Download presentation
Presentation is loading. Please wait.
Published byЗоја Зорић Modified over 6 years ago
1
The Predictors and Consequences of Early Parenthood
PhD Student: Dylan Kneale Supervisors: Professor Heather Joshi & Dr Jane Elliott Centre for Longitudinal Studies, Institute of Education, University of London
2
Background UK demography witness to decreasing fertility rates.
Age at first birth consistently increasing – in just 15 years, between 1985 and 2000, the age at first birth rose from 25.9 to 29.1 years¹. One of a set of demographic changes characteristic of the ‘Second Demographic Transition’ (SDT) A number of explanatory theories for demographic transition that revolve around either economic or cultural change or both. Particular recent interest in young age at first birth as it is a recognised marker of disadvantage. However, confusion if young age at first birth purely a marker of existing disadvantage (Geronimus 2003; Goodman 2004) or a cause of future disadvantage (SEU 1999). ¹United Nations Economic Commission for Europe 2002 The total fertility rate – the synthesised number of children a woman would bear if she was subject to all the age specific fertility rates of that year – dropped from 2.93 births per woman in 1964 to 1.80 births per woman in 2006. This puts the UK, together with Spain, as having the joint highest mean age at first birth in Europe. However, in somewhat of a transition from this, the UK also has the highest levels of teenage pregnancy in Western Europe. Despite this, the UK is following patterns characterised by the Second demographic transition. Where the first demographic transition describes the move from high mortality and birth rates; the second demographic transition describes the continuation of this with more components of social change. These include higher rates of divorce, cohabitation and individuals acting outside of the collective – greater individualism. In addition, the link between sex and procreation is weaker and there is greater sophistication in contraceptive strategies. In general, demographic theories of change and processes revolve around several key areas but these can be summarised as economic theories that see greater emphasis placed on the cost of children such as Notestein’s classical theory and a second branch that places emphasis on cultural changes such as the diffusion theory that sees demographic behaviour as adaptive to both the economic and social and cultural structures in an individual’s environment. Generally in the second demographic transition, economic factors are thought to operate through the increased labour force participation of women with room for cultural influences through the greater role of the media in influencing fertility decisions.
3
Background Chart 1: Age at first birth and household income (Age 16)
These are purely descriptive charts showing the mean household income and the correlation with age at first birth. These are for men and women combined and have no value above descriptive – have some dot to dot just to illustrate the correlation between general rise in income with age at first birth. The smoothing and dip at age 46 may have something to do with period effects. Anyway, so this is illustrative of the problem that early parenthood and early motherhood poses and a number of investigations have focussed on the issue of whether early parenthood leads to poorer outcomes or are poorer outcomes an artefact of existing disadvantage. Chart 1: Age at first birth and household income (Age 16) Chart 2: Age at first birth and household income (Age 46)
4
Background In line with these theories, the timing of parenthood can be viewed as an economic and cultural adaptation to local circumstances. A wide body of evidence exists on the predictors of early motherhood. Less interest in the timing of fatherhood. Known predictors include educational attainment, socioeconomic measures of family background and personal characteristics among others. However, according to some demographic theories, contextual influence is highly significant in the timing of parenthood, although this area comparatively unexplored. However, my interest, for the time being is on the earlier stages and identifying the predictors of early parenthood. I want to try to address the predictors of early parenthood.
5
When should parenthood be thought of as being early?
Research Questions When should parenthood be thought of as being early? How many people experience early parenthood? What are the factors that predict early parenthood? To what extent do contextual factors predict early parenthood? Do these contextual factors affect the outcomes of parents and children?
6
Data Utilise strengths of the birth cohort studies to examine factors affecting fertility. National Child Development Study (NCDS) – Originally included all live births in G.B from one month in 1958. British Cohort Study (BCS70) – Originally included all live births in U.K from one month in 1970. In both studies, the addition of immigrants has been offset by attrition: Cohort Original Sample No. ever involved Previous (2004) Sample NCDS 17,634 18,558 9,534 BCS70 16,572 17,287 9,665
7
Data Collection NCDS Data Collection BCS70 1974 NCDS (Age 16)
1958 NCDS Birth 2000 NCDS (Age 42) 1975 BCS70 (Age 5) 1986 BCS70 (Age 16) 2000 BCS70 (Age 30) 1980 BCS70 (Age 10) 1996 BCS70 (Age 26) 1970 BCS70 Birth 2004 BCS70 (Age 34) Data Collection BCS70
8
What is early parenthood?
Early parenthood synonymous with teenage parenthood in media Is this meaningful distinction or just convention? Age specific fertility rate for teenagers decreased from 30 births per 1000 women in 1985 to 26 births per 1000 in 2005 (ONS) Is focus reactive to media and political constructions or real differences?
9
What is early parenthood?
Despite the hike in articles published on teenage motherhood, actual age specific birth rates for mothers under 20, stayed fairly constant from the 1980s onwards and actually declined slightly after This does throw into question the justification of examining ‘teenage’ motherhood. Furthermore, studies that claim to examine ‘early motherhood’ may not differ from those that examine ‘teenage motherhood’. In reviewing studies published after 1999, among those studies that explicitly included analysis that examined early motherhood or the effect of early motherhood from a chronological perspective, half used the conventional teenage cut off point.
10
Thinking about defining parenthood…..
Here I will run through some alternative definitions of early motherhood. There include: Socio-legal definition of 16 years and nine moths. I’ve used this because sex is illegal under the age of 16 and this definition is in order to catch those that entered motherhood through ‘illegal conception’. In the NCDS, this applies to 91 girls, or 1.7% of mothers. In comparison this applies to only 17 men, or 0.3% of fathers. Sociological construction. This is to capture the teenage parents. There were 915 teenage mums in NCDS, or 16.9% of mothers. This proportion was almost three times greater than the comparison for teenage fathers, who comprised only 5.8% of all fathers. Socio-biological construction – This would look at the interval between menarche and childbirth (or puberty and childbirth for men) to determine when is an early birth. The evidence for the importance of this is conflicting. Some have found a link between early menarche and early parenthood, for example Udry and Cliquet (1982) and more recently Kiernan and Hobcraft (1997). However, there is conflicting evidence. In particular, early menarche is also associated with poorer fertility in early years and heavy foetal loss (McKibben & Poston). This construction would attempt to just take into account what is biologically possible and differentiates between a woman who became a mother at 19 and was menstruating at 10 and a mother at 19 who only began when she was 16. In the NCDS it is possible to study this and there is a menarche variable that is measured in three ways, as well as for boys another indicator – voice breaking. At age 11, this was a question that was answered by the parent, usually the mother, that was grouped into No, Before 5 years, 5 to 8 years, Aged 9 years, 10 to 10.5 years, 10.5 to 11 years and 11 years & over. It was also collected in the same way in Wave 3 at age 16, where the groups had changed to No, not yet, Before 11th year, Aged 11 years, Aged 12 years, Aged 13 years, Aged 14 years, Aged 15 or more and the same don’t know categories. This variable was also collected through a medical questionnaire at the same age which collected single year information up from age 5 to 16. This variable was constructed through giving precedence to the information collected at the earliest age in order to attempt to minimize recall bias, with the Medical Questionnaire taking precedence after because of higher data quality and because it may reduce respondent bias (because the cohort member was instrumental in providing the information). Finally, where these measures were missing, the parental information given at age 16 was used. The mid points to all time periods were also used, and girls who had not gone through puberty at age 16 were treated as missing as this question was not repeated at age 23. This meant that the average age at puberty for 6,158 women was 13 years 2½ months. Behavioural construction – No information collected on this in the NCDS. Obstetric/medical construction – Evidence is also conflicting on this topic and I have yet to find papers that quantify a definite age. Some papers; e.g. Lao and Ho (1997) find no negative impact of early childbirth, and therefore do not give a definition of early (or too early) as such. In other papers, there is strong evidence of negative consequences and in Jolly et al’s paper, early, or too early, is defined as 17 and under. Mirowsky (2005) finds that the correlation of motherhood with health problems and mortality hazard switches from detrimental to beneficial in comparison with non-mothers from about age 22. What he basically is saying is that having a child before 22 is detrimental to health of women and a child after 22 beneficial compared to being childless. In the NCDS, according to this definition, 1660 women made decisions detrimental to their health – or 30.6% of mothers at 42. Mirowsky’s paper (2005), evaluates that 34 is the optimum age for childbirth, where economic gains are at an advantage that does not seriously hamper fecundity. After 34 fecundity takes a serious downturn. Delaying as close to this age as possible maximises economic benefit – so does this mean that a birth before 34 is actually too early? Socio-demographic construction calculating the interval between first marriage and childbirth. As this may involve negative survival times if calculated as date of first child’s birth – date of marriage does this mean two different groups with two different termination events – date of marriage and date of cohabitation. What this type of analysis would look at and quantify is whether experiencing a life course event speeds up the process of experiencing another. Socio-demographic construction calculating the interval between first cohabitation and childbirth. This is similar to the criteria above, although would involve a larger sample size as more women had been cohabitating. Again this poses the same question of whether these should be treated as two differing groups. A statistical construction could see the full spectrum of timing broken down by percentile, and births below say the 5% mark being classed as early births. However, while this would be an adequate measure for a single cohort such as NCDS, it may not be feasible for a wider application. Culturally specific or neighbourhood definitions basically examine whether the same definition of early can be used for a Bangladeshi mother as for a White mother for example. Using Labour Force Survey data, they found substantial differences in the mean age at first motherhood, for example the mean age for White mothers being 25.4 and for Bangladeshi 21.8 years. This does pose the question of whether the same definitions can be used for Bangladeshi and White mothers. In a similar vein, neighbourhood differences may mean that neighbourhood specific definitions could be used. Alternatively these differences could be minimal, and neighbourhood is merely a factor to examine and not a defining factor of the timing of early motherhood. These constructions may be important, but is it feasible to vary definitions according to such diverse factors.
12
How should early parenthood be defined?
Using absolute cut off points, survival analysis and piecewise regression to derive definitions of early parenthood: The literature supports the idea that the effects of the timing of parenthood are felt up to a certain point, changing after this. Piecewise regression methods are used to test this proposition. The Kaplan-Meier Survival function gives the probability that a subject will have a survival time greater or equal to time t conditional on survival up to time t. This function takes censoring into account through continually altering the number of subjects at risk by taking into account previous patterns of survival and censoring. The formula for the Kaplan-Meier survival function for the probability of becoming a first time parent at time t () is given below; and is a product of the number of cohort members at risk at time t (ni) and the number of cohort members becoming first time parents (di). Piecewise regression is used when we believe that a relationship is linear up to a certain point, after which it changes and the slope of the model changes. The point at which the relationship changes are known as the knots and the separate slopes known as the splines. The use of splines is theoretically consistent with the literature on the outcomes, and to a lesser extent, predictors of early parenthood. This literature essentially states that the effects of the age at first parenthood are consistently negative up to a certain point, usually 19 years, after which age becomes insignificant. This method is used in this case to identify clusters based on parenthood and not to identify causal effects. This is consistent with strategies used in Robson and Berthoud. The model shows that the Expected value of y, say income in this case; is the product of the age at first birth (X1) + the second term represents how much the slope increases or decreases at the spline point.
13
How should early parenthood be defined – example output…..
Household Income Age 46 Coef. (£) P>|t| [95% Conf. Interval] age1 1201.2 0.000 884.2 1518.2 age2 0.001 -632.0 _cons 7515.0 In the above output, a spline has been identified at age 32. Up to this age, delaying fatherhood corresponds to an increase in household income of value age1 (However, other factors do mitigate the relationship) Age2 corresponds to the value of the difference between slopes up to 32 years and ages afterwards – significantly different This is some example output from the definitions work – talk through it – blah blah blah
14
How could early fatherhood be defined?
27 years spline identified from NCDS cm’s mother’s age at first birth 21 years 2 months, first 12.5% of NCDS cohort are fathers 24 years 11 months, first 25% of NCDS cohort are fathers 32 years spline identified from NCDS income 14 years 5 months, first NCDS father Teenage Birth I support a proportional definition of early parenthood Talk through this slide No splines were found in BCS70 – there was no big point of differentation 15 years 2 months, first BCS70 father 23 years 11 months, first 12.5% of BCS70 cohort are fathers 27 years, first 25% of BCS70 cohort are fathers 28 years spline identified from NCDS test scores Conclusion: Teenage definition is a poor definition of early fatherhood. Given the data, a proportional definition is satisfactory
15
How could early motherhood be defined?
22 years 2 months, first 25% of NCDS cohort are mothers 27 years spline identified from NCDS cm’s mother’s age at first birth 19 years 11 months, first 12.5% of NCDS cohort are mothers 33 years spline identified from NCDS income 13 years 2 months, first NCDS mother Teenage Birth I support a proportional definition of early parenthood Talk through this slide No splines were found in BCS70 – there was no big point of differentation 23 years 11 months, first 25% of BCS70 cohort are mothers 15 years 2 months, first BCS70 mother 20 years 8 months, first 12.5% of BCS70 cohort are mothers 28 years spline identified from NCDS test scores Conclusion: Teenage definition is a poor definition of early motherhood. Given the data, a proportional definition is satisfactory
16
However, both studies have suffered from attrition – how representative are the cohort fertility patterns with other sources? Can’t actually test this because don’t have the background national statistics data but it does appear that the data is highly representative, although the NCDS is slightly more fertility than would be expected and BCS70 the reverse.
17
How many early parents feature in NCDS and BCS70
How many early parents feature in NCDS and BCS70? – the definitions to be used in the remainder of the research…. Gender Definition NCDS BCS70 No. Age ♂ Teenage 278 <20 188 First 12.5%* 892 21y 2m 735 23y 11m First 25%* 1684 24y 11m 1466 27y ♀ Teenage** 903* 615 885 19y 11m 761 20y 8m 1763 22y 2m 1519 Parents at 46 Childless at 46* (%) Parents at 34 Childless at 34* (%) Timing 4646 21% 2997 44% 5304 16% 4060 30% *Based upon KM estimates; **because this definition does not vary significantly from 12.5%, it will not be used extensively in further research
18
What do we know about the predictors of early parenthood?
Know from the literature that socioeconomic, educational, demographic and personal variables are significant predictors of the timing of parenthood in addition to contextual effect. These variables will form controls in models that test the residual contextual effects Neighbourhood data available as census variables with diverse indicators such as % married women working and % with indoor bathroom at 2 levels of geography Know from the literature that socioeconomic, educational, demographic and personal variables are significant predictors of the timing of parenthood. These variables will form controls in models that test the residual contextual effects The actual variables I use for this stage are father’s social class at age 16, standardized test scores at age 11, household tenure, social class, parental structure (divorced, lone parent, adopted etc) and Neighbourhood data available as census variables with diverse indicators such as % married women working and % with indoor bathroom at 2 levels of geography. While most analysis that uses ecological level information hopes to capture a neighbourhood effect, in actual fact the level of information used often has little connection to traditional notions of neighbourhood and often more to do with cartographic constructions. Information attached to cohort members’ records at age 16 years is no exception and is based upon both the Enumeration District level as well as the local authority level. Enumeration Districts possibly provide the closest available approximation to traditional notions of neighbourhood, defining in 1971, on average, 430 people and reaching up to 3,265 people. Local authority level data provides higher level information, with the size ranging from 6,423 in Kinross to 7,452,343 in London and averaging 567,343 people (494,411 excluding London) and may give a wider insight into a regional context (Office for National Statistics 1971) Contextual Factors Individual Factors Timing of Motherhood
19
Neighbourhood factors as predictors of early parenthood (I)
Due to the nature of the birth cohort studies, no complex modelling needed to capture neighbourhood effects in BCS70 and NCDS (such as MLM). Models below could be used with inclusion of individual level measure to avoid ecological fallacy However, neighbourhood data is only available in the NCDS at age 16 and is not available at all at BCS70. HOWEVER…….
20
Neighbourhood factors as predictors of early parenthood (II)
Despite limitations, after controlling for known confounders and individual measurements, some conclusions can be drawn: Neighbourhood found to exert a small but significant effect on fertility Some more cultural measures of neighbourhood such as % of immigrants; % of population aged 0-4 years and % of married women working highly insignificant. In general, neighbourhood effects appear to exert a stronger influence on the timing of fatherhood than motherhood, although significant results found for the timing of motherhood Social class measurements on the local authority level more likely to influence the timing of fatherhood Tenure measurements on the enumeration district level more likely to influence the timing of motherhood Some indication of gender-occupation interactions found with certain variables (e.g. % of workforce in mining and manufacturing) However, neighbourhood data is only available in the NCDS at age 16 and is not available at all at BCS70.
21
Neighbourhood factors as predictors of early parenthood (III)
Covariate Early Fatherhood vs Later Fatherhood Very Early Fatherhood vs Later Fatherhood Odds Ratio Professional and Managerial (% in Local Authority) 0.96** 0.95** Social Class (Baseline Social Class V) Social Class I 0.41* 0.22* Social Class II 0.60 0.49* Social Class III 0.77 0.57 Social Class IV 0.92 0.80 Tenure (Baseline Owner Occupation) Council Housing 1.91** 2.41** Privately Rented 0.85 0.62 Tied and other forms 2.28** 2.15* Educational Achievement Age 11 0.74** 0.80** *p<0.05; **p<0.01; BS = Borderline significant (p>0.05 P<0.07) r² =
22
Early Motherhood vs Later Motherhood
Covariate Early Motherhood vs Later Motherhood Teenage Motherhood vs Later Motherhood Very Early Fatherhood vs Later Fatherhood Odds Ratio Unskilled (% at Local Authority) 1.067* 1.087** 1.087* Social Class (Baseline Social Class V) Social Class I 0.180* - 0.216** Social Class II 0.460 0.470* 0.476** Social Class III 0.493 0.682 0.565 Social Class IV 0.188 0.788 0.644 Tenure (Baseline Owner Occupation) Council Housing 1.63** 1.387 1.970** Privately Rented 0.875 1.404 0.600 Tied and other forms 0.952 2.977 2.103* Educational Achievement Age 11 0.770** 0.724 0.782** Parental Age at First Birth 0.968** X Parental Structure (Baseline No Natural Parents) 1 Natural Parent 4.134* 2 Natural Parents 2.264 Parental Interest in Education (No Parent with Healthy Interest) Both Parents Healthy Interest 0.427** 0.569** 1 Parent with Healthy Interest 0.866 1.062 *p<0.05; **p<0.01; BS = Borderline significant (p>0.05 P<0.07); Standard errors in brackets; X = Covariate insignificant and not included in parsimonious model
23
Other Contextual factors as predictors of early parenthood
Because of the limitations in exploring neighbourhood (census) variables further, a broader definition of context adopted will be adopted. In particular this will explore the concept of resilience by looking at the environment provided by parents not directly associated with socioeconomic factors. Preliminary work into parental interest in education reveals this to be significant predictor of fertility patterns net of socioeconomic factors. A number of measures of family life remain unexplored.
24
Future Directions Immediate:
Conclude neighbourhood measurements section and change modelling strategy for the timing of parenthood (survival) variables: theoretically and empirically problematic. Continue exploratory analysis of parental environment variables and explore their significance as predictors. Mid-term: Decide on final modelling strategies for predictors Conclude the effects of contextual factors on the timing of parenthood Long-term: Begin to examine the effects of these contextual factors and interactions with age at first parenthood on the outcomes of mothers and children
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.