Postponement and childlessness: Evidence from two British cohorts

This paper starts by reviewing existing projections of childlessness among British men and women. Low current fertility implies high eventual childlessness unless the postponement of parenthood is taken into account. Such re-timing of first births appears to be occurring differentially across social groups. Exploiting the disaggregated evidence of two British cohort studies, the 1958 National Child Development Study and the 1970 British Cohort Survey, this paper investigates the extent of postponement across cohorts and projects its impact on eventual levels of childlessness. Men and women are considered separately in our models of a population stratified by educational attainment. We find the most striking postponement occurring among graduate men. Among graduate women, after taking family building intentions into account, we estimate that about a quarter of 1970 born graduate women will remain childless, rather than something nearer 40 per cent as had been conjectured elsewhere. 1 Institute of Education, University of London. E-mail: [email protected] 2 Institute of Education, University of London. E-mail: [email protected] Kneale & Joshi: Postponement and childlessness: Evidence from two British cohorts 1936 http://www.demographic-research.org 1. Background Increasing numbers of people in Britain are either postponing parenthood, or not experiencing it at all. Since the 1960s, decreases in Total Fertility have been accompanied by rises in the median age of women at first birth, as fewer women have first births in their teens and early twenties. This partly involves a re-timing of the initiation of parenthood to the late twenties and thirties, but it also involves a gradual rise in rates of eventual childlessness, not necessarily intended. Graduate women have been a focus of scholarly and popular attention as being in the vanguard of this trend. Male fertility histories, on the other hand, have largely been overlooked. In this paper, we examine the pattern of postponement and childlessness in Britain across the range of social groups in two cohorts for both genders. We use the (near) complete observed fertility schedule of the earlier 1958 cohort (NCDS) to estimate a full fertility schedule for the later 1970 (BCS70) cohort, observed only up to age 34. 2. Existing estimates of childlessness The past four decades have witnessed low and late fertility over much of the developed world. Discussion of the Second Demographic Transition includes both the outcomes of these changes and their determinants (Lesthaeghe and Neels 2002; van de Kaa 1987). The outcome has been below replacement level fertility and rising age at first birth. Social changes in patterns of female labour market and educational participation and changes in marital and cohabitation patterns have been offered as drivers, and explanations for low and late fertility may account for postponement, voluntary childlessness and involuntary childlessness. While these are not stable or mutually exclusive ‘states’ one may lead to another, and they may respond to different stimuli each has very different implications for the level of childlessness that could be predicted, and the extent to which it may be a cause for concern for policy, society and individual couples. Recent trends in births in Britain are visible through the period Total Fertility Rate (TFR), which (in England and Wales) dropped from a peak in 1964 of 2.95 children per woman, and remained below 1.80 into the twenty-first century, rebounding somewhat to 1.86 in 2006 (Office for National Statistics 2007). At the same time, mean age at first birth in England and Wales rose from 26.7 to 27.6 between 1996 and 2006 (Office 3 Office for National Statistics estimates refer to England and Wales alone. However, given that England and Wales account for nearly 90% of the UK population, and that Scottish fertility trends have mirrored those of England and Wales in terms of rising age at birth, it is suggested that these estimates do reflect national trends. Demographic Research: Volume 19, Article 58 http://www.demographic-research.org 1937 for National Statistics 2007). While not being directly indicative of rising levels of childlessness, these period indicators do demonstrate that fertility is becoming lower and later among British women. In a similar way, lower and later rates of entry to first birth cannot directly illuminate trends in completed family size; but it is easy to speculate that higher levels of childlessness and late entry to first parenthood will have a negative impact on higher order births. Withdrawal from parenthood during the twenties may be a postponement or permanent. While entry to parenthood is irreversible, deferring it increases the risk of it never happening, due to biological and other constraints. Indicators of cohort fertility show that the proportion of eventually childless women doubled from 9% for those born in 1946 to 18% for those born in 1958 (Office for National Statistics 2007). Projections for women born in 1970 suggest that 18% also will be childless by the age of 45 years (Bray 2008). These latest projections for the 1970 cohort represent a downward revision of projected childlessness for this cohort, compared with previous estimates (Bray 2006; Office for National Statistics 2004), and in fact the 1970 cohort represents a small dip in projected childlessness compared to adjacent cohorts (Bray 2008). These projections are also lower than projections from other sources that placed childlessness for the 1970 cohort around 21.5%-22% (Sobotka 2004). In fact, the 1970 cohort appears to be experiencing something of a rapid recuperation of first births in the mid thirties so that rates of parenthood become comparable with the 1958 cohort (Office for National Statistics 2007). This belies some commentators who consider recent estimates to be optimistically low (Dixon and Margo 2006). Although there may not be much of a difference in the proportion of the 1970 and 1958 cohorts who are childless at 45, at age 34 there was a 4 percentage point lead in motherhood for the 1958 cohort. In short, recent cohorts of women have been postponing first births, although the pace of this postponement, and consequences as reflected in levels of childlessness, appears to be slowing, and most women may still eventually become mothers. Despite the widespread interest in childlessness, few studies project ultimate childlessness for relatively young cohorts of women. There is little sound evidence (or questioning) as to whether the retreat from parenthood in the twenties is a delay or a cancellation. Therefore, what has been termed postponement thus far could turn out to be permanent avoidance. Furthermore, it is unclear whether ‘postponement’ implies a full or a partial recuperation of rates of parenthood (Lesthaeghe and Willems 1999). Postponement can be used to describe the state where the proportion of parents in a later cohort is lower than in an earlier one, at the same age (e.g Simpson 2006). However, comparisons are often made of full with partially observed fertility schedules. 4 Childlessness in this instance refers to childlessness at the end of reproductive capacity – typically between ages 45-50 years. 5 Sobotka’s estimates are based upon the LS (Longitudinal Study) survey for England and Wales. Kneale & Joshi: Postponement and childlessness: Evidence from two British cohorts 1938 http://www.demographic-research.org Without additional evidence, there is no basis for assuming that parenthood is either postponed or is never going to happen. However, as we have seen, the ONS projections for later cohorts (including the 1970) do assume some recuperation. This paper looks within social groups to document retreat from parenthood during the twenties, and offers some basis for the notion of recuperation. 3. Childlessness and educational class in the literature Increases in childlessness among post-war cohorts have been associated with changes in the composition of the female population in terms of educational attainment and labour market participation, and an exodus of women from ‘traditional’ confines as housewives (Joshi 2002; Sobotka 2004; Simpson 2006). Given the relationship between educational attainment and labour force attachment, graduate women have found themselves at the forefront of these rises, both in real terms and as a focus of popular interest. Even among graduate women, abandonment of the ‘traditional’ and the embrace of male academic subjects involves additional difficulties in balancing careers and motherhood (Blackwell and Glover 2007; Hoem et al. 2006), which constitute a particular incentive to defer childbearing or avoid it altogether. Low and late fertility among recent cohorts of graduate women reflects these difficulties. One estimate in circulation in the media suggested as many as 40% of British graduate women born in 1970 might expected to enter their forties remaining childless (Wolf 2006a). Other estimates (including our own preliminary ones) placed the level at approximately thirty percent for graduate women born in 1965 (Ratcliffe and Smith 2006) and 1970 (Joshi 2007). A weakness of our own preliminary estimates was the failure to fully account for postponement, something which is addressed in this paper. In fact, once we do examine the notion of postponement, the ‘Monstrous Regiment’ of child-free career women (Wolf 2006b) turns out to be something of a myth. The transition to parenthood is one of several which mark the onset of adulthood; others include completing education, finding a stable partner and a permanent place to live. These events do not necessarily occur in an orderly sequence, but the passage into becoming the mother or father tends to be more prolonged for those who have prolonged their education. Graduate women have particular reasons to postpone. Firstly there is a delay while accumulating educational capital, and then there is reason to delay while accumulating career experience and avoiding the loss of earnings (and/or childcare costs) associated with motherhood (Joshi 2002). They also have more knowledge of how to avoid childbearing before they are ready for it. However the arguments for delay do not apply only to graduate women. Asset accumulation prior to embarking upon parenthood may be of greatest concern to those who are able to acquire Demographic Research: Volume 19, Article 58 the most assets (usually the tertiary qualified), and may also operate entirely differently between groups. Housing is one example, where delaying parenthood is of benefit to those who are to become owner occupiers (especially graduates), but delay may be of little benefit to those who will enter social housing. Given that these benefits are shared with the partner, these factors are replicated among graduate men especially if they have graduate partners. Partnership operates more consistently, and all groups will benefit from delaying entering parenthood until the right partner has been found. In this paper, we analyse three groups for both genders – those in a high, medium and low educational class, where economic reasons to postpone parenthood will tend to be aligned within groups. We use education to proxy for a complement of socioeconomic factors influential in transition to parenthood. We also allow for transition to parenthood to depend on partnership (and partner’s characteristics), as well as family forming expectations, which are tempered by fecundity. These are incorporated into our study and outlined in Figure 1. Figure 1: Model of factors leading to childlessness at end of reproductive years Parenthood status in early-mid adulthood (NCDS age 33) Parenthood status in early-mid adulthood (BCS age 34) Family forming expectations Educational class Partnership factors Partnership factors Family forming expectations Educational class Parenthood status at end of reproductive years (BCS70 age 46) Parenthood status at end of reproductive years (NCDS age 46) http://www.demographic-research.org 1939 Kneale & Joshi: Postponement and childlessness: Evidence from two British cohorts 1940 http://www.demographic-research.org In addition to examining the anticipated level of childlessness among graduate women, this paper will compare the situation for graduate women with graduate men, and also with other education groups. In many cases, the fact that these populations differ in their characteristics over time has not been incorporated into projections. In particular, differences that occur according to their family forming intentions – a very potent measure of ‘postponement’ – have been neglected. Conceptually, family formation intentions may be one of a variety of interrelated factors that determine the transition to parenthood from the early thirties onwards. Few of the estimates of childlessness discussed earlier had used any predictors of transition to parenthood beyond educational class to refine projection estimates. This paper introduces a range of projections of levels of childlessness for the 1970 born BCS70 cohort using an observed partial fertility schedule for the 1970 cohort and a complete fertility schedule from the earlier 1958 born NCDS cohort. The inclusion of male accounts of fertility, as mentioned, is uncommon in the literature. Much of the reason for this lack of analysis of male fertility histories has rested on the inaccuracy associated with male fertility information (Rendall et al. 1999; Greene and Biddlecom 2000). Usually the bias is assumed to be negative, with underreporting of the fathering children, and it is unknown whether underreporting varies by educational group. While this underreporting is a potential consideration of all the information presented in this paper, and specifically any estimate of childlessness we create may be too high; this bias can be assumed to be equal among both our datasets, and our findings will still reflect changes between cohorts as well as being a close approximation of the actual level. Given that parenthood is usually a joint decision between couples, our conceptual chart in Figure 1 reflects factors that apply equally to both genders. After introducing our data sources and methods, the paper: • describes entry to parenthood in both cohorts at observed ages • projects childlessness for BCS70 assuming that transition to parenthood follows the age specific transition observed in NCDS with and without a downward shift based upon its lower rates up to 34 • incorporates reported family building intentions to project rather more recuperation of parenthood. 6 This has usually been due to data constraints. The breadth of the British Birth cohort studies allow for exploration of a number of determinants of transition. Demographic Research: Volume 19, Article 58 http://www.demographic-research.org 1941 4. Data The British Birth Cohort studies comprise of four prospective longitudinal studies following the lives of individuals born during periods in 1946, 1958, 1970 and more recently in 2000. Ferri and colleagues and Bynner and Joshi outline the history of these studies and some of their most prominent findings (Ferri et al. 2003; Bynner and Joshi 2007) Two of them are used in this research – the National Child Development Study (NCDS) that followed individuals born in 1958 and the British Cohort Survey (BCS70) which began in 1970. Both studies prospectively followed individuals and were essentially a census of all born in one week in 1958 and 1970. For the NCDS, further periods of data collection have occurred at ages 7, 11, 16, 23, 33, 41-42 and 46 years. While this last sweep may represent the end of female reproductive capacity, men may still continue to father children. Data from BCS70 cohort members were collected at ages 5, 10, 16, 26, 29-30 and 34 years. This paper examines live births, excluding stillbirths and other fertility outcomes. It excludes those still pregnant or those who have fathered a pregnancy not carried to full term. 4.1 Data considerations: Attrition among the cohort studies As is the case for all longitudinal studies, both cohorts have witnessed a decline in the size of samples under observation, either through death, emigration, non-response or untraceable cohort members (Plewis et al. 2004; Young et al. 2006). Cohort members who attrit from studies are thought to differ systematically from those who continue to participate (Young et al. 2006; Niccoletti and Peracchi 2005), and studies are more likely to lose men and those of low socioeconomic status. This could affect estimates of entry into parenthood. In addition, not only are men more likely to lose contact with the survey, but they are also more likely to underreport births they have previously fathered (Greene and Biddlecom 2000; Rendall et al. 1999). When comparing birth cohort estimates with national statistics estimates (Office for National Statistics 2007); for women at least, the picture is relatively close (Figure 2) showing that entry into motherhood, as estimated from the cohort studies, is representative. Additionally, entry patterns into first parenthood disaggregated by highest qualifications also mirror those 7 Both studies cover the three countries of Great Britain England, Wales and Scotland. Although the 1970 cohort initially included Northern Ireland in collection of data at birth, cohort members from Northern Ireland were not followed-up. Immigrants were added to both studies in childhood waves. Ethnic minorities are not analysed separately because of the low numbers present in both cohorts. 8 Further details of the inclusion criteria and construction of fertility information can be found in Appendix 1. Kneale & Joshi: Postponement and childlessness: Evidence from two British cohorts from other sources (Rendall and Smallwood 2003). Despite these positive indications, attrition means the cohort studies no longer reflect a true census of people born in 1958 and 1970, but are a close approximation. As such all the results presented in this study are likely to contain some margin of error. Figure 2: Comparison of national statistics cohort estimates of rates of entry into first motherhood with NCDS and BCS70 rates 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 Age Pr op or tio n en te rin g m ot he rh oo d 1958 National Statistics