Geography and Geographical Analysis using the ONS Longitudinal Study Christopher Marshall & Julian Buxton CeLSIUS
Aims of the Presentation What is the ONS LS and what data does it contain? What geographical information is in the LS and at what level? What does this allow us to do? The strengths and weaknesses of using geographical data in the LS. Examples of using geographical data in the LS. The Role of CeLSIUS (Centre for Longitudinal Study Information and User Support).
The ONS Longitudinal Study Census data for individuals with one of four birthdates enumerated at the 1971 Census (c. 1% of population) Maintained i through h addition of immigrants i and new births with LS birth date Information from later censuses (1981, 91 & 2001) added and linked to that already there. Event data including deaths of LS members, cancer registrations, death of spouse, births to female members, and now, under test, the Claimant Count Cohort. Birth Birth of of Child Child Census Emigration Re-entry Death of Spouse Census Cancer Death 1971 1976 1981 1986 1991 1996 2001 2006 2011
Study Structure Entrants New Births 214,000 Immigrants 107,000 1971 1981 1991 2001 Original sample: 530,000; selected from 1971 Census 536,000 sample members found at 1981 Census 543,000 sample members found at 1991 Census 545,894 sample members found at 2001 Census Plus members Plus members Plus members Plus members of household of household of household of household Births to sample women 201,000000 Events 1971-2001 Widow(er)hoods 66,000 Cancer registrations 70,000 Deaths 189,000 Embarkations 30,000 1. Census. High N. 2. Linking 3. Non-members 4. Entrance & Termination 5. Events
Geographical Location of LS Members From Census Information Based on Address of Usual Residence on Census Day Visitors flagged prior to 2001 and usually excluded from studies Fully coded to the Administrative boundaries and also to Health boundaries (e.g. Regional Health Authority)
Main Time Points Available 1. 1970 from address 1 year ago in 1971 2. Census day 1971 (25 th /26 th April) 3. 1980 from address 1 year ago in 1981 4. Census day 1981 (5 th /6 th April) 5. 1990 from address 1 year ago in 1991 6. Census day 1991 (21 st /22 nd April) 7. 2000 from address 1 year ago in 2001 8. Census day 2001 (29 th /30 th April) 9. Also possible: 1966 from address 5 years ago 1971 10. 1939 (in theory)
LS Geography (England & Wales) 1971-2005 Census 1971 1981 1991 2001 Region SR 1971 & SR 1974 SR 1974 SR 1974 GOR 2001 County 1971 & 1974 1974 1974 2001 & 2003 County District 1971 & 1974 1974 1974 2001 & 2003 Ward 1971 & 1974 1974 1974 2001 & 2003 Lower Grid Enumeration Enumeration GR, PC, ED Reference District District & & Output Levels Postcode Areas
What does this allow us to do? Migration: Mobility Geographical & Social Categories Use existing definitions Create new analytical definitions Attach and analyse ecological data Create new geographies Analyse specific areas
Social and Geographical Mobility 1971 - Living in NE Social Class: Skilled Non-Manual Tenure: Social Housing 2001 - Living in SE Social Class: Managerial Tenure: Owner Occupier
Geographical Migration Patterns North East 1971 South East 1981 South West 2001
Use of existing definitions - Geographical Continuity for 30 years at 1974 geography despite changing boundaries. Standard Region / Government Office Region 2001 can be mapped to preceding years and lookup tables can bring earlier geographies forward to 2001 County and County Districts can be treated similarly.
Use of existing definitions: Geographical (2) Administrative boundaries Environmental boundaries Ecological Deprivation Indices Area Classifications (Urban / Rural) Population densities Small area statistics (Aggregate level variables)
Create new definitions iti - Geographical Urban Rural Travel to Work Craig Webber Other valid divisions
Create new definitions: Socio-economic Social Class by Sex Social Class by Age Group Working Status Age Group Social Class by Tenure
Attach ecological or social data Any data can be attached to individual LS members if it is produced in the form of a look-up table with a valid geographical reference code (e.g. Ward, County District) attached. Air pollution indicators Average rainfall in 10 years Ecological Deprivation indices
Create new geographies It could be that for your specific purpose the Geographies within the LS are inappropriate. p Define the geography you want based on LS wards or county districts and this can be attached to LS members and used for analysis. e.g. We have regions but you want to divide each into two or three separate areas not wholly based on counties.
Look at specific areas The purpose of your analysis is to look only at a specific area of the country and compare it with one or two others. Depending on parameters chosen the analysis can run into disclosure control restrictions keep the analysis simple with a limited number of parameters. Analysis at Ward level or below would require aggregation of results, while at county district level, outputs do not usually require aggregation. g
Strengths of geographical data in the LS 1. Consistency over time 1971, 1981, 1991 all coded to same base (1974 geography). 2001 can be produced on the same base down to county district level with confidence fairly easily. 2. If necessary earlier data can be brought forward to 2001 by the use of look-up tables.
Strengths of geographical data in the LS 3. 30 years of continuous follow-up of individuals 4. 9 Time points (1966 2001) 5. Consistency of geography through this time period.
Strengths of geographical data in the LS 6. Flexibility of study design 7. Individual and Area data 8. Can add data using geographical g identifiers (e.g. Carstairs deciles) 9. High level of detail available for later data.
Weaknesses of geographical data in the LS 1. While County and County District Codes have remained fairly consistent the Ward codes needed to attach additional data have changed significantly over time. 2. Data are for England and Wales only 3. Members who move to Scotland classed as embarkations (migrants)
Weaknesses of geographical data in the LS 4. Data codings and data detail not consistent. 5. It is not possible to back transfer all geographies. 6. Ward code history many changes and manipulations i lay traps for the unwary.
Scotland Members in 1971 found with a Scottish NHS number were incorporated into the LS. Events to LS members (e.g. deaths) that occur in Scotland are traced and do get linked to LS members. LS members who migrate to Scotland are treated as Emigrants and this is recorded in the LS. Earlier data remain within the LS.
Disclosure Control Rules Researchers should design their projects such that it would never be possible to identify an individual from the output data generated (Population uniques). Output cell counts of 1 or 2 are considered potentially disclosive (although most 2s will be released to users), and for publication purposes some aggregation of data would be required. Exposure times for single events are a Exposure times for single events are a considered a risk and have to be disguised.
Disclosure Control Rules Tables containing data with a mix of any of the following types of variable will be examined more scrupulously: l Occupation Country of birth Industry Ethnicity Cause of Death Higher education levels Sub regional geographical fields
Examples of Using LS Geography within a project 1. Do people move out of London when they retire? 2. Have people moved from Urban to Rural areas between 1991 and 2001?
Research Question: 1. Do people move out of London when they retire? Main Study Population: LS members present at 1991 & 2001, Males: 55-65 in 1991 & Females 50-60 in 1991 All resident within Greater London in 1991 from county code 1991
1. Do people move out of London when they retire? GOR of residence in 2001 by Sex for those living in London in 1991 GOR of residence Sex 2001 Male Female Total % North 12 9 21 0.38 York&Humb 11 10 21 0.38 EMids 24 31 55 0.98 W Mids 9 11 20 0.36 E of Eng 130 130 260 4.64 SE 147 192 339 6.06 SW & Wales 69 67 136 2.43 London-not moved 1,755 2,144 3,899 69.65 London-moved 369 478 847 15.13 Total 2,526 3,072 5,598 100.00 Source: ONS Longitudinal Study
Research Question: 2. Has the population distribution between Urban and Rural areas changed between 1991 and 2001? Main Study Population: LS members present at 1991 & 2001 Age 16-55 in 1991 Report: resident in Urban / Rural Ward classification in 1991 and in 2001 using 1974 boundaries.
2. Distribution ib ti of Urban / Rural residency between 1991 and 2001. Area of residence 1991 v Area of residence 2001 Simple Rural Simple Rural classification 2001 classification 1991 < 400 / sq km >400 / sq km Total (people per sq km) (Rural) (Urban) < 400 / sq km rural 32,985 13,606 46,591 >400 / sq km Urban 17,093 181,639 198,732 Total 50,078 195,245 245,323 Source: ONS Longitudinal Study
The Role of CeLSIUS (Centre for Longitudinal Study Information and User Support) An interface between academics and the Office for National Statistics. Provide - through our Web site - information on: The structure of the ONS LS. How to decide if it is for you. Training modules to assist your understanding of the data and how it can be manipulated. All the documents needed to apply for permission to All the documents needed to apply for permission to use the ONS LS and access to the LS datasets.
The CeLSIUS training modules: Socio-economic indicators LS Outputs Households and families Defining a study population Ethnicity Geography NEW just released
www.celsius.lshtm.ac.uk A Free service for UK academic users General enquiries: celsius@lshtm.ac.uk 020 7299 4634 Emily Grundy Lynda Clarke Christopher Marshall Andy Sloggett Julian Buxton Jo Tomlinson