Abstract
This issue of the National Institute Economic Review presents a selection of research findings on the economics of the family, education and the labour market, linked by their use of evidence from the British Birth Cohort Studies. In addition to providing insights, these papers serve also to draw attention to this rich resource available for research. The collection throws light on: the relationship across generations between parents and offspring; developments within a generation between early and later life during individual lifetimes; patterns of inequality within a generation; and on changes over historical time between people born at different points.
These studies of contemporaries have acquired a specific label from the Roman army: cohorts – people setting out, from a common point, marching together through time. Even if they have nothing else in common, members of a birth cohort have a common birth date or birth year. Following each member through time makes their study longitudinal. The British Birth Cohort studies differ from the British Household Panel Study and its successor, Understanding Society, which follow panels of people of a whole range of ages at their starting point. The latter are grouped in households, and collect data more frequently, but have fewer subjects of a given age. The Cohort Studies are multi-purpose, multi-disciplinary longitudinal databases with a host of applications, separately or in comparison with each other. It is just a few of these uses which are illustrated in the contributions to this issue, but it is hoped that other research will be stimulated, adding to the already large corpus of work taking advantage of this heritage.
All of the articles here use longitudinal data from one of the three studies, on the cohorts born in 1958, 1970 and 2000, which are managed by the Centre for Longitudinal Studies (CLS), in the Institute of Education, and available for secondary research through the UK Data Archive. Most of the articles are in some way concerned with the transmission of economic circumstances from parent to child. These studies have also provided most of the British evidence used in a recent cross-national review of the subject edited by Ermisch, Jänti and Smeeding (2012). Another major area in which cohort studies provide material for economists is the study of gender issues in the labour market, represented here by the paper by Dex and Bukodi on women's occupational histories. 1 After an outline of the contributors' results, the reader will find more information about the history, content and future of the studies.
Research questions and results
In the opening article, NIESR's Director of Macroeconomics plunges in to the microeconomics of the family. Angus Armstrong develops the classic theory of parental investment in their children, including the role of credit constraints, to incorporate family culture, in the form of parental beliefs. The concern is whether or not the family believes that effort is rewarded, ‘belief in a just world’. The argument is that parents' beliefs about whether effort is worth making will be absorbed by the child, influence how they choose to behave and then be reflected in childhood attainment, regardless of intentional parental investment in supplying the child with material or time resources. The empirical test of this hypothesis combines the cognitive (vocabulary) scores obtained in 1991 from the children of the 1958 National Child Development Study (NCDS) cohort members and the answers from these adults to questions reflecting belief in a ‘just world’. These measures of family culture are more robust predictors of child scores than family income (though not the parent's own cognitive ability). How the second generation sub-sample fits into the context of the wider structure of the 1958 study can be seen in table 1. Although the children who are included in the analysis are necessarily restricted to those with relatively young parents, and have not been followed up longitudinally, the second generation study offers a rare opportunity to compare childhood circumstances and outcomes across generations within the same family.
National Child Development Study (NCDS): main sources of information (all births in Great Britain born in a week of March 1958)
Notes: (a) Cross-sectional achieved sample of cohort members – immigrants with appropriate date of birth included from NCDSI-3.
(b) Assessments of cognitive skills directly administered to cohort members.
(c) For a random sample of one in three cohort members. Information was collected directly from 3,008 children of cohort members.
(d) This could be the Cohort Member, their Spouse, or Partner (same 1 in 3 random sample). 2,588 mothers completed the ‘mother’ questionnaire, giving information on 4,278 children.
• Geographical identifiers permit splicing in of external data from 1974.
• For full details of response, sub-studies and questionnaire content see www.cls.ioe.ac.uk/ncds.
Armstrong's findings add a new perspective to the debate about how and why children of higher income parents ‘do better’ than the children of the poor. Others have stressed parenting behaviour (Field, 2010; Allen, 2011), the complementarity of parental resources and home learning (Ermisch, 2008, Kiernan and Mensah 2011), or the parent's willingness to give care (Michael, 2011, also based on the NCDS second generation study). Armstrong's paper is unusual in this literature for recognising that children are not merely recipients of ‘transmissions’. They also learn from their parents and choose to make an effort or not. The final model in the paper finds that in the even smaller sample where there is some evidence about the child's own sense of agency (locus of control), this variable trumps the predictive power of family culture. It is far from obvious that this paper identifies any easy policy levers, or that the manipulation of these aspects of home life is either feasible or desirable (Henricson, 2012). Armstrong shows how parents' disaffection may impact on their children, and possibly over generations, in a way which is not easily solved by money alone. Although he argues that beliefs change slowly over time, readers may wonder how the intergenerational cycle might be intensified by a prolonged macroeconomic recession.
The second article, by Shirley Dex and Erzsébet Bukodi, is on the role of part-time employment in the occupational mobility of women. It also uses the 1958 cohort, but looks longitudinally over the course of their adult lives. Labour market trajectories are followed for up to 30 years from school leaving in the 1970s to age 46 in 2004. They have painstakingly constructed and cleaned month-by-month job histories, with consistent dates and consistent occupational coding, for both men and women, and have made the histories available through the UK Data Archive. The ranking of occupations according to average occupational earnings reveals that both men and women experienced both upward and downward moves in their careers to mid-life (Bukodi, Dex and Joshi, 2012). Among women, downward mobility into part-time employment upon child-bearing has been a particular feature, and a major source of gender wage inequality. For NCDS women from ages 16 to 46, the event history analysis shows a clear association between transition into part-time work and occupational downgrading, much less likely in switches into full-time jobs. Such switches did not normally restore original status after earlier downgrading. The chances of retaining a high rank were greater in occupations offering opportunities to work part-time. It is likely that the combined disadvantages of long employment breaks and low level occupation on return have reduced somewhat for recent cohorts of women, but the penalties of part-time employment have not been eliminated. They are all too real and irreversible for the generation of individuals studied here, now approaching (a receding) retirement age, many of whose lifetime earnings have been dominated by part-time episodes in low paying occupations.
The third paper in the collection, by Kathryn Duckworth and Ingrid Schoon, turns to the 1970 cohort (BCS70), outlined in table 2. The BCS70 experience as teenagers is compared with another group of young people, born some twenty years later – the Longitudinal Study of Young People in England (LSYPE) – who can for these purposes stand in for the ‘missing’ birth cohort that did not happen in the 1980s. LSYPE has been followed from age 13-14 up to age 18 in the recessionary conditions of October 2008. The authors look at young people who at 18 are neither employed nor in education or training (NEET). The first cohort was 7 per cent NEET at 18 in 1988 versus 16 per cent in LSYPE in 2008. In each cohort it was children from relatively disadvantaged homes who were more likely to be NEET, but not all of them. The authors identify ‘protective’ factors which reduce the chances of NEET outcomes for those apparently at risk, given their home background, or raise those of ‘Beating the Odds’. Notable among these are the young person's motivation (resonating with Armstrong's finding), cognitive ability, part-time employment at 16, and for LSYPE, a favourable social mix at the school. The experience of earlier cohorts suggests that those who are NEET at 18 face poor prospects of employment in later years, whereas those who ‘beat the odds’ have a better chance of an upwardly mobile career before them. The datasets used by Duckworth and Schoon are far from unique in suffering from missing data. This paper tackles the problem by adopting multiple imputation of missing data on predictors.
British Cohort Study 1970 (BCS70): main sources of information (all births in Britain, born in a week of April 1970)
Notes: (a) Cross-sectional achieved sample – immigrants with appropriate date of birth included from age 5 and 10. Numbers for Great Britain only as births in Northern Ireland not followed up.
(b) Assessments of cognitive skills directly administered to cohort members.
(c) For a random sample of one in two cohort members. Information was collected directly on 3881 children of cohort members.
(d) This could be the Cohort Member, their Spouse, or Partner (same 1 2 random sample). 2,846 parents completed the ‘mother’ questionnaire, giving information on 5207 children.
• Geographical identifiers permit splicing in of external data from 1986.
• For full details of response, sub-studies and questionnaire content see www.cls.ioe.ac.uk/1970cohort.
The last two papers use the Millennium Cohort Study (see table 3). Denise Hawkes and Heather Joshi are concerned with the role of the age at which a woman becomes a mother in the intergenerational transmission of advantage and disadvantage. (Among other things this has implications for how much it matters that the NCS and BCS70 second generation studies are confined to young parents). They focus on the 43 per cent of the Millennium Cohort children who were their mother's first-born. Child developmental outcomes at age 5 (MCS3) are analysed in terms of retrospective information on the mother's earlier life and circumstances at the first survey (MCS1). The paper investigates differences between women entering motherhood at extreme ages – teenagers versus new mothers in their late thirties and early forties. There are relatively few either below 20 or over 35, 12 per cent each, but there is interest in whether the age of these women in itself makes a difference to their capacity as mothers, or reflects other correlates of early and late childbearing which may also affect the offspring. The analysis found very little difference in child outcomes between the entrants to motherhood in their late thirties or forties and those who had had their first child in their early thirties. There were differences for the children of younger mothers, particularly the small number of ‘teenage mothers’, under 20 at the cohort child's birth. These were related to social disadvantages recalled from the woman's childhood, and also to socioeconomic disadvantages current in the child's first year. It varied by child outcome how far these factors accounted for the apparent disadvantage of having a young mother. It came nearest to being explained by socioeconomic circumstances for behavioural difficulties and least for the verbal cognitive test. The results are consistent with modest ‘returns’ in child development to delaying having children, up to a point, but even these are not conclusive as the information on the mother's early life is limited. In any case, most recent evidence (and that presented by Duckworth and Schoon) suggests that teenage childbearing is on the decline, despite the recession.
UK Millennium Cohort Study (MCS): main sources of information (born 2000/1)(a)
Notes: (a) Children eligible were eligible if born between Sept 1, 2000 and August 31, 2001 (England & Wales), or between Nov 23, 2000 and Jan 11, 2002 (Scotland and Northern Ireland), and were resident (at 7 months) in a stratified sample of electoral wards. Over-sampled areas had high child poverty rates, high minority ethnic populations in England, or lay outside England.
(b) Unweighted achieved sample of children. The number of families is about 300 fewer due to twins or triplets. Numbers were boosted at Sweep 2 by some 700 who were eligible for Sweep 1, but missed. No recruitment of immigrants.
(c) Assessments of cognitive skills directly administered to cohort members.
(d) Conducted in 2005 for a small subsample (632) of children attending group settings in England at MCS2.
(e) Includes measures of anthropometry, links to routine health records and various auxiliary collections of health data.
(f) Routine records of the Foundation Stage Profile in England, postal survey of teachers in Scotland, Wales and Northern Ireland.
(g) Includes National Pupil Database for England, accessible via the Secure Data Service, and OFSTED ratings.
(h) England and Wales only.
• Full details of sample design, response, sub-studies, questionnaire content, geographical identifiers and descriptive reports of initial findings from www.cls.ioe.ac.uk/mcs.
The last paper, by Anitha George, Lucy Stokes and David Wilkinson, focuses on early education and outcomes at primary school for the MCS children. The very latest data are featured, linked from the 4th survey into the educational records, Key Stage 1 results, from English state schools (see table 3). They are analysed in terms of material from the three previous sweeps, and the OFSTED inspectors' rating of the school attended at age 7. The Millennium Cohort was among the first to experience the free nursery school provision for three- and four-year-olds, and this study is the first to relate outcomes at age 7 to the new regime of pre-school education. Perhaps because the early education was almost universally taken up, and may not have been of a particularly high quality when it was first rolled out, it has proved difficult to detect much impact in the average Key Stage 1 results for seven-year-olds. However attendance at early education does appear to be associated with better school attainment among children from families who were in poverty. This finding may reflect poorer home learning environments in the poorest families. It may also be an encouraging piece of evidence that the poorest were able to benefit most from a universal intervention. However, one should remain cautious, as the study does not consider the quality of early education, the measurement of which is complex and costly. A multi-purpose non-experimental study like the MCS, sampling homes rather than settings, has not proved ideal for evaluating this policy.
What are the Birth Cohort Studies?
National birth cohort studies were pioneered in Britain by the cohort of 1946, the Medical Research Council's National Survey of Health and Development (Wadsworth, 2010). That remarkable study, still going strong with subjects in their sixties, is not represented here. Medical funding has limited (though not ruled out) its application to economic topics. The earliest of the national birth cohorts to be supported by the ESRC and based at the Centre for Longitudinal Studies, is the National Child Development Study (NCDS). This started out as a study of maternal and perinatal health in a cohort born in Great Britain in a week of March 1958, as in the 1946 study. The study became longitudinal in 1965 to inform the Plowden enquiry on primary education. Unlike the 1946 study, there was no sub-sampling of the original cohort (over 17,000 births) and the follow-ups in childhood actually recruited immigrant children with the study birth dates (see table 1 which also summarises the types of data collected sweep by sweep). This shows that the fifth follow-up (NCSD5) at age 33 was particularly complex, including a supplementary survey of the children and their mothers of one cohort member in three (used by Armstrong, below). This included cognitive tests designed to be comparable with those administered to the children of the US National Longitudinal Study of Youth, NLSY79. The numbers in the bottom row of table 1 show that NCDS, like indeed most other longitudinal surveys, suffered attrition through the contacts at ages 7, 11, 16, in childhood and up to 50, so far, in adulthood, with the next survey in preparation for age 55 in 2013.
The second cohort managed by CLS is the British Birth Cohort Study of 1970, now known as BCS70. It also started with all births in a week across the country. This time, from the outset there was the intention to follow up. Although births in Northern Ireland were included, it proved impossible to continue the study there, so the longitudinal study is also based on Great Britain. Even on the mainland, the funding was not always secure. The surveys at age 5 and age 10 were run from Bristol University, the one at age 16 was put in the field by an independent charity. The ESRC funded a postal survey at age 26, which ensured that the follow-up continued into adulthood, at four-year intervals, as set out in table 2, up to the interviews currently in progress at age 42.
No new national birth cohort study was initiated until the Millennium Cohort Study (MCS). The subjects of this fourth national cohort were born shortly after the turn of the millennium (see table 3). Although it was inspired by the model of the first three cohort studies, and provides comparisons with them, it has several new features. As it was no longer feasible to use National Health Service staff in the increasingly sophisticated data collection process, the practice of taking an unclustered national sample in one week was replaced by a clustered sample design, stratified to over-represent areas (electoral wards) with ethnic minorities, high child poverty and the three smaller countries of the UK including, this time, Northern Ireland. The complex survey design necessitates complex analysis (Plewis, 2007). It also covers all seasons of birth. Other innovations include the interviewing of fathers, the incorporation of data from administrative records, and a greater emphasis on social and economic information than in the previous maternity surveys. As shown in table 3, the MCS families were followed up when the cohort child was 3, 5 and 7. The fifth sweep (MCS5) when the children are 11, in their last year of primary school, is still ongoing. Plans for a sixth survey at age 14 in 2015 are in train, and the intention is to keep following this cohort too into adulthood.
Future developments
The existing cohort studies are well placed to continue following the lives of their members and a new cohort study is joining the portfolio on a larger scale than anything done so far. Resources have come from the Large Facilities Capital Fund and it will be based at the Institute of Child Health in University College London, under the direction of Professor Carol Dezateux. Life Study, as it has been named in 2012, will recruit the mothers of 100,000 babies in pregnancy after a pilot in 2013. It aims to be a multi-purpose interdisciplinary resource, and has an economist, Professor Alissa Goodman, as its deputy director.
Government investment in cohort studies via the Large Facilities' Capital Fund is also being reinforced by the establishment of an infrastructure to strengthen collaboration between cohorts and comparison and dissemination of results. The Cohorts and Longitudinal Studies Enhancement Resource programme (CLOSER) started in October 2012. It is a network of nine research resources, bringing together the three studies featured here (NCDS, BCS70 and MCS), two other national birth cohort studies, the 1946 birth cohort (MRC NSHD and Life Study); the new enlarged household panel study, Understanding Society, and three locally based cohorts. One of these, the Avon Longitudinal Study of Parents and Children (ALSPAC), with its large sample of children born in the early 1990s, is also used to stand in for the ‘lost generation’ of cohorts between 1970 and 2000 (for example, Goodman, Gregg and Washbrook, 2011). The others are the Hertfordshire Cohort Study and the Southampton Women's Survey. The facility is funded for five years in the first instance, and may include more surveys in due course. For more information consult www.closer.ac.uk. In its role to support use of the cohort studies presented here, the Centre for Longitudinal studies welcomes questions about how to use the existing resources, and advice on how the quality and utility of future data collection may be improved.
