National Strategy for Longitudinal Studies – by chenmeixiu


									              University of Essex

                   INSTITUTE FOR SOCIAL
                   & ECONOMIC RESEARCH

National Strategy for Longitudinal Studies

    Report to ESRC Research Resources Board

                February 2002
                                                           TABLE OF CONTENTS

   EXECUTIVE SUMMARY ....................................................................................................................................... 1
   INTRODUCTION ................................................................................................................................................... 2
   THE DEVELOPMENT OF THE NATIONAL STRATEGY ............................................................................................ 3
   THE CENTRAL ROLE OF THE EXISTING ESRC STUDIES ...................................................................................... 4
   RESEARCH RATIONALES AND RESEARCH IMPACTS OF LONGITUDINAL STUDIES ................................................ 8
       Birth Cohort Studies...................................................................................................................................... 8
       BHPS ............................................................................................................................................................. 9
   PRIORITIES FOR THE DEVELOPMENT OF LONGITUDINAL STUDIES .................................................................... 11
   A FIFTEEN YEAR CONCERTED PLAN FOR NATIONAL LONGITUDINAL DATA SETS ............................................ 13
       Cohort studies ............................................................................................................................................. 15
       BHPS ........................................................................................................................................................... 16
   THE QUALITY PROFILES ................................................................................................................................... 18
   A METHODOLOGICAL RESEARCH PROGRAMME ............................................................................................... 19
   TRAINING ......................................................................................................................................................... 21
   OUTREACH/USER ENGAGEMENT ...................................................................................................................... 22
       Inter cohort comparison .............................................................................................................................. 24
       Intergenerational Studies ............................................................................................................................ 25
   EXEMPLARY USES OF THE BHPS...................................................................................................................... 25
       Poverty ........................................................................................................................................................ 25
       Family ......................................................................................................................................................... 25
       Labour market ............................................................................................................................................. 26
       Migration .................................................................................................................................................... 26
       Health.......................................................................................................................................................... 27
       Politics and values ...................................................................................................................................... 27
       Area effects.................................................................................................................................................. 27
       Wealth ......................................................................................................................................................... 27
       Social exclusion........................................................................................................................................... 27

                     National Strategy for Longitudinal Studies
                        Report to ESRC Research Resources Board

                                       Nick Buck
                            ESRC UK Longitudinal Studies Centre
                                   University of Essex

                                         February 2002

Executive Summary

This paper presents a National Strategy for longitudinal studies, developed by the ULSC with
the ESRC‟s National Longitudinal Strategy Committee. Its aim is to support the future
development of longitudinal research in the UK by ensuring the optimal portfolio of
longitudinal research resources is available within available budgets, along with other
measures to support longitudinal research.

The national strategy works with six key principles:

   Longitudinal studies are long term investments with increasing returns to duration.
   Gaps in the portfolio must be identified, understanding that current users may not always
    be able to identify these gaps.
   Complementarity between studies meeting distinct needs is critical to the strategy
   The strategy must depend on complementarity and collaboration between funders
   High standards of data quality are essential for longitudinal research.
   Data usage is central to the strategy, prioritising data usability and training

The existing studies largely funded by the ESRC, the BHPS and the Birth Cohort Studies
remain at the centre of the strategy. They provide a core of resources which makes the UK
one of the international leaders in longitudinal research, and they complement each other in
research uses.

The strategy also identified a set of gaps in current provision:

   Longitudinal data on ethnic minorities with adequate sample sizes;
   Better data on youth transitions
   Bridging the gap between 1970 and 2000 in the cohort studies
   Longitudinal data on ageing

Priorities for the development of longitudinal studies up to 2014 are presented, agreed with
the National Strategy Committee, based on two assumptions about resources.

Further sections of the report briefly discuss:

   The development of quality profiles and quality standards for longitudinal studies
   Longitudinal methodology research requirements
   Training needs for longitudinal research
   Other mechanisms for outreach and user engagement


The provision of a range of high quality longitudinal research resources in the social sciences
is now critical to a vibrant national research culture that has a strong international visibility.
The UK has taken a prominent role in the development of longitudinal studies, especially
with the unique portfolio of birth cohort studies, the ONS Longitudinal Study of the Census
and more recently with the internationally respected British Household Panel Survey. This
portfolio of studies has underpinned major developments in social science over the last two
decades. Indeed, these studies have provided an understanding of social change, the
trajectories of individual life histories, and the dynamic processes which underlie social and
economic life. These have already led to a step change in both subject matter and
methodology in many areas of the social sciences, and to a substantial increase in the capacity
of social to contribute to the understanding and development of many areas of public policy.
These changes are still working themselves through. Moreover, for all the justified concern
about the training of skilled social science researchers, it is clear that there is a large
community of skilled and active users of these studies.

These successes and the critical importance of longitudinal data were recognised in a set of
major new investments around the start of the current decade, by the ESRC and by others.
These saw the establishment of a new birth cohort study, major new sweeps of BCS70 and
the NCDS, the continuation of the BHPS and its enhancement by new survey extensions to
permit independent analysis of Scotland, Wales and Northern Ireland. One of the significant
gaps in the portfolio, longitudinal studies focused on issues around the ageing population is
now on the way to being filled by the establishment of the English Longitudinal Study of
Ageing (ELSA). It will necessarily be several years before the impacts of these investments
on research are fully realised, but major opportunities are clearly opening out.

Alongside these investments, the ESRC also established the UK Longitudinal Studies Centre.
The ULSC has the role of overseeing collection of longitudinal data on the main ESRC
funded studies, encouraging the development of longitudinal research by promoting
awareness of longitudinal studies, improving the usability and documentation of longitudinal
studies, and providing training for longitudinal research1. Central to its mission has been the
development of a National Strategy for Longitudinal Studies, jointly with the National
Strategy Committee. The role of the national strategy is to consolidate this success, to
establish priorities for investments in longitudinal studies, and to ensure that ESRC
investments are co-ordinated with those of other funders. In a larger and richer system such
as the USA a diverse portfolio of resources might be developed with essentially
uncoordinated activity. However the fact that longitudinal studies required sustained funding
plans over long time scales raises greater questions in a smaller system such as the UK. So
the essential mission of the National Strategy is to keep this aspect of Britain‟s role in global
social science research adequately resourced and encouraged, in the context of national
research goals. Thus, the core task for the strategy is to propose answers to the question: what
can the ESRC and other funders do to promote this area of research in a co-ordinated

 These activities which constitute the major part of the work of the ULSC, in terms of staff time, are reported
more fully in the ULSC Annual Reports to the ESRC.

The Development of the National Strategy

This document sets out in brief the current view of the National Strategy, after four meetings
of the National Strategy Committee, and a range of user consultations. As previous reports to
the RRB have made clear the National Strategy has had to be extremely cautious in balancing
user needs for longitudinal studies with available resources. It could have focussed simply on
identifying a set of enhancements to existing studies and new studies which would „complete‟
the portfolio of longitudinal studies. This could either have no result because no resources
were available to implement it, or even more damagingly, could lead new investments which
were unsustainable over the longer term.

At the heart of the National Strategy are a set of basic understandings and principles.

The first of these is the rather obvious point that longitudinal studies are by their nature long
term investments. There are generally increasing returns to extending the life of studies since
this generates new research opportunities. There are exceptions, but such fixed length
individual studies are normally seen as part of a longer term strategy of regularly starting new
comparable studies to track the impact of social change on the life courses of individuals and
groups between studies – youth cohorts are a good example. Moreover windows of
opportunity for collecting data at a particular age and stage in the lives of participants in a
longitudinal study, once closed, can never be re-opened, weakening the integrity and
scientific potential of the whole data set. Aside from the scientific arguments for continuity,
there are also pragmatic arguments, since the risks of continuing an existing successful study
are less than those associated with starting a new study, and the scientific returns to new
investments in an existing study are much more rapid than those on a new study. Even in an
annual panel study it typically takes five or more years for a significant body of research to
build up and get into scholarly print. The strategy must guard against these pragmatic
considerations becoming too dominant.

The national strategy has therefore, secondly, been concerned to look widely at potential gaps
in provision, in the light of the understanding that current perceptions of needs within the
social science community will be heavily shaped by the existing provision. Gaps may either
be defined in terms of population groups or subject matter associated with particular research
agendas. For any potential gaps a central issue is whether they be dealt with more effectively
by new studies or enhancements to existing studies.

Thirdly, the strength of the overall strategy depends on complementarity between studies,
meeting different scientific purposes, not on studies competing to meet the same scientific
needs. It is important that the strategy respects the identities of individual studies, and within
an overall strategy those responsible for individual studies should be encouraged to focus on
the comparative strength of their study (as well as the possibilities for comparisons, with
British and International longitudinal sources). This means that evaluations of studies must
be placed in the context of their purposes, and are likely in the end to centre on the
importance of those scientific purpose. However, within a context of a diverse range of
studies, there are clearly research benefits to be drawn from facilitating linkage and
comparison between the studies, so that their complementary strengths may used to address
the same research problems.

Fourthly, there is another sort of complementarity, between funders. The ESRC has been key
to the development of longitudinal studies of relevance to social scientists, and through the
national strategy its central role will continue. Social scientists do, however, use longitudinal

data produced under the auspices of other bodies. Medical research funders have an
important role in the production of longitudinal data, and though there are often access
problems, these resources are important to many social scientists. The 1946 cohort study, the
first of the four British birth cohort studies, is an important example here, though unlike the
other birth cohorts it is not deposited in the Data Archive. This complementarity goes both
ways. The longitudinal studies with a primary social science focus have also been most
important as a research resource for medical researchers, and medical funders have a major
potential role in supporting the production of longitudinal data on these studies. It is
important here that common standards of data access are assured. Government departments
have taken an increasing role in support for longitudinal studies, as policy interest in
longitudinal research has increased. Government departments established a Longitudinal
Data Co-ordination Group, chaired by the ONS, leading to important initiatives, including
support for the Millennium Cohort Study and the English Longitudinal Study of Ageing
(ELSA). They are thus important co-funders of a number of studies, and there are some
government data sets which are important in longitudinal research. Continued co-ordination
with the activities of these other bodies is critical to the development of the national strategy.

Fifthly, data quality is critical to the delivery of longitudinal research which meets the highest
international standards. It is not necessarily the case that data quality issues actually have a
larger impact on longitudinal than on cross-sectional research, but they are certainly much
more visible. However, social scientists understanding of data quality issues is to some
extent deficient, and the quality of longitudinal research will be improved as this
understanding develops. The National Strategy has therefore devoted considerable efforts to
developing a framework for the understanding of data quality, which is discussed below. It is
proposed that this framework will be adopted as a template for future data collection
exercises. A further consequence follow from the importance of quality. This concerns data
collection costs. High quality longitudinal surveys are necessarily expensive. The efforts to
maximise response and tracing of study members in order to minimise attrition will raise
costs above those of a cross-sectional survey. This means that it is very risky to under-invest
in a longitudinal survey, since in the extreme case research based on it will not be publishable
in the best journals.

Finally the strategy needs to recognise the importance of data usage. Data is not worth
collecting if it is not going to be effectively used. One might want to argue that this should
set limits on the level of investment in longitudinal studies. However it is also clear that good
investments generate usage. Well designed longitudinal studies which inspire researchers to
believe that their research questions can be answered will encourage the effort to acquire the
skills to use them, provided barriers to usage are not too great. This suggests that critical to
the strategy are efforts to promote longitudinal studies, to ensure ready access to the data, to
ensure that it is as easily usable as possible, and well documented. Training in longitudinal
research methods is also critical as is follow-up support to ensure that the detailed technical
and scientific knowledge of the study collected by the team managing the study over the
years is available to potential and actual users.

The Central Role of the Existing ESRC Studies

The discussion above emphasises the importance of the current heritage of UK longitudinal
studies. By accident and by design, the UK has chosen to invest in certain key types of study:
birth cohort studies, household panel studies and census linkage studies. The fact that other
countries have made different choices should not necessarily lead to a change in priorities.
Indeed the fact that some studies, such as the set of birth cohort studies, are close to unique

internationally makes them important to social scientists throughout the world. It must also
be recognised that in terms of breadth and quality, the UK is second only to the USA.
Internationally, household panel studies tend to be the most common form of nationally
representative studies, and here the UK has in BHPS what is recognised as one of the three
top studies (alongside PSID in the USA, and GSOEP in Germany). Apart from this no
country other than the USA has such a diverse and high quality array of national studies.
Even in the USA, there has been no nationally representative longitudinal study of the US
population following up a cohort from birth until very recently.

In essence this is to argue that these past choices had already led to a national strategy for
longitudinal studies. Moreover given that many of the choices and decisions which produced
this over the last two or three decades were independent of one another, the strategy was
remarkably coherent. The studies are to a very considerable degree complementary with one
another. The studies are also rather heavily used. Recent Data Archive figures suggest that
BHPS is on one of its measures the most heavily used data set, not excluding the major large
scale government cross-sectional surveys, and there is very substantial use of both BCS70
and NCDS which will undoubtedly grow now that the 1999/2000 sweeps are available. All
the studies are central to the research programmes of major social and economic research
centres, including IFS, CASE, CEP, the IER at Warwick, and new DfES Centres. They also
major components in the research of leading social science academic departments. For
example, BHPS has an average of 20 registered users in each of the economics departments
rated 5* at the 2001 RAE. BHPS is also heavily used in cross-national comparative data sets,
including the European Community Household Panel, and the Cross-National Equivalence
File, combining BHPS, PSID, GSOEP and the Canadian SLID. The existing studies have
called into existence a very substantial body of users – amounting to close to 1,000
individuals and research groups, mostly in the UK but also and increasingly elsewhere in the

This of course generates its own problems, and perhaps the most difficult responsibility
facing the ULSC is ensuring that the issues of heritage, while given their rightful
consideration, do not unduly dominate the ESRC future decision-making in this area. The
heritage has a substantial influence on users‟ perceptions of data needs, and of the degree to
which these are currently satisfied. Any new data collection reflects a current research
agenda, but subsequently the very existence of that data set may to some degree constrain the
research agenda subsequently.

The large body of users has been conditioned to expect certain types of data, and certain
standards of data quality. And as a result, when the ULSC conducts a consultation about
future needs, what emerges is to a considerable degree a reflection of the past. Some new
requirements do emerge, but these are, precisely because the community has emerged to use
the existing data sets, often rather conservative requirements. We find demands for
“interstitial” studies (eg for the “missing” birth cohorts), or for increases in the sample sizes
of existing studies to improve coverage for newly emerging foci of academic or policy
interest (new sub-national or ethnic groups). We find demands for improvements in data
standards or accessibility – in the existing studies. We find less evidence of strong demand
for radically new sorts of studies.

The reasons for the largely conservative results of the enquiry are straightforward. Even the
simplest of longitudinal studies is considerably more complex to design and to use than
conventional cross-sectional studies. Longitudinal studies require a degree of specific
expertise that is not widespread. And this expertise is typically acquired ex post – as a

consequence of the use of a particular existing study for a particular substantive purpose. So
the body of users we consulted was, almost of necessity, a group already committed to
particular existing studies, or at least to well-established paradigms of survey and analysis.
There are of course potential users of longitudinal data resources, with substantive research
interests that would benefit from longitudinal evidence, who are currently unacquainted with
the general field of longitudinal data analysis. And it is also the case that some of those
current users whose specialities lie with particular existing studies, might, if exposed to the
wider perspective of the full range of possible longitudinal study designs, as a result develop
new research requirements. This has clear implications for training and information
dissemination activities that fall clearly within the locus of a national strategy for longitudinal
studies (we return to these in the recommendations following). Of course this first
consultation exercise cannot be the end of the process, and we outline below mechanisms for
continuing to establish new needs.

Nevertheless the review did identify a small number of needs which could not be met from
the existing studies. One of these recommendations, a study of ageing, is in fact now being
met through the English Longitudinal Study of Ageing (though there might be legitimate
concern about coverage in Scotland, Wales and Northern Ireland). The National Strategy
Committee has considered papers on three areas, in order to assess possible mechanisms for
developing new studies, and how social interests might be secured:

1. Longitudinal data on minority ethnic groups. The sample sizes in existing studies for
   these groups tend to be much too small for independent analysis. Moreover, any
   longitudinal study on ethnic minorities will require at least some focussed questions to
   address ethnic minority research agendas – e.g. ethnic identity, the experience of racial
   harassment and discrimination. There are different views of the priority associated with
   these questions, and the NSC came to the conclusion that the highest priority was for
   comparable data with the population as based on comparable designs, though mechanisms
   such as the supplementation of the BHPS sample.

2. A comprehensive Longitudinal Study of Young People, perhaps on the model of the US
   National Longitudinal Study of Youth. This would both be broader in terms of subject
   domains, and follow subjects for longer periods than the current Youth Cohort Studies. It
   would provide high quality data on the processes of transition from school to work across
   a wide age range, of a kind not currently available for large samples from other sources.
   The DfES has interests in such a development with wide data coverage and embracing a
   substantial ethnic minority sample, but following the pattern of a single birth/age cohort
   first contacted at 14 and followed up annually to the age of 25.

3. In the analysis of social change, there is also clearly a need to bridge the very long gap in
   the series of birth cohort studies between the 1970 cohort and the Millennium Cohort.
   This would involve taking a sample born in the late 1980s and with the intention of
   following them in parallel with the current studies. There would be some use of
   administrative records to fill the early life gap. There turns out to be a significant
   convergence between this need, and the need for a cohort study of young people, by
   taking the LSYP as the basis for follow-up into later life.

In addition to these needs for data about groups who may not be sufficiently well covered by
the existing studies, there are also some needs for coverage of research domains which the
current studies do not sufficiently support, either because of pressure of questionnaire space,
or because of sensitivity. One significant area here is criminology, with interests in both
victimisation and pathways into crime. New youth longitudinal data may provide some
support here, and there is also interest within the Home Office in using both administrative
records and new survey data collection. The National Strategy will need to keep this area
under review.

There are other opportunities in the area of linkage of administrative records, though
developments here are by no means straightforward. Cross government interest is such
studies is likely to grow, not least because of the increasing interest in linkage of
administrative data to the ongoing longitudinal studies. Data Protection issues are, however,
also becoming prominent, as a potential obstacle to this development. A forthcoming report
from the Cabinet Office Performance and Innovation Unit on Privacy and Data Sharing may
provide some clarification. The ONS convened LDCG is an important forum for addressing
such matters and would appear to provide a valuable vehicle for ESRC to make the case for
data linkage. It is possible that the user community has not so far voiced a demand for record
linkage because it has not seemed conceivable, and there may also be some inaccurate beliefs
about the inflexibility of analysis possibilities in the ONS LS, recently greatly improved, but
the Safe-setting model for analysis cannot be avoided, implying the need for resources in user

The National Strategy also has a number of continuing mechanisms by which the evolution of
user needs could be kept under review, including for example:

   a continuous review of new survey developments in other countries,

   regular trawl of the international scientific literature for innovative examples of
    longitudinal analysis, and

   regular discussions with government departments and others to identify new policy issues
    with a longitudinal dimension.

However, the central conclusion of the strategizing process, unanimously supported by the
National Strategy Committee, is that support for the core of the existing studies with which
ESRC is already associated, and for future developments to that core are the most effective
way of sustaining the UK position in longitudinal research, and must therefore remain the
ESRC‟s highest priority. Support for new studies must depend on substantial new resources,
or co-ordination with other funders.

There are still major issues and major opportunities associated with the future of the existing
studies, and we can see ways in which enhancements to those studies could contribute
enormously to the development of the social science base over the next ten to twenty years.
We can also see important opportunities for investments in new activities over that period.
Below we set out in more detail the broad strategy over the next ten years:

   To develop and sustain the BHPS as a core longitudinal research tool across the social
    sciences, to ensure that it can meet both needs for research on short term dynamics, and
    the opportunities which its extended life provides to analyse intergenerational and life
    course effects, and to ensure that it remains at the heart of international comparative
    longitudinal research.

   To develop the set of cohort studies to ensure that they are capable of providing well into
    the future a coherently organised research resource, which allows inter-cohort comparison
    of long-term life course effects. The strategy must ensure that for the future cohort

    planning has the very long term perspectives which are required to ensure that critical
    data for later research are collected at each stage.

   To make progress towards filling the current gaps in the portfolio of studies, particularly
    in relation to youth transitions and ethnic minorities, to ensure that studies of ageing and
    retirement can be sustained.

   To improve co-ordination between distinct funders of longitudinal research resources, so
    that taking account of their different priorities, collaborative funding over the longer term
    can be assured and maximum use can be made of scarce resources. Enhancing and
    protecting ESRC‟s investment in longitudinal studies will require a clearer framework for
    decision making where multiple funders are involved.

Research Rationales and Research Impacts of Longitudinal Studies

We have stressed that the ESRC studies need to be seen as a complementary package, with
different studies meeting different needs. This section spells this point out in rather more
detail. In essence it suggests that the birth cohort studies address major scientific interests in
the processes of mobility across the life course, the influence of long-term antecedents on
subsequent outcomes. Through the comparison between different cohorts it provides an
important basis for the understanding of social change. The BHPS focuses initially on short
term processes of change, for a general population, and allows understanding of individual
change within the social context of the household. As the study lengthens, high quality
longer term measures are built up from the annual interview, allowing analysis of
accumulation processes. The prevalence of the household panel design also makes it an
important vehicle for international comparative research.

The studies are discussed separately below. Appendix A presents examples of research based
on the studies which illustrate their key uses.

Birth Cohort Studies

Birth cohort studies offer a means of charting the development of the human life course
through the analysis of the biographies of large numbers of individuals born at a particular
point in time. The data can be used prospectively to make predictions about the outcomes of
particular circumstances and experiences in life occurring at particular points in time. The
data can also be used retrospectively to identify the circumstances and experiences in earlier
life that underpin a given outcome later. The data collected for any single birth cohort
confounds age, period and cohort effects at any particular point in time. Comparison of more
than one cohort enables the investigator to hold constant one of these three “extrinsic”
factors, for example, comparing cohorts at a given age to establish a cohort effect or cohorts
at the same age at different times to establish a period effect.

One major advantage of the large representative samples in the British birth cohort studies is
the capacity to focus on sub-samples with characteristics with a relatively low prevalence in
the total population and to follow them up to chart in detail the way their life courses develop.
Examples here are in childhood those with disabilities and those living in disadvantaged
circumstances, children who are adopted and later in life, people exhibiting various kinds of
medical symptoms such as respiratory illness or poor basic skills.

Longitudinal studies are described by some methodologists as quasi experiments, in the sense
that although there is no randomized allocation of individuals to treatments, the temporal
sequencing of longitudinal data does offer a powerful means of control in comparison with a
cross sectional survey. We can at least say with certainty from the statistical analysis of
biographies what events preceded others, even though we still have problems in deciding
which event in relation to another event was the underlying cause.

The longer the studies are continued the richer the potential of the data for analysis of lifetime
outcomes becomes. For example, in the medical arena, sequelae of early childhood
conditions leading to particular outcomes can be tested through morbidity and mortality
outcomes in adulthood. Similarly in the social domain, the consequences of poverty in
different places and at different times, and those experiences in education and the
community, which appear to alleviate its effects, can also be followed through across the
whole of the life cycle. Such key issues of current policy concern as social exclusion lend
themselves perfectly to analysis through the birth cohort studies because of the long-term
nature of the processes involved.

Currently, for the adult cohorts, the priority areas for data collection and research are labour
market attainments, income and welfare, health - both physical and mental - housing and
residential mobility, family formation and childcare, personality and cognitive characteristics
and social participation, attitudes and values. These are set against the well-established
factors of demographic variation such as gender, social class, urban/rural location,
geographical region, highest qualification, partnership and family status. Each new sweep
will add further details of the picture of how the life courses of individuals, grouped by
different combinations of these characteristics reveal continuities and discontinuities and how
these vary from those in the past. Such key theoretical questions as the development of
individualization (the breakdown of class values and norms) accumulated risk under
conditions of modernity and post modernity, and growing polarization with its consequence,
marginalisation of particular groups all make the case for collecting data through the whole
life course of individuals. They also underline the need to extend the birth cohort series with
new cohorts such as the millennium birth cohort study (MCS).

Appendix A presents examples of cohort uses focusing on economic research, research on the
family and on education, health, inter-cohort comparisons and intergenerational effects.


In contrast to the Birth Cohort Studies, the BHPS is based on a sample of the whole
population of Great Britain, rather than single years of age, and its aim is to understand the
dynamics of change of the whole population, and its evolution over the lifetime of the study.
Given that any individual wave of BHPS must address issues of relevance to the whole age
range, the scope must be somewhat broader than that of any individual sweep of the cohort
studies. . It also collects data, and, usually, follows, all the people living in the sample
household, not just a reference individual.

The first and central purpose of the BHPS is to provide high quality data on the short-term
processes of change at the individual and household level for the domains with which it is
concerned: labour markets, income, savings and wealth, household and family organisation,
housing, consumption, health, social and political values, education and training. Through the
collection of this short-term data, they also allow construction of longer sequences of high
quality biographical information across a range of domains. The origin of the household
panel studies was in the need to explore the dynamics of poverty and income. It involved an
understanding that these could not be explored through separate snapshots, but rather

required an approach to collecting a continuous record about income in particular. This
implied frequent interviews to minimise recall problems. These general statements can be
broken into a range of research uses:

      The analysis of the incidence of states and events such as poverty or unemployment
       over time, which provides a very different understanding of their distribution in
       society from that provided by cross-national data.

      The measurement of the rates of transition between states, and the factors associated
       with these transitions. This analysis may be based either on repeated annual measures
       or on the construction of complete histories based on monthly calendars for the
       periods between waves. This short-term retrospective data is much more reliable than
       that collected from longer period life histories.

      The design in which all household members are interviewed and followed permits the
       analysis of associations between the life course of different household members, and
       how their individual decisions may impact on each other. It also makes the household
       panel study particularly suited to the analysis of the dynamics of household formation
       and dissolution, and associated events and outcomes.

      The analysis of the association between change in the different domains (eg health
       and the labour market), in order both to understand causal ordering, and to understand
       the wider social impacts key events and processes.

      The analysis of associations between measures in a modelling context which take
       account of unobserved heterogeneity through the use of repeated measures fixed and
       random effects models.

      The accumulation of life history data, both within the panel itself, and in the
       retrospective life histories collected in the early waves makes the panel particularly
       suited to the analysis of the long-term accumulation of resources (personal and
       financial), and their impact on later outcomes.

By continuing into the over the next decade and beyond it will start to provide data on period
differences in some of the short-term transition processes (for example, whether jobs and
families are becoming still more unstable - something difficult to explore with retrospective

Certain new advantages start to emerge as the panel increases in length. For example it
becomes possible to analyse some of the longer spells and sequences with full data on
antecedents and the evolution of events during the spell. It becomes possible to analyse the
impacts of earlier life stages, including for example the impacts of childhood poverty an
family disruption on later life circumstances, with information from multiple measurement
points, allowing interference about impacts and instability. It also becomes possible to
analyse other lifetime acquisition processes, especially of wealth accumulation, allowing
prediction of economic circumstances in retirement. It is also becoming an important
resource for the analysis of mortality.

Analyses of within household relationships and influences are also informed by better
historical information on all household members. It also becomes possible to undertake
analyses of inter-generational influences.

The existence of an extensive network of household panel studies, throughout Europe, in
North America, means that there is already a substantial level of international comparative
research using the BHPS, and one key goal is to expand the opportunities for such research.

Appendix A provides examples of BHPS uses, focussing on poverty transitions and income
dynamics, family formation issues, labour market transitions, migration, health, political
values, and international comparative research.

Priorities for the Development of Longitudinal Studies

Given the priority to secure the return on the current investments in longitudinal studies, we
start with clarification on current plans, and the issues they raise, before going on to future

The following table summarises the state of plans for the major longitudinal studies which
are, or are immediately likely to come, within the scope of the National Strategy, and
reviewed by the National Strategy Committee. For the period up 2004, the end of the first
period of funding of the ULSC, the table shows approximate spending, both from ESRC and
from other sources. In the period beyond letters indicate timings of activities.

The main aim of this document is to discuss priorities beyond the end of the first five year
funding period of the ULSC, (i.e. September 2004). This also corresponds to the end of the
period of committed funding for the MCS. However it is worth clarifying the funding
situation of the ESRC funded studies up to that point.

A Fifteen Year Concerted Plan for National Longitudinal Data Sets

                        1999-    2000-   2001-   2002-   2003-    2004- 2005- 2006- 2007- 2008- 2009- 2010- 2011- 2012- 2013-
                        2000     2001    2002    2003    2004     2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

NCDS          ESRC      £1.30M                           £0.10M                       A(I)                    A(P)
              Other     £0.23M                   MRC     £0.77M
BCS70         ESRC      £1.30M                           £1.10M                       A(P)                    A(I)
              Other     £0.23M
MCS           ESRC                    £2.35M        £2.23M                        A           B                 A
              Other                   £1.14M        £0.55M
CLS support   ESRC      £0.10M £0.10M £0.10M £0.10M £0.10M             A    A     A     A     A     A     A     A     A     A

BHPS       ESRC         £1.35M £1.35M £1.35M £1.35M £1.35M             A    A     A     A     A     A     A     A     A     A
BHPS/ECHP Other         £0.16M £0.16M £0.16M
BHPS/S&W ESRC           £0.60M £0.60M £0.60M £0.60M £0.60M             A    A     A     A     A     A     A     A     A     A
BHPS/NI    ESRC                       £0.15M £0.15M £0.15M             A    A     A     A     A     A     A     A     A     A
           Other                      £0.15M £0.15M £0.15M
BHPS2                                                                  B    B     B     B     B     B     B     B     B     B
ULSC other ESRC         £0.27M £0.27M £0.27M £0.27M £0.27M             A    A     A     A     A     A     A     A     A     A

              ESRC      £4.92M £2.32M £4.82M £2.47M £5.90M

Cohort studies

Sweeps of the NCDS and BCS70 were funded jointly by the ESRC and government
departments in 1999/2000, as indicated in the table, under a contract with the Joint Centre for
Longitudinal Research (JCLR). ULSC funding included around £500K to support for general
costs in support of the National Strategy and the maintenance of the cohorts. This funding
was somewhat more front-loaded than is shown in the table, in order to support completion of
work on the 1999/2000 sweeps. ULSC funding also contains £1.2 million earmarked for
fieldwork costs in 2003/2004, concentrated on a face-to-face interview survey with the
BCS70, to produce age 33 data comparable with Sweep 5 of the NCDS. It is recognised that
this is insufficient to complete the sweep.

The MRC has agreed to fund a “medical” sweep of NCDS in 2002/3. The long term plan for
the “old cohorts” originally envisaged a sweep of data collection in 2003/4; the MRC funded
sweep could under certain conditions substitute for the planned NCDS sweep; though the
social data envisaged for collection in the MRC survey may be very limited for social science
purposes, and there remain issues of access to these data. Various options for a social science
driven contact with cohort members are under consideration. One factor in deciding between
these is the availability of £770K for a basic skills assessment in the two cohorts in 2003/04.
Medical funding might also be available for BCS70 through an application that is being made
to the Wellcome Foundation. The addition of an intergenerational component through a study
of BCS70 cohort members also appeals because 33 is exactly the age at which a comparable
study was done of one third NCDS cohort members and their children. This might usefully be
a focus of ESRC support for the 2003/04 surveys. Whatever the outcome of these funding
possibilities, the social science interest in the BCS70 must be the priority for the use of ESRC

The first “9 month” sweep of the Millennium Cohort has been funded by the ESRC and a
consortium of Government Departments via a contract with the Centre for Longitudinal
Studies (CLS). The ESRC contribution to the second wave of the Millennium Cohort, to take
place at 36 months, has been agreed at the same level as for the first sweep. Negotiations are
still continuing on government funding, and there may still be gaps in the funding,
particularly given survey cost inflation.

In relation to future plans for the birth cohort studies, the frequency of waves or sweeps is not
fundamental to the design, as it is with household panel studies, and is the main variable as
the level of resources change. Clearly there are returns to more frequent data collection, and
there is certainly value in maintaining regular contact with respondents. The rationale for
frequency of contact must also vary with stage in the life course. At some stages
development and change occur more rapidly. This is the main rationale for the more frequent
interviews for the MCS. The NSC broadly supported the cohort‟s forward plan, which
involved full interviews at eight year intervals, with postal sweeps in the mid-point of the
gap. The plan involves an alternating cycle for NCDS and BCS70, which is designed to
ensure that the two cohorts will have full interviews at the same ages. Being able to analyse
cohort members at the same age is clearly critical to inter-cohort comparison.

However the NSC also recognised that there were some economies carrying out both NCDS
and BCS70 at the same time. This also had some advantages in preserving period
comparability between sweeps. There are also clearly some issues of overload on the work of
the CLS if cohort survey work is too heavily concentrated. One way of securing both the

advantages of age and period comparability would be to carry out full interviews with both
sweeps every four years. However, the cost of this, estimated by the CLS as £3.5 every four
years, if substantially met from ESRC sources would imply a significant increase in ESRC
resources going to longitudinal studies.

However, as the MRC funding for NCDS in 2002-2003 demonstrated, there is considerable
interest in further use of the cohorts for health related research. This can only be of mutual
benefit to the social science and medical research communities, provided that issues
concerning data access are treated in an even-handed way, and data collection plans are not
disrupted. In addition to this there is substantial interest in other research uses of the cohort
studies from government departments, for example in relation to the basic skills agenda. It
might then be possible to plan to use the four year point where a postal sweep is planned for
medical or other survey work. It would be helpful then if the, relatively limited, social
science needs from the planned postal survey could be collected as part of this work.
Reasonable security of funding from the ESRC for the core part of the data collection
programme who make it much easier to plan other data collection, and safeguard social
science research interests in the study.

It is planned to carry out a third sweep of the Millennium cohort study in 2006, as cohort
members reach the age of five and, when members have all entered primary school. This
must be a high priority for funding to ensure that childhood data collected which provides a
baseline for future research. Plans for the MCS beyond age five have not yet been fully
articulated by the study team, but it is clear that the highest priority is a sweep at age 10 or
11, broadly in line with the pattern for BCS70 and NCDS, and if possible, an additional
sweep at age seven.

Although MCS will provide a much larger sample of children with evidence on their own
development than the NCDS second generation study did, there are limits to the amount of
intergenerational analysis that can be done of it. The limits concern the amount of
information that can in principle and in practice be obtained about their parents‟ early life.
Certain sorts of variable, such as behaviour or cognitive scores, could never be retrieved.
Other details of childhood circumstances might be obtained, subject to recall problems and
the limits on the time available to collect the data. A second generation study of BCS70 has
the potential to use much richer data, on one parent, and the information on the other parent
could be as good as what can be collected from MCS parents, subject to co-operation of the
BCS70 partner. Results could be available sooner if the BCS70 second generation is tested
before MCS children old enough to do pencil and paper assessments. This opportunity has
not been fully debated by the NSC, but it must not be allowed to pass without some


Core funding for the original BHPS up to the conclusion of the September 2003 wave is
provided as part of the ULSC contract. Funding for the “Low Income” add-on sample to the
BHPS drawn from the original UK-ECHP, is provided by Eurostat to the end of the
September 2001 wave. ECHP will replaced by a new instrument (EU-SILC), for which
BHPS is likely to provide the longitudinal data. However it now appears most unlikely that
the Low-income add-on will be seen as a useful contribution to this. It is however likely that
other sample supplementation may be required (see 10 below). The ULSC do not propose to
seek refunding for this add-on. The ESRC has also funded from 1999 extension samples in

Scotland and Wales, and from 2001 has co-funded an extension sample in Northern Ireland,
along with NI government departments.

Beyond 2003 the highest priority is the maintenance of annual measures from the original
core sample. However there are some other emerging issues, especially concerning overall
sample size, and re-sampling. The BHPS is still maintaining high data quality and low
attrition, comparable with the attrition rates on the PSID in the USA, and rather lower than
the GSOEP. There is therefore no doubt about its ability to continue to provide high quality
data. However, discussions within the academic community, and with government
departments suggest a need to consider carefully the future of the sample. There are in effect
two distinct lines of argument (respectively achieving adequate sample size, and evaluating
and combating attrition-related bias) that converge on a single conclusion: the desirability of
starting a new full sample to run in parallel with the existing sample, and eventually establish
a very long period rotation.

BHPS is intended as a multi-purpose study with a sample size sufficient to permit analysis of
key policy relevant socio-economic groups. In addition to its academic user base, the BHPS
meets most of the basic requirements for government research needs, but its sample size
(around 5000 households in the core sample) is seen as too small. Both academic and
government users would like to increase the resolution of the study, to allow further regional
breakdowns, and to permit analysis of smaller subgroups. As part of the Government
departments review of data needs, Payne, White and Lakey (1999) reported that „The sample
size becomes more of a constraint when the focus is a minority sub-sample: people with
disabilities, those in further and higher education, women with school-age children or using
childcare services and so on. Also, the proportions experiencing particular kinds of transition
in any one year may be small.‟

The Panel Study of Income Dynamics in the USA (PSID, now in its 34th year) has
demonstrated that a very long panel can remain viable, and continue to provide nationally
representative analyses of relevance to policy makers (the PSID is currently supporting
important research on for example wealth accumulation across the life-cycle, and the impacts
of family disruption and poverty on long term outcomes for children). However at this age
the panel may have become less useful for the analysis of short term dynamics. We would
like to avoid this situation and bring in substantial new refreshment samples at a reasonably
early stage. This has an additional advantage that the new samples can be used to calibrate
the impacts of attrition on transition measures in the first panel. Without such new samples it
is difficult to tell whether attrition is for example leading to a sample which is more stable
than the general population. By assessing the scale of these possible effects, the new sample
allows the first panel to be used with more confidence. It should be noted that the German
Socio-Economic Panel has recently had large new samples added.

The proposed long term strategy is to start a new sample (BHPS-2) of approximately the
same size as the initial sample around wave 13. This sample would be run in tandem with the
first sample for 10-12 years, providing a overall sample for analysis of around 10,000
households. After this a third sample could be started, with the first sample continuing with a
lighter interview, which focused on collecting data for the longer term analyses. This can
only work with very substantial co-funding from government departments, and they are
currently considering funding proposals in relation to this strategy. However, we believe that
the ESRC will want to preserve a stake in BHPS over the very long term, and as the funding

position with government departments becomes clearer we must put in place mechanisms to
secure a co-ordinated approach to the future of the study.

Alongside this we must consider the future of the extension samples in Scotland, Wales and
Northern Ireland, which the ESRC has very largely funded. These are of major importance in
supporting comparative public policy research and the long term impacts of changes in
government of the UK, and in developing research capacity in different part of the UK.
However, while these samples increase the total sample size including the original sample, to
around 9,000 households, there are major differences in selection probabilities in different
parts of the UK. Weighting to remove these biases would cause an inflation of variances in
estimates of sample means or proportions of almost 60%, reducing effective sample size in
UK analyses to under 6,000. The extra samples clearly do not help with analyses of England.
The ESRC, together with government departments need to consider the appropriate overall
pattern of support for BHPS.

In relation to other studies, there do not appear to be any immediate funding issues. Funding
for the English Longitudinal Study of Ageing (ELSA) has been provided by a consortium of
the US National Institutes for Health and various UK government department. The support
appears to be adequate at present but if the US element is reduced in the future the NSC
might think it appropriate to recommend that the ESRC becomes involved.

If the DfES Longitudinal Study of Young People is launched, the National Strategy should
certainly develop proposals to provide for its continuation beyond the planned end, to form
the basis for the „Missing cohort‟.

Overall, our baseline priorities are indicated by the letter A‟s in the table above: alternation of
BCS70 and NCDS face-to-face and postal sweeps, MCS sweeps at age 5 and 10, continuation
of current BHPS samples. At current cost, the costs per annum at 2002 prices over the ten
years from 2004 would be somewhat less than the ESRC spend on longitudinal studies per
annum over the five years 1999-2004 – approximately £3.4 million compared with
approximately £4 million. These number are at constant prices, and plans must allow for
survey cost inflation in particular.

If one was able to assume continuation of current ESRC spending levels, and assuming that
the costs of the Scotland and Wales extensions of BHPS, and for BHPS2 were shared equally
with government departments, it would possible to also support the BHPS2 extension, the
MCS age 7 sweep, and the beginnings of ESRC participation in the new Cohort Study of
Young People sufficient to form the basis for the „missing cohort‟.

The Quality Profiles

As we have already suggested data quality is a fundamental issue for both longitudinal
surveys and longitudinal research. There needs to be a regular process of quality assurance
for longitudinal surveys, and UK social scientists need further consciousness raising around
quality issues. Better understanding of data quality issues is fundamental to better research.
The assessment, monitoring, maintenance and systematic reporting of data quality are
important components of the National Strategy for Longitudinal Studies.

The ULSC has developed a framework for longitudinal survey quality and proposes a
structure for a standard quality profile. This also suggests how quality standards might be

developed and maintained. Quality standards and profiles will help to achieve the following
important aims:

   Documenting in an accessible manner the main aspects of data quality for each study
    (and, as a result, enabling users to make more appropriate uses of the data);

   Identifying priorities for methodological research, to fill gaps where critical aspects of
    data quality cannot be properly documented;

   Identifying priorities for remedial work on the study, where the quality profile suggests

   Identifying priorities for future developments of the study, where the quality profile
    suggests particular strengths and opportunities;

   Raising awareness of data quality issues, and hence contributing to quality improvements.

In developing this framework, we have drawn upon the literature concerning survey quality,
adapting and expanding concepts used by others (mainly National Statistical Institutes) in
ways that make the framework most appropriate for academic longitudinal studies. The
conceptual framework (included as Appendix B) provides the crucial foundation necessary as
the basis upon which to develop quality profiles and quality standards. Without this, profiles
and standards run the risk of being atheoretical “shopping lists”. The framework focuses on
six dimensions of quality.

   relevance
   accuracy
   timeliness
   accessibility
   interpretability
   coherence

A second document (included as Appendix C) outlines the components and structure of a
generic longitudinal survey quality profile. It raises a series of questions which should be
answered for most longitudinal studies. The profile includes standard statistical issues and
issues relating to content, data availability and usability.

A draft BHPS quality profile was presented to the January 2002 meeting of the NSC, and
equivalent documents for the cohort studies are expected to be developed. When complete the
quality profiles will provide important information both for users of the studies, and for the
planning of their future development to maximise survey data quality and optimal use of
resources. We propose that the preparation of such quality profiles would be expected of all
longitudinal studies funded by the ESRC. This may have resource implications for those
producing the data.

A Methodological Research Programme

The collection and use of longitudinal data are underpinned by an active community of social
scientists interested in research methodology. The National Strategy must ensure key needs
for methodological research related to longitudinal data are met. The strategy cannot work

against the development of methodology as a science, with priorities set by what appears to
be the exciting potential new development within an international literature. However,
methodology also acts to support other activities, both research and data collection, and
insofar as there are clear needs, they should be articulated. Three broad areas can be

1. Analytical methods which allow researchers to exploit the particular, and often rather
   complex structures of longitudinal data. We should have a reasonable expectation that
   the ESRC programme can make a significant contribution here.

2. Research which contributes to an understanding of how longitudinal data quality issues
   may influence the research which can be carried out using the studies.

3. Research which contributes to the improvement of longitudinal data collection methods
   and longitudinal study design.

The quality profile exercise will contribute important information towards the second and
third of these, and most importantly, it will help to identify gaps in our knowledge.

As a next step the ULSC is undertaking a scoping study on the current state of knowledge
across the main areas of data quality concern. In the first instance this is to establish whether
systematic reviews already exist, take a broad overview of existing work and identify where
the main gaps in knowledge appear to be. The following are main headings:

   Attrition/Unit Non-Response: Patterns of attrition/bias in panels of different duration,
    interval, target population, etc, Longitudinal weighting, methods for bias assessment,
    response maximisation techniques

   Item Non-Response: Imputation methods, minimisation techniques

   Measurement errors: Validation studies, reliability studies, studies of memory effects,
    alternative questioning methods

   Panel Conditioning: Effects on knowledge, Effects on reporting, respondent burden

   Dependent interviewing: effects on micro-validity and statistical error, effects on
    respondent perceptions and reactions

   Interviewer Effects: on response rates, on measurement error, on standard errors

   Longitudinal Sampling Issues

In addition the ULSC and the NSC will be reviewing the outcomes of the ESRC Research
methods programme, particularly with a view to whether it adequately supports priorities for
developments in longitudinal analysis methods

Over and above the research funded from the current call under the ESRC Methods
programme we may identify the following ways in which further research may be supported:

   The producers of longitudinal data should be encouraged to report methodological
    findings of relevance to data quality and data collection methodology, and substantive

    researchers of these studies should similarly be encouraged to publish similar relevant

   Other standard funding sources should be tapped to support research which meets
    international publication standards.

   The ESRC should be made aware of major strategic needs, which might for example be
    met through the second round of funding under the Methods programme.

Our own thinking at the ULSC suggests that one major need which will prove relatively
expensive to meet is for properly designed validation studies, which would almost certainly
require new data collection.


Longitudinal data are relatively complex to use compared with cross-sectional data, and in
order to ensure that they continue to be used various steps are required. This includes efforts
to improve usability and accessibility of longitudinal data sets. The ULSC is working with
CLS on the cohort studies to this end, and the ESRC‟s invitation to tender for archiving and
data dissemination functions promises further investments in this area. However, this in no
way reduces the need for training in longitudinal research methods (including issues
concerning both analysis and data collection) where current provision is inadequate.

The ESRC has identified the need to improve training in quantitative methods for UK social
scientists. This is particularly salient to longitudinal studies where the UK has recognised the
importance of these studies and made substantial investment in data collection. However, if
these studies are to be fully exploited it is also necessary to invest in the training
infrastructure. This requires a recognition of the cost of providing training and careful
consideration of where the costs should fall.

Whilst training is provided by the ULSC and the CLS, it is important that training is not
confined to these sites. A range of training needs to be available, designed for different sets of
users, at different levels of expertise, across a range of disciplines and geographically
dispersed. It is also important to recognise that the need for training in longitudinal studies is
not confined to academia. There are important savings to be made through joining with
government in a joint training programme. The Committee therefore recommended talks with
ONS and the Chief Economists in various government departments in order to secure joint
funding for a common programme. Committee members are able and willing to facilitate this
if needed.

A earlier paper on behalf of the ESRC‟s National Longitudinal Strategy Committee set out a
framework for training in longitudinal studies and made specific recommendations to the
ESRC, which are followed up in part through the Research Methods programme. Following
funding decisions under the Methods Programme the Committee will want to ensure that
there is a comprehensive listing of training available in longitudinal studies with links to
appropriate sites. It will also want to identify gaps in training that remain and make
recommendations to ESRC over how these can be filled.

Outreach/User Engagement

Wider support for the programme of longitudinal studies in the UK depends in broadening
knowledge of the opportunities and results of longitudinal research. This depends in part on
the dissemination of research results, to policy makers, to the wider academic community,
and through the media to the public. This activity will mainly happen through the
dissemination programmes of individual research centres (as for example the ISER
dissemination strategy). Other steps include the conferences held around particular data
resources, such as the BHPS research conference in July 2001. This was the first of a series
of conferences which will alternate biennially with the GSOEP conference in Berlin. Other
steps include commissioning publications based on new data sets. Networks of authors in
both Scotland and Wales have agreed to co-ordinate volumes based on the first waves of the
BHPS extension samples.

There is also room for broadening of the intellectual horizons of the academic community as
a whole (ie both current and prospective users of longitudinal materials). There are in
existence many more longitudinal studies, outside the ESRC portfolio, of substantial interest
to the academic research community, both in the UK, and in other countries. And there is
considerable scope for the sharing and enhancement of longitudinal analytic and data
management skills across the community. Here the ULSC has a major role. It is recently
completed (in collaboration with the ONS) a major exercise to improve the dissemination of
information about the full range of longitudinal studies through an easily accessible on-line
catalogue, providing study descriptions, documentation, contact and access details.

A website is also crucial to disseminating information about the National Strategy and the
activities of the NSC. It should provide a channel for both consultation and dissemination of
activities related to longitudinal studies. Users need to be able to register their interests in
longitudinal studies and find information about conferences. The website also needs to hold a
programme of training activities with links to the relevant providers and a list of publications
based on the studies. The current website,,
which contains basic information on the activities of the NSC, is due to be re-launched
shortly with many of these additional features. This as an important step in establishing an
over-arching framework for longitudinal studies in the UK.

                                         Appendix A

                         Exemplary Uses of the Existing Studies
                       Exemplary Uses of Birth Cohort Study Data

Examples of excellent uses of birth cohort study data, which were indispensable to the
research question pursued are numerous and can be found in all the main domains to which
the birth cohort studies relate.


Returns to qualifications and returns to education more generally have been pursued through
a large number of studies, most notably in research undertaken at the Institute of Fiscal
Studies. Outcomes in adulthood such as occupation and earnings are set against
qualifications, taking account of ability as tested in childhood and numerous other
circumstances and experiences earlier in life which might be confounded with them.
Statistical modeling of this kind, is not a perfect substitute for the controls offered in
randomized experiments, but goes some of the way to producing the most plausible accounts
of micro economic processes. Similar information, with measures of human capital drawn
form education, childhood test and earlier adult employment experience has been used in the
analysis of gender differences in wages (Joshi and Paci 1998)


Family formation and dissolution is a longitudinal process which reflects the changing nature
of human partnerships and their outcomes in child bearing and child rearing. There is much
debate about the short term and long term consequences of such dynamics, particularly for
individual children. Classic studies have been undertaken using birth cohort study data by
such researchers as Kath Kiernan at the London School of Economics and Martin Richards,
at the Family Studies Unit at the University of Cambridge and Elsa Ferri at CLS, which have
pointed to the critical role of economic conditions in shaping child outcomes. For example,
separation and divorce of parents may not be a problem under conditions of unchanging
affluence, as far as most children are concerned. Similarly quality of relationships before
separation and divorce is as critical factor in determining the impact on children of marriage
breakdown as the divorce itself. The mediating role of family resources was also apparent in
the study of children in the NCDS second generation who experienced family change,
according to the CLS team (Joshi, Wiggins, Clarke et al) funded under the ESRC Programme
on Children 5-16.

Escape from Disadvantage

The birth cohort studies offer one of the few opportunities to identify those individuals and
the factors affecting them who buck the trend in failing to bear out the predictions of earlier
studies. For example in one project using NCDS conducted by Doria Pilling people were
identified at age 23 whose early childhood had predicted very poor consequences in
adulthood; yet the individuals concerned had actually been successful. These people were
followed up at age 27 in a supplementary interview to compare them with individuals whose
life courses had conformed with prediction. From analysis of the data collected it was

possible to identify the key role of family aspirations and positive evaluations by teachers as
overriding poor economic conditions in providing the support children needed to succeed.


Some of the main evidence on school organization was produced from analysis of birth
cohort study data. Jane Steedman demonstrated the negligible effect on children‟s progress of
a segregated system offering grammar schools, technical and secondary modern schools
compared with a fully comprehensive system. Bright children were shown to thrive in either.
Only those who were vulnerable to risk of failure in various ways might benefit from one
from of school organization rather than another. Studies of literacy and numeracy conducted
for the Basics Skills Agency at CLS have been greatly informed by birth cohort study data.
Examples include: establishing the importance of numeracy, over and above literacy, in
modern employment; identification of the antecedents of poor basic skills in early childhood,
especially the failure to develop visual motor skills; the decline in numeracy with
unemployment duration; the consequences of basic skills difficulties for employment, for
social and political participation and for psychological health. Much of the work on social
exclusion has been fuelled by the growing recognition of the significance of basic skills in
modern society and the problem of marginalisation for those who fail to acquire them.


Numerous examples from the birth cohort studies literature affirm the huge value medical
researchers have gained from analysis of the cohort studies data. The first identification of
smoking as a hazard in pregnancy for later child development came from analysis of NCDS
data. Similarly some of the evidence of the value of breast milk as opposed to artificial milk,
for babies was also identified through this route. The relative value of eye tests and hearing
tests at different ages was also demonstrated. More recently evidence has been produced for
the role of air pollution in asthma and wheezing. The positive and negative consequences of
vaccination have all been pursued through birth cohort studies as have the aetiology of
threatening conditions such as cancer. The growing field of health inequalities research has
also been greatly informed by birth cohort studies data as witnessed by the recent Acheson

Inter cohort comparison

There has been an increasing tendency for research to embrace the analysis of data from more
than one birth cohort together. Recent examples include the research carried out by the CLS
in collaboration with colleagues in other institutions (LSE, IFS, City University, University
of Warwick) in policy driven research on the benefits to the individual and society of
improving adult basic skills; on obstacles and opportunities on the route to adulthood;
accumulated risk in the life programme and young people‟s pathways to employment and
independence. A book in preparation, arising from the most recent cohort studies surveys,
compares cohort members‟ statuses and their relationship to other variables across all three
British birth cohorts: the 1946, 1958 and 1970 birth cohort studies. Cohort comparisons point
to the heightened importance of qualifications in relation to entry into and engagement with
the labour market and increasing evidence of social exclusion among those without them.
Such studies are particularly valuable to policy because of the facility they provide for
„benchmarking the long-term effects on the life course of social (and policy) change. From

the social science perspective they also enhance explanatory power in modelling life course
transitions with the additional conditioning on cohort effects.

Intergenerational Studies

The 33 year NCDS follow-up one third of cohort members and their children, with
comparable cohorts in the US National Longitudinal Study of Youth have led to a growing
body of research in the UK and the USA using on returns of investment of parental time such
those done by Heather Joshi and colleagues on the effects on child development of mothers‟
working in the child‟s early years and of changes in family living arrangements: by Paul
Gregg on poverty and child development, and by Bob Michael on international comparisons
of measures of cognitive development and on gender differences in parenting effects on

Eliabeth Cooksey, (Ohio State University) collaborated with Joshi and others at CLS to
compare data on the children on NLSY and NCDS. The fitted parameters, particularly those
of interest for family structure and income were remarkably similar. US and UK children
showed about the same modest level of disruption from father absence and or low income,
not as might have been expected a greater degree of adaption in the higher divorce setting of
the US.

Exemplary Uses of the BHPS


Research on the dynamics of poverty has identified that while there is considerable mobility,
much of it is short distance churning, rather than long distance movement. Cross-national
comparative research also suggests higher child poverty risks associated with key life events
in Britain, compare with Germany, and that is not because the greater prevalence of economic
shock in Britain, but because the German welfare state is substantially better at protecting
children from the consequences of these shocks.

Other research has focused on the sources of income risk, and the determinants of income
mobility, and on the association between unemployment risk and poverty, suggesting which
types of people are most risk of poverty arising from changes in the macro-economic cycle.

Research on pensioner poverty related to wealth accumulation has identified impacts of
gender, lifetime employment patterns, and household disruption. Comparative research has
shown similar overall patterns of income risk of widows after husband‟s death in US, UK and
Germany, but different causal pathways – e.g. greater role of the wife‟s earnings in UK, of
the husband‟s in US.


BHPS research has provided much of the recent evidence on the implications of the growth
of cohabitation. In particular it has shown how it is associated with a growth in the single
parent families. Research on the socio-economic influences on the size and attributes of the
lone parent family population has been used to improve medium term forecasting of the lone
parent family population. Other related research has focused on the education and
employment outcomes for children from lone parent families; on the stability of child support
receipts from non-resident fathers, and on the influences job mobility rates for lone parents.

Research on the impacts of parental circumstances on child outcomes has identified impacts
from cohabitation and partnership disruption and influences on parental labour market
participation and the availability of child care.

Comparative research across four European countries using the BHPS has suggested least
postponement of maternity in the UK and quickest return to work after first birth, controlling
for education.

Research on how the division of domestic labour adjusts to the employment participation of
husbands and wives suggests rather slow and incomplete adaptations to change.

Labour market

Research on the labour market has suggested an increasing turbulence, which is two sided.
On the one hand there are an increasing number of jobs with poor career and pay prospects,
but on the other hand individuals are also using the flexibility of the labour to find
employment which better suits their skills and requirements, and hence promote career

Research on the interaction of work and family suggests substantial long term impacts on
human capital acquisition associated with different patterns of female labour market
participation around the birth of children.

Research has also focused on the scarring effects of unemployment, on mobility between low
paid jobs and unemployment (suggesting that getting people into just any job does not
necessarily lead to long term improvement in their situations). Other research has examined
the long term consequences of temporary employment, the impacts of minimum wage
legislation, and the impacts of insecure employment on mental health.

A range of other research has examined gender differences in the labour market, suggesting
some convergence of male and female careers, but continuing differences in promotion
prospects and pay growth.

Other recent research has used the BHPS to assess the impact of the national minimum wage,
and found no evidence of reduced job opportunities for lower wage workers.


Research on migration and residential mobility, which for the first time has been able to use
high quality information on the situation before the move has shown the importance of
income and human resources in influencing mobility probabilities, and suggested that the
previous emphasis on the role of housing tenure may have been over-stated.


Analysis of repeated measures of stress such as the GHQ has shown how increases in levels
are associated with changes in other life-domains, and are conditioned by poverty and other
family background characteristics.

Politics and values

Research on changes in political attitudes and values have provided evidence on the
associations between political values of different members of the same household both in
terms of cross-sectional patterns and patterns of change. Other research has investigated the
association between political attitudes and beliefs and political behaviour. It has also
suggested considerable instability over time in many social attitudes.

Area effects

Research on neighbourhood effects on life chances has identified associations with
employment and poverty transitions, and neighbourhood variations in social networks and
social disorder. However it also suggests individual characteristics are more important, with
implications for neighbourhood focussed policy.


Comparative analysis of wealth data from BHPS and the PSID suggests great differences in
the portfolios of wealth holding, with much greater wealth in the UK held in housing,
compared to a greater share of wealth in stocks in the US. It is suggested that one source of
this is earlier investment within the life-cycle in housing in the UK.

Social exclusion

Research investigating social exclusion as a multi-dimensional longitudinal phenomena has
suggested that persistent exclusion on multiple dimensions is a extremely rare in the UK.

                                Appendix B

          A Quality Framework for Longitudinal Studies


To top