or by telephone and will be able to use any local registration service point to access registration services; • events relating to the active population will be recorded on a central database, which will be available locally and centrally using web technology;

National Statistics

2

•

•

data from the NHS will be used to create a ‘skeleton’ entry for births and deaths on the database, which will be completed with the information received from the informant; and the system for marriages will move to a ‘celebrant’ model rather than a place-based one as at present.

There will be new models for access to historic and modern registration records that will balance openness and privacy. Historic records (those relating to people over 100 years of age) will be fully available to everyone. For records relating to the active population the majority of the record will be available, as at present, but access to information on addresses, occupations and cause of death will be restricted. This restricted information will be available to the individual, their families, those who are given access by the individual/family and to those organisations where the law so provides. Medical and social researchers will continue to be able to use civil registration records within a framework that maintains the undertakings on confidentiality given to the public. There will be increased opportunity for data sharing and record linkage. The use of certificates for official proposes will be phased

out over time as public and private sector organisations increasingly access the central database to verify the particulars of an event. It is proposed to link the birth, marriage and death records that occurred in England and Wales and that relate to an individual to create a ‘through life’ record. Other records such as divorce records will be added to enhance the quality of the ‘through life’ record. The White Paper recognises the importance of the civil registration system in providing information about society. It sets out proposals for providing a more flexible model for the collection of statistical information at the time of registration in order to meet changing needs. There will also be greater use of informed consent and follow-up surveys. Users of the statistical information will be consulted about the potential changes. These proposals for reforming civil registration form part of the wider Government work on modernising public services. The changes to the legislation will be made using the order making powers of the Regulatory Reform Act 2001. This Act contains a wide power, matched by tough safeguards, for reforming burdensome legislation such as that relating to civil registration. It will involve thorough consultation with stakeholders as well as Parliamentary scrutiny. The changes to the legislation are likely to take until early 2004. Notes and references 1. Office for National Statistics (2002) Civil Registration: Vital Change, TSO: London. Also available from www.officialdocuments.co.uk/document/cm53/5355/ 5355.pdf 2. Office for National Statistics (1999) Registration: Modernising a Vital Service. Available on www.statistics.gov.uk

3. The introduction of on-line registration for deaths will be dependent on the outcomes of the Shipman Inquiry and the Review of Coroners.

Population Trends 107

Census 2001 With an estimated response rate of 98 per cent, the field operations aspect of the Census are now in their final stages and the processing of the Census forms is in progress.

CURRENT SITUATION • • • • •

Response rate estimated to reach 98 per cent of household. 88 per cent of forms are estimated to have been returned by post. 7 per cent are estimated to have been collected by enumerators. 3 per cent are estimated to be vacant dwellings or second homes. 65 per cent of the forms have now been scanned and coded.

Spring 2002

QUALITY

EVALUATION

The evidence from Census forms processed so far in England and Wales and from information gathered during the enumeration phase indicates that the quality of the results will be as least as good as in the 1991 Census and that achieved in recent Censuses in other countries.

Evaluation reports on each aspect of the Census will be made available as soon as they have been completed and brought together in a General Report in 2003. In addition, all information on quality will be gathered together and published in a Census Quality report, again in 2003.

TIMETABLE FOR •

•

PUBLICATION

The first results – mid-2001 population estimates for each local authority area by age and sex – will be published at the end of August 2002. Main results will be made available in a series of releases through a variety of media over a six month period scheduled to start around the end of 2002.

Further information is available on the National Statistics website at www.statistics.gov.uk/census2001/ default.asp

Recent Publications Living in Britain 2000 (The Stationery Office, £37.50, December, ISBN 0 11 621479 1) Mortality Statistics: cause 2000 (December, available on the National Statistics website at: www.statistics.gov.uk/products/p618.asp) Birth Statistics 2000 (December, available on the National Statistics website at: www.statistics.gov.uk/products/p5768.asp) Family Spending 2000/2001 (The Stationery Office, £39.50, January, ISBN 0 11 621478 3) Annual Abstract of Statistics 2002 edition (The Stationery Office, £39.50, January, ISBN 0 11 621473 2) Social Trends 2002 edition (The Stationery Office, £39.50, January, ISBN 0 11 621472 4)

Civil Registration:Vital Change (Cm 5355) (The Stationery Office, £12.50, January, ISBN 0 10 153552 X). A Welsh translation of this White Paper is also available (The Stationery Office, £12.50, January, ISBN 0 11 621547 X) Health Statistics Quarterly 13 (The Stationery Office, £20, February, ISBN 0 11 621531 3) Cancer Statistics, Registrations 1998 (February, available on the National Statistics website at: www.statistics.gov.uk/products/ p8843.asp)

All publications listed can be downloaded free of charge from the National Statistics website, www.statistics.gov.uk .Where indicated, publications from The Stationery Office can be ordered on 0870 600 5522 or online at www.clicktso.com.

3

National Statistics

P opulation Trends 107

Spring 2002

Demographic indicators Figure A

England and Wales

Population change (mid-year to mid-year)

Thousands 300 200 100 0 Natural change -100 7 19

1–

72

Net migration and other changes

Total change

0 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 9 –7 3–7 4–7 5–7 6–7 7–7 8–7 9–8 0–8 1–8 2–8 3–8 4–8 5–8 6–8 7–8 8–8 9–9 0–9 1–9 2–9 3–9 4–9 5–9 6–9 7–9 8-9 200 9 9– 9 9 9 9 9 9 9 9 8 8 8 8 8 8 8 8 8 8 7 7 7 7 7 7 7 72 9 Mid-year

Figure B

Total period fertility rate

TFR (average number of children per woman) 2.5 2.0 1.5 1.0 0.5 0.0 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Year

Figure C

Live births outside marriage

Percentage of all live births 40 35 30 25 20 15 10 5 0 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Year

Figure D

Infant mortality (under 1 year)

Rate per 1,000 live births 20 15 10 5 0 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 Year

National Statistics

4


Spring 2002

2000-based national population projections for the United Kingdom and its constituent countries Chris Shaw Government Actuary’s Depar tment

INTRODUCTION The Government Actuary’s Department produces national population projections for the United Kingdom and its constituent countries at the request of the Registrars General of England & Wales, Scotland and Northern Ireland. The assumptions are agreed in consultation with the statistical offices of the four countries. This article presents the main results from the 2000-based national population projections, which replace the previous 1998-based national projections.1 The main focus of these projections is on the next twenty-five years, i.e. up to 2025. However, the results of longer-term projections are included in the graphs in this article and discussed where appropriate.

BASE POPULATION The 2000-based projections are based on the mid-2000 population estimates produced by the Office for National Statistics (for England & Wales), the General Register Office for Scotland and the Northern Ireland Statistics and Research Agency. 2 The estimated population of the United Kingdom at mid2000 was 59.8 million. As Table 1 shows, this was just 6 thousand (0.01 per cent) higher than envisaged in the 1998-based projections. However, net inward migration to the UK between mid-1998 and mid-2000 was slightly higher than projected. This underprojection of net migration was largely balanced by there being nearly 30 thousand (2 per cent) fewer births than projected between mid-1998 and mid-2000.

The 2000-based national population projections, carried out by the Government Actuary at the request of the Registrars General, show the population of the United Kingdom rising from 59.8 million in 2000 to nearly 65 million by 2025. Longer-term projections suggest the population will peak at nearly 66 million around 2040 and then gradually start to fall. The population will become gradually older with the median age expected to rise from 37.4 years in 2000 to 42.4 years by 2025. In 2000, there were 1.3 million (12 per cent) more children aged under 16, than people of state pensionable age. However, by 2007, the population of state pensionable age is projected to exceed the number of children.

UNDERLYING ASSUMPTIONS The assumptions used in the 2000-based national population projections are shown, for the United Kingdom as a whole, in Figure 1, while those for individual countries are summarised in Table 2.

5 National Statistics

P opulation Tr ends 107

Spring 2002

Figure 1

Box one SUBNATIONAL PROJECTIONS

United Kingdom (a) Total fertility rate (TFR) and average completed family size (CFS)*, 1951–2025

The most recent long-term subnational projections for England are 1996-based. 3 The next set of long-term subnational projections will be 2002-based. (Short-term subnational projections are produced for the Department of Health in the intervening period between long-term projections. Although these are consistent with the national projections, the local migration assumptions used are those from the 1996based long-term subnational population projections. These short-term subnational projections are not used for local authority or other planning purposes.)

3.0

Children per woman

2.6

2.6

2.4

2.4 CFS*

2.2

Replacement level

2.2

2.0

2.0

1.8

1.8 1.6 1951

1961

1971

1981

1991 Year

2001

2011

2021

* CFS relates to cohort born 28 years earlier - 28 years being roughly the mean age at childbearing. Assumed CFS is given for cohorts who have not yet completed childbearing.

United Kingdom (b) Expectation of life at birth, 1981–2040 84

84 Assumptions

➝

82

Females

82

EOLB (years)

80

Thousands 1998-based projections

3.0 2.8

1.6

Population change 1998–2000: actual change compared with 1998-based projected change

Mid-year estimates

➝

2.8

Website addresses Office for National Statistics: www.statistics.gov.uk General Register Office for Scotland: www.groscotland.gov.uk National Assembly for Wales: www.wales.gov.uk Northern Ireland Statistics and Research Agency: www.nisra.gov.uk

United Kingdom

Assumed TFRs

TFR

2000-based subnational population projections for Scotland, consistent with the new projections described in this article, were recently published by the General Register Office for Scotland. 4 1998-based subnational population and household projections for Wales were published at the end of 2001. 5 2000-based subnational projections for Northern Ireland will be published later this year by the Northern Ireland Statistics and Research Agency.

Table 1

Assumptions for 2000-based national population projections

Difference 000s %

80 Males

78

78

76

76

74

74

72

72

70

70 1981 1986 1991 1996 2001 2006 2011 2016 2021 2026 2031 2036

Year

Population at mid-1998 Components of change (1998–2000) Births Deaths

59,237

59,237

1,399 1,262

1,428 1,264

-29 -3

-2.0 -0.2

Natural change Net migration and other changes

137

163

-26

-

382

350

32

-

Total change

519

513

6

-

59,756 49,997 2,946 5,115 1,698

59,750 49,993 2,946 5,111 1,700

6 4 0 4 -2

0.01 0.01 0.01 0.1 -0.1

United Kingdom (c) Total net migration, 1982–83 to 2010–11 200

200 Assumptions

➝

Population at mid-2000 England Wales Scotland Northern Ireland

Thousands

175

175

150

150

125

125

100

100

75

75

50

50

25

25 0

0 1981 -25

National Statistics

6

1986

1991

1996 Year

2001

2006

2011 -25


Table 2

Spring 2002

Summary of assumptions for individual countries

Total fertility rate

2000–01

2005–06

2010–11

England Wales Scotland N. Ireland

1.64 1.69 1.48 1.76

1.71 1.74 1.55 1.78

1.74 1.75 1.58 1.80

UK

1.64

1.70

1.73

Net annual migration (thousands) from 2014–15

1998-based projection from 2011–12

2000–01

2001–02

from 2002–03

1998-based projection from 2001–02

1.75 1.75 1.60 1.80

1.80 1.80 1.75 1.85

157.5 5.0 -2.0 -0.5

144.0 7.0 -0.5 -0.5

128.0 7.5 0.0 -0.5

91.0 5.5 -1.0 -0.5

1.74

1.80

160.0

150.0

135.0

95.0

Expectation of life at birth (years) Males

Females 1998-based projection 2024–25

for 2024–25

1998-based projection

2000–01

2010–11

2020–21

England Wales Scotland N. Ireland

76.1 75.6 73.6 75.5

77.6 77.1 75.0 76.8

78.8 78.3 76.2 78.0

79.2 78.6 76.5 78.3

79.1 78.5 76.9 78.4

80.8 80.2 78.8 80.5

81.7 81.2 79.9 81.3

UK

75.8

77.4

78.6

78.9

78.9

80.6

81.5

Fertility Fertility assumptions are formulated in terms of the average number of children that women born in particular years will have. This cohort measure of fertility is more stable than the analogous calendar year or period measure (the total fertility rate), as it is affected only by changes in the total number of children women have and not by the timing of births within their lives. Period rates, in contrast, may rise or fall if births are brought forward or delayed for any reason. The assumed average completed family sizes and resultant total fertility rates (TFRs) are both shown in Figure 1, while the TFRs for individual countries are summarised in Table 2. The assumptions about completed family size are based on family building patterns to date and other relevant evidence. For the United Kingdom as a whole, completed family size has been falling steadily from an average of around 2.45 children for women born in the mid 1930s. The family sizes to be achieved by younger cohorts are highly conjectural, but for this projection it has been assumed that average completed family size, for the United Kingdom as a whole, will continue to decline until around the 1985 cohort and eventually level off at 1.74 children per woman. This is a slight reduction compared with the long-term average of 1.80 children per woman assumed for the 1998-based projections.6 Compared with the previous projections, the long-term fertility assumptions for both England and Wales have been reduced from 1.80 children per woman to 1.75. In Scotland, where current fertility levels are particularly low (the TFR fell below 1.5 in 2000), and differentials with the rest of the UK are now long established, the long-term assumption was reduced further - from 1.75 to 1.60. In Northern Ireland, fertility, although falling, remains higher than elsewhere in the UK and the long-term assumption was reduced from 1.85 to 1.80.

Mortality Future assumed improvements in mortality rates are based on the trend in mortality rates in the years up to 1999. It has been assumed that annual rates of reduction in mortality rates, which currently vary considerably from age to age, will tend towards a common reduction of 0.75 per cent a year by 2025. Thereafter, reductions are assumed to halve every ten years, i.e. to 0.375 per cent in 2035, etc.

2000–01

2010–11

2020–21

2024–25

for 2024–25

83.0 82.5 81.2 82.5

83.4 82.9 81.6 82.9

83.2 82.9 81.5 82.8

82.8

83.2

83.0

For the UK as a whole, life expectancy at birth, based on the mortality rates for the given year, is assumed to rise from 75.8 years in 2000–01 to 78.9 years in 2024–25 for males, and from 80.6 years to 83.2 years for females. However, as shown by Figure 1b, the rate of improvement is assumed to slow down gradually. Assumed expectations of life to 2024–25 for the individual countries are shown in Table 2. Further increases in expectation of life are projected for the longer-term, but in line with the gradually reducing improvements in mortality rates assumed after 2025. Current mortality levels differ between the individual countries. But, with one exception, the same future improvements in mortality rates have been assumed for all countries and, therefore, relative differences in levels of mortality between the four countries are maintained throughout the projection period. However, mortality for Scottish males at some ages (notably those in their 30s and, to a lesser extent, those over 50) has been improving more slowly than elsewhere in the UK in recent years and, for these projections, slightly smaller improvements in the period to 2025 have been assumed at the ages concerned. The assumed increase in expectation of life to 2025 for Scottish males is therefore slightly less than in the 1998-based projections. Otherwise, long-term assumptions of expectation of life at birth are generally very similar to those used for the previous projections.

Migration Figure 1c shows the recent trend in total net migration to the United Kingdom. Although there are considerable fluctuations in the series, the underlying trend has clearly been upward. There has been a particularly large rise in the last two years with net inward migration estimated in both 1998–99 and 1999–2000 to have been around 190,000. These were years of record net migration inflows into the United Kingdom. In the light of these recent data, the long-term assumption of net migration into the United Kingdom has been raised from 95,000 persons per year to 135,000. Although a very substantial increase, the new long-term assumption is well below the levels of the last two years. Partly taking account of provisional data for the latter half of 2000, a net inward flow of 160,000 has been assumed for 2000–01, reducing to 135,000 per year from 2002–03 onwards. 7 National Statistics


Spring 2002

The total net migration assumptions include international migration as measured by the International Passenger Survey (IPS) and those sources of migration not covered by the IPS (net migration from the Irish Republic, asylum seekers allowed to remain in the UK and people entering the UK as short-term visitors who are subsequently granted an extension of stay for a year or longer - “visitor switchers”). The increased long-term assumption mainly reflects an increase in the IPS component. The allowance for the number of asylum seekers allowed to remain in the UK is unchanged and is consistent with the latest assumptions made by the Home Office to estimate future public expenditure on asylum support. 7 Long-term annual net migration assumptions 2000-based projections Total +135,000 IPS +100,000 Irish Republic -5,000 Asylum seekers +20,000 Visitor switchers +20,000 Table 3

1998-based projections +95,000 +65,000 -5,000 +20,000 +15,000

The net migration assumptions for the constituent countries of the United Kingdom are shown in Table 2. These combine assumptions regarding the distribution of international migration with assumptions about cross-border migration between the countries of the United Kingdom. Compared with the previous projections, higher net inflows are assumed for both England and Wales and a long-term net outflow is no longer assumed for Scotland. The long-term assumption for Northern Ireland is unchanged.

RESULTS OF THE 2000-B ASED NATIONAL POPULATION PROJECTIONS Total population The results of the new projections are summarised for the constituent countries of the United Kingdom in Table 3 and Figure 2. The projections for the individual countries are carried forward for forty years, i.e. until 2040, but projections to 2070 for the UK as a whole are shown in Figure 2 and in some of the remaining graphs in this article.

Components of change: five-year summary, 2000–2025

Annual averages (Thousands) 2000–2001 United Kingdom Population at start Births Deaths

2001–2006

2006–2011

2011–2016

2016–2021

2021–2025

59,756 674 602

59,987 676 622

60,946 679 612

61,956 694 613

63,038 707 628

64,105 702 654

72 160

54 138

67 135

81 135

78 135

48 135

Total change Population at end

232 59,987

192 60,946

202 61,956

216 63,038

213 64,105

183 64,836

England Population at start Births Deaths

49,997 568 498

50,225 572 515

51,165 576 506

52,151 590 507

53,207 603 520

54,262 603 543

70 158

57 131

69 128

83 128

83 128

60 128


228 50,225

188 51,165

197 52,151

211 53,207

211 54,262

188 55,013

Wales Population at start Births Deaths

2,946 31 33

2,949 31 34

2,971 32 33

3,000 33 33

3,034 33 34

3,067 32 35

-2 5

-3 7

-2 8

-1 8

-1 8

-3 8


3 2,949

4 2,971

6 3,000

7 3,034

7 3,067

5 3,085

Scotland Population at start Births Deaths

5,115 53 57

5,109 52 58

5,078 50 57

5,047 50 57

5,014 50 58

4,973 48 60

-4 -2

-6 0

-6 0

-7 0

-8 0

-12 0


-6 5,109

-6 5,078

-6 5,047

-7 5,014

-8 4,973

-12 4,926

Northern Ireland Population at start Births Deaths

1,698 22 14

1,705 21 15

1,732 21 15

1,759 21 16

1,783 21 16

1,803 20 17

7 -1

6 -1

6 -1

5 -1

5 -1

3 -1

7 1,705

6 1,732

5 1,759

5 1,783

4 1,803

2 1,813

Natural change Net migration




Natural change Net migration Total change Population at end

National Statistics

8


The population of the United Kingdom is projected to increase gradually from 59.8 million in 2000 to reach nearly 65 million by 2025. This is equivalent to an average annual rate of growth of 0.3 per cent. Around two thirds of the projected 5.1 million increase between 2000 and 2025 is attributable to net inward migration. The remainder is due to projected natural increase (more births than deaths). Longer-term projections suggest the population will peak around 2040 at nearly 66 million and then gradually start to fall. Figure 2

Actual and projected population of the United Kingdom and constituent countries, 1951–2070

Millions 70

70

Projected

➝

65

65

Northern Ireland United Kingdom

60

Scotland

Wales

55

The population of Scotland is projected to continue to decline slowly from 2000, while the populations of Wales and Northern Ireland are projected to peak in around 30 years’ time and then start to fall. The population of England is still projected to be increasing at 2040, but at a very low rate of growth.

Births and deaths Projected numbers of births and deaths are shown in Figure 3. With the single exception of 1976, the United Kingdom gained population through natural increase (births less deaths) throughout the 20th century. However, it is projected that deaths will begin to outnumber births in about 30 years’ time. By 2040, this natural deficit is projected to exceed the assumed net gain to the population through net migration and so the population then begins to decline. Of course, projections so far ahead are subject to considerable uncertainty. In particular, the projected trend in births depends on the assumed future level of fertility (including that for women not yet born) and has much greater uncertainty attached to it than the projected trend in deaths which is largely determined by the age structure of the population alive today.

60

Age distribution

55

Table 4 and Figure 4 summarise the projected age structure of the population. The age structure will become gradually older with the median age of the population rising from 37.4 years in 2000 to 42.4 years by 2025. Longer-term projections show continuing ageing with the median age eventually stabilising around 44.5 years from 2040.

England 50

Spring 2002

50

45

45

40

40 1951 1961 1971 1981 1991 2001 2011 2021 2031 2041 2051 2061 Year

Figure 3

The number of children aged under 16 is projected to fall by 8.5 per cent from 12.1 million in 2000 to 11.0 million in 2011 and then to remain around this level to 2025. The number of people of working age (currently defined as 16 to 64 for men and 16 to 59 for women) is projected to rise by six per cent from 36.9 million now to 39.0 million in 2011. Allowing for the planned change in women’s state pension age from 60 to 65 between 2010 and 2020,8 the working age population would then rise further to 40.8 million by 2021. Figure 4

Actual and projected births and deaths, 1951–2070

Actual and projected age distribution, 1981–2070

United Kingdom

United Kingdom

Millions 1.1

% of total population 1.1

100

100 Projected

Projected

➝

1.0

➝

90

75+ 90

1.0

60–74

80

80

70 0.9

70

0.9

45–59 60

0.8

0.8

50

60 Median age

50 30–44

40

40

Births 0.7

0.7 30

30 15–29

20 0.6

20

0.6 Deaths 10 0.5

0.5 1951 1961 1971 1981 1991 2001 2011 2021 2031 2041 2051 2061 Year

10

0–14

0

0 1981

1991

2001

2011

2021 2031 Year

2041


2051

2061


Table 4

Spring 2002

Projected population by age, 2000–2025

United Kingdom

Thousands

Age Group

2000

2001

2006

2011

2016

2021

2025

All ages 0–14 15–29 30–44 45–59 60–74 75 & over

59,756 11,322 11,400 13,751 11,083 7,801 4,399

59,987 11,240 11,341 13,912 11,262 7,780 4,452

60,946 10,707 11,696 13,701 12,009 8,243 4,590

61,956 10,312 12,106 12,714 12,763 9,313 4,749

63,038 10,252 11,996 12,161 13,668 9,927 5,034

64,105 10,410 11,467 12,511 13,461 10,653 5,602

64,836 10,512 11,117 12,910 12,672 11,234 6,392

37.4

37.7

39.3

40.6

41.6

41.9

42.4

12,076 36,890 10,789

12,003 37,154 10,830

11,515 38,192 11,239

11,045 38,963 11,948

10,933 40,026 12,079

11,093 40,760 12,251

11,201 40,517 13,118

327 292 620

323 291 615

301 294 596

283 307 590

273 302 575

272 301 573

276 324 600

Median age (years) Under 16 Working age* Pensionable age* Dependants per 1,000 persons of working age Under 16 Pensionable age* Total*

* Working age and pensionable age populations based on the state pension age for given year. Between 2010 and 2020, state pension age will change from 65 for men and 60 years for women, to 65 years for both sexes.

Even ignoring the change in women’s state pension age, the working age population will become much older as the baby boom generations of the mid 1960s age. So, during the period to 2016, the number of adults aged under 45 is projected to fall slightly. However, the 45 to 59 age group is projected to increase by nearly 25 per cent over the same period, from 11.1 million in 2000 to 13.7 million in 2016.

Figure 5

Actual and projected dependency ratios, 1971–2051

United Kingdom Dependants per1,000 persons of working age 900

The number of people over state pension age is projected to increase by 11 per cent from 10.8 million in 2000 to 11.9 million in 2011. However, with the increase in women’s state pension age, the population of pensionable age will rise only slightly further (to 12.3 million) by 2021. A faster increase will then resume with longer-term projections suggesting the number of pensionable age peaking at just over 16 million around 2040. Without the change in women’s state pension age, the population of pensionable age would have risen to 14.3 million by 2021, eventually peaking at 18 million.

900

Projected

➝

800

800 Total 700

700

600

600

500

500

In 2000, there were nearly 1.3 million more children aged under 16, than people of state pensionable age. But the number of children is projected to decline, while those of pensionable age are projected to rise. As a result of these changes, from 2007 the population of state pensionable age is projected to exceed the number of children and by 2025 is projected to exceed it by nearly 2 million.

Dependency ratios

Pensionable age 400

400

300

300 Children under 16 200

200 1971

These changes in age structure will, in time, have a marked effect on the future proportion of dependants in the population. Figure 5 shows projected dependency ratios, i.e. the number of children under 16 or the population of state pensionable age (or the sum of the two) expressed as a proportion of the working age population. These are, of course, somewhat arbitrary boundaries as, in reality, full-time education ends, and retirement starts, at a range of ages. Figure 5 shows that the total dependency ratio will fall gradually from 620 dependants per 1,000 persons of working age in 2000 to about 570 per 1,000 in 2020 when the increase in women’s state pension age is complete. It will then increase rapidly, with longer-term projections suggesting a levelling off at nearly 700 per 1,000 from the mid-2030s. However, this would actually be slightly lower than the ratio in the early 1970s, although then it was children who comprised the majority of

National Statistics

10

1981

1991

2001

2011 Year

2021

2031

2041

2051

Note: The ‘working age’ population is that aged between 16 and state pension age, and the population of ‘pensionable age’ is that over state pension age. Between 2010 and 2020, state pension age will change from 65 years for men and 60 years for women, to 65 years for both sexes. The dotted lines in the graphs show what the dependency ratios would have been, had the present state pension ages applied throughout.

dependants. Of course, without the planned change in women’s state pension age, the proportion of dependants would rise earlier and further as indicated by the dotted lines in Figure 5. Population ageing will be experienced to a greater or lesser extent in all Western countries. Indeed, the latest Eurostat projections9 show that in the year 2020, compared with the EU as a whole, the UK will have proportionately fewer older people. However, the proportion of younger


Spring 2002

people is projected to be slightly higher and, therefore, the overall dependency ratio will be around the EU average.

where there is no increase to the long-term migration assumption to offset the fertility change, projected populations at 2025 are lower.

COMPARISON WITH 1998-BASED NATIONAL PROJECTIONS

The effect of the reduced fertility assumption on births is partly offset by the increased net migration assumption, which raises the number of women of childbearing age. Overall, there are just over 400 thousand (2.3 per cent) fewer births projected for the UK for the period 2000 to 2025 than in the 1998-based projections. The changes to the mortality assumptions are comparatively small, and there is therefore relatively little change to the projected number of deaths.

The projected total population of each country is compared with the 1998-based projections in Table 5 and the difference between the two projections is broken down into changes in the base population and changes in the projected number of births, deaths and migrants. Reductions in the projected numbers of deaths (as compared with the 1998-based projections) are shown as positive numbers in the table as they contribute to an increase in the size of the population.

The change in the projected size of the population of the United Kingdom in particular age groups is shown in Table 6. As migrants are heavily concentrated at younger ages, the increased net migration assumption leads to an increase in the projected population of working age. For example, the population aged 30–44 in the UK at 2025 is nearly 4 per cent higher than previously projected. The reduction in the fertility assumption leads to the projected population of children aged under 16 being nearly 2.5 per cent lower than in the previous

The projected population of the United Kingdom at 2025 is about 600 thousand (0.9 per cent) higher than in the previous projections. In England and in Wales, the effect of the higher net migration assumptions outweighs the effect of the reduced fertility assumptions and the projected total populations at 2025 are increased. In Scotland where assumed fertility has been reduced most, and in Northern Ireland

Table 5

Change in projected population compared with the 1998-based projections

Thousands Change due to Country

2000-based

1998-based

base

projected

projected

projected

projections

projections

Total change

population*

births

deaths**

migrants

Population at 2011 England Wales Scotland Northern Ireland United Kingdom

52,151 3,000 5,047 1,759 61,956

51,922 2,993 5,087 1,771 61,773

229 7 -41 -13 183

4 0 4 -2 6

-210 -11 -46 -11 -277

2 -1 -7 0 -6

433 19 9 0 460


54,262 3,067 4,973 1,803 64,105

53,715 3,047 5,058 1,821 63,642

547 20 -85 -18 463

4 0 4 -2 6

-262 -17 -93 -17 -388

1 -2 -14 1 -14

803 39 19 0 860


55,013 3,085 4,926 1,813 64,836

54,316 3,060 5,026 1,833 64,235

697 26 -101 -20 602

4 0 4 -2 6

-262 -19 -110 -19 -411

5 -3 -16 1 -13

951 47 23 0 1,020

* Difference between the estimated population at mid-2000 and the 1998-based projection of the population at mid-2000. ** Reductions in the projected number of deaths (compared with the previous projections) are shown as positive numbers as they contribute to an increase in the size of the population.

Table 6

Change in projected population by age compared with the 1998-based projections

United Kingdom

Thousands/percentages 2000

2011

2021

2025

Age group

000s

%

000s

%

000s

%

000s

%

Under 16 16–29 30–44 45–59 60–74 75 & over

-21 -35 41 13 7 2

-0.2 -0.3 0.3 0.1 0.1 0.0

-277 164 236 28 13 18

-2.4 1.5 1.9 0.2 0.1 0.4

-212 48 393 189 11 34

-1.9 0.4 3.2 1.4 0.1 0.6

-151 -37 473 252 21 44

-1.3 -0.4 3.8 2.0 0.2 0.7

6

0.0

183

0.3

463

0.7

602

0.9

All ages



Spring 2002

projections at 2011, although this differential later reduces as the higher projection of the number of women of childbearing age increasingly offsets the lower fertility assumption. In contrast, there is comparatively little change in the projected size of the population aged over 60 in the period to 2025.

SENSITIVITY ANALYSIS The one certainty when making population projections is that, due to the inherent unpredictability of demographic behaviour, they will turn out to be wrong as a forecast of future demographic events or population structure. One way of giving users an indication of uncertainty is by considering the performance of past projections. An analysis of the accuracy of national population projections made between 1971 and 1989 was published in Population Trends 77.10 Another way of illustrating uncertainty is by preparing variant projections based on alternative assumptions of future fertility, mortality and migration. The official 2000-based variant projections were not ready in time for inclusion in this article but will be published in early 2002 on the GAD website: www.gad.gov.uk. For the first time, these will include variant projections at individual country level in addition to the usual UK level variants. These will be discussed in an article in a future issue of Population Trends. To give an indication of uncertainty in the present article, Figures 6 and 7 show the total population of the UK and the percentage of the population aged over 65 under the variant assumptions used in the 1998-based national projections. Compared with the principal 1998based projection assumptions, the fertility variants assumed long-term family sizes of ± 0.2 children per woman. Projected life expectancy at birth at 2025 differed by about ± 1 year from the principal projection, and long-term annual net migration inflows were assumed to be 50,000 persons above and below the principal projection. These variant

Figure 6

Population of the United Kingdom according to principal and variant 1998-based projections, 1971–2051

assumptions were intended as plausible alternative scenarios and not as upper or lower limits for what might occur in the future. It is clear from Figure 6 that there is considerable uncertainty regarding the future size of the population. Under these alternative, but still plausible, fertility and migration assumptions, the population at 2025 differs from the principal projection by around ± 2 million (±3 per cent). The uncertainty widens with time and by the middle of the next century, the total population under the high and low fertility assumptions differs by over ± 5 million (± 8 per cent) from the principal projection. Figure 6 also shows that future population decline is not inevitable. Indeed, previous work has shown, for example, that a longterm family size of 2.0 children per woman coupled with net migration of 100,000 per year would lead to continuing population growth throughout the 21st century.11 However, while population decline is not inevitable, Figure 7 demonstrates that population ageing will occur under any plausible set of future assumptions. In 2000, under 16 per cent of the population were aged 65 and over. But, although higher fertility levels would reduce population ageing, even the high fertility variant in the 1998based projections produced an increase to 20 per cent by 2025 and 22 per cent by 2050. And in the low fertility or high life expectancy variants, the proportion would be over one in four by 2050. The inevitability of population ageing is a consequence of the current age structure of the population. This, in turn, is a result of changes in the past numbers of births. Thus, during the first half of this century, the number of elderly people will rise as the relatively large cohorts born after the Second World War and during the 1960s baby boom replace at older ages the much smaller cohorts born before 1945. Conversely at younger ages, the relatively small cohorts born since the mid 1970s will replace the baby boomers.

Figure 7

Proportion of the population aged 65 or over according to principal and variant 1998-based projections, 1981–2051

United Kingdom Millions 70

Projected

➝

High fertility

68

Percentage 27.5

Projected

➝

High migration

Low fertility High life expectancy Low migration Principal High migration Low life expectancy High fertility

25.0 66

High life expectancy

64

Principal

22.5

Low life expectancy

62

Low migration

20.0

60 Low fertility

17.5

58

15.0

56

54 1971

1981

1991

2001

2011 Year

2021

National Statistics

2031

2041

2051

12.5 1981

12

1991

2001

2011 2021 Year

2031

2041

2051


NATIONAL P OPULATION PROJECTIONS ON THE INTERNET Full details of the results of the 2000-based national population projections for the United Kingdom and its constituent countries are available on the GAD website www.gad.gov.uk. Tables and charts for the individual countries, corresponding to Figures 1, 3 and 4 and Tables 4 and 6 in this article, are also available on the website.

REVIEW OF METHODOLOGY FOR PROJECTING MORTALITY GAD has recently completed a review of the methodology used for projecting mortality in national population projections. 12 This review was undertaken as part of the National Statistics Quality Assurance programme. Some of the recommendations from this review were applied in determining the mortality assumptions for the 2000-based projections.

Key findings ●

The United Kingdom population is projected to increase gradually from an estimated 59.8 million in 2000 to reach nearly 65 million by 2025, equivalent to an average annual rate of growth of 0.3 per cent. Longer-term projections suggest the population will peak at nearly 66 million around the year 2040 and then gradually start to fall.

●

The population of Scotland is projected to continue to decline slowly from 2000, while the populations of Wales and Nor thern Ireland are projected to peak in around thirty years’ time and then start to fall. The population of England is still projected to be increasing in forty years’ time, but at a low rate of growth.

●

Around two thirds of the projected 5.1 million increase in the UK population between 2000 and 2025 is attributable to the assumed level of net inward migration. The remainder is due to projected natural increase (more births than deaths).

●

The population will gradually become older with the median age projected to rise from 37.4 years in 2000 to 42.4 years by 2025.

●

By 2007, the population of state pensionable age is projected to exceed the number of children.

●

The number of people of state pensionable age is projected to increase by 11 per cent from 10.8 million in 2000 to 11.9 million in 2011. Allowing for the change in women’s state pension age, the population of pensionable age will rise only slightly further (to 12.3 million) by 2021. However, a faster increase will then resume with longer-term projections suggesting the number of pensionable age peaking at just over 16 million in 2040.

●

The projected total population of the United Kingdom in 2025 is about 600 thousand (0.9 per cent) higher than in the previous (1998-based) projections. This is a result of an increased assumption of future net inward migration, partly offset by a lower assumption of future fertility.

Spring 2002

REFERENCES 1. Office for National Statistics (2000). National population projections: 1998-based. Series PP2 no. 22. The Stationery Office: London. Can be downloaded from www.gad.gov.uk 2. Available from the National Statistics website at www.statistics.gov.uk/downloads/theme_population/PENo3/ PENo3_v3.pdf 3. Office for National Statistics (1999). 1996-based subnational population projections: England. Series PP3 no. 10. The Stationery Office: London. Can be downloaded from www.statistics.gov.uk/ onlineproducts 4. General Register Office for Scotland (2002). Population Projections Scotland (2000-based). GROS Edinburgh. 5. National Assembly for Wales (2001). 1998-based Population and Household Projections for Wales. NAW: Cardiff. Available from the NAW website at www.wales.gov.uk/ keypubstatisticsforwalesheadline/content/population/2001/ headline_population.htm 6. Shaw C (2001). Assumptions for the 2000-based national population projections. Population Trends 105, pp.45–47. 7. Asylum seeker support - estimates of public expenditure (1999). Home Office: London. 8. Pensions Act 1995 Chapter 26 Part II, Section 126 and Schedule 4. 9. Shaw C, Cruijsen H, De Beer J and De Jong A (1997). Latest population projections for the European Union. Population Trends 90, pp. 18–30. 10. Shaw C (1994). Accuracy and uncertainty of national population projections for the United Kingdom. Population Trends 77, pp. 24– 32. 11. Shaw C (2001). United Kingdom population trends in the 21st century. Population Trends 103, pp. 37–46. 12. Government Actuary’s Department (2001). National Population Projections: Review of Methodology for Projecting Mortality. National Statistics Quality Review Series, Report No.8. Available from the National Statistics website at www.statistics.gov.uk/ methods_quality/quality_review/population.asp



Spring 2002

Improving ONS spatial referencing – the impact on 2000 births and deaths data

Elspeth McVey, Allan Baker Office for National Statistics

This article describes the improvements in the way that ONS will be allocating births and deaths data to administrative and health areas in England and Wales, beginning with data for 2001. Births and deaths registered in 2000 are used to examine the differences in area allocations at ward, local authority and health authority levels, which this change will introduce. Differences in the geographical characteristics of births and deaths are also described.

OVERVIEW The Office for National Statistics (ONS) routinely aggregates vital events such as births and deaths to administrative areas and health areas in England and Wales. The postcode of usual residence recorded on the birth or death certificate is used to assign an event in an electoral ward, local authority, health authority area or other standard area. Recent improvements in the accuracy and availability of postcode grid references mean that ONS can take advantage of more automated and reliable techniques of assigning events to defined areas. However, there will be resulting discontinuity in time series, beginning with data for 2001. In order to assess the impact of this change, analysis was carried out on dual coded data for live births and deaths registered in the year 2000 in England and Wales. The main findings show that smaller areas, such as electoral wards, have the greatest differences in the number of events allocated to them compared to the larger local authority or health authority areas. There is less chance for differences to cancel themselves out over smaller areas. Each data source has a distinctive profile in the types of differences observed. Births tend to use more recently introduced postcodes, and under the old system these were less reliable in their allocation. However, this only affects a small proportion of births overall. Deaths tend to cluster in their postcode allocation more than births, and this is most likely to be explained by nursing and residential homes being recorded as the place of usual residence. Some of the issues, are discussed in more detail to help users of births and deaths data to understand the differences that may be observed at local level.

INTRODUCTION Geography is a key dimension in National Statistics. Administrative areas, such as local authority districts and health authorities provide a structure for the aggregation, analysis and publication of many kinds of National Statistics National Statistics

14


social and economic data. However, the boundaries of these areas keep changing for various reasons, making the management of comparable statistics over time very difficult. Tying data to points based on grid references rather than fixed areas allows greater flexibility in managing boundary change. In addition to providing a precise location, it also allows the possibility of aggregating data to areas other than standard administrative ones. The availability of the Gridlink postcode data set, which gives an Ordnance Survey grid reference accurate to 1 metre resolution for every live postcode in the UK, makes this possible. Boundary change can be managed separately using digital boundary data and Geographical Information Systems (GIS) techniques for aggregating data to areas.

The usual method of assigning vital events data to administrative areas involves taking the postcode from the birth or death certificate and comparing it against a central postcode directory: an example is shown in Figure 1. For each current or historical postcode in the UK, there is a record containing all the codes for the current administrative areas in which it falls. The electoral ward is usually the first area to be found for a postcode, as in most cases higher level administrative areas can be made up from groups of electoral wards. Postcode to ward assignment is done on a best fit basis, as postcodes are designed for mail delivery and form a collection of delivery points, rather than having definitive boundaries. It is possible that a part of a postcode may straddle a boundary, but it will be assigned according to the location of the first address in the postcode. On average there are 15 addresses in one postcode, 1 however, in some cases there may be many more. Since the introduction of the central postcode directory in the early 1980s, extensive checks have been conducted on data quality and a high proportion of the assignments of postcodes to administrative areas have been clerically checked using paper maps and street directories. However, with over 2.2 million records to manage, the extent of change over time in the postcode geography of the UK has made it necessary in recent years to assign new postcodes to administrative areas by automated imputation based on neighbouring postcodes. New postcodes are regularly introduced by the Royal Mail due to new properties being built, or as a result of postcode reorganisation. As

Example of postcode lookup for a death record

Deceased’s Postcode

Date of death

Sex of Deceased

Cause of death

DOB of deceased

PO16 7EA

04102000

1

12345

17021919

EH12 3PG

04102000

2

23456

30071911

SO31 6DE

04102000

1

34567

12061976

LA Ward

Health Area

PO16 6DZ

ABCDEF

1234

› PO16 7EA

ABCDEF

1234

PO16 7EB

ABCDEF

1234

The examples below show how a new postcode is assigned in the traditional system. In the first example, the postcodes before and after the new postcode have been clerically assigned and therefore have a status indicator of 0. The existing postcodes are both in the same ward. The new postcode ‘PO12 1QB’ will also be assigned to this ward, and because the imputation has been based on clerically assigned postcodes, the status indicator will be 1 (most reliable imputation). In the second example, the postcodes before and after the new postcode have themselves been imputed. The new postcode ‘PO12 1QF’ will inherit the ward assignment of the neighbouring postcode with the more reliable imputation, in this case ‘PO12 1QD’. The status indicator of the new postcode assignment will be 2, because it has been based on a previously imputed postcode with a status indicator of 1. Example 1: Assign postcodes either side of the new postcode Before imputation: postcode date of introduction PO12 1QA 198001 PO12 1QB 200107 PO12 1QC 198001

Before imputation: postcode date of introduction PO12 1QD 198001 PO12 1QF 200107 PO12 1QM 198001

PO16 7EA

04102000 1

17021919

EH12 3PG

04102000 2

23456

30071911

SO31 6DE

04102000 1

34567

12061976

Health Area

ABCDEF 1234

›

Sex of Cause of DOB of LA Deceased death deceased Ward 12345

USING IMPUTATION

ward code 24UFFA

status indicator 0

24UFFA

0

ward code 24UFFA 24UFFA 24UFFA

status indicator 0 1 0

Example 2: Imputed postcodes either side of a new postcode .

add area codes to database Deceased’s Date of Postcode death

Box one

After imputation: postcode date of introduction PO12 1QA 198001 PO12 1QB 200107 PO12 1QC 198001

Look up the postcode Postcode

part of the regular postcode maintenance, some postcodes may be terminated, but area assignments are maintained for these postcodes by ONS to enable historical data to be aggregated for time series analyses. Box 1 shows how a new postcode is assigned to an electoral ward based on the postcode immediately before and after it in the directory. When a postcode has been assigned to an electoral ward by imputation, a status indicator is added to the record to indicate how reliable it is believed to be.

ASSIGNMENT OF A NEW POSTCODE TO A WARD

Traditional postcode directory

Figure 1

Spring 2002

After imputation: postcode date of introduction PO12 1QD 198001 PO12 1QF 200107 PO12 1QM 198001

ward code 24UFFA

status indicator 1

24UFFC

4

ward code 24UFFA 24UFFA 24UFFC

status indicator 1 2 4



Spring 2002

THE G RIDLINK POSTCODE DIRECTORY A directory based on postcode location data made available via the Gridlink® Consortium is replacing the traditional postcode directory. The aim of this Consortium is to provide a reliable and consistent postcode location data set that meets agreed national standards and can be used in the core products of its members. These include Ordnance Survey (OS), Ordnance Survey Northern Ireland (OSNI) and the General Register Office for Scotland (GROS), ONS and the Royal Mail.2,3 The OS maintains the Gridlink Master Database, which contains core postcode data based on Royal Mail’s Postcode Address File (PAF®) and Postzon products. National Health Service and administrative area names and codes are provided by ONS. Postcodes are allocated to these areas using Boundary-Line and Land-Line® data and GIS Point in Polygon techniques.1 The new ONS Gridlink directory will therefore be based on the Gridlink Master database. A list of the fields available in Gridlink is shown in Box 2.2 It will include an accurate grid reference for all postcodes, which existed from the year 2000 onwards, and a positional quality indicator field to show the accuracy of the location (a full list of the indicators is shown in Box 3).2 The postcode grid references will be based on the postcode mean, which is explained in Box 4, with the most accurate possible to 1 metre resolution.

Box four DERIVING THE POSTCODE MEAN The grid references are derived from the Ordnance Sur vey Address-Point database, which provides a precise geographic reference for each of the 26 million addresses in Great Britain to a resolution of 0.1 metres. 1,4 As there can be a number of delivery points in a postcode, the grid reference used for the postcode is calculated as the arithmetic mean of all the Address-Point grid references and then “snapped” to the closest Address-Point grid reference. This is to ensure that the grid reference remains within the postcode in cases where the area covered is irregular. The map below shows the mean locations of postcodes in a collection of streets. Mean postcode grid reference in each street

Box two GRIDLINK® C ORE DATA Unit postcode Grid reference to 1m resolution Positional quality indicator for grid reference UK country codes Administrative Electoral ward codes Administrative Local authority district codes Administrative Unitary authority codes Administrative County codes Health regional office area codes Health area codes

Box three GRIDLINK POSITIONAL QUALITY INDICATORS 1 Automatically calculated to be within the building of the postcode mean 2 As for status 1 except for visual inspection of Landline maps 3 Approximate to within 50 metres of true position 4 Postcode unit mean (mean of matched address with the same postcode, but not snapped to building) 5 Postcode imputed by ONS by reference to surrounding known postcodes 6 Postcode sector mean (mainly PO Boxes) 8 Postcode terminated (no postcodes of this type provided by Gridlink) 9 No co-ordinates available Note: No. 7 not in use.

National Statistics

16

The ONS will still maintain any postcodes that were terminated before the Gridlink data was introduced. Where possible, Gridlink information has been mapped back to postcodes that were terminated as a result of name change only. This will leave a small proportion that will continue to be assigned clerically or by imputation due to the unavailability of accurate grid references. Any postcodes terminated since Gridlink was set up will be included in the data set with the last known grid reference. The underlying method of referring to a postcode directory to allocate each birth or death record to a ward, local authority, etc, has not changed, but the accuracy of the area assignments of each postcode, and especially for new postcodes, has improved. The OS have a program of continual survey where new properties and corresponding postcodes are continually kept up to date. In addition to providing a precise location which is not affected by boundary change, the advantage of grid referenced postcodes is that they can be plotted and aggregated to areas automatically using Geographical Information Systems (GIS) and digital boundary data as shown in Figure 2. This has major benefits in saving time involved in clerical checking and providing a consistent and reliable approach to area management.


Figure 2

Grid referenced postcodes and digital boundaries - GIS ‘point in polygon’ selection

Spring 2002

RESULTS Table 1 shows the breakdown of births and deaths by the respective quality indicators (the positional quality indicator of the grid reference in the Gridlink directory and the reliability of the assignment indicator in the Traditional directory). Around 99 per cent of the postcodes for both births and deaths had the highest score for positional accuracy in the Gridlink directory, i.e. the grid reference had been automatically calculated to be within the building closest to the postcode mean. It was not within the scope of this analysis to test this. Rather, the Gridlink postcodes are accepted as providing an accurate location for almost all of the postcodes referenced by the data. The same postcodes referenced in the traditional directory showed that 86 per cent of births records used clerically assigned postcodes with 14 per cent imputed. Deaths had a higher percentage of postcodes that had been clerically assigned, 91 per cent, with the remaining 9 per cent imputed. Comparing the two sets of ward codes (assigned to the births and deaths records from both postcode directories) showed that there was a good match where postcodes had been clerically assigned in the traditional directory. The match rate was 97 per cent for both births and deaths. This dropped to 79 per cent where postcodes had been imputed for births and 88 per cent for deaths.

2000 BIRTHS AND DEATHS DATA Methods Gridlink postcode data first became available in November 2000 and the first opportunity to test it was using births and deaths registered in 2000. Using the postcode from each record in these two data sources, the area codes for electoral wards, local authorities and health authorities were found in both the Traditional and Gridlink postcode directories and added to each birth and death record. The total live births and total deaths were then found for each area according to the traditional directory and compared with the totals for the same areas according to the Gridlink directory. The numeric differences and percentage differences were calculated for each area. The best way of finding explanations for the differences observed was to plot the Gridlink grid references from the data and visually inspect their locations against the boundaries of the areas concerned using MapInfo GIS. Background raster maps showing detail such as housing, land use and road layouts were also used to provide a geographical context for the data. Once the data had been plotted it was possible to interpret the main cause of the differences. However, it was only practical to do this for a small selection of areas. Additional information on each postcode including the year that it was introduced, whether it had been clerically assigned or imputed in the traditional directory, and the positional quality indicator of the grid reference from the Gridlink directory were also included to help in the analysis. As may be expected with an average of 15 addresses to a postcode, each postcode attached to a birth or death record is not used exclusively. A total of 604,130 live births records registered in 2000 used 385,409 postcodes; a ratio of 1.6 births to a postcode. There were 536,527 deaths records and these used 332,120 postcodes, with the same ratio of 1.6. However the number of times a postcode was referenced varied in some circumstances.

Tables 2 and 3 show how the differences affect the allocations at health authority, local authority and ward level, for births and deaths respectively. Differences were calculated by subtracting the traditional total from the Gridlink total; positive differences indicate that the traditional system has under-allocated events and negative differences that it over-allocated events. Percentages differences were calculated as a percentage of the traditional allocation to reflect changes from the current system. Health authorities had the smallest percentage differences and were the largest in size of the three levels. Percentage differences were largest at ward level, which was the smallest of the three areas. At all levels, positive and negative differences were approximately equal. For births, almost half of all health authorities (48.6 per cent) had no difference and none had more than 2 per cent difference in either direction. Of the 377 local authorities, 17.0 per cent had no difference and only 1.3 per cent had more than 5 per cent difference. At ward level, over half (58.8 per cent) had differences of less than 5 per cent, but for a minority there were differences as large as 50 per cent or over. For deaths, more than half of health authorities (55.2 per cent) had no difference and none had more than 2 per cent difference. Among the local authorities, 24.4 per cent had no difference and only 0.3 per cent had a difference greater than 5 per cent. At ward level, two-thirds (69.8 per cent) had differences of less than 5 per cent. At all levels, therefore, there was less overall difference in the allocation of deaths than for births. Table 1

Percentage of postcodes referenced by births and deaths having highest quality indicator, traditional and Gridlink systems, 2000

England and Wales

Births Deaths

Records

Gridlink: highest positional quality indicator

Traditional: clerically assigned

Number

Per cent

Per cent

604,130 536,527

99.5 98.7

85.6 90.7



Table 2

Spring 2002

Percentage of distribution of differences1 between traditional and Gridlink allocation of birth registrations to health authorities, local authorities and wards2, 2000 Per cent3

England and Wales Difference

Health authorities

50 and over Less than 50 Less than 25

0.0 0.0 0.0

0.0 0.0 0.0

0.2 2.0 8.7

Less than 10 Less than 5

0.0 25.7

0.5 40.6

9.3 20.1

No difference

0

48.6

17.0

18.8

Positive difference (Gridlink greater)

Less than 5 Less than 10 Less than 25 Less than 50 50 and over

25.7 0.0 0.0 0.0 0.0

41.1 0.8 0.0 0.0 0.0

19.9 9.4 7.8 2.7 1.0

Negative difference (traditional greater)

Local authorities

Wards

1

Difference expressed as a percentage of traditional allocation. Excluding wards with less than 20 births allocated by either system. 3 Percentages may not sum to 100 because of rounding. 2

Table 3

Percentage of distribution of differences1 between traditional and Gridlink allocation of death registrations to health authorities, local authorities and wards 2, 2000 Per cent3

England and Wales Difference

Health authorities

Local authorities

Wards

Negative difference (traditional greater)

50 and over Less than 50 Less than 25 Less than 10 Less than 5

0.0 0 0.0 0 25.7

0.0 0 0.0 0 37.9

0.1 1.2 5.6 8.2 21.9

No difference

0

55.2

24.4

25.6

Positive difference (Gridlink greater)

Less than 5 Less than 10 Less than 25 Less than 50 50 and over

19.0 0.0 0.0 0.0 0.0

37.4 0.3 0.0 0.0 0.0

22.3 7.6 5.4 1.6 0.6

1 2 3

Difference expressed as a percentage of traditional allocation. Excluding wards with less than 20 deaths allocated by either system. Percentages may not sum to 100 because of rounding.

Tables showing the number of births and deaths allocated to each health authority and local authority by the two systems, and the numeric and percentage differences, are available on the National Statistics website www.statistics.gov.uk

Effects of the postcode geography of births Closer examination of the patterns and differences in allocation of vital events shows the effect of the different postcode geography of births and deaths.

The overall differences reported are net differences. A small net difference may hide a larger number of changes (gains and losses between neighbouring areas) which have cancelled each other out, and even if an area has the same number of events allocated by both systems, these are not necessarily exactly the same records. For example, if a ward contained 10 events according to the traditional allocation and 10 events according to the Gridlink allocation, there might only be 8 postcodes common to each both. The 2 postcodes that could not be matched in either allocation (involves 4 postcodes altogether) have come from other areas.

The greater difference between traditional and Gridlink systems for births reflects the fact that a higher proportion of the births postcodes were created in the 1990s, when they were more likely to be less reliably imputed, compared to the postcodes used by deaths. Table 4 shows a summary of the year in which postcodes referenced by births and deaths were introduced and whether they were imputed (with high or low reliability) or clerically assigned. The year 1980 is shown separately as this is when the postcode file was started and also when the majority of the postcodes date from; 75 per cent for births and 82 per cent for deaths use postcodes introduced in this year.

Although some wards had no differences in the net number of births, only three-quarters of those wards contained exactly the same records. Similarly, in wards that had no difference in the net number of deaths, only four-fifths of those wards contained exactly the same records. In larger areas, such as health authorities, most changes occurred between wards within the overall area, so that net differences at this level are caused mainly by changes in the allocation of postcodes that lie close to the boundaries.

Table 5 shows that there is a higher proportion of imputed postcodes in wards with larger differences for births than deaths. For example, the percentage of imputed postcodes in wards with no difference was 11 per cent, while the percentage of imputed postcodes in wards with the largest differences was 33 per cent. For deaths the proportion of imputed postcodes in wards with larger differences did not increase so much. For example, in wards with no difference there were 8 per cent

National Statistics

18


Table 4

Spring 2002

Percentage of new postcodes which were imputed, 1980–2000, and percentage of births and deaths 2000 by year of postcode introduction Per cent1

England and Wales Births Proportion clerically assigned

1980 1981–1985 1986–1990 1991–1995 1996–2000 1

99.75 88.82 85.52 49.43 0.73

Deaths

Proportion of births with imputed postcodes

Proportion clerically assigned

High reliabilty

Low reliability

Percentage distribution of births by year of postcode

0.2 10.83 12.91 25.89 76.51

0.05 0.35 1.57 24.68 22.76

75 3 3 11 8

Percentage of imputed postcodes in ward1 by percentage difference2 between traditional and Gridlink allocations, births and deaths, 2000

England and Wales

Per cent Postcodes which were imputed

Difference at ward level 0 less than 10 less than 20 less than 50 50 and over

2

High reliabilty 99.66 83.89 81.56 55.78 0.62

Low reliability

Percentage distribution of deaths by year of postcode

0.07 0.68 1.42 23.05 8.61

82 3 3 7 5

0.27 15.42 17.02 21.17 90.77

Percentages may not sum to 100 because of rounding.

Table 5

1

Proportion of deaths with imputed postcodes

Births

Deaths

11 13 19 27 33

8 9 11 12 12

All wards included. Positive and negative differences counted together.

Figure 3

imputed postcodes whereas this only rose to 12 per cent for wards with over 50 per cent difference. Postcodes with a more recent date of introduction indicate areas of new building development or postcode re-organisation as a consequence of pressure on existing postcodes. The association between births and more recent postcodes, and hence the pattern of differences between traditional and Gridlink allocations for births, is likely to be because young families are more likely to occupy areas of new housing. An illustration of this is found in the neighbouring wards of Sarisbury in Fareham and Wickham in Winchester, both in the County of Hampshire. Wickham ward had 149 births allocated using the traditional method, but only 70 using the Gridlink method. Sarisbury ward had 56 births allocated using the traditional method but 150 using the Gridlink method. Figure 3 shows the distribution of births postcodes around the border of Fareham and Winchester. The births have been plotted using the grid references from the Gridlink system, but the codes are from the

Births around the boundary between Fareham and Winchester, plotted using Gridlink grid references and coded by traditional allocation, 2000

WINCHESTER DISTRICT (Wickham ward)

Births in Wickham (according to traditional allocation) 4 to 8

FAREHAM DISTRICT (Sarisbury Ward)

3 2 1

Births in Sarisbury (according to traditional allocation) 3 to 5 2 1

Ward boundary District boundary



Figure 4

Spring 2002

Area of new housing development in Fareham and Winchester

traditional system: it can be seen that a large number of births correctly located in Fareham by Gridlink have been allocated by the traditional system to Winchester. Figure 4 is a 1:10,000 map at the centre of which is an area of new housing development near Swanwick, which straddles the boundary between the two districts. The 32 postcodes representing this area accounted for 79 births. Of these postcodes, 27 had been introduced since 1995 and 21 had status indicators indicating imputation of the lowest reliability. All of these postcodes had the highest quality positional indicators in the Gridlink directory. The remaining difference in Sarisbury was due to improved locations in other parts of the ward.

Effects of the geography of deaths

This greater geographic concentration of deaths points to the influence of communal establishments, primarily nursing and residential homes. Where the change from traditional to Gridlink allocation moves a communal establishment between wards, a large number of deaths may be re-allocated between wards and higher level aggregations accordingly. An illustration of this is found in Worcestershire health authority, which was the health authority with the largest negative difference (that is the number of deaths allocated by Gridlink was smaller than that allocated by the traditional method) between the two systems of allocations. Figure 5 shows deaths in Worcestershire: the size of the symbols reflects the number of deaths at one postcode. Circles in black represent

A second difference between the geography of births and deaths is that relatively large numbers of deaths may occur in the same postcode. Table 6 shows the number of birth and death records and percentage distribution of births and deaths, by the number of events sharing one postcode. A similar proportion of birth and deaths occurred in post-

Table 6

codes where there was only one event (42 per cent and 44 per cent respectively). However, 16 per cent of deaths compared with only 4 per cent of births were in postcodes shared by more than 5 events, and some 3 per cent of deaths were in postcodes where more than 20 deaths occurred in the year.

Vital events sharing a postcode, births and deaths, 2000

England and Wales Number of events sharing a postcode

Births

1 up to 5 up to 10 up to 20 up to 30 up to 40 up to 50 more than 50 1

Number

Per cent1

Number

Per cent1

250,662 330,390 22,079 999

41.5 54.7 3.7 0.2

236,840 213,115 37,680 34,159 10,408 2,815 758 752

44.1 39.7 7.0 6.4 1.9 0.5 0.1 0.1

Percentages may not sum to 100 because of rounding.

National Statistics

Deaths

20


Figure 5

Spring 2002

Deaths in Worcestershire health authority, plotted using Gridlink grid references and coded by traditional allocation, 2000 South Staffordshire

Dudley Birmingham

Solihull Coventry

Shropshire

Warwickshire Worcestershire

Deaths 'lost' to neighbouring HAs 27 1 to 2

Herefordshire

Deaths 'gained' from neighbouring HAs 2 1 Deaths allocated to Worcestershire HA (no difference) 5 to 34

Gloucestershire

deaths that were allocated to Worcestershire by both systems. Circles in dark colour show the location of deaths at postcodes that which were ‘lost’ to neighbouring health authorities by changing to the Gridlink system, while circles in lighter colour represent the smaller number of deaths ‘gained’ in the change. The largest concentration of deaths ‘lost’ occurred in a postcode at the border with Birmingham health authority. There were 27 deaths allocated to this postcode, 24 of which occurred in a nursing home.

DISCUSSION Implementing the Gridlink postcode directory results in more accurate allocation of vital events to geographical areas. However, it does introduce some discontinuity in the time series, which affects smaller areas such as wards more than larger areas such as health authorities. This makes it important to be aware of whether the change in an area from one year to the next can be due to improved location information or to some other cause requiring more detailed analysis. Analysing change over time is not straightforward in any case, as boundary changes are frequent, especially at ward level. Ward boundaries are altered to reflect changes in the population, and hence differences caused by moving to the Gridlink system comprise only one component in change over time. The extent of differences at local authority level or health authority is not expected to have an impact on the production of standard rates. The results presented in this paper only show the first year in which it was possible to compare the two methods of assigning vital events data. The differences observed for births and deaths in 2000 do not in themselves provide a basis for a ‘conversion factor’ or other means for a simple conversion of historical data to the new basis. What is clear though, is that the time spent clerically checking the original directory has been worthwhile, as shown by the agreement of the postcodes used by births and deaths with the improved Gridlink version. This gives some assurance for historical allocations of events data.

4 3 2 1

The main problems lie with postcodes that have been introduced in more recent years and have not been able to be clerically checked. However, this only affects a very small proportion of records, both for births and deaths. It is not always the case that areas with a high number of new postcodes will have discrepancies, as it is only those postcodes occurring near to a boundary that will give rise to differences. Within larger areas in particular, the chance for differences to cancel each other out is much greater. The number of events occurring at one postcode is also an important consideration. It has been shown that deaths occur in the same postcode many more times than births, due to the effect of nursing and residential homes. Improving the location of just one postcode, especially if it crosses a boundary, can therefore have a considerable impact on the number of events seen to occur in an area. However, the postcode will not always be used exclusively by one communal establishment; some other premises in the same street may share the postcode. The only way to reliably identify individual properties is by the full address including street number and name. This would be possible using the Address-Point™ which is available from Ordnance Survey and contains a grid reference and unique identifier for every property in Great Britain. Address-Point™ enables events to be located individually rather than by using the mean location for several addresses and therefore overcomes the problem of postcodes straddling boundaries. As a next step in improving spatial referencing, ONS plans to use AddressPoint™ as the means of allocating individual events to areas for data from 2003. This will introduce another methodological change, requiring further analysis to determine its impact on births and deaths data. The availability of accurate area directories, based on postcodes or addresses will enable historical data to be aggregated to current boundaries. Since postcode is held on most records, this will remain as the key spatial reference. Vital events records with historical postcodes



Spring 2002

which do not have Gridlink data, will not be as accurate, however, and it will be necessary to rely on the traditional means of assigning them to areas. Postcode and corresponding grid references may be stored with each vital events record to manage boundary change more easily in future. Grid references together with digital boundary data can be used to aggregate data to new geographies, for example, environmental areas or neighbourhoods. Confidentiality is a very important factor to consider when data becomes more flexible and, although not dealt with in this article, will not be compromised by ONS in any of these improvements.

CONCLUSIONS The implementation of the Gridlink postcode directory results in more accurate allocation of vital events to geographical areas, as well as greater flexibility to allow for boundary changes. Further improvements in both accuracy and flexibility can be expected with the introduction of Address-PointTM in 2003. The accuracy of allocation takes on greater importance with the increasing demand for small area statistics, for example through the Neighbourhood Statistics Service. The impact of the changes can be readily observed in differences to the allocation of births and deaths at ward level. Where there are large net changes, these tend to be associated with concentrations of births in areas where there are more recent postcodes, and concentrations of deaths in postcodes which include residential and nursing homes. However, both of these effects are important only where they occur near a boundary between areas. At the level of health and local authorities, ward-level differences tend to cancel each other and the overall effect of the change is relatively small. No health authority had more than a 2 per cent difference between the traditional and Gridlink systems in the allocation of either births or deaths. A difference of more than 5 per cent between the two

Key findings ●

●

●

●

●

●

●

2001 bir ths and deaths data will be allocated by ONS to administrative and health areas using Gridlink, a more accurate system of postcode location than was previously available . The more accurate geographical location of vital events at addresses near the boundaries between areas will in some cases cause them to be allocated to a different area than under the previous system. Analysis of births and deaths in 2000 showed large differences between the two systems of allocation at ward level, but much less for local and health authorities. No health authority had more than 2 per cent difference for either births or deaths. Less than 2 per cent of local authorities had more than 5 per cent difference for either births or deaths. Differences in the area allocation of births were particularly associated with areas of recent building development, which might house a disproportionate number of young families. Differences in the area allocation of deaths were particularly associated with nursing and residential homes located near area boundaries.

National Statistics

22

systems was found in only 1.3 per cent of local authorities for births, and 0.3 per cent for deaths. The use of local knowledge to examine observed differences in the light of the effects described in this article is likely to be a fruitful approach in determining whether change at small area level is due to improved spatial referencing or some other cause.

REFERENCES 1. Ordnance Survey. Code-Point® user guide, v1.3 - 11/99 available from www.ordnancesurvey.co.uk 2. Gridlink ® based version of the ‘All fields’ postcode directory. 2001 Userguide available from ONS Geography Customer Support Unit [email protected] Tel. 01329 813243 3. More information on Gridlink is available at www.ngdf.org.uk/uksgb/gridlink 4. Ordnance Survey. ADDRESS-POINTTM user guide v.30 - 7/2001. Available from www.ordnancesurvey.co.uk


Spring 2002

Implications of changes in the UK social and occupational classifications in 2001 for vital statistics Angela Donkin, Yuan Huang Lee, Barbara Toson Office for National Statistics

I NTRODUCTION Registrar General’s Social Class (RGSC) has been the principal classification of socio-economic status used in the UK since its first appearance in the Registrar General’s Annual Report for 1911.1 Analysis by RGSC has consistently shown social gradients in health, and particularly in mortality at working ages, infant mortality, and birthweight.2 From 2001, RGSC will be replaced by the new National Statistics Socio-Economic Classification (NS SEC) in all official statistics. 3,4,5 NS SEC also replaces Socio-Economic Group (SEG) which has also been used in official statistics. These socio-economic classifications are based on occupation, in combination with employment status and in some circumstances size of workplace. Revision of the Standard Occupational Classification (SOC) takes place at decennial intervals, and the latest version – SOC2000 – was introduced in 2001. Changes to harmonise the coding of employment status for births and deaths with the 2001 Census have also taken place. There have therefore been three simultaneous changes: a change in the classification of occupations; a change in the coding of employment status; and a change in the broader classification of socio-economic status. This article describes the implications of these changes for the coding and analysis of vital records in England and Wales. The classification changes particularly impact on the ability to monitor change by social group over time. One approach to aid the comparison of data before and after the introduction of NS SEC is to estimate NS SEC prior to 2001 using the previous Standard Occupational Classification (SOC90) and employment status codes. The degree to which estimated NS SEC, known as NS SEC90, matches final NS SEC is described using an analysis of the dual coded 2001 mortality data.

In 2001, three simultaneous changes were introduced which influence the reporting of vital statistics by socio-economic status. Registrar General’s Social Class was replaced by the new National Statistics SocioEconomic Status (NS SEC); the Standard Occupational Classification (SOC) was updated from SOC90 to SOC2000; and the coding of employment status on vital statistics changed to stay in line with census coding and SOC2000. This article reports on these changes and describes a method for coding data prior to 2001 to NS SEC (named NS SEC90) in order to facilitate the analysis of health inequalities over time. NS SEC90 allocates 90 per cent of deaths to the same analytical category as NS SEC.This figure varies according to sub group.

23 National National Statistics Statistics


Spring 2002

CHANGES TO C LASSIFICATIONS AND CODING i. Standard Occupational Classification In 2001, SOC20006,7 replaced the previous version of the occupational classification, referred to as SOC90,8 which had been in use since 1991. The main features of the revision of SOC90 to SOC2000 include: a tighter definition of managerial occupations; expansion in the field of computing and related occupations; the introduction of specific occupations associated with the environment and conservation; changes linked to the upgrading of skills but the de-skilling of manufacturing processes; and the recognition of the development of customer service occupations and the emergence of remote service provision through the operation of call centres. Most of the major groups of occupations were renamed, and the composition of all major groups changed.9

Box one THREE VERSIONS OF EMPLOYMENT STATUS CODING Registrars employment status code At the registration of a birth or death, the Registrar records the job title of the parent or deceased person, and whether they are self-employed with or without employees, an employee or a super visor/foreman on a registration form. Old rules employment status code Prior to 2001, under the old rules for coding employment status, supervisor status was only coded if there was evidence to support this decision in the textual information on occupation on the registration form, irrespective of the Registrar’s recording of employment status. New employment status code From 2001, under the new rules for coding employment status, the Registrar’s indication of employment status is given precedence over the textual information on occupation on the registration form and so births and deaths can be allocated a ‘supervisor’ status without additional supporting evidence from the textual information on occupation.

Table 1

While the majority of individual occupations are included in both classifications, there is no overall one-to-one equivalence between the two coding structures. The relationship between SOC90 and SOC2000 codes has been examined using the Labour Force Survey (LFS). Data on occupation in the LFS for one quarter of 199710 and one quarter of 2000 were coded with both classifications. The 1997 and 2000 data showed agreement rates of 78 per cent and 82 per cent respectively, the agreement rate being the proportion of records to which correctly matching SOC90 and SOC2000 codes could be assigned.

ii. Employment status Categorisations of employment status distinguish between people who work as managers or supervisors, employees without supervisory responsibilities, self-employed workers, and those who are not in gainful employment for various reasons. The information collected on employment status in the 2001 Census is more detailed than the 1991 Census, in that it allows the separate identification of supervisors or foremen, and the rules for coding employment status in births and deaths data are being changed in 2001 to harmonise with this. Consistency between Census and vital events data is desirable to reduce numerator-denominator bias in the calculation of rates. There are three employment status codes that this article will refer to and these are listed in Box 1. Table 1 shows, for all deaths under the age of 75 in England and Wales in 2000, a comparison between employment status coded according to the old rules, and employment status as recorded by the Registrar. The latter is an approximation to the effect of coding according to the new rules. The shaded cells show the number of deaths, which were assigned, to an equivalent status by the two methods. The most significant change is that 17,275 (9 per cent of total) deaths would have been allocated to ‘employees’ under the old rules, but are ‘supervisors’ under the new rules. The effect of the change of rule regarding supervisor status which accompanied the introduction of SOC2000 can be estimated using NS SEC90, and Figure 1 compares the percentage distribution of deaths in 2000 across NS SEC90 classes using the two different rules. Note that some recoding of the Registrar’s employment status code was necessary so that managers and those without a given employment status were allocated correctly. The effect of the change on the nine class version of NS SEC, was to increase the proportion of deaths allocated to lower managerial and professional occupations (class 2) and to employees in small organisations and own account workers, lower supervisory and technical occupations (classes 4 and 5) at the expense of the proportion

Comparison of employment status 1 recorded by the Registrar and coded employment status, deaths of persons aged under 75 years, 2000

England and Wales Coded employment status (according to old rules)

Numbers Employment status recorded by Registrar Employee Not super-vising Super-vising

Self-employed No employees With employees

No gainful employment

Not known or not stated

Other (no code from registrar)

Total

Employee Manager Foreman Self-employed, without employees Self-employed, with employees No gainful occupation or not stated Other

63,997 1,324 284

17,275 9,860 3,416

236 206 12

83 572 3

5,733 628 138

12 2 0

28,449 3,576 691

115,785 16,168 4,544

12

6

9,049

6

35

0

95

9,203

96

75

279

4,433

81

0

546

5,510

438 1

129 1

51 0

48 0

21,864 0

2 0

21,076 1

43,608 3

Total

66,152

30,762

9,833

5,145

28,479

16

54,434

194,821

1

Employment status of the mother in the case of childhood deaths, and of the deceased for all other ages.

National Statistics

24


Figure 1

Distribution of deaths by NS SEC 90 class, using old and new coding rule for employment status, 2000

England and Wales 25

Percentages

Old rule 20

New rule

15

Spring 2002

change at three levels: the assignment of occupations and employment statuses to classes, the underlying classification of occupations and its implementation, and the actual composition of the UK workforce.3,13 In 1994, the then Office of Population Censuses and Surveys commissioned a review of government social classifications, which resulted in the development of NS SEC. The deliberations and conclusions of the review have been reported elsewhere3,4,5,14 and full details of NS SEC have been published on the National Statistics website www.statistics.gov.uk. NS SEC was developed from a sociological classification of occupations known as the Goldthorpe Schema15,16,17, which has been widely used in research and is well validated. Following Goldthorpe, NS SEC differentiates occupations in terms of their dominant form of employment regulation, including source of income, economic security and prospects of advancement (see Box 3) and typical work situation including location in systems of authority and control at work. The complete version of NS SEC in vital events data has forty basic categories, which allow flexible aggregation for analytical purposes. From these basic categories, fourteen ‘operational categories’ are derived which can then be aggregated to produce nine, eight, five or three classes, as shown in Figure 2. The three class level of NS SEC is likely to be used as a replacement for the common division of RGSC into ‘manual’ (IIIM, IV, V) and ‘non-manual’ (I, II, IIINM).

10

5

0 1.1

1.2

2

3 4 5 NS SEC 90 class

6

7

8

allocated to intermediate occupations (class 3) and to semi-routine and routine occupations (classes 6 and 7) respectively. At the three class level, this represented an increase of 1.72 per cent in deaths allocated to the Managerial and Professional class, accompanied by a decrease of 0.18 per cent in deaths to the Intermediate class and 1.54 per cent to the Routine and Manual class.

iii. Socio-economic classification RGSC (shown in Box 2) is an ordered scale which has been described at different times in its history as portraying ‘standing within the community’11 or ‘level of occupational skill’.12 Although RGSC has proved robust over time in its relationship to mortality and other health measures, it has been criticised for this lack of a clear theoretical rationale and for other limitations, such as its failure to deal adequately with women’s occupations and with individuals not currently in employment. RGSC has been revised on several occasions and the size and composition of the classes has varied over time as a result of

Box two REGISTRAR GENERAL’S SOCIAL CLASS (SOCIAL C LASS BASED ON OCCUPATION) I II IIIN IIIM IV V

Professionals Managerial and technical Skilled Non-manual Skilled Manual Partly skilled manual Unskilled Manual

Source: OPCS 1980

As RGSC and NS SEC are both classifications of socio-economic status based on occupation, it may be expected that both will show similar relationships to measures of health outcome. Table 2 shows standardised mortality ratios (SMRs) for men aged 20–64, based on 1991 Census data and deaths in 1991–93 coded to interim NS SEC90.18 Overall the gradient between the ‘top’ and ‘bottom’ of each classification is similar in direction and magnitude. The SMR for the class with the highest mortality is 2.8 times that in the class with the lowest mortality in NS SEC90, and 2.9 times in RGSC. However, the different structure and rationale of the two classifications mean that there is no

Box three FORMS OF EMPLOYMENT REGULATION IN THE NS SEC

CONCEPTUAL BASIS OF

Service relationship The employee renders ‘ser vice’ to the employer in return for ‘compensation’ in terms of both immediate rewards (e.g. salary) and long-term or prospective benefits (e.g. assurances of security and career opportunities). The service relationship typifies NS SEC class 1 and is present in a weaker form in class 2 of the eight-class version. Labour contract The employee gives discrete amounts of labour in return for a wage calculated on amount of work done or by time worked. The labour contract is typical of NS SEC class 7, and in weaker forms, classes 5 and 6 of the eight-class version. Intermediate Forms of employment regulation which combine aspects of ‘ser vice relationship’ and ‘labour contract’ are typical of NS SEC class 3 of the eight-class version.



Figure 2

Spring 2002

NS SEC operational categories and analytical classes

Operational categories

Analytical classes Nine-class version

Eight-class version

2 Higher managerial occupations

1.1 Large employers and higher managerial occupations

1 Higher managerial and professional occupations

3 Higher professional occupations

1.2 Higher professional occupations

1 Employers in large Establishments

Five-class version

Three-class version

1 Managerial and professional occupations

1 Managerial and professional occupations

4 Lower professional and higher technical occupations 2

Lower managerial and professional occupations

2 Lower managerial and professional occupations

7 Intermediate occupations

3

Intermediate occupations



8 Employers in small establishments 9 Own-account workers

4

Small employers and own-account workers

4 Small employers and own-account workers

3 Small employers and own-account workers

5

Lower supervisory and technical occupations

5 Lower supervisory and technical occupations

4 Lower supervisory and technical occupations

6

Semi-routine occupations

6 Semi-routine occupations

5 Lower managerial occupations 6 Higher supervisory occupations


10 Lower supervisory occupations 11 Lower technical occupations 12 Semi-routine occupations 13 Routine occupations

7

Routine occupations

7 Routine occupations

14 Never worked and long-term unemployed

8

Never worked and long-term unemployed

8 Never worked and long-term unemployed


3 Routine and manual occupations


method, which is the derivation used in births and deaths data, does not require size of organisation. Instead, records with an employment status of self-employed or manager are allocated to the appropriate class based on the typical size of organisation for that occupational code. The ‘simplified’ method uses occupation only, and NS SEC class is based on the dominant employment status for that occupation.

simple relationship between them, with the self-employed and intermediate occupations in particular being treated in very different ways. For example, members of RGSC class IIINM may be assigned to any one of five NS SEC90 classes, with SMRs ranging from 99 to 122.

CODING TO NS SEC AND NS SEC90

Classification to RGSC has been based on the version of the Standard Occupational Classification applicable at the relevant time, which from 1991 to 2000 was SOC90. Classification to NS SEC is based on SOC2000. However, during the development of NS SEC interim

There are three methods for deriving NS SEC, which can be applied to different data sources depending on the extent of the information available (see Box 4). The ‘full’ method requires occupation, employment status and size of organisation the person works in. The ‘reduced’ Table 2

5 Semi-routine and routine occupations

Comparison of standardised mortality ratios for men aged 20–64 by Registrar General’s Social Class and NS SEC classes, 1991–93

England and Wales Registrar General’s Social Class Interim

I

II

IIINM

IIIM

IV

V

NS SEC90 1.1 1.2 2 3 4 5 6 7 All

Armed

All

Forces 65 68 46 66

53 74 73 79 84 250 99 137 72

99 96 100 125 122 487 100

251 94 81 116 132 136 117

Shaded cells indicate the RGSC class making up the majority of the interim NS SEC90 class. Figures in italics are based on a small number of events. Source: Fitzpatrick (forthcoming)

National Statistics

26

98 79 96 116 138 116

67 120 112 216 189

97 143 -

55 69 76 96 83 114 120 155 -


Table 3

Spring 2002

Percentage of first quarter deaths 2001 allocated to the same NS SEC90 category given their NS SEC category

England and Wales

Percentages Percentage agreement between NS SEC and NS SEC90 NS SEC90 calculated using SOC90 old rules employment status code

NS SEC90 calculated using SOC90 and Registrar’s employment status code1

NS SEC90 calculated using new rules employment status code2

All (NS SEC 9 class)

83.9

89.9

90.9

By NS SEC: 9 class 1.1 Large employers and higher managerial 1.2 Higher professional 2 Lower managerial and professional 3 Intermediate 4 Small employers and own-account 5 Lower supervisory and technical 6 Semi-routine 7 Routine 8 Never worked and long term unemployed

73.9 79.8 67.5 87.4 86.2 56.2 88.6 92.8 96.3

73.4 80.4 80.3 87.1 95.3 88.1 88.6 92.9 96.6

74.0 80.3 83.6 87.5 97.6 89.8 89.2 93.4 96.7

By NS SEC: 5 class 1 Managerial and professional 2 Intermediate 3 Small employers and own-account workers 4 Lower supervisory and technical 5 Semi-routine and routine Never worked and long term unemployed

70.3 87.4 86.2 56.2 91.0 96.3

79.5 87.1 95.3 88.1 91.0 96.6

81.9 87.5 97.6 89.8 93.4 96.7

By NS SEC: 3 class 1 Managerial and professional 2 Intermediate 3 Routine and manual Never worked and long term unemployed

70.3 86.8 82.2 96.3

79.5 91.1 90.3 96.6

81.9 92.3 91.2 96.7

By gender Males Females

80.3 89.4

87.6 93.2

88.9 93.8

By age 16–25 26–35 36–45 46–55 56–65 66+

91.6 85.2 84.7 83.6 84.4 83.1

91.9 87.2 88.2 89.5 90.3 90.0

92.4 88.2 89.0 90.7 91.1 91.1

By SOC2000 major group 1 Managers and senior officials 2 Professional occupations 3 Associate professional and technical occupations 4 Administrative and secretarial occupations 5 Skilled trades occupations 6 Personal service occupations 7 Sales and customer service occupations 8 Process and customer service occupations 9 Elementary occupations

78.4 87.6 67.9 72.6 79.5 77.0 79.8 82.8 87.5

83.3 87.8 73.8 87.3 91.7 86.7 85.6 89.5 92.1

86.2 88.1 74.5 88.9 93.5 88.5 86.5 90.4 92.8

By SOC90 major group 1 Managers and administrators 2 Professional occupations 3 Associate professional and technical occupations 4 Clerical and secretarial occupations 5 Craft and related occupations 6 Personal and protective service occupations 7 Sales occupations 8 Plant and machine operatives 9 Other occupations

71.6 86.5 79.1 73.1 80.5 77.5 78.6 81.2 89.5

77.0 86.7 82.6 87.3 91.2 91.1 84.3 88.2 92.7

82.1 86.9 82.8 87.7 92.0 93.1 85.4 88.9 93.4

1 Where Registrar’s code was supervisor, and coded employment status was manager, manager status was coded, if manager status was not coded then a supervisor code was given. Where Registrars code was not stated, the code given by the coded employment status was taken. 2 It will not be possible to use this on data prior to 2001, but this is included to illustrate the influence that the change in the Standard Occupational Classification alone has on the figures.



Spring 2002

Table 4

Box four

Absolute numbers of deaths allocated to NS SEC90* and NS SEC categories

DERIVATION OF VERSIONS OF NS SEC 8 class category

NS SEC90

NS SEC

1,073 1,515 6,342 3,728 3,554 5,344 6,703 8,899

1,107 1,503 6,562 3,574 3,382 5,393 6,635 8,859

Data required Derivation

Occupation

Full NS SEC Reduced NS SEC Simplified NS SEC

Employment Size of status organisation

Yes Yes Yes

Yes Yes No

Yes No No

1.1 Large employers and higher managerial 1.2 Higher professional 2 Lower managerial and professional 3 Intermediate 4 Small employers and own-account 5 Lower supervisory and technical 6 Semi-routine 7 Routine

1 Note NS SEC90 calculated as described in text, with Registrars employment status.

versions based on SOC90 were used, an interim version has already been used for analysis of mortality in order to validate the classification.19 For the purposes of historical comparison, this interim version of NS SEC based on SOC90 codes has been revised and this version of NS SEC is referred to here as ‘NS SEC90’. Although the categories of NS SEC90 are the same as those of NS SEC based on SOC2000 codes, allocation to them is not completely equivalent because of the differences between SOC90 and SOC2000 and the changes in employment status coding rules described above. A code may exist in SOC2000 for a particular occupation but not in SOC90, or vice versa, or may be located differently within the structure of SOC90 compared to SOC2000 resulting in a different class allocation. Within the 78 per cent of the 1997 LFS sample which could be assigned valid codes in both SOC90 and SOC2000, agreement between allocation to NS SEC and NS SEC90 classes was around 95 per cent for all age and sex sub-groups.14 However, the rate of agreement may vary between data sources because of differences in population characteristics, sample selection, data collection, derivation methods and coding variance. In addition it is not possible to just work with those data that have matching SOC90 and SOC2000 codes. The following section outlines the agreement between allocation of NS SEC and NS SEC90 on the dual coded mortality data for 2001.

COMPARISONS OVER TIME The changes to classifications and coding rules described above make a direct comparison in terms of socio-economic status of births and deaths data for years up to 2000, and from 2001 onwards, problematic. To allow continued analysis of trends in health inequalities, it is necessary firstly to recode data for the years preceding 2001 to NS SEC90 to provide a consistent set of categories over time, and secondly to make adjustments to the recoded data to take account of the differences between NS SEC and NS SEC90, so that real changes in the health of the population can be distinguished from artefacts caused by these changes in methods. For this purpose, births and deaths data for 2001 have been dual coded using SOC90 and SOC2000 and old and new employment status coding rules. From these data, NS SEC90 and NS SEC have been derived. This article reports on preliminary analysis of a sub-set of these data-deaths in the first quarter of 2001. Similar analyses of births data for the same period will be published separately. Table 3 shows agreement rates between NS SEC90 and NS SEC, that is, the proportion of deaths, which were, assigned to the same class at eight class, five class and three class levels. There are three columns presenting agreement rates between NS SEC90 and NS SEC, the first calculates NS SEC90 using the old rules employment status code, the second calculates NS SEC90 given the Registrar’s employment status code and the third calculates NS SEC90 using the new rules employ-

National Statistics

28

ment status code. The third column therefore shows what the agreement rates would be if employment status were kept constant. The numbers in this column show the highest agreement rates. The first column uses old employment status rules and the numbers in this column are the lowest. The influence of the change in employment status coding is very marked between the two columns. For instance, if we were to calculate NS SEC90 using the old employment status variable, only 56.2 per cent of those in the lower supervisory and technical class in NS SEC would be allocated the same code using NS SEC90. If we used the same employment status code as was used to derive NS SEC in these data, the agreement rate increased to 89.9 per cent. Another noticeable difference is seen in the percentage of men correctly allocated to NS SEC90: using the old employment status code only 80.3 per cent are correctly allocated, compared to 88.9 per cent using the new employment status code. Table 3 shows that 9.1 per cent of deaths were assigned to different NS SEC90 and NS SEC categories, due to the influence of the change in the occupational classification. A further 7 per cent would be allocated to different classes if the old employment status code was used. However this can be reduced to just a 1 per cent additional misallocation if the Registrar’s employment status code is used to approximate to the new employment status code. These data can be used in conjunction with the old employment status variable, in order to differentiate managers from others. One of the disadvantages of using the Registrar’s code is the higher number allocated to a ‘not stated’ code as this field is often not filled in. Information from the coded employment status, when this is the case, can be used to decrease the number ‘not stated’ as this code incorporates textual information from the job description. Using the Registrar’s employment status code, as shown in column two of Table 3, there was 89.9 per cent agreement between the NS SEC90 and NS SEC overall. The percentage agreement varied according to sub group, and was particularly low for those in managerial and professional occupations: at the three class level just 79.5 per cent matched. Although SOC2000 incorporates, a tighter definition of managerial occupations than SOC909 there were more deaths allocated to managerial and professional classes under NS SEC than NS SEC90 (Table 4). This may be due to some technical or supervisory occupations in SOC90 being reclassified in SOC2000 to lower professional, and likewise from self employed to professional or managerial status.

DISCUSSION This article highlights changes in social and occupational classifications which took place in 2001 as they apply to mortality data, and describes an approach to the analysis of trends over time in mortality by NS SEC. The approach described does not produce 100 per cent matching between NS SEC90 and NS SEC and therefore time trends comparing


data using the approximation to NS SEC should be treated with caution. 10 per cent of deaths would be allocated to different classes under NS SEC and NS SEC90 as a result of coding and classification changes. This should be taken into account in mortality analysis. Further work on births will be published at a later date. It should be noted that the agreement rates for births may be different, mainly because the nature of people’s work recorded at younger ages may be different from that at older ages. New occupational classifications are introduced to reflect the changing nature of employment. One limitation on the reliability of this approach is therefore that the proportion of people whose employment is accurately described by an occupationally-based classification will tend to diminish over time. As a result, agreement rates between NS SEC and NS SEC90 cannot be expected to have been constant over the whole decade 1991 to 2000. The ability to analyse historical births and deaths data by NS SEC is important, not only because it allows the monitoring of trends in health inequalities, but also because NS SEC has a tighter conceptual foundation than RGSC. NS SEC was constructed to improve the potential for explaining causation and not simply to maximise statistical associations with particular outcomes. The nature of employment relations and conditions have been found to be central to the issue of socio-economic inequalities, and so an explicit rationale based on employment relations theory means that causal pathways can more readily be identified to link NS SEC to observed outcomes.5,14 Hence, being able to analyse data prior to 2001 using an approximation to NS SEC might help users to more fully understand associations between social status and health.

Key findings In 2001, three simultaneous coding and classification changes were made which impact on the measurement of health inequalities over time: the decennial update of the Standard Occupational Classification (SOC), changes to the coding rules for employment status, and the replacement of Registrar General’s Social Class with the new National Statistics Socio-Economic Classification (NS SEC). ● It is possible to create an approximation to NS SEC for use with vital events data prior to 2001. This is known as NS SEC90 and is based on the 1990 occupational classification (SOC90) and an approximation to the new employment status coding rule. ● Analysis of deaths for the first quarter 2001 illustrated that 90 per cent of all deaths were allocated to the same NS SEC90 and NS SEC categories. ● Of the 10 per cent of deaths allocated to different categories, nine percentage points resulted from the change in occupational classifications (SOC90 to SOC2000) and one percentage point to the use of an approximation to the new employment status coding rules. ● The level of agreement between NS SEC90 and NS SEC varies by NS SEC categor y, gender, age and occupational group. ●

Spring 2002

Finally, although the need to harmonise employment status coding of vital events with the Census has been mentioned, there are other issues affecting the comparability of vital events and Census data. For example, at death registration, occupation is necessarily reported by a third party and so might differ from the occupation the deceased would have given on a Census form. An additional consideration relating to the routine 2001 Census tables is that occupation will not be coded for people aged over 65 or for those who have not been in work for five years or more. This will limit the availability of denominators for mortality data, where the last gainful employment is recorded for those who have died aged 75 or under.

REFERENCES 1. Registrar General (1913). Seventy-fourth report of the Registrar General. HMSO: London. 2. Drever F and Whitehead M (eds.) (1999). Health Inequalities. The Stationery Office: London. 3. Rose D and O’Reilly K (1997). Constructing Classes. ESRC & ONS: Swindon 4. Rose D, O’Reilly K and Martin, J (1997). The ESRC Review of Government Social Classifications. Population Trends 89, pp.49–59. 5. Rose D and O’ Reilly K (1998). The ESRC Review of Government Social Classifications. ESRC & ONS: Swindon. 6. Office for National Statistics (2000a). Standard Occupational Classification 2000: Volume 1. Structure and descriptions of unit groups. The Stationery Office: London. 7. Office for National Statistics (2000b). Standard Occupational Classification 2000: Volume 2. The coding index. The Stationery Office: London. 8. Office of Population, Censuses and Surveys (1990). Standard Occupational Classification: Volumes 1–3. HMSO: London. 9. Beerten R, Rainford L and Jones A (2001).Changing to Standard Occupational Classification (SOC) 2000 – dual coding on the Labour Force Survey. Labour Market Trends Vol. 109 No. 7, pp. 357–366. 10. Rose D and Pevalin D (2000). ESRC Review of Government Social Classifications: A Report to ONS on Phase 4 of the Review. University of Essex: Colchester 11. Office of Population, Censuses and Surveys (1970). Classification of Occupations. HMSO: London. 12. Office of Population, Censuses and Surveys (1980). Classification of Occupations. HMSO: London 13. Szreter S R S (1984). The genesis of the Registrar General’s social classification of occupations. British Journal of Sociology XXXV, pp. 522–546. 14. Rose D and Pevalin D J (with O’Reilly K) (forthcoming). The National Statistics Socio-economic Classification: Origins, Development and Use. ESRC & ONS: Swindon. 15. Goldthorpe J H (with C. Llewellyn) (1980 and 1987). Social Mobility and Class Structure in Modern Britain. Clarendon: Oxford. 16. Goldthorpe J H (1997). The ‘Goldthorpe’ class schema: some observations on conceptual and operational issues in relation to the ESRC review of government social classifications. In Rose D and O’Reilly K (eds.) Constructing Classes: Towards a New Social Classification for the UK. ESRC & ONS: Swindon. 17. Erikson R and Goldthorpe J H (1992). The Constant Flux. Clarendon: Oxford. 18. Fitzpatrick J (forthcoming). Examining mortality rates by NS-SEC, using death registration data and the 1991 Census. In Rose D and Pevalin D J (eds.) Confirming classes: introducing and validating the National Statistics Socio-economic Classification. Sage: London. 19. Fitzpatrick J and Dollamore G (1999). Examining adult mortality rates using the National Statistics Socio-Economic Classification. Health Statistics Quarterly 02, pp. 33–40. 29 National Statistics


Spring 2002

The Millennium Cohort Study

Kate Smith and Heather Joshi Centre for Longitudinal Studies Institute of Education

The Millennium Cohort Study is the latest in the line of British birth cohort studies. MCS resembles its predecessors which follow people born in 1946, 1958 and 1970 in the intention to become multi-purpose longitudinal data resource charting many aspects of individual’s lives over time.The families of a sample of around 20,000 babies are being interviewed during 2001–02, when eligible babies reach 9 months, to establish the conditions from which they set out in life. The survey contrasts with the previous cohort studies in various ways. Instead of taking all births in one week, the sample of births is spread over a year; the births are from a selection of electoral wards, thereby enabling eventual analysis by neighbourhood characteristics; it also over samples children living in deprived areas, wards with high ethnic minority populations and samples have been boosted in Scotland, Wales and Northern Ireland. The latter UK country has not been covered by the other studies. It interviews fathers as well as mothers, and given that its initial funding comes via the ESRC, puts a greater emphasis on socio-economic data than in early parts of the other studies. MCS has been enhanced by additional Government funding. The research team, based at the Institute of Education, aims to deposit a multi-purpose dataset for public use at the ESRC data Archive in the Spring of 2003.

National Statistics National Statistics

INTRODUCTION The Millennium Cohort Study, known in the field as the Child of the New Century, was launched in 2000, ending a 30-year gap in the sequence of national longitudinal birth cohort studies for which the UK is renowned. These studies all aim to gather information on a group of people born around the same time and to link information from infancy and early life with subsequent developments, in principle into adult life. While maintaining the essential design of its predecessors, the studies of the 1946 (Douglas, 1964; Wadsworth 1991), 1958 (Butler and Bonham, 1963; Ferri (Ed), 1993) and 1970 cohorts (Chamberlain et al, 1975; Bynner et al, 1997), the Millennium Cohort has some new features. It takes a sample of births from a whole year, rather than a week, as is the case with the other studies. It covers the whole of the United Kingdom rather than just the three countries of Great Britain, with boosted samples in Scotland, Wales and Northern Ireland to allow adequate sample sizes for within country analysis. The sample is geographically clustered, with an over-representation of deprived areas, which will increase the scope for including community-level information in analysis and for the study to be used in the national evaluation of Sure Start. The Government decided that the millennium was a good occasion to add a new birth cohort study to the British heritage of longitudinal data resources, and made funds available to the ESRC to initiate the Millennium Cohort. Co-funding from government departments, in a consortium led by ONS, has subsequently augmented the budget for the first and second sweeps. The other members of the consortium are: Department for Employment and Skills, Department of Health and Department for Work and Pensions, along with the devolved administrations of Scotland, Wales and Northern Ireland. The Principal Investigator contract for the first sweep of the MCS was awarded to an academic consortium based in three institutions: the Centre for Longitudinal Studies (CLS) at the Institute of Education; the Department of Community Epidemiology and Public Health and the Institute of Child Health, both at University College London; and the

30


Department of Psychology, City University. The scientific management of the project is based at CLS. The fieldwork contract has been awarded to the National Centre for Social Research (NatCen), who also had recent experience in interviewing the 1958 and 1970 cohorts. Work at the Centre for Longitudinal Studies was funded from August 2000 by the ESRC following a tendering process announced at the end of February. Both before and after the announcement of the award, the timetable was very tight, if children born in the year 2000 were to be included. Accordingly, the sample for the survey has been defined as children aged 9 months in England and Wales born between September 2000 and August 2001, and in Scotland and Northern Ireland between December 2000 and November 2001, living in selected wards. The sample is being provided by the Department of Work and Pensions (DWP, formerly DSS) from Child Benefit Records. The target sample is 20,646 children. Interviews for the first sweep take place when each child is aged 9 months. The fieldwork therefore started in June 2001. NatCen is conducting the interviews in England, Wales and Scotland. In Northern Ireland interviewing has been sub-contracted to the Central Survey Unit of the Northern Ireland Social Research Agency (NISRA).

RATIONALE Understanding the social conditions surrounding birth and early childhood is increasingly appreciated as fundamental to the study of the whole life course. This applies across a wide spectrum, from looking at the origins of social exclusion through investigation of the influence of early circumstances on health over the life course to providing evidence for major policy initiatives such as “Sure Start”. The initiation of the MCS presents an exceptional research opportunity to investigate the allimportant first year of life and potentially resolve many of the research questions about its long-term impact. These include issues of central policy interest such as the foundations of social capital and cohesion. Major questions about the prospects for children born in 2000–2001 concern poverty and wealth, the quality of family life and its support by public policy and the broader community. The health and wellbeing of parents and infants will be located in the context of the rich socioeconomic data to be collected in the study. Issues to emerge for future sweeps of the cohort will include: advantage and disadvantage in education, health, employment and the parenting of the next generation. There is much interest in the consequences of changing family structures, and ongoing social and economic changes in the labour market, as well as changes in technology, social polarization, gender roles, and the ideology of individualism which are likely to affect different children differently, and make the unfolding lives of the new cohort different from those of their predecessors.

O BJECTIVES The study’s broad objective is to create a new multi-purpose longitudinal dataset, describing the diversity of starting points from which children born in the new century are setting out on life. Government funding has extended the objectives to provide control cases for the national evaluation of Sure Start by adding 35 wards in deprived areas of England to the sample – adding about 2,600 extra children. The devolved administrations of Wales, Scotland and Northern Ireland have also invested additional funds to boost the samples in these countries of the UK. The aim of this is to achieve adequate numbers for analysis within the individual countries. The particular objectives of the first MCS survey are: 1. To chart the initial conditions of social, economic and health advantages and disadvantages facing new children in the new century, capturing information that the research community of the future will require.

Spring 2002

2. To provide a basis for comparing patterns of development with the preceding cohorts. 3. To collect information on previously neglected topics, such as fathers’ involvement in children’s care and development, and the effects of season of birth within a year. 4. To investigate the wider social ecology of the family, including, social networks, civic engagement and community facilities and services. Additional objectives, such as the recording of adaptation of other family members to the new baby and aspirations for the baby’s future have not been implemented in the first sweep. Neither are grandparents well covered. These intended areas of questioning have had to be sacrificed, in the first sweep at least, due to the limits of interview time. The length of the total contact time (for mother and father) has been extended to an hour and three quarters. Any more would become burdensome. The objective of studying the wider social environment of the family is being approached through assembling data from ONS Neighbourhood Statistics to add to the database, and through a survey of Health Visitors on local services. The enhancement of health information has started with a series of feasibility studies. These enhancements include a supplementary postal survey to mothers who have received assisted reproduction treatment and the linking of hospital records of the delivery of the MCS baby for all mothers who have given consent. Further details of these activities are outlined below. The government consortium is funding these enhancements of ecological and health data as well as the extension on Sweep 1 interview time and boosts to the sample and also, in due course the preparation, of reports and analyses, as well as making contributions to future sweeps.

SAMPLE DESIGN Timing of Cohort births and age at interview As outlined above, the population for the study is being drawn from a sample of a full year’s births with interviews taking place when each baby reaches 9 months of age. Eligible birth dates for England and Wales range from September 2000 to August 2001, so that the children in the sample will belong to the same academic year. In Scotland and Northern Ireland the eligible birthdays started 3 months later (from December 2000) to avoid overlap with the Infant Feeding Survey (already taking large samples of September-October births in those countries). Fieldwork started in June 2001 in England and Wales, when babies born at the beginning of September 2000 reached 9 months of age. The survey in Scotland and Northern Ireland was launched in September, when the babies born at the beginning of December turned 9 months. Interviewing in all four countries will continue for a year from commencement.

Sampling: the selection of wards The sample is geographically clustered, and based upon electoral wards, with overrepresentation of wards with a high minority ethnic population in England; wards with high child poverty rates in all countries; and over-sampling in Scotland, Wales and Northern Ireland generally. The ward boundaries are those obtaining in 1998 for which the Oxford index of (child) deprivation was available. Sparsely populated wards were amalgamated with their neighbours before the sample was drawn, with help from ONS. All babies with eligible birth dates resident in the sampled wards at 9 months of age are eligible for the survey. The size of the sample is therefore determined by the sampling fractions of the wards (and expected response rates). Disadvantaged areas were defined as those children living in wards which fell into the upper quartile (i.e. the poorest 25 per cent of wards) of the ward-based Child Poverty Index. 31 National Statistics


Spring 2002

The original target sample size for the study was 15,000 babies. Subsequent decisions to boost the sample in Wales, Northern Ireland and Scotland as well as deprived areas of England increased the final target sample of babies to over 20,000 (see Table 1).

instrument (CASI) in order to assess the timing of the instrument before the major work to convert the interview schedule into CAPI format. The pilot sample size was boosted from 30 to 60 thanks to the ONS consortium funding.

Sampling: the use of Child Benefit Records

The overall response by parents to the pilot survey was positive, however the instruments proved to be substantially over the target time of 105 minutes. As a result reductions were made to the interview and self-completion in consultation with scientific advisors.

The study has taken advantage of the offer from DWP to make Child Benefit records available to find the eligible families and draw the sample. It had originally been assumed there was no alternative to use of birth registration records for this purpose. The DWP route was favoured because of concern about response to the postal opt-in involved in drawing a sample from birth registrations. The representation of socially excluded groups, including those with poor literacy or poor grasp of English, could be compromised with an opt-in. DWP ask people if they want to opt out, which gives marginal cases and people with low literacy the chance to agree to join in when they meet the interviewer. Because the Child Benefit records will not reveal all families who have moved into the sample wards as the child approaches 9 months of age, a supplementary strategy for recruiting movers-in to sample wards through the help of Health Visitors has been devised.

INSTRUMENT DEVELOPMENT The survey is being carried out via CAPI (Computer Aided Personal Interview) and self-completion (also computer aided) by both the mother and (where resident) father of the cohort baby. There is no handling or measuring of the baby.

A second pilot took place during April 2001 on 91 families with the revised instruments. The interview for this pilot was fully computer based (CAPI and CASI). As a ‘dress rehearsal’ for the main stage, all the contact and administrative processes were also tested. Thirteen wards were selected for this pilot, including one in each of Wales and Scotland. The main impression of the pilots was that respondents and interviewers had, on the whole enjoyed the experience. However, the interview, and particularly the self-completion, still proved to take too long. Interviews where translation was needed took an exceptionally long time to complete. An experiment transcribing data from the Personal Child Health Record was not sufficiently successful to take into main fieldwork, but the experimental consent form for health record linkage was well received. Loss of interviews with families who move out were found to be non-trivial, and it proved difficult to interview all the partners even where present due to the short field period the dress rehearsal was operated within.

Giving information to informants

First Pilot Following consultations with academic and government longitudinal data users, the teams at NatCen and CLS jointly developed a draft questionnaire for use in the first pilot in January 2001. This pilot was conducted as a paper interview and computer aided self-completion

Table 1

Dress Rehearsal Pilot

In addition to the initial invitation letter and leaflet sent by the DWP, use of a second advance letter sent by the interviewer, shortly before their first visit to the family, proved very successful in the dress rehearsal pilot and has been adopted in the main stage. This letter serves to remind the family of the survey and introduce their interviewer by name.

Millennium Cohort Sample Structure by stratum and country: babies aged 9 months and wards

United Kingdom Country

Original target sample

Boost from original number

Final target sample

Total

Non-disadvantaged wards

Disadvantaged wards

High Minority Ethnic wards

England Babies Wards

10,500

2,646*

13,146 200

5,250 110

5,271 71

2,625 19

Wales Babies Wards

1,500

1,500*

3,000 73

750 23

2,250 50

Na

Scotland Babies Wards

1,500

1,000

2,500 62

1,250 32

1,250 30

Na

Northern Ireland Babies Wards

1,500

500 *

2,000 63

750 23

1,250 40

Na

15,000

5,700

20,646 398

8,000 188

10,021 191

2,625 19

Total United Kingdom Babies Wards ● ●

Boost to sample all in deprived wards. Note ‘ward’ includes some amalgamations of small electoral wards.

National Statistics

32


In order to comply with recommendations made by the Medical Research Ethics Committee, a simplified leaflet has also been produced for interviewers to give to respondent families. Translations of both the advance letter to be sent by the interviewer, and the new simplified leaflet into common non-English languages are also provided to all interviewers. The list of languages for translation is: Bengali, Gujerati, Kurdish, Punjabi, Somali, Turkish and Urdu. Where there is no-one accessible in the household to translate the interview, an interviewer speaking the required language is provided where available. The DWP letter and the leaflet sent with it are also translated into Welsh.

CONTENT OF SURVEY The survey attempts to establish, mostly from both parents a wide range of information about their own backgrounds, circumstances and health, and their attitudes on a number of issues, many directly relevant to parenthood. From one parent, usually the mother, details of the baby’s health and development are also collected. A list of topics included in Sweep 1 instruments is summarized in Table 2.

Table 2

Elements and content of the first survey

Respondent

Element

Content

Mother or Father

Interview

Household & family

Mother

Interview

Ethnicity & language Baby’s father Lone parenthood Pregnancy, labour & delivery Baby’s health & development Childcare Grandparents, friends & social support Parental health Education & training Employment & earnings Housing, local community & services Interests, time with & without the baby

Father

Self-completion

Baby’s temperament & behaviour Relationship with partner Previous relationships Domestic tasks Previous pregnancies Mental health Attitudes to relationships, parenting, work, etc

Interview

Ethnicity & language Father’s involvement with baby Lone parenthood Baby’s mother (if not resident) Grandparents & friends Parental health Education & training Employment & earnings Interests, time with & without the baby

Self-completion

Baby’s temperament & behaviour Relationship with partner Previous partners Previous children Mental health Attitudes to marriage, parenting, work, etc

the first wave covered births between 1 September 2000 and 28 September 2000. This rhythm of recruiting the sample has been dictated by the cycle of DWP procedures, scanning the Child Benefit database every four weeks. The process for drawing each wave of DWP sample is as follows: Prior to fieldwork, the DWP send opt-out letters to all parents of children with an eligible birth-date who are registered (for child benefit purposes) as living within one of the sampled wards. Batches of these letters are sent every four weeks, to families whose babies are then approximately seven months old. The letter invites parents to take part in the study and gives them the opportunity to opt-out of the study by telephoning or writing to the DWP. An information leaflet explaining the study is enclosed with these letters. Any parents who opt-out of the study are then removed from the sample, as are a few cases (ca. 1 per cent) deemed by the DWP to be ‘sensitive’. The data is sent by the DWP to CLS in two stages, a week apart, in order to ensure that any late exclusions or change of addresses can be notified as near to the start of fieldwork as possible. After the final data has been received, serial numbers are assigned to each valid case and the data is then sent to NatCen for issuing to the field. The aim is that the fieldwork for each four-week wave should be as self-contained as possible, with the minimum amount of overlap. Interviewers have been briefed to aim to interview families when the individual baby is 9 months and 15 days old in order to standardize the data being collected as far as possible. There will be exceptions where interviewing is delayed because of tracing problems, but the window of opportunity to interview is brief; up to 11 months of the baby’s age for the main interview and up to 12 months for the partner (as long as a main interview has been completed).

FIELDWORK PROGRESS In the early waves of interviewing so far completed, the project has on the whole been very well received in the field. The experience has, again on the whole, been a happy one for both families and interviewers alike. In the vast majority of cases, parents eligible for interview have been content to participate. The implications for achieving the sample target are still being assessed, at the time of writing, awaiting the completion of at least one full wave in Scotland and Northern Ireland. It is also, of course too early for any results. Fieldwork on Sweep 1 is due to finish in June 2002 in England and Wales and September 2002 in the case of Scotland and Northern Ireland.

DATA ENHANCEMENTS Supplementing Data on Health As mentioned above, alongside the development of the main survey a set of feasibility studies have taken place designed to enhance the collection of health data. Besides developing instruments for health questions at later Sweeps, most of these are intended to bring in information that has been collected elsewhere to reduce the burden on the interview. They include: ●

FIELDWORK Fieldwork Timetable The fieldwork is being carried out in 16 consecutive waves. Each issued wave of fieldwork will contain babies born in a 4-weekly birth cycle,

Spring 2002

Abstracting data from the Personal Child Health Record. It had been hoped to supplement the interview and reduce its time by taking information routinely entered in the child’s record book issued by the Health Service. The pilot study found that this was impractical in the time available and given the varying forms of records held in different parts of the country. Questions on weights, immunisations etc., have therefore been put in the main questionnaire, asking the mother to refer to her PCHR for the answers.



●

●

Spring 2002

Linking survey data to routinely collected health data. The successfully piloted approach to obtaining informed consent for collecting and linking information from health records has been taken on in the main survey. A virtual working group has been set up to advise the study and exploration of the extent and quality of routinely held data has begun in England and Scotland. Postal follow-up of infertility treatment. A pilot study conducted to investigate the feasibility of mother’s recall of infertility treatment has been successfully completed. The questionnaire for a postal survey of MCS mothers who report having treatment for infertility has been developed.

Ecological Data – Neighbourhood Statistics To enhance the interview data with information on the wider social ecology of the family, work has started on creating look up tables containing Neighbourhood Statistics and other data including metadata, which can be matched into the MCS files at a later stage. Thus far official Neighbourhood Statistics data have been matched into files containing the English, Welsh and Scottish MCS wards. There is little information about Northern Ireland at the moment. The next stage will be to add in other data such as Local Authority level information. The possibilities of adding in other contextual data, which may require copyright permission to be sought and approved before releasing it to users are also being investigated.

Survey of Health Visitors The study has made contact with most of the health visitors involved in the survey wards. There are plans to ask them, via a postal survey, for additional information about these localities not contained in the standard statistics. The possibilities and feasibility are under discussion and subject to consultation. They are likely to focus on the health and other services available locally to families with young children.

DATA DELIVERY AND DISSEMINATION OF SWEEP 1 RESULTS Because the data collection is spread over the best part of 18 months, there will be no results to report to any audience, be they informants, sponsors or the general public until well into 2003. Data from the first wave is expected to be available to the ESRC Data Archive in spring 2003. A large multi-purpose data set will then be available to any interested researcher. CLS will produce preliminary reports and analyses, which should encourage further use of the data, and also inform preparations for the next sweep. The research team will be making preliminary reports at the earliest opportunity, again with the help of government funding, but the full range of analysis possibilities, even of the cross-section at the first sweep remain open for the wider research community to plan. They should be aware of the data destined for the Data Archive in spring 2003 and the long-sighted should be anticipating the analysis of further follow-ups. Progress will be reported in Population Trends.

FEEDBACK TO INFORMANTS Adequate and frequent communication with the cohort families is essential to maintaining good relations and promoting future co-operation. Experiences from the 1958 and 1970 cohorts have shown that cohort members rate feedback and communication very highly. With the MCS we are in the relatively fortunate position of being able to plan for the next sweep, which means we can also plan the timing of feedback. In addition to a dedicated website, we are also planning to send an annual celebration card to each study family asking them to confirm

National Statistics

34

their address or give us details of any move. The precise form of this has yet to be decided, but will provide a vital tool in panel maintenance. Other planned communications include a summary of initial findings, although this will not be available until after receipt of the final data.

FUTURE SWEEPS With the next follow-up planned for when the children are aged 3, the Millennium Cohort will become a truly longitudinal survey. It is likely that the third sweep will take place when the children are aged 5, and all in school. The content and scope of these follow-up surveys is already under discussion by an interdisciplinary group of researchers, again co-ordinated at CLS, in consultation with other academics and government departments. The aim will be to see where the conditions of the first sweep have led, and also to lay further foundations for use in the understanding of the cohort’s later lives.

CONCLUSION The Millennium Cohort Survey aims to give an account of the progress and setbacks of its subjects as they embark on the race of life in the twenty-first century. Although it is far too early to relay the full story, the survey teams can report that getting the study off to its start has itself been a bit of a race, against time and over many hurdles. Although the enterprise remains challenging, progress so far augurs for a successful and fascinating series of outcomes. This is thanks to the enthusiasm and co-operation of all concerned, not least the cohort babies’ parents. This human document promises to prove a durable contribution to the national Heritage in commemoration of the new millennium. Further information can be obtained from the Millennium Cohort Team at CLS: E-mail: [email protected] Internet: www.cls.ioe.ac.uk/Mcs/mcsmain.htm

REFERENCES 1. Butler N R and Bonham D G (1963). Perinatal Mortality. E & S Livingstone: Edinburgh. 2. Bynner J et al (1997). Twenty-something in the 90s: Getting On, Getting By, Getting Nowhere. Aldershot: Dartmouth Press . 3. Chamberlain R et al (1975). British Births 1970, Volume 1, The first week of life. Heinemann Medical Books. 4. Douglas J W B (1964). The Home and the School. MacGibbon & Kee. 5. Ferri E. (Ed) (1993). Life at 33: the fifth follow-up of the National Child Development Study. National Children’s Bureau, City University, Economic and Social Research Council: London. 6. Wadsworth M E J. (1991). The Imprint of Time: Childhood, History and Adult Life. Oxford University Press: Oxford.


Spring 2002

Marriages in approved premises and register offices in England and Wales: the proportions of couples who marry away from home John Haskey Population and Demography Division Office for National Statistics

I NTRODUCTION The Marriage Act 19941 introduced two major new facilities for civil marriage: the option to marry with a civil ceremony in a place other than a register office – in approved premises, in fact; and the ability of couples to marry by civil ceremony – either in an approved premises or in a register office – outside their district of residence (see Box 1). Both these facilities were unavailable for civil marriage before the Act was implemented; previously, all civil marriages took place in register offices, and couples could marry with a civil ceremony only in the register office in the district in which one or other, or both, marriage partners resided (see Box 2). The first part of the Act – which enabled couples to marry away from home – came into force on 3rd January 1995, and the second part – which introduced the ability to marry in approved premises – came into force on 1 April 1995. An earlier article considered the outcome of implementing the second part of the Act, charting the growth in the number of marriages in approved premises during the first 1 3/4 years from 1 April 1995, and analysing the characteristics of couples marrying in them.2 The present article assesses the effect of the first part of the Act, and, therefore, together with the previous article, evaluates the way in which these two new facilities of the Marriage Act 1994 have been used in practice. In particular, this article considers the numbers, proportions and characteristics of marriages solemnised in registration districts in which neither party resides – undoubtedly the less widely known of the two new provisions.

This article analyses how one of the two new facilities of the Marriage Act 1994 has been used in practice; the extent to which couples marry with a civil wedding in registration districts in which neither party resides - “away marriages”. The Act made these “away marriages” possible for the first time from the beginning of 1995. Using two samples of marriages in England and Wales in 1998, the numbers and characteristics of away marriages are contrasted between those solemnised in approved premises and those in register offices. Comparisons are also made between “home marriages” and “away marriages”.

The Act introduced the ability to marry away from home for register office marriages as well as for those in approved premises. However, approved premises include many special places which attract couples

35 National National Statistics


Spring 2002

Box one The Marriage Act 1994 amended the Marriage Act 1949 (Section 35) to permit marriages to be solemnised in registration districts in which neither party to the marriage resides. In par ticular, two new sub-sections were inserted: “A superintendent registrar may issue a certificate or, if the marriage is solemnised by licence, a certificate and licence, for the solemnisation of a marriage on approved premises, notwithstanding that the office is not within a registration district in which either of the persons to be married resides.” “A superintendent registrar may issue a certificate or, if the marriage is solemnised by licence, a certificate and licence, for the solemnisation of a marriage in the office of another superintendent registrar, notwithstanding that the office is not within a registration district in which either of the persons to be married resides.”

from a distance, and so, by their very nature, marriages in approved premises might be expected to consist of proportionately more “away marriages” than register office marriages. Indeed, the provision of the ability to marry away from home was undoubtedly a necessary pre– condition for marriages in approved premises to be a viable new institution. As a result, this new provision is fundamental to the understanding of the growth in marriages in approved premises. Whilst the main attention in this article will inevitably be focussed on “away marriages” in approved premises, comparative analyses are also included of “away marriages” in register offices, not only to provide comparisons but also to complete an appraisal of the effect of this particular provision of the Marriage Act 1994. Religious marriages were unaffected by the Act, and the situation regarding residence has not changed (see Box 3). The Anglican Church has recently been considering these issues 4, in the light of both the Marriage Act and also of prospective changes in registering civil marriages, but at the time of writing, no final decisions have been made (see Box 4). For all these reasons, this article concentrates upon civil marriages.

Box three RESIDENCE QUALIFICATIONS FOR

(Note: underlining added here to indicate the distinction – approved premises and register office)

Box two RESIDENCE QUALIFICATIONS Before the Act, for civil marriages, couples had to marry in the district of residence of one of the marriage partners. (A residence qualification was defined in terms of the registration district in which the person concerned was residing; for example, a London Borough, or a Unitary Authority. Shire counties usually comprise between three and five districts, or sometimes more (see Figure 7c).) If a marriage partner moved into a district and resided there for 7 days, they could give notice that, for the past 7 days, their residence had been in that district, and use that address. There is, of course, no way of estimating how many couples may have established such a residence qualification to marry in the registration district - and Register Office - of their choice. However, others may have wrongly given the address of a relative or friend, rather than their own, and again, it is impossible to ascertain the extent to which it occurred. In contrast, since the implementation of the Act, couples can marry with a civil ceremony in any registration district of England and Wales, regardless of their districts of residence. It may reasonably be assumed therefore that, since the Act, the addresses of the par tners before their marriage, as recorded in their marriage entr y, are more likely to be their longer-term (and true) residential addresses before marriage. Another relevant consideration – which has applied both before and after the Act – is the fact that a large proportion of couples premaritally cohabit – and therefore have identical residential addresses 6,7 before marriage.

National Statistics

36

RELIGIOUS

MARRIAGES

Marriages in the Church of England and Church in Wales remain subject to the provisions of the Marriage Act 1949 whose basic qualification is that of actual residence. That is, one of the marriage partners must live in the parish and the registration district in which the church or chapel is situated, or be on the electoral roll of that building. Archbishops of Canterbur y have the power to issue a special licence for marriage – which is, in effect, the granting of a derogation from the marriage laws – which dispenses with the normal residence requirements and enables couples to be married in, say, parish churches or unlicensed buildings. In practice, such licences are only issued where it can be demonstrated that there is a genuine, substantial and continuing connection with the church or chapel concerned. The vast majority of applications received are for those who wish to return to a family home to be married, and, according to the Faculty Office, such applications are almost always granted. 3 All marriages which are not solemnised according to the rites of the Church of England or Church in Wales must be preceded by civil preliminaries. This involves giving notice to the Superintendent Registrar in the district of residence. The Superintendent Registrar can issue a certificate (or a certificate and licence before 2001) for a marriage in a registered building in a different registration district only if one or both of the marriage partners can show it is their place of usual worship, or if there is no building in the district(s) in which they live where people of their faith can marry.


Box four RESIDENCE QUALIFICATIONS FOR MARRIAGES IN THE CHURCH OF ENGLAND In 1997, the General Synod of the Church of England agreed that there were a number of aspects of marriage law which should be reviewed – including residence qualifications, and the possibility of universal civil preliminaries. Shortly afterwards, the Government announced a review of the civil registration system including the registration of marriages. At the star t of 1999, the Archbishops’ Council set up a Working Group to review marriage law, allowing for previous proposals and views, and the Government’s proposals – with the aim that any legislative changes which the Church wished to make could be tied in with the Government’s own legislative programme. The Working Group’s Report4 was published in October 2001. From their consultation, a clear majority was in favour of relaxing the present legal requirements on marriage venues. However, the Group did not think it appropriate to allow couples a completely free choice as to where they could be married according to the rites of the Church of England, or to extend any relaxation of the general rules on places of marriage beyond parish churches and other parochial places of worship. The Working Group made a series of recommendations including the following: •

Parishioners are entitled to marry in their parish church or place of worship, or in a church on whose electoral roll one or both of the couple are entered.

•

Flexibility should be introduced to allow couples who do not meet the above residence qualifications to marr y in such parochial places of worship if they can demonstrate a connection with the building in question.

•

The Working Group recognised that occasions arise where a couple might wish to marry in a non– parochial place of worship and recommend that such marriages should take place only on the authority of a Special Licence so that the Faculty Office can consider the case.

•

A Special Licence should be used in exceptional cases where there are very strong pastoral reasons for permitting a dispensation from the normal rules.

•

The Working Group expressed the hope that the criteria under which a Special Licence is granted could be reviewed to take account of the Group’s recommendations for a “demonstrable connection”.

•

The standard cases which should be accepted as establishing a ‘demonstrable connection’ should be where the couple can show that:

(i) one or both parties have present family homes in the parish (parents’, grandparents’ homes, etc.); (ii) one or both parties have been resident in the parish in the past for a substantial period; (iii) parents, grandparents, etc of one or both of the parties are resident in the parish, have their names on the electoral roll, or were resident for a substantial period; (iv) one of the parties was baptised in the church/place of worship; (v) one or both parties have been regular worshippers in the past at the church/ place of worship. It is expected the Church of England will take decisions on these issues at the next General Synod – in July 2002.

Spring 2002

BACKGROUND AND UPDATE The previous article2 described the growth of marriages in approved premises in the early months of the new Act up to the end of 1996. During the last 9 months of 1995, there were just under 21/2 thousand such marriages, but during the subsequent year, the number exceeded 15 thousand; a dramatic increase, resulting in one in 18 marriages being solemnised in an approved premises. Since 1996, the numbers have continued increasing year by year, so that during 2000, almost 46 thousand marriages were celebrated in approved premises, representing one in 6 of all marriages. The growth may be appreciated from Figure 1, which shows not only the monthly number of marriages in approved premises, but also the corresponding underlying trend, using a 12–month moving average. This trend closely follows that in the number of premises approved for solemnising marriages, estimated from copies of the list of approved premises published by the General Register Office, based on notifications by local authorities (see Box 3 in previous article).2 It may be seen that there is a pronounced seasonal pattern in the monthly numbers of marriages in approved premises, the peak month being August, with marriages in adjacent months also being high. There is a minor peak, too, in May each year. Of course, there is a seasonal pattern to marriages as a whole, so the question naturally arises whether marriages in approved premises show a greater degree of seasonality than other marriages. Figure 2 throws light on this matter, giving the proportions which marriages in approved premises form of both civil marriages and of all marriages. It may be seen that a similar seasonal pattern – with peaks in both May and August – occurs in the proportion of civil marriages, indicating, not surprisingly, that marriages in approved premises are more seasonal than those in register offices. As a proportion of all marriages, however, the seasonal pattern is much less, though still evident, so marriages in approved premises are slightly more seasonal than all marriages. Interestingly, the peaks are in May and October, rather than May and August. However, the important results from Figures 1 and 2 are that marriages in approved premises – both in numbers and in proportions – have grown uniformly since the first full year of the Act, with the number of marriages slightly outpacing the number of approved premises. It was to be expected that the number of marriages in approved premises would increase in the very early years of the new Act as new applications to local authorities for premises to be approved were granted after the implementation of the Act. Also, it is not surprising there was a build–up of marriages in approved premises in the first few years, since couples may have been unaware of all the possibilities for marrying there, and approved premises may not initially have been able to cater for all the marriages – or be able to accept bookings for the dates the couples requested. What is perhaps somewhat surprising is that the increase should have continued unabated into the fifth year of the Act – not only in terms of numbers, but also the growing proportions these marriages represent of all marriages. That is, marriages in approved premises have continued growing in relative importance, without, as yet, any signs of slowing down or reaching a plateau. It is tempting to interpret this phenomenon as a decisive social change in which couples view marriage differently – certainly not as a religious or sacramental act – indeed one of decided secularity – but as one which permits an essentially social occasion in an informal or leisure setting. Also given the place and character of some approved premises, the marriages in them may also be thought as having a special “theme” or interest. Of course, self–evidently, the undoubted change which has taken place has only been possible because legislation has made it so. That



Figure 1

Spring 2002

Monthly number of marriages in approved premises, and underlying trend, and number of approved premises, 1995–99

England and Wales

Possibly too much emphasis has been placed on the advent of approved premises as such – after all, the form of civil marriage celebrated in them is of exactly the same type as Register Office ones. It might be that it is the greater freedom to marry where one wishes, in particular away from home, or in a new place of one’s choosing, which has been the key factor in the growth in the popularity of marriages in approved premises. It is this feature which is the subject of the present article.

Number 6,000

DATA SOURCES AND INFORMATION AVAILABLE FROM THEM

Number of marriages in approved premises

Two sources of data have been used in this article: copies of marriage entries – commonly called “marriage certificates” – received from Superintendent Registrars; and a number of lists of approved premises, produced by the General Register Office for sale to the public.

Number of approved premises

5,000

4,000

ONS receives a copy of the marriage entry for each marriage in England and Wales – and extracts certain statistical information5 to add to the ONS marriages database. A fictitious copy of a marriage entry – showing an away marriage – is shown in Figure 3. Details such as the manner of solemnisation of the marriage – whether with a religious or a civil ceremony, and, if the latter, whether officiated in an approved premises or a register office, are recorded for each marriage. For the purposes of the present study, the registration districts of three addresses were required:

3,000

2,000

1,000

0 1995

1996

1997

1998

1999

1. of the bride’s residential address before marriage; 2. of the bridegroom’s residential address before marriage; 3. of the address of the approved premises/register office where the marriage was solemnised.

Month/year

Figure 2

Marriages in approved premises as a percentage of all civil marriages, and all marriages, 1995–99

England and Wales

Only the third registration district was available from the marriages database, i.e. that of the approved premises or the register office. So two separate systematic samples of marriages in England and Wales in 1998 were taken; one for marriages in approved premises, and the other for marriages in register offices. For each of the marriages in the two samples, the addresses before marriage of both the bride and the bridegroom were postcoded and written on the paper copy of the sample marriage entry. In addition, for the sample of marriages in approved premises, the type of approved premises – for example hotel, stately home, etc – was also added in the form of a code to the paper copy of the sample marriage entry. (Further details of the types of approved premises – and their coding – were given in the previous article.)2

Percentage 30 As a percentage of all civil marriages 25 As a percentage of all marriages 20

In the final stage, the sample marriage computer records were extracted from the marriages database, and the extra details – the two postcodes and the type of approved premises, where applicable – were added to each record. Finally, the registration districts corresponding to the two postcodes were generated and added to the two files of sample marriages. Fuller details are given in Appendix 1.

15

10

DIFFERENT TYPES OF HOME AND AWAY MARRIAGES 5

0 1995

1996

1997

1998

1999

Month/year

is, the potential demand for the kind of marriage setting which approved premises allow could well have been present, if unfulfilled, for many years. Also, the introduction of the possibility of marrying in approved premises may have coincided with – or reflected – the growth in secularity in society generally, and in the secularity of marriage in particular.

National Statistics

38

There are 5 different ways in which the registration districts containing the addresses of the bride, the bridegroom and the place of marriage can be all the same, or all different, or two the same and the other one different. These are set out in Table 1 which gives the proportions of marriages in each of the five types. In the present context, the types of marriages which are away marriages are of greatest interest – the remaining types together forming what will be termed “home marriages”, for brevity. It is important to emphasise that away marriages – as referred to in this article – are those which have been made possible for the first time by the new Act – and consist of marriages in which either the three addresses mentioned above all fall in different registration districts, or where the registration district of the place of marriage is different from


Figure 3

Spring 2002

Fictitious copy of a marriage entry showing an away marriage

bride’s and bridegroom’s addresses before marriage were identical

Approved premises ➤

➤

type: hotel

postcode WD7 9ZZ 532 Hatfield

registration district code 536 atford Watf Watford

Bar net Barnet 219

536 232 Harro w Harrow

236

Ealing 226

registration district code (from marriages database) 227

Bro xbourne Broxbourne 529 Enfield 227 255

Hendon 235 221 Brent

Hillingdon

➤

➤

St Albans 535

Har ingey Waltham Haringey idge Redbridge altham Redbr 233 247 orest Forest ering ver Havering Bar king Ha Barking Hac kney 250 Hackney 234 and Camden 238 230 Ne wham Dagenham Newham 257 218 243 246 258

Thurroc Thurrockk 477

260 239

Southw ark Greenwich Southwark Hounslo w Hounslow 229 251 Be xley Bexley 237 andswor orth Wands Wandsworth 220 242 Richmond 256 Lambeth Le wisham Lewisham upon Thames 241 248 Mer ton Merton 244 Bromle Bromleyy Kingston upon Thames 222 Cro ydon Croydon Sutton 240 225 254

Medw Medway ay 561 Gr esend Gravesend avesend 562



Spring 2002

the registration district in which both partners were living before their marriage. Conversely, home marriages – which were the only ones possible before the Act came into force – consist of all those marriages where the registration district of the place of marriage is the same as either that of the address of one or other, or both marriage partners.

THE CHARACTERISTICS OF COUPLES MARRYING WITH HOME AND AWAY MARRIAGES

Age at marriage

Table 1 shows the profile of the different types of home and away marriages separately for marriages in approved premises, register offices and for all civil marriages. For each set of marriages, the majority fall in the first two types, A and B in Table 1; either type A, where all three registration districts are the same – the traditional form of home marriage – or type B where the bride and bridegroom had both been living in the same registration district, but married in another – the numerically most important kind of away marriage.

A feature of couples marrying with an away marriage in approved premises is that their ages tend to be more in the peak marrying ages than those marrying with a home marriage – see Figure 4a. More particularly, amongst all brides marrying with an away marriage, there are relatively more who are aged between their mid–twenties and their mid–thirties than there are amongst all those marrying with a home marriage. About two thirds of brides marrying with an away marriage in an approved premises are aged from 25 to 34, whereas the corresponding proportion for home marriage brides is only just over one half.

Two thirds of all civil marriages are the traditional home marriages, type A, and more than one quarter are away marriages of type B. Overall, about three in every 10 civil marriages are estimated to be away marriages. However, amongst marriages in approved premises, over one half, 55 per cent, are away marriages, whereas the corresponding proportion for register office marriages is less than one quarter, 23 per cent.

A somewhat similar phenomenon – though less pronounced – also occurs for couples marrying in away marriages in register offices (Figures 4c and 4d). However, in this case the brides and bridegrooms are slightly more likely to be aged within a broader range of peak marrying ages, 25 to 44, than those marrying with a home marriage.

It is perhaps remarkable that the contrast between these two proportions should be so large. It might have been expected that a majority of marriages in approved premises would be away marriages, but relatively few register office marriages are away marriages. Most likely, those which are away marriages consist largely of those where the bride and bridegroom have chosen to marry in a register office in an immediately neighbouring registration district. Nevertheless, register offices have become more attractive places for weddings and have responded not only to the growing challenge which approved premises represent, but also of the competition from other register offices. It is possible, therefore, that the proportion of away marriages amongst register office marriages will rise. Table 1

Profile of marriages by combination of registration districts of the bride’s and bridegroom’s addresses before marriage and of the place where the marriage was solemnised, 1998

England and Wales

Percentages

Types of home/away marriages Registration districts of addresses of:

Approved premises marriages

Register office marriages

A. bride, groom and place of marriage all the same

42

B. bride and groom the same, but place of marriage different

51

22

27

C. bride and place of marriage the same, but groom different

2

4

4

D. groom and place of marriage the same, but bride different

1

3

2

E. bride, groom and place of marriage all different

4

1

2

55 45

23 77

29) 71)

100 2,406

100 1,999

100 4,405 *

(F. All away marriages; F = B+E (G. All home marriages; G = A+C+D Total - percentage - sample no. *

70

All civil marriages*

40

Marital status before marriage The patterns of home and away marriages do differ between those who were single and those who were previously divorced – see Table 2 – as might be expected from the different age profiles discussed above. Amongst all marriages in approved premises, just over one half, 51 per cent, involved a marriage which was the first for both partners, but the corresponding proportion was larger amongst away marriages, 55 per cent, and lower amongst home marriages, 45 per cent. That is, first–time couples marrying in approved premises figure relatively more frequently amongst away marriages than they do amongst home marriages – an understandable finding. Conversely, couples in which both partners are remarrying after a previous divorce account for relatively fewer amongst all away marriages than amongst all home marriages; only one in 7 of away marriages, but over one in 5 of home marriages.

65

Estimated by weighting the two sets of sample numbers to accord with the national numbers of AP and RO marriages in 1998.

National Statistics

It is noticeable, too, that for older couples marrying in approved premises – those where the brides and bridegrooms are in their fifties or older – that away marriages are relatively only about half as popular as home marriages. Possibly older brides and bridegrooms, including some who have had previous marriages, do not necessarily want their marriage to involve a large social gathering, or a romantic setting, and so settle for a quieter ceremony in a convenient venue close to where they live. Conversely, those in the peak marriageable ages, especially for those whose marriage is their first, are perhaps understandably keen to invite many relatives and guests and make a real occasion of their wedding by choosing a special setting away from home.

The corresponding results for register offices are also given in Table 2. There is much less difference in the profile of the marital statuses of the couples between home and away marriages than with approved premises marriages. Couples where both are remarrying after a previous divorce account for a larger proportion of all register office marriages than of all marriages in approved premises – and these couples also account for larger proportions of both home and away marriages amongst register office marriages than the corresponding proportions amongst marriages in approved premises. The final two columns of Table 2 provide estimates of the proportion of, say, first marriage couples who married with an away marriage. For marriages in approved premises, couples where both were marrying for the first time were the most likely to have an away marriage – six in every 10 did so. Couples in which the bride was single and the bridegroom divorced were the next most likely, and so on, with couples where both were remarrying after a previous divorce being the least likely to have an away marriage.


Figure 4

Spring 2002

Profile of ages at marriage of brides and bridegrooms for home and away marriages, 1998

England and Wales Percentage 40

Percentage

(a) Age of Brides

Away Home

Marriages in approved premises

35

30

25

25

20

20

15

15

10

10

5

5

0

0

Percentage

(b) Age of bridegrooms

Away

Home

(d) Age of bridegroom

Away

40

Home Marriages in approved premises

35

Away

Marriages in Register Offices

35

30

Percentage 40

(c) Age of brides

40

30

30

25

25

20

20

15

15

10

10

5

5

0

Home

Marriages in Register Offices

35

0 16–19

20–24

25–29

30–34

35–39

40–44

45–49

50 and over

16–19

20–24

25–29

Age at marriage

Table 2

30–34

35–39

40–44

45–49

50 and over

Age at marriage

Marital statuses before marriage of bride and bridegroom, for home and away marriages, 1998

England and Wales

Percentages

Previous marital status in combination Bride

Bridegroom

single single divorced divorced

single divorced single divorced

other combinations All combinations - percentage - sample no.


Percentage of away marriages amongst all marriages in:


Approved premises

Register offices

43 14 15 24

60 56 51 46

23 26 18 26

5

5

46

20

100 1,535

100 1,999

55

23

Away

Home

Total

Away

Home

Total

55 16 11 14

45 16 14 21

51 16 12 17

43 16 11 26

43 14 16 23

3

4

4

4

100 1,334

100 1,072

100 2,406

100 464

Note: single means never-married.



Spring 2002

The apparent pattern to these results is that one partner being single increases the likelihood of marrying with an away marriage, and both partners being single increases the likelihood still further. In addition, the bride being single has more effect than the bridegroom being a bachelor. Such an explanation is quite plausible – first marriage is understandably regarded as a major event, and brides might be expected to influence the choice of marriage venue more than the bridegrooms. For register office marriages, the relationship between previous marital status and the proportion of away marriages is quite different; the largest proportion of away marriages occurs for couples where both partners were previously divorced – who were the group of couples with the lowest proportion of away marriages amongst couples marrying in approved premises. However, the proportions of the register office marriages which are away marriages are very similar between the different groups of couples – and are all much smaller than the corresponding proportions of away marriages in those of approved premises. Quite irrespective of whether the marriage is a first or a subsequent one, there is perhaps less incentive for choosing a register office which is not the “home one” of either the bride or the bridegroom, unless there are strong practical reasons due to work or family. However, it is possible that the nearest register office geographically is not the “home one”.

Living together before marriage Previous research6,7 has provided evidence that many marriage partners who have given the same residential address before marriage have cohabited pre–maritally. That is, identical addresses of the bride and the bridegroom is a good indicator that the couple lived together before marriage. Quite apart from estimating the prevalence of living together, these addresses are of interest in the present context of home and away marriages – and throw additional light on the types of home and away marriages shown in Table 1.

reason, may be indicative of an underlying causative factor. Possibly away marriages are arranged where couples will live after their marriage – and they moved in together before the marriage and before moving. Alternatively, it could simply be a case that the organisation of an away marriage is easier if the couple are already living together, particularly, say, if one of the partners is changing jobs. A further possibility is that of a selection effect; couples who are already living together may be more predisposed to choose an away marriage.

MONTHLY AND DAILY PATTERNS OF MARRIAGE There is some monthly variation in the proportion of away marriages amongst both register office marriages and those in approved premises as Figure 5 indicates. Somewhat surprisingly, the variation is larger for register office marriages, with the proportions being comparatively large during the spring and early summer – April, May and June – and during the early autumn – August and September. To a large extent, the same peaks also occur amongst the proportions for marriages in approved premises, although they are not so pronounced. There is an interesting contrast in that the proportions of away marriages in both December and January are particularly large amongst register office marriages but particularly small amongst marriages in approved premises. Undoubtedly the attractions of approved premises are enhanced in the better–weather months, and an away marriage in a register office in one of the two months close to Christmas provides the ideal format for a quiet and private wedding. The overall profiles of home and away marriages by month of the year are broadly similar – whether one considers those in approved premises or those in register offices – see Figure 6. However, the seasonality is slightly more pronounced amongst away marriages than home marriages – for both approved premises marriages and register office ones. In particular, the proportions marrying with away marriages in May and August are larger than those with home marriages in these months.

Overall, the marriage partners addresses were identical in 88 per cent of all approved premises marriages, and in 84 per cent of all register office marriages (Table 3). (The corresponding proportion for religious marriages – not shown in Table 3 – was much lower at 51 per cent.) Consequently, we might expect a slightly larger proportion of couples marrying in approved premises to have lived together before marriage, compared with those marrying in register offices. Using this measure, couples where both partners were remarrying after divorce were the most likely to have lived together beforehand – both for register office marriages and also for those in approved premises. The results for home and away marriages show that the proportions of couples who lived together were almost always larger for away marriages than for home marriages – for both register office and approved premises marriages.

Previous work8 has shown the present day popularity of Saturday weddings – and also that the profile of marriages by day of the week has changed considerably over the centuries. More recently, the current analyses provide the additional opportunity of comparing the daily marriage patterns of home and away marriages – and these are shown in Table 4.

Although the differential between home and away marriages in living together before marriage is not large, it is consistent, and, for that

The most important result is that one in ten marriages in approved premises are solemnised on Sundays (whereas there are none on

Table 3

The difference between register office marriages and those in approved premises in the early months of the year can also be seen in Figure 6; there are relatively fewer marriages in approved premises than in register offices, during the months from January to April, inclusive, for both home and away marriages.

Percentage of marriages in which the bride’s and bridegroom’s addresses before marriage were identical, for home and away marriages, by marital status before marriage, 1998

England and Wales

Percentages

Previous marital status in combination Bride

Bridegroom

single single divorced divorced

single divorced single divorced



Away

Home

Total

Away

Home

Total

89 90 93 93

88 86 79 87

89 88 86 90

88 84 96 91

80 84 84 89

82 84 86 90

other combinations

70

78

74

84

67

70

All combinations

90

86

88

89

82

84

Note: single means never-married

National Statistics

42


Figure 5

Percentage of marriages which were away marriages, by month, 1998

England and Wales Percentage 60


40


20

10

0 Jan

Feb

Mar

Apr

May

Jun Jul Month

Aug

Sep

Before the Act, civil marriages on a Sunday were technically possible, although ONS advised Superintendent Registrars that they were not obliged to open their register offices on Sundays, since to do so would mean staff working seven days a week. Hotels and other similar kinds of approved premises naturally view weekends as prime business time, and argued for the availability of wedding staff. Since the new law allowed local authorities to recover the cost of providing such staff – which the statutory fee for a register office wedding would not meet – Sunday weddings have blossomed. However, although the emergence of Sunday weddings represents one very tangible effect of the Marriage Act 1994 (there were previously only a very few religious ceremony weddings on Sundays), Saturdays remain the overwhelmingly most popular day, followed by Friday. About six in every 10 couples marrying in approved premises choose a Saturday, and about five in every 10 couples marrying in register offices. Overall, the proportions marrying on the different days of the week are very similar for both home and away marriages in approved premises – and the same is true for marriages in register offices.

50

30

Spring 2002

Oct Nov Dec

The proportion of approved premises marriages which are away marriages shows an interesting variation by day of the week (penultimate column in Table 4). The proportion is largest for Mondays weddings at 67 per cent (not shown in Table 4), followed by Sunday weddings at 62 per cent, and smallest for Friday and Saturday marriages. Possibly Friday and Saturday weddings are easiest to arrange as home weddings so reducing the proportion of away marriages on those days.

REGIONAL AND COUNTY PATTERNS OF MARRIAGE Figure 6

Profile by month of home and away marriages, 1998

The regional patterns of home and away marriages are broadly similar, albeit with a few exceptions as Table 5 indicates. Almost one quarter of all away marriages in approved premises are solemnised in the South East, and a further one in 8 in the North West, the region with the second largest proportion. In Yorkshire and the Humber, away marriages in approved premises account for only 7 per cent of all such marriages in England and Wales, whilst the region’s corresponding proportion of all home marriages is 15 per cent. In North Yorkshire many former register offices have been redesignated as approved premises for marriages. Consequently a number of home marriages in what would have formerly been register offices are now counted amongst those in approved premises. Also, as may be seen from Figure 7c, North Yorkshire consists of a single, particularly large, registration district, rather than a number of smaller ones. Therefore marriages are more likely to be home marriages than away marriages.

England and Wales Percentage 18 Approved premises away marriages

16 14 12 Register Office home marriages

10 8 6

Approved premises home marriages

4 Register Office away marriages

2 0 Jan

Feb

Mar

Apr

May

Jun Jul Month

Aug

Sep

Oct

Nov Dec

Sundays in register offices in the sample, as they tend to be closed). The proportion of Sunday weddings is even larger – one in 8 – amongst away marriages in approved premises, but slightly smaller amongst home marriages, at one in 11, so couples taking advantage of the provision of away marriages have chosen Sundays more frequently than those marrying with a home marriage.

As may be seen, the corresponding proportions of register office marriages – both home and away – which took place in Yorkshire and the Humber are smaller than the corresponding proportions for the marriages in approved premises. A further contrast between register office marriages and those in approved premises concerns those solemnised in London; over one in 4 of all register office away marriages and one in 5 home marriages take place in London – each about three times the corresponding proportion for those in approved premises. Civil marriages have always been in the large majority in London, but it also has some prestigious register offices, and ones in prime locations, as well as many who have made efforts to meet the challenge of the new possibilities for civil marriages in approved premises. In addition, London has relatively few approved premises, considering its large population, so fewer marriages in approved premises might be expected there – whether home marriages or away marriages. Another contrast between register office marriages and those in approved premises concerns the share of the national number solemnised in the South East. Whereas at least one in 5 of all away and home marriages in approved premises are celebrated there, the corresponding proportions are only about one in 7 of all register office marriages. Part of the explanation is likely to be the converse of that for



Table 4

Spring 2002

Profiles by day of the week of home and away marriages, 1998

England and Wales

Percentages

Day of the week of marriage

Approved premises marriages Away

Monday – Wednesday* Thursday Friday Saturday Sunday All days - percentage - sample no. *

Home

Percentage of away marriages amongst all marriages in:

Register office marriages Total

Away

Home

Total

Approved premises

Register offices

2 4 22 57 12

2 4 21 61 9

2 4 22 59 10

6 8 21 55 0

6 8 23 50 0

6 8 22 51 0

57 58 56 54 62

21 23 22 25 -

100 1,334

100 1,072

100 2,406

100 464

100 1,536

100 1,999

55

23

Average for these three days; total includes three times this average.

Table 5

Regional profiles of home and away marriages, and percentages of away marriages, by region in which solemnised, 1998

England and Wales

Percentages

Government Office Region in which marriage solemnised

Approved premises marriages Away

Home

Register office marriages Away

Home

Percentage of away marriages amongst marriages in region in: Approved premises

Register offices

North East North West Yorkshire and the Humber

4 13 7

3 14 15

4 9 6

5 10 9

65 53 38

20 20 16

East Midlands West Midlands East of England

8 7 12

7 6 12

7 11 8

8 10 10

59 60 56

22 24 19

London South East

7 23

6 20

26 15

20 15

60 59

28 23

South West Wales

11 7

11 6

10 3

8 5

54 59

29 16

100 1,333

100 1,073

100 461

100 1,538

55

23

England and Wales - percentage - sample no.

London; the South East has relatively many approved premises. (Also, many register office marriage rooms in Kent were redesignated as approved premises.) Table 5 also provides estimates of the proportion of marriages in each region which are away marriages (final two columns). For marriages in approved premises in England and Wales as a whole, just over one half, 55 per cent, are away marriages. This proportion is exceeded in several regions – and Wales – with the North East, London and the West Midlands having the largest relative numbers of away marriages, and Yorkshire and the Humber, the North West, and the South West having the smallest. The regional pattern of the proportions of away marriages in approved premises must reflect the numbers and locations of these premises, and their attractiveness – including their relative attractiveness to register offices. However, it could also be the case that there are different preferences and patterns of choice in the different regions with regard to both home and away marriages – and also to marriages in approved premises and register offices. The numbers of home and away marriages in the two samples are sufficient to examine the proportions of away marriages in each county of England – and these are depicted in Figures 7a and 7b, for marriages in approved premises and register offices, respectively. Of course, many away marriages involve the addresses of the bride, bridegroom and place of marriage all falling within the same county, albeit within different registration districts. Figure 7c shows the registration districts of England (and Wales); with a total of 352 in England in 1998, there

National Statistics

44

were approximately 8 per county, on average. In Figures 7a and 7b, counties are indicated by whether they fall in the top, or bottom, third of the range of proportions of away marriages – which correspond to boundaries of about one fifth above or below the national, England and Wales, proportion. (Results in Figures 7a and 7b are shown only for the counties of England – that is, the pre–1995 areas before local government reorganisation5 took place; a number of 1998 registration districts in Wales straddled the pre–reorganisation Welsh county boundaries, which, together with smaller sample numbers, made estimation impractical. Also, the number of marriages solemnised in each Unitary Authority which existed in 1998 – comprising of one or more registration districts – was included in the total for the county from which the Unitary Authority had been created.) There are two main areas where away marriages amongst marriages in approved premises are relatively more numerous than the national average – the group of counties usually described as the “home counties” ranging from Gloucestershire through to West Sussex, and the northern remoter rural counties of Cumbria, Northumberland and Durham (Figure 7a). Perhaps because of their relatively sparser population, couples who wish to marry in an approved premises are much less likely to have one within the districts where they both live, other factors being constant. In contrast, amongst the counties in which the proportion of away marriages is lower than the national average, is the set of more populated


Figure 7a

Spring 2002

Percentage* of marriages in approved premises which were away marriages, by county of England in which solemnised, 1998

* estimated

Percentage of away marriages (England and Wales=55) 65