Presentation on theme: "Introduction to longitudinal studies in the UK Nick Buck, Institute for Social and Economic Research, University of Essex."— Presentation transcript:
Introduction to longitudinal studies in the UK Nick Buck, Institute for Social and Economic Research, University of Essex
Summary Overview of UK Longitudinal Studies Rationale for longitudinal research methods Introduction to the research contribution of longitudinal studies Taxonomy of longitudinal studies
UK Longitudinal Studies UK portfolio is particularly rich. Around the world we can identify five main types of LS: –Household Panel surveys –Birth Cohort Studies –Longitudinal studies of ageing –Studies of the transition from school to work –Census linkage and other admin based studies The UK almost unique in being well represented in all these.
UK Longitudinal Studies (2) Household Panel surveys: British Household Panel Survey Birth Cohort Studies: National Child Development Study, Birth Cohort Study 1970, Millennium Cohort Study and more Longitudinal studies of ageing: English Longitudinal Study of Ageing Studies of the transition from school to work: Youth Cohort Study, Longitudinal Study of Young People in Education Census linkage and other admin based studies: ONS Longitudinal Study, Scottish Longitudinal Study... plus others, e.g. Families and Children Survey, British Election Study, Wealth and Assets Survey
What are Longitudinal Studies? Most generally, longitudinal studies collect data about the same subjects relating to multiple time points Subjects may be individual people or other entities, e.g. organisations such as firms – but here mostly concerned with studies of people Normally LS restricted to studies which collect data at several times (contrast retrospective life history studies which collect data about multiple time points on single occasion). But LS often use retrospective methods
Why longitudinal research? Longitudinal research provides an understanding of social change, of the trajectories of individual life histories and of the dynamic processes that underlie social and economic life, not possible from research based on cross-sectional data. The recent development of LS in the UK has underpinned advances in social science method and in understanding of major social changes and policy interventions We can identify a range of issues and types of research where longitudinal seem especially appropriate.
Rationale for longitudinal research (1) Focus on change Longitudinal approaches are essential when phenomena of interest directly concerned with individual level change – time is part of the definition, e.g.: –the dynamics of poverty, –employment instability, –social mobility –most demographic events: births, marriage, divorce, death, migration
Rationale for longitudinal research (2) Investigating casual processes Use longitudinal approaches when want to infer causation from temporal ordering, e.g.: – the effects of unemployment on mental health –of childhood circumstances on later life chances; Or to investigate effects of events, –e.g. what happens to incomes of husbands and wives after divorce
Rationale for longitudinal research (3) Controlling for fixed characteristics Can use longitudinal methods to investigate associations where an unmeasured characteristic may affect the outcome, provided this can be assumed to be fixed over period of observation Use repeated measures data to model differences (i.e. if effect if B on A is contaminated by unmeasured C, can investigate effect of change in B on change in A, if C fixed). Examples of unmeasured fixed characteristics: ability, motivation, preferences. E.g. effects of wage rates on hours worked
Rationale for longitudinal research (4) Separating age, period and cohort It is hypothesized that a persons characteristics, attitudes and behaviour, change as they get older, but also that cohorts will differ including both the effects of transmission from their parents, and period specific events experienced. Understanding social change of requires disentangling this – normally needing long period longitudinal data. E.g. are changing social norms with respect to marriage/cohabitation or gender roles a consequence of cohort succession, or all cohorts changing together?
Rationale for longitudinal research (5) (quasi-)experimental design Longitudinal methods often used to establish the effect of a treatment in formal random control trail Established LS may also be used to assess policy effect e.g. impact of the introduction of public policy, (methods developed to deal with non-random selection into policy) LS may be able to take advantage of natural experiments – e.g. devolution in UK government
Examples of Longitudinal Research Disentangling the effects on children of school and family background in order to understand social mobility and the effectiveness of educational interventions – and to identify the key points for intervention. Examining the effects of changing patterns of marriage, cohabitation and childbirth on the time children are likely to spend in lone parent families – and the effects on their later lives. Understanding the defining characteristics of people who experience repeated spells of unemployment and poverty – and their scarring effects, which make it difficult for people to find work and/or escape poverty in the future.
Taxonomy of longitudinal studies Distinguish surveys from other types (e.g. using administrative data) Prospective (repeated surveys) versus retrospective Key dimensions of longitudinal study design Research rationale for different designs Most surveys discussed here designed as longitudinal from the start – not always the case
Surveys versus administrative data Administrative data (collected for bureaucratic/management purposes) an alternative to the costly process of collecting sample surveys Advantages of admin data: –comprehensive coverage of clientele of admin agency –provide authoritative statement of behaviour/circumstances related to the agencys activities (eg benefit records) Problems: –limitations on access and restrictive confidentiality requirements –limited scope of data (since limited to the agencys own purposes) But the model in which admin records are used as the sampling frame for further survey activity, or supplement survey data may provide promise for the future.
Retrospective data as alternative to prospective data Retrospective Advantages: Quick, in the sense that all the data arrives at the same time, and therefore cheap, requiring only a single measurement cycle (contrast prospective respondents narrative have internal consistency Disadvantages: Recall error Internal consistency may be a consequence of recall error and reinterpretation Subject to survivor bias – not everyone lives long enough to be interviewed. Prospective Advantages: Higher quality of current reports over retrospective Full sample at start of process Disadvantages: Slow, since repeated measures take time to accumulate into a longitudinal narrative Costly – repeated data collection Attrition Measurement error issues
Key dimensions of longitudinal study design Demarcation of population universe: –Whole country/ smaller area –Whole age range/ particular cohort –Population sub-group e.g. minority ethnic group –Relevant for particular programme/policy Unit of analysis: individual, family grouping, individual-in-household, organisation Study duration: indefinite, whole life, planned fixed duration Measurement frequency: fixed frequency, key life stages
Research aims of design Repeated measurement using same questions to establish rates of change and patterns of association over time Assessment of rates of development through questions tailored to particular ages at different waves/sweeps Accumulation of life history data from multiple waves/sweeps Pre- and post-intervention measurement Respondents own accounts of change NB: individual studies combine several of these
Household Panel Surveys (e.g. BHPS) Universe: whole population Unit of analysis: individual in household Study duration: indefinite Measurement interval: fixed, annual Research aims of design: repeated measures, accumulation of life history data Many international comparator studies: e.g. PSID, German SOEP
Birth Cohort Studies (e.g. NCDS, BCS70, MCS) Universe: sample of births from particular year (1958, 1970, 2000-2001) Unit of analysis: individual, with associated others Study duration: whole life Measurement interval: key development points, then regular intervals Research aims of design: developmental measures, some repeated measures, accumulation of life history data Few international comparator studies, but growing number around millennium
Ageing Studies (e.g. ELSA) Universe: whole population above e.g. 50 at time of sampling – possible refreshment Unit of analysis: individual plus partner Study duration: whole life Measurement interval: fixed, biennial Research aims of design: repeated measures, accumulation of life history data (focus on measures of relevance to older population) International comparator studies in USA (HRS) and Europe (SHARE)
Youth Studies (e.g. YCS, LSYPE) Universe: individual school years Unit of analysis: individual plus data from e.g. parents Study duration: fixed planned duration, up to end of process under study; start new cohorts to replace Measurement interval: fixed, annual Research aims of design: developmental measures, some repeated measures Limited international comparator studies, e.g. NLSY in USA
Census link studies (e.g. LS, SLS) Universe: whole population Unit of analysis: individual (some linked measures for household) Study duration: indefinite Measurement interval: census fixed, decennial, plus other continuous input sources - registration Research aims of design: repeated measures, accumulation of life history data Limited international comparator studies
Conclusion This presentation has shown the value and variety of longitudinal studies Next presentation discusses individual studies in more detail – how to chose a particular study for research.