Webinar: Introduction to the National Child Development Study. Matt Brown, Brian Dodgeon, Tarek Mostafa

Transcription

1 Webinar: Introduction to the National Child Development Study Matt Brown, Brian Dodgeon, Tarek Mostafa

2 Agenda Introduction to the NCDS and what s new in the age 55 survey (Matt Brown, NCDS Survey Manager) How to access the data (Brian Dodgeon, CLS Research Fellow) Documentation available (Brian Dodgeon) About the NCDS sample, including issues of non-response and attrition (Tarek Mostafa, CLS Research Officer) Using the mixed-mode data from the age 55 survey (Matt Brown) The Future (Matt Brown) Questions

3 Introduction to the NCDS and what s new in the age 55 survey Matt Brown (NCDS Survey Manager) Centre for Longitudinal Studies Institute of Education University College London

4 Centre for Longitudinal Studies Centre for Longitudinal Studies (CLS) is responsible for the conservation, development and dissemination of three British Birth Cohort Studies (1958, 1970, Millennium) and also Next Steps (formerly, the Longitudinal Study of Young People in England) We are funded as a resource centre by ESRC

5 Timeline of the studies National Child Development Study (1958) British Cohort Study 1970 Next Steps Millennium Cohort Study

6 Birth cohort studies Fully representative samples of the GB/UK population Based on one week s/12 months births c17k babies Followed up from birth into adulthood Multi-purpose & multidisciplinary Resources for the research and policy communities

7 Resource for research and policy communities Long term outcomes of experiences and decisions in early life Medium and short-term outcomes & links between different life domains (e.g. health and employment) Descriptions of individual trajectories careers, relationships, fertility, poverty and disadvantage Intergenerational transmission of advantage and disadvantage and the processes involved The links between social change and the changing experiences of different cohorts

8 About NCDS Began as Perinatal Mortality Study Data collected about births and families of just over 17,000 babies born in England, Wales and Scotland in one week in 1958 Early focus was medical but with each sweep the scope has broadened: Childhood sweeps: educational and physical development Transitions into adult life: Family formation, employment etc Formation and maintenance of adult identity. Examine the impact of experiences on later life outcomes and achievements Funded by the ESRC

9 NCDS 58 A study of everyone born in one week in Birth main respondent mother parents parents cohort member / parents subject subject subject subject subject subject subject secondary respondent medical medical / school medical / school medical / school partner mother children survey instruments cognitive assessments cognitive assessments cognitive assessments cognitive assessments linked data area of residence (census) area of residence (census) response rate 17,415 15,425 15,337 14,654 12,537 11,469 11,419 9,377 9,534 9,790 9,137

10 NCDS 58 A study of everyone born in one week in Birth main respondent mother parents parents cohort member / parents subject subject subject subject subject subject secondary respondent survey instruments linked data medical medical / school medical / school medical / school In 1965, 1969 and 1974 the cohort cognitive was augmented cognitive cognitive by the addition assessmentmentments of immigrants assess- assess- to Britain who were born in the target week in 1958 area of residence (census) area of residence (census) partner mother children tests response rate 17,415 15,425 15,337 14,654 12,537 11,469 11,419 9,377 9,534 9,790 9,137

11 NCDS 58 A study of everyone born in one week in Birth main respondent secondary respondent survey instruments mother parents parents medical medical / school cognitive assessments medical / school cognitive assessments cohort member / parents subject subject subject subject subject subject 1971 and 1981 census data medical describes / area of school 8 residence in 1974 and 1981 cognitive assessments partner mother children tests linked data area of residence (census) area of residence (census) response rate 17,415 15,425 15,337 14,654 12,537 11,469 11,419 9,377 9,534 9,790 9,137

12 NCDS 58 A study of everyone born in one week in Birth main respondent mother parents parents cohort member / parents subject subject subject subject subject subject secondary respondent survey instruments medical medical / school cognitive assessments medical / school cognitive assessments medical / school cognitive assessments partner mother children Mother and Child Survey a sample of 1 in 3 cohort members tests linked data area of residence (census) area of residence (census) response rate 17,415 15,425 15,337 14,654 12,537 11,469 11,419 9,377 9,534 9,790 9,137

13 Topics covered by life stage Birth School years Adult Family Parental employment Obstetric history Smoking in pregnancy Pregnancy (problems, antenatal care) Labour (length, pain relief, problems) Birth (problems, sex, weight, length) Family Parental employment Financial circumstances Housing Health Behaviour School Views and expectations Attainment Family (partners, children) Employment Income Housing Health Health-related behaviour Courses and qualifications Basic skills Cognitive ability Views and expectations

14 Biomedical survey at age 45 AIM: To obtain objective measures of ill-health/biomedical risk Physical measurements: Standing & sitting height, weight, waist & hip Blood pressure & pulse rate Lung function Blood collection for a range of assays and DNA Vision - Near (reading) and distant (2m) vision, stereopsis and refractive error Hearing Psychiatric diagnostic interviews: CIS-R Saliva collection for early-morning cortisol level (stress)

15 Cross cohort comparisons Age Year

16 Cross cohort comparisons 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Smoking during pregnancy 29% 41% Breast-fed 38% 69% Measles by age 10/11 50% 92%

17 Some findings Poor school attendance and truancy is associated with problems in adult employment and marital breakdown. Being bullied at school has long term psychological and other effects. A key predictor of children s educational attainment is interest shown by parents. There are substantive lifetime wage returns for men and women to attending higher education. Youth unemployment exerts a long-term wage scar.

18 Some findings (2) Psychological problems in childhood can have substantial long-term economic and social consequences. Mothers who smoke during pregnancy are more likely to have children with low birth weight and, by age 16, reduced height and lower reading and mathematics attainment. Breast feeding is associated with reduced risk of heart disease in adult life. Chronic lower respiratory tract illness in adults is associated with poor home circumstances in early life

19 Age 55 Survey - Context Time of change: At 50 most were in employment by 60 a third of men and half of women may no longer be working o The well off who can retire early o Those with low incomes who cannot work because of poor health Changing family circumstances - children leaving home, arrival of grandchildren, loss of parents Disease and ill health still fairly rare but poor health will begin to rise

20 Age 55 Survey - Content Updating event histories (household composition, children, housing, economic activity, qualifications) Help and care provided to parents and grandchildren Earnings/income/ housing wealth Retirement plans / pensions Self-reported health and health conditions Smoking and drinking Well-being

21 Example findings Caring responsibilities

22 Example findings Health 0% 20% 40% 60% 80% 100% Age 50 19% Age 55 14% 33% 29% 13% 6% Self-rated health 34% 32% 14% 6% Excellent Very good Good Fair Poor 80% 70% 60% 50% 40% 30% 20% 10% 0% Overweight/Obesity 24% 16% 23% 16% 46% 47% 34% 26% Men Women Overweight Obese 60% 40% 20% Smoking 0% Men Women

23 Expectation of working past 60/66 By lifetime income 100 How likely on a scale of is it that you will be working at the age of 60, and after the age of 66? % reporting >50% likely Poorest 2nd 3rd 4th Richest

24 CLS Website

25 Accessing NCDS data from the UK Data Service Brian Dodgeon Centre for Longitudinal Studies Institute of Education University College London

26 NCDS Data How do I download it? Register with UK Data Service Give details of research use Types of access End User Licence (normal access to anonymised data) Special Licence sensitive and disclosive data (including area data) Secure access (very disclosive data) - cannot be downloaded, instead accessed remotely Choice of formats for download: SPSS STATA TAB-delimited

27 NCDS data: Files Flat file contains most of the data from the Age 55 survey 1 record per case Hierarchical datasets multiple records per case: Relationships - Relationship history since last interview / 2000 Person grid Details of people living with respondent (past and present) 1 record per person Housing - Housing history since last interview / 2004 Employment Economic activity history since last interview / 2004 Qualifications Qualification history since last interview / 2004 Children who have died - Details of any children of the respondent, born alive but since died and not already recorded in the person grids Unfolding brackets respondents unable/unwilling to provide a precise amount to some income variables are asked a series of questions which define a band in which the answer lies. Data included in a separate dataset. Derived variables a set of useful key variables which have been constructed from the data

28 NCDS documentation on the CLS website Brian Dodgeon Centre for Longitudinal Studies Institute of Education University College London

29 NCDS what documentation is available? User Guide Survey details, making sense of the variables available, conventions used Full Questionnaire Online and telephone versions Index of variable names linked to questions Technical Report Sample design, development work Conduct of fieldwork, survey response Coding, editing and data preparation Explore variables at UK Data Service NESSTAR software Shows frequency counts of all variables on request Create user account to explore variables further: Cross-tabulations Charts

30 Attrition in NCDS Tarek Mostafa Centre for Longitudinal Studies Institute of Education University College London

31 The NCDS Sample and Attrition over Time.

32 To be covered: The NCDS sample. Attrition in NCDS. What do we know about attrition in NCDS What can we do about it?

33 The NCDS Sample Design The NCDS follows the lives of all people born in England, Scotland and Wales in one week of March The target sample was augmented to include immigrants born in the same week, in the first three sweeps (ages 7, 11 and 16).

34 Problems of non-response/attrition Distinction between unit (respondents ) non response and item non-response (focus on former here) Types of non-response (have separate reasons) Non-contact Refusal Inability Out of scope/ineligible Non-response on increase in all surveys Non-response may not be permanent (in panel survey) Effects of non-response/attrition

35 Attrition Attrition is the discontinued participation of some individuals in a longitudinal survey for reasons that are unknown and/or beyond the control of the researcher

36 Response in NCDS 18,000 Response in NCDS 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 Wave 0 Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7 Wave 8 Wave 9

37 Types of nonresponse Wave 0 Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7 Wave 8 Wave 9 Age Birth Non-contact 223 1, ,867 1,529 1, Not issued ,415 4,248 3,553 4,698 Refusal ,151 1,160 1,776 1,148 1,448 1, Other unproductive , Not issued - emigrant ,196 1,335 1,268 1,272 1,293 1,287 Not issued - dead ,050 1,200 1,324 1,460 1,503 Ineligible Total 1,143 3,133 3,221 3,904 6,021 7,089 7,139 9,024 8,768 9,225

38 What happens to the sample 60.0 Sample composition: percent Wave 0 Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7 Wave 8 Wave 9 CM is a man CM was not breastfed Social class: unskilled manual

39 What happens to the sample (2) Sample composition: percent Wave 0 Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7 Wave 8 Wave 9 Mother was married at birth Mother not living in South-East

40 Non-response bias - implications Missing data: Smaller samples, fewer transitions, incomplete histories. Loss of statistical power due to smaller samples. Biases in results Disproportionate to some groups (mobile, disadvantaged, men, long working hours) Problem if linked to survey topic focus/variables

41 What can we do about attrition? Attrition weights. The weights are constructed using binary logit response models. They are equal to the inverse of the probability of participating in the study. Multiple imputations: used to deal with item nonresponse. FIML method: run the imputation simultaneously with the substantive model. It deals with item nonresponse.

42 CLS Missing Data Strategy Currently we are working on the CLS strategy to deal with missing data. Objective: identify the key predictors of unit non-response in all CLS cohort studies (NCDS, BCS70, MCS, Next Steps). So far the analyses were carried out for NCDS and a report will be published soon. Main message: both variables from the earlier and the later waves do predict response. These include: social class and circumstances, Region of residence, Health status, literacy

43 Using the mixed-mode data from the age 55 survey Matt Brown Centre for Longitudinal Studies Institute of Education University College London

44 Response rate 83% 21% 62% 77% Mixed Mode Web Tel Telephone Only

45 Who responds online? Who responds online? Those who had participated in prior survey Regular computer / internet users White British Professional / managerial occupations Homeowners Higher net pay Those with better self-reported health Non-smokers Non-problematic drinkers Those with a partner Those with better cognitive function

46 Item non-response (Don t know / Refusal rates) 13% 15% 11% 8% 8% 8% 2% 2% Expected value of property Gross weekly income Party voted for in general election Units of alcohol consumed in last 7 days MM TO

47 Mode effects on measurement 63% 69% 51% 40% 40% 30% 19% 15% 11 8 Self-rated financial situation - living comfortably Feel the future looks good to me - Often Voted Liberal Democrats Sport at least once a week Units of alcohol per week MM TO

48 Advice If variables affected by mode are independent variables or dependent variables (or both) then mode of completion (N9SURVER) should be used as an additional control in multivariate analyses This is a precaution. In most cases this will make little difference to estimates (either the variables of interest or other co-variates).

49 Linear regression dependent variable = units of alcohol per week (Dependent variable affected by mode) No mode indicator B With mode indicator B Mode Web 1.30** Difference Sex Male 6.31** 6.28** 0.03 General ability score age 11 (standardised) 1.52** 1.38** 0.14

50 Logistic regression dependent variable = fair or poor self-rated health (Dependent variable affected by mode) No mode indicator Exp(B) With mode indicator Exp(B) Mode Web 0.58** Difference Sex Male Childhood SES (Ref=Unskilled Manual) Skilled Manual Lower grade technicians Self-employed Routine non-manual 0.74** 0.78** 0.04 Lower grade managers 0.57** 0.60** 0.03 Professionals 0.50** 0.54** 0.04 Missing

51 Logistic regression dependent variable = fair or poor self-rated health (Dependent variable and both independent variables affected by mode) No mode indicator Exp(B) With mode indicator Exp(B) Mode Web 0.54** Difference Units of alcohol per week 0.99** 0.99** 0.00 Plays sport at least once a week 0.39** 0.37** 0.02

52 Further guidance forthcoming: Mixed mode user guide will shortly be available via UKDS CLS currently conducting a more detailed examination of the impact of mode. Findings to be made available via CLS website

53 Future Plans Matt Brown Centre for Longitudinal Studies Institute of Education University College London

54 Future plans Future sweeps every 5 years Next survey at Age 60 (2018) 75 minute face to face survey (with funding being sought for additional biomarkers) New data deposits Work history file - up to 55 Partnership history file up to 55

55 Please respond to the NCDS Age 60 Survey Consultation! Finances and employment: including topics such as work, income, wealth, retirement plans and education Family, relationships and identity: including topics such as social networks, relationships, neighbourhood, social capital, social and political participation, attitudes and values, and religion Health, wellbeing and cognition: including topics such as physical health, mental health, medical care, health behaviours, cognitive function Deadline for written submissions: 10 June 2016 Consultative conference: Monday 11 July 2016 Website:

56 Questions?

57 Feedback