Download Impact of Socioeconomic Disadvantage on Academic Achievement & Occupation: Dual Cohort Stu and more Study notes Psychology in PDF only on Docsity!
LIFE-SPAN
DEVELOPMENTAL THEORY
39
L
ife-span developmental theory concerns the study of individual development, or ontogenesis, from conception to death. A key assumption of this theory is that develop- ment does not cease when adulthood is reached (Baltes, Lindenberger, & Staudinger, 1998, p. 1029) 1. Life-span researchers and theorists assume that each major period of life has its own developmental challenges and accomplishments, and that adaptive processes are at work within all periods of the life span. This theoretical approach is clearly focused on individual devel- opment rather than on family development. It is concerned with comparing an individual’s devel- opment with that of others and with the individ- ual’s own status at various points in time. We include it in this volume on family theories for several reasons. First, since the purview of the theory is “womb to tomb,” it of necessity touches upon all of the family-related issues that characterize family life course theory, such as the birth of a child, the development of Alzheimer’s disease, the macro-level political and economic slings and arrows that impact both individuals and families—the latter impacted either directly or indirectly through one or more of its members.
Of course, as noted, regardless of its study of development across the human life span, the focus is still on individuals rather than groups, and this makes the theory decidedly distinct from the family life course perspective that focuses on the family as the unit of analysis. Second, we believe that this approach has much to teach others within subdisciplines of the family and human development sciences. For example, this theoretical approach is exception- ally rigorous in the way it links theory with methodology across the life span (you’ll get a taste of this from the two readings we have selected for this chapter) as well as in the breadth of substantive issues that are studied onto- genetically. In addition, it has progressed from a largely descriptive focus on developmental stages (predominant 40–50 years ago) to a more recent focus (beginning in the 1970s) on trying to determine the mechanisms that determine developmental change or consistency. (Family life course theory is undergoing a similar transi- tion as it has evolved from an emphasis on the family life cycle to a more comprehensive com- mitment to examining the causes of family development and change [cf. Chapter 3; Elder, 1996; White & Klein, 2002].) Overall, for the
reasons mentioned as well as because it passes with flying colors most tests (White & Klein, 2002, p. 232, Table 9.1) of how to evaluate a theory’s usefulness, life-span developmental theory can serve as an important model for new students as well as for scholars. What does it mean to say that the focus of life-span developmental theory and research is the individual? Given the identification of a sub- stantive focus (e.g., personality development or cognitive abilities), this approach aims to study one or more of the following things:
- Normative developmental change (For example, what is the typical course of personal- ity development, or of cognitive capabilities? What are the mechanisms for this typical course of development or developmental change?)
- Interindividual differences in develop- mental change (For example, are there differences between people over time, within age periods or across age periods, in their outgoingness or in their visual acuity? What are the mechanisms that produce such differences between people?)
- Intraindividual change and consistency in development (For example, what is the course of development for self-esteem or for the tendency to take personal risks? Are these things consistent in individuals over time or are they characterized by “plasticity”? [Developmental plasticity is a general term used in the study of ontogeny to indicate how malleable or changeable something is.] What are the mechanisms producing consis- tency or change?)
It is important to note that these three emphases apply to individual periods of the life span (e.g., infancy, adolescence, or old age) as well as to the life span as a whole. In asking the above questions, one also begins to get a flavor of the kinds of methodological implications of life-span developmental theory. How does one investigate developmental plasticity or con- sistency? Does one compare, say, a group of 9-year-olds with a group of 23-year-olds to see if interindividual changes in cognitive functioning
occur? Or, does one follow longitudinally the development of cognitive functioning over a pro- tracted period of time? Further, if one substitutes the word family or families for people or individ- uals in the above questions, it becomes clear that life-span developmental theory and family life course theory are both fundamentally time-related theoretical approaches. (Do note, however, that time is really just a shorthand descriptor for all that happens during it. Like age, it doesn’t explain anything.) This develop- mental perspective—what happens and why, over time—is an integral part of both conceptu- alizations, and it is one thing that distinguishes them from the other theoretical perspectives in this book. Although some of the other theories (e.g., social learning theory or social exchange theory) may imply or make assumptions about change over time, the concept of developmental change is not an integral component of these other perspectives.
THE READINGS
Schoon et al. (2002) use the life-span develop- mental perspective in combination with concepts from ecological theory (Bronfenbrenner & Ceci, 1994; Chapter 10) and developmental contextu- alism (Ford & Lerner, 1992) to examine the rela- tionships between early and continuing social risk and academic achievement in childhood/ adolescence as well as between earlier risk and adult attainments. Developmental contextualism views human development as the dynamic inter- action between a changing individual and a changing context. This reading uses fairly com- plex data analysis procedures (path analysis and multiple regression) to explicate a complex development trajectory over a significant portion of the life span. (You can safely ignore the statis- tical details since the article’s text is very clear in stating what the analyses mean and how they can be interpreted.) Another key feature of the Schoon et al. reading is what we mentioned earlier in this chapter: the link between theoreti- cal perspective and methodological choices. In
40 • READINGS IN FAMILY THEORY
THE INFLUENCE OF CONTEXT,
TIMING , AND DURATION OF RISK
EXPERIENCES FOR THE PASSAGE FROM
CHILDHOOD TO MIDADULTHOOD
INGRID SCHOON , JOHN BYNNER, HEATHER JOSHI , SAMANTHA PARSONS , RICHARD D. W IGGINS , AND AMANDA SACKER
42
Abstract
This study investigated the long-term effects of social disadvantage on academic achieve- ment and on subsequent attainments in adult- hood. The study drew on data collected for more than 30,000 individuals born 12 years apart, following their development from birth to adulthood. The pathways that link social dis- advantage to individual development across the life course were analyzed in a developmental– contextual systems model. The results showed that the influence of risk factors associated with socioeconomic disadvantage depended on the developmental stage of the individual, the experience of long-term or continuous disad- vantage, and the overall sociohistorical con- text. Early risk had a moderate influence on the formation of individual competences. The greatest risk was associated with persisting and accumulating experiences of socioeco- nomic disadvantage throughout childhood and adolescence. Material conditions improved for the later-born cohort, yet pervasive social inequalities existed that affected outcomes dur- ing childhood and were consequently reflected in adult attainment.
Introduction
Children raised in socioeconomically dis- advantaged families are at risk for a variety of adjustment problems, including increased risk for poor academic achievement (Bolger, Patterson, & Thompson, 1995; Campbell & Ramey, 1994; Duncan, Brooks-Gunn, & Klebanov, 1994; Felner et al., 1995; Pungello, Kupersmidt, Burchinal, & Patterson, 1996; Ramey & Ramey, 1990; Walker, Greenwood, Hart, & Carta, 1994) and adjustment problems in later life, as reflected in occupational attainment or social position (Blau & Duncan, 1967; Bynner, Joshi, & Tsatsas, 2000; Caspi, Wright, Moffitt, & Silva, 1998; Rutter & Madge, 1976; Schoon & Parsons, 2002b; Sewell, Hailer, & Ohlendorf, 1970). Socioeconomic background is one of the main predictors of cognitive devel- opment, which provides the underpinnings of academic achievement on which much success in later life depends. The experience of socioeco- nomic disadvantage may severely strain adapta- tional abilities of children, and is thus a potential risk factor for development. The consequences of growing up in a disadvantaged family environ- ment can continue into adulthood or even into the next generation (Birch & Gussow, 1970;
Garmezy, 1991). Yet, most research on the influence of socioeconomic disadvantage on developmental outcomes has been cross-sectional in nature, and has assessed the impact of episodic rather than persistent economic difficulties. The aim of this study was to examine the long-term effects of socioeconomic disadvantage on aca- demic achievement and consequent adult attain- ments. Taking a longitudinal perspective stretching from birth to adulthood, this study examined the impact of early and persistent social disadvantage (which was termed “social risk”) on academic achievement during childhood and adolescence and on adult outcomes in a developmental- contextual framework. Fundamental to the idea of risk is the pre- dictability of life changes from earlier circum- stances. As expressed through the concept of a “risk trajectory,” one risk factor reinforces another, leading to increasingly restricted out- comes in later life (Rutter, 1990). In this study the question of the relation between early life experiences and consequent adjustment patterns was recast in terms of a testable model of con- tinuities in social disadvantage and individual adjustment and their interactions over time. By analyzing data from two cohorts of children born in 1958 and 1970, the investigation also took into account the changing sociohistorical context.
Development and Context The processes by which the socioeconomic background influences individual development are not yet fully understood. It has been argued that socioeconomic status (SES) at the time of the child’s birth is an indicator of the social con- text, but may also reflect parental genetic char- acteristics, which are assumed to have some role in determining the level of academic functioning of their children (Plomin & Bergeman, 1991; Plomin & McClearn, 1993; Scarr, 1992). Propo- nents of behavioral genetics see the course of human development as a function of genetically controlled maturational sequences (Scarr, 1992). However, findings show that in no case is the
genetic determination so strong that there is no room for environmental effects (Plomin & Daniels, 1987; Scarr, 1992). There is now increasing skepticism about the usefulness of approaches formulated within behavioral genet- ics, on scientific as well as social and ethical grounds (Baumrind, 1993; Hoffman, 1994; Jackson, 1993; Lerner & von Eye, 1992). Human development takes place in a social context, and is therefore influenced by a person’s interactions within that context. Genetic factors are only one of a much larger series of possible explanations for human behavior. More recent developmental approaches have conceptualized genes and other biological variables as contributors to reciprocal, dynamic processes that can only be fully under- stood in relation to sociocultural environmental contexts (Bronfenbrenner & Ceci, 1994; Gottlieb, Wahlsten, & Lickliter, 1998; Horowitz, 2000). Human beings differ in their capacity for realiz- ing individual talents, and it is important to understand under what circumstances individual potentials find expression. This study was not designed to examine the heritability of certain traits, but rather to investi- gate the long-term effects of social risk on acad- emic attainment and consequent adult outcomes in a changing sociohistorical context. Especially useful for this type of analysis are approaches developed by proponents of an ecological per- spective of the life course that conceptualizes human development as the dynamic interaction between a changing individual and a changing context (Baltes, 1987; Bronfenbrenner, 1979; Elder, 1985; Featherman & Lerner, 1985; Lerner, 1984, 1996; Sameroff, 1983). For example, in their bioecological theory of nature-nurture effects, Bronfenbrenner and Ceci (1994) argue that the long-term interactions between children and their environments are a necessary condition for the expression of any trait. They differentiate between the proximal environment, which is directly experienced by the individual (e.g., the family environment), and more distal cultural and social value systems that have an indirect effect on the individual, and are often mediated by the more proximal context. Proximal
Life-Span Developmental Theory (^) • 43
(Clarke & Clarke, 1981). Early adversity may be overcome by improved circumstances, but may, nevertheless, leave the individual potentially more vulnerable to any disadvantage experi- enced at a later stage (Cicchetti & Tucker, 1994). Recent studies have shown that persistent socio- economic disadvantage has stronger effects than intermittent adversity on individual outcomes (Ackerman, Schoff, Levinson, Youngstrom, & Izard, 1999; Bolger et al., 1995; Duncan et al., 1994; Pungello et al., 1996) and it has been argued that chronically stressful environments hinder the development of successful adaptation (Hammen, 1992). On the other hand, there is evi- dence that adversity during early childhood, as opposed to during later developmental periods, has a crucial impact on later adjustment, espe- cially for academic attainment (Axinn, Duncan, & Thornton, 1997; Duncan, Yeung, Brooks- Gunn, & Smith, 1998; Haveman & Wolfe, 1994). There is, however, also evidence that current contextual adversity determines current adjust- ment (Campbell, Pierce, Moore, Marakovitz, & Newby, 1996; Feiring & Lewis, 1996; Tizard, 1976). Differences in findings can be explained by methodological variations in the studies that involve different developmental periods, differ- ent indicators of adversity and adjustment, and different analytical strategies. The relative contribution of early, concurrent, or persistent effects can only be elucidated by drawing on longitudinal studies, which provide detailed information of individuals followed over time.
Assessing Socioeconomic Risk Socioeconomic disadvantage is associated with a variety of cofactors, such as poor living conditions, overcrowding, or lack of material resources that pose risks for adaptive develop- ment (Ackerman et al., 1999; Conger et al., 1993; Duncan & Brooks-Gunn, 1997; Fergusson, Horwood, & Lawton, 1990; Fitzgerald, Lester, & Zuckerman, 1995; Huston, McLoyd, & Coll, 1994). Individual risk factors do not exert their effect in isolation, but rather in interaction with other influences. It has been suggested that it is the
number of these factors and their combined effect that shape development (Sameroff, Seifer, Baldwin, & Baldwin, 1993; Rutter, 1979). The relation between any single risk factor and sub- sequent outcomes tends to be weak, and usually many variables are involved in determining an outcome (Ackerman et al., 1999; Rutter, 1990; Sameroff et al., 1993; Szatmari, Shannon, & Offord, 1994). Serious risk emanates from the accumulation of risk effects (Robins & Rutter, 1990). In comparison with single-risk models, multiple-risk models have been shown to be good predictors of individual outcomes (Ackerman et al., 1999; Caprara & Rutter, 1995; Fergusson, Horwood, & Lynskey, 1994; Sameroff et al., 1993). Most multiple-risk studies use a single index, summing the number of risk factors present. Summing the number of risk factors in a single index, however, gives equal weight to all risk factors and does not take into consideration the relative contribution or overlap in risk factors (Greenberg, Lengua, Coie, & Pinderhughes, 1999; Szatmari, Shannon, & Offord, 1994). There is now an increasing awareness that the processes that link socioeconomic disadvantage to individual development operate at varying levels of specificity, and that there is a need to distinguish between economic disadvantage per se and other associated aspects of environmental adversity (Ackerman et al., 1999; Duncan et al., 1994; McLoyd, 1990; Szatmari et al., 1994). Furthermore, Ackerman et al. (1999) argue that a single multiple-risk index aggregates a set of variables that may relate differently to child functioning, or that may function differently for advantaged and disadvantaged families, and that it does not distinguish between persistent and transitory experiences. Thus, they recommend the use of discrete groupings of indicators, narrowing the focus of the variables involved, or isolating specific factors that pose risks for indi- vidual adjustment. The usefulness of such an approach has been demonstrated in a number of studies (Ackerman et al., 1999; Deater-Deckard, Dodge, Bates, & Pettit, 1998; Szatmari et al.,
- and was adopted in the present study. Instead of aggregating various cofactors of
Life-Span Developmental Theory (^) • 45
socioeconomic disadvantage, only indicators of the socioeconomic family background were included to assess the unique effect of socioeco- nomic disadvantage on individual development. The study drew on indicators of socioeconomic family status as well as indicators of living con- ditions and material resources available to the family. This approach thus went beyond studies that used social status or income as sole indica- tors of socioeconomic risk and more accurately reflected the everyday experiences within the proximal family context.
Effects of the Wider Sociohistorical Context Another issue to be addressed concerns the impact of the wider sociohistorical context in shaping individual development. For example, Elder (1999) demonstrated the crucial impact of the Great American Depression and the outbreak of World War II on the developmental pathways of individuals born between 1920 and 1921 in Oakland, CA. The present study compared the development of two British birth cohorts born 12 years apart, thereby taking into account contextual effects that might help to explain differences in response to disadvantage. The study drew on data collected for two birth cohorts born in 1958 and 1970, respectively. Changes in social, economic, and education policies between 1960 and 1980 resulted in the cohort members growing up in very different environments. Between 1979 and 1986, and again between 1989 and 1993, the sharpest rise in unemployment since World War II took place in the United Kingdom. The mid- 1980s saw the virtual disappearance of the youth labor market in Britain (Banks et al., 1992; Bynner, Elias, McKnight, & Pan, 1999). Many have argued that children born in the 1970s expe- rienced a major shift in life expectations across the generations. This generation “X” grew up at a time when the prospects of achieving employ- ment directly after leaving school and main- taining a continuing career were increasingly in question, especially for those young people with- out qualifications (Bynner, Ferri, & Shepherd, 1997; Schoon, McCulloch, Joshi, Wiggins, &
Bynner, 2001; Schoon & Parsons, 2002a). In response to the changing nature of labor markets and employment opportunities, young people are under increasing pressure to continue full- time education beyond the age of 16 years, and to acquire formal qualifications. Poor academic achievement, which presented no significant barrier to employment in the past, now predicts real difficulties in finding employment, and ulti- mately exclusion from the labor market (Bynner, Joshi, & Tsatsas, 2000).
A Developmental-Contextual Model of Cumulative Risk Effects The aim of this study was to examine the extent of continuity of socioeconomic disadvan- tage from birth to midadulthood and the mainte- nance of academic adjustment in the face of that risk. To better understand the long-term effects of social risk, the extent to which risk effects per- sist and how they interact with individual adjust- ment was considered. Using Structural Equation Modeling (SEM; Bollen, 1989), the long-term influence of socioeconomic disadvantage on individual development was assessed, taking into consideration the influence of context, timing, and duration of risk experiences. The pathways through which the experience of socioeconomic risk influences the development and mainte- nance of individual adjustment (i.e., academic attainment) during childhood and adolescence as well as the pathways that link childhood conditions to adult outcomes were investigated. Figure 3.1 gives a diagrammatic representation of the developmental-contextual model for assessing the long-term impact of socioeco- nomic disadvantage. The model is an explicit developmental model that assesses the timing and the duration of the interactions between individual and context. The variables shown are all latent or unobserved variables. The model specifies that conditions at birth (parental social class) influence the consequent experience of socioeconomic risk (i.e., low social class of parents and material disadvantage in the family home), as well as the level of individual
46 • READINGS IN FAMILY THEORY
reports of early life events were not influenced by knowledge of the subsequent personal history; the cohort included appropriate controls; and the conjoint impact of different factors experienced at different time points could be analyzed in a multi- variate, multicausal model.
Method
This study used data collected for the National Child Development Study (NCDS) and the British Cohort Study (BCS70), two of Britain’s richest research resources for the study of human development. The participants of the NCDS included all persons born in Great Britain between March 3 and March 9, 1958. In five follow-up studies, data were collected on the physical, psychosocial, and educational develop- ment of the cohort at ages 7, 11, 16, 23, and 33 years (Shepherd, 1995), and, most recently, at age 42 (Bynner, Ferri, Shepherd, & Smith, 2000). At each sweep between 1958 and 1974, wide-rang- ing information was collected from parents, teachers, school medical officers, and, at later stages, from the cohort members themselves via personal interviews. Satisfactory response rates have been reported for each sweep, and compari- son of data has shown that the achieved samples did not markedly differ from the target samples, or from other survey samples of the British population (Shepherd, 1993, 1995). The BCS7 0 has followed children born in the week April 5–11, 1970, from birth through to adulthood. Data sweeps took place when the cohort members were ages 5, 10, 16, and 26 years (Ekinsmyth, Bynner, Montgomery, & Shepherd, 1992), and most recently at age 30 (Bynner, Ferri, Shepherd, & Smith, 2000). In the birth survey, information was collected by means of a question- naire that was completed by the midwife present at birth, and supplementary information was obtained from clinical records. In 1975 and 1980, parents of the cohort members were interviewed by Health Visitors, the cohort members themselves undertook ability tests, and the school health service gathered medical information on each child. This was supplemented at ages 10 and 16 by
information gathered from head and class teachers who completed questionnaires. A low response rate at age 16 occurred because of a teacher strike, which coincided with the fieldwork. All school children were afflicted in the same way, and the demographic characteristics of the sample at age 16 remained representative of the target population (Shepherd, 1997). The follow-up study at age 26 was conducted via a mail survey. At age 30, data were collected by interview with the cohort members (Bynner, Ferri, et al., 2000). An analysis of response bias showed that the achieved samples did not differ from the target samples across a number of key variables (social class, parental edu- cation, and gender), despite a slight underrepre- sentation of the most disadvantaged groups (Butler, Despotidou, & Shepherd, 1997; Davie, Butler, & Goldstein, 1972; Shepherd, 1997).
Sample The sample consisted of all individuals for whom complete data were collected at birth. The following analyses were based on a total sample of 16,994 cohort members for the NCDS and 14,229 cohort members for the BCS70. In both cohorts, there were 52% men and 48% women. Data collected between birth and age 16 were linked with data collected at age 30 (BCS70) and age 33 (NCDS) when the cohort members reached midadulthood. Potential bias due to missing variable information in both cohorts was addressed in the section on estimating the model.
Measurement of Socioeconomic Risk Risk factors associated with socioeconomic disadvantage were indicated by parental social class and material conditions in the family household. The indicator variables were mea- sured at ages 7, 11, and 16 for the NCDS, and ages 5, 10, and 16 for the BCS70.
Parental Social Class
In both the NCDS and the BCS70, social class was measured by the Registrar General’s mea- sure of social class (RGSC). The RGSC is
48 • READINGS IN FAMILY THEORY
defined according to job status and the associated education, prestige (OPCS, 1980), or lifestyle (Marsh, 1986), and is assessed by the current- or last-held job. The RGSC is coded on a 6-point scale: I, professional; II, managerial and technical; IIINM, skilled nonmanual; IIIM, skilled manual; IV, partly skilled; and V, unskilled (Leete & Fox, 1977). The occupational categories used in the U.S. census and other European countries are similarly based on the skills and status of different occupations (Krieger & Williams, 1997). Class I represents the highest level of prestige or skill and Class V represents the lowest. In cases in which the father was absent, the social class (RGSC) of the mother was used in the BCS70. The same applied to the NCDS at ages 7, 11, and 16; how- ever in cases in which there was no father at birth, the mother’s father’s social class was used.
Material Conditions Material conditions in the family environment were assessed on the basis of a summative index, in which the presence or absence of four indica- tor variables (listed below) is summed. The scale gives an overall score of material disadvantage that ranges between 0 and 4.
Overcrowding. This is a dichotomous variable based on the ratio of people living in the house- hold to the number of rooms in the household. One or more persons per room was coded “1,” and less than one person per room was coded “0.”
Household amenities. This is a dichotomous scale based on the cohort member’s family having sole use of a bathroom, toilet, and hot water. The same three questions were asked in both studies. Sole access to all of these amenities was coded “0,” and shared use or no access to any of these amenities was coded “1.”
Housing tenure. The tenure of the home was defined as owner-occupier (0) or other (1).
Receipt of state benefits. Receipt of state benefits is an indicator of financial hardship within the
family environment (Fogelman, 1983). The assessed benefits include payment of unemploy- ment benefit, income support, and housing bene- fit, but exclude payment of pension or child benefit. Parents were coded as either not in receipt of benefits (0) or in receipt of benefits in the last 12 months (1).
Individual Adjustment Individual adjustment was measured by the child’s academic attainment at each measure- ment point. To reflect the changing competencies of the growing child, academic attainment was assessed differently during early childhood (age 5 or 7), middle childhood (age 10 or 11), and adolescence (age 16).
Academic Attainment at Age 5 (BCS70) and 7 (NCDS) The Human Figure Drawing Test used in the present study was a modified version of the Draw- a-Man test originally devised by Goodenough (1926) and developed further by Harris (1963). The Harris-Goodenough Test has good reliability (.94; Osborn, Butler, & Morris, 1984). It has been evaluated as a measure of intelligence and signif- icant correlations, r s averaging between .4 and .5, with conventional IQ tests (Binet, Wechsler) have been reported (Scott, 1981). The scoring of the drawings produced by the children is based on 30 developmental items suggested by Koppitz (1968) and uses the Harris (1963) point system of scoring. One point is scored for each item represented in the drawing, giving a maxi- mum possible score of 30. In both cohorts the children had to draw two figures. In the BCS70, however, only one figure was coded. Thus, the maximum score in the NCDS was 60, and in the BCS70 was 30. The achieved scores ranged from 0 to 53 in the NCDS, and 0 to 23 in the BCS70. The Copy-a-Design test (Davie et al., 1972) assesses the cohort member’s perceptual-motor ability. The ability to copy designs or geometric shapes is included as one element of assessment in many standard intelligence tests. The test used
Life-Span Developmental Theory (^) • 49
with a range from 1 to 5 (or A to E). Generally, GCE grades of D and E are classified as failures. There is an accepted equivalence between the two examination systems with a grade 1 on the CSE examination being seen as equal to at least a grade C pass on the GCE examination. For both cohorts, the highest level of qualifications obtained at age 16 was recorded, ranging from none (0), CSE grade 2–5 (1), and CSE grade 1 or 0 level (2). An overall “exam score” could also be calcu- lated from the examination performance at age
- The actual examination results of the NCDS cohort were collected from schools in 1978, whereas the BCS70 cohort members self- reported their examination results in a follow-up study in 1986. The examination system was the same for both cohorts, with the BCS70 being one of the last cohorts to sit the two-tiered examina- tion structure of 0 levels and CSEs. A simple scoring technique was applied to the results, in which a score of 7 was given to a grade 10 level and a score of 1 was given to a grade 5 CSE. Scores ranged from 0 to 106 in the NCDS and from 0 to 97 in the BCS70.
Attainments in Adulthood Adult attainment was indicated by two measures of social position: the RGSC and the Cambridge Scale (CS) assessed at age 30 for the BCS70 cohort members and at age 33 for the NCDS cohort members. The 6-point RGSC scale, developed by the Office of Population and Census Surveys (OPCS; 1990), was described above. For ease of interpretation the coding was reversed, so that a high score indicated a high social position. The CS was conceptualized as an indicator of general social advantage and lifestyle (Prandy, 1990). It is based on the analy- sis of friendship choices, judged to be the most accurate indication of perceived and experienced social distance between members of different occupations. The scale is measured on a 100- point continuum, whereby high scores indicate a higher level of social advantage.
Statistical Analysis:
Modeling Cumulative Risk Effects
Structural equation modeling (Bollen, 1989) was used to formulate theoretically derived hypotheses about variable relations and to test postulated pathways between the variables and the assumed mediating processes involving latent variables with multiple indicators. Latent vari- ables represent hypothetical concepts that cannot be observed or measured directly. Instead, a set of observed variables are hypothesized to be imper- fect indicators of the latent variable. Because the study involved a cross-cohort comparison of data that had been collected for different surveys, some of the data were similar but not necessarily identical for the two cohorts. Great care was taken to define the latent constructs in as similar a way as possible in the two cohorts. Table 3. summarizes the selection of comparable indica- tor variables in both cohorts. All analyses were carried out using the SEM program AMOS 4.01 (Arbuckle, 1999). The AMOS program uses maximum likelihood estimation that can be based on incomplete data, known as the full information maximum likelihood (FIML) approach. This approach is preferable to estimation based on complete data (the listwise deletion [LD] approach) because FIML estimates tend to show less bias and be more reliable than LD estimates, even when the data deviate from missing at random and are nonignorable (Arbuckle, 1996). In the LD approach, the complete data covariance matrix is the data source for the latent variable analysis. In the FIML approach, estimation is based on the many covariance matrices between observed variables for all patterns of missing data in the other observed variables. Thus, it is not possible to present a single correlation matrix for the observed variables. Instead, Appendices A and B give the FIML estimates of the correlations between the observed variables, the means and SDs for the observed variables, and the FIML estimates of means and SDs. In line with current practice, several criteria were used to assess the fit of the model to the
Life-Span Developmental Theory (^) • 51
data. The x 2 statistic is overly sensitive to model misspecification when sample sizes are large or the observed variables are non-normally distributed. The root mean square error of approximation (RMSEA) gives a measure of the discrepancy in fit per degrees of freedom (Steiger, 1990). It is bounded below by 0, only taking this value if the model fits exactly. The RMSEA is useful because it encompasses the idea that a model is only expected to provide an approximation to the data rather than an exact fit. If the RMSEA is <.05, the model is considered a close fit to the data. Another advantage of the RMSEA is that confidence intervals may be calculated, which give further information on the reliability of the goodness of fit. The Consistent Akaike Information Criterion (CAIC) is a mea- sure of parsimonious fit recommended for large samples (Bollen & Long, 1993). The CAIC con- siders both the fit of the model and the number of estimated parameters whereby smaller values indicate a more parsimonious fit (Bollen & Long, 1993; Bozdogan, 1987). The final index of choice was the comparative fit index (CFI) whose values are restricted to lie on a 0 to 1
continuum, with higher values indicating a better fit (Bentler, 1990). The CFI is a population- based index that compares the model to a “null model.” The null model is a model in which there are no relations between any of the observed variables, but their variances are not constrained and are free to be estimated by the fitting proce- dure. The CFI of a model is normally tested against a minimum criterion value of .95.
Modeling Strategy
Four separate models were run for each cohort. Model 1 was the Full Developmental– Contextual Model, which included all paths indi- cated in Figure 3.1. In addition to Model 1 three other models were fitted to the data, to test the increase of fit depending on the pathways included, and to identify the most parsimonious model for describing the long-term influences of socioeconomic adversity on academic adjust- ment. Model 2 was the Developmental–Con- textual Model without path “g” (which assessed the direct effect of parental social class on an individual’s adult social status independent of
52 • READINGS IN FAMILY THEORY
Table 3.1 Variables and Observed Indicators in Both Cohorts
Variables Observed Indicators
Socioeconomic risk (RISK) Parental social class Father’s social class Mother’s social class Material conditions No housing tenure, overcrowding (>1 person per room), household amenities (shared use of bathroom), receipt of state benefits Academic adjustment (ACA) Academic attainment Ages 5 and 7: Human Figure Drawing, Copy-a-Design, Reading and vocabulary (NCDS and BCS70), and Arithmetic (NCDS only) tests Ages 10 and 11: Reading and mathematics tests (NCDS and BCS70) Age 16: Exam scores (NCDS and BCS70) Social position in adulthood (SOC) Social position at age 30 (BCS70) Registrar General’s measure of social class (RGSC) and age 33 (NCDS) Cambridge Scale (CS)
NOTE: NCDS = National Child Development Study; BCS70 = British Cohort Study.
54
Table 3.
Comparative Goodness of Fit of Structural Equation Models Run for the 1958 National Child Development Cohort (NCDS) and the1970 British Cohort Study (BCS70)
Goodness of Fit
Test of Close Fit
Step-Down Goodness of Fit
df
(^2) x
CFI
CAIC
RMSEA
90% CI
∆^
df
∆^
(^2) x
p
NCDS Model 1: The full model
93
1,221.
.
1,
.
.025–.
Model 2: Excluding path “g”
94
1,229.
.
1,
.
.025–.
1 (M2-M1)
<.
Model 3: Excluding paths “d”
95
1,886.
.
2,
.
.032–.
2 (M3-M1)
<.
Model 4: Excluding correlations “c”
96
1,716.
.
2,
.
.030–.
3 (M4-M1)
<.
BCS70 Model 1: The full model
77
.
1,
.
.023–.
Model 2: Excluding path “g”
78
.
1,
.
.023–.
1 (M2-M1)
<.
Model 3: Excluding paths “d”
79
.
1,
.
.024–.
2 (M3-M1)
<.
Model 4: Excluding correlations “c”
80
1,022.
.
1,
.
.027–.
3 (M4-M1)
<.
NOTE:
df
=
degrees of freedom; CFI
=
comparative fit index; CAIC
=
Consistent Akaike Information Criterion; RMSEA
=
root mean square error of approximation;
CI
= confidence interval.
In both cohorts Model 2 was judged to be the best model for describing the relations in the data. Because the x 2 statistic is overly sensitive to model misspecification when sample sizes are large, the x 2 statistic in conjunction with the CAIC were used to identify the best-fitting model for the data. Model 2 was run without path “g” suggesting that there were no cohort effects independent of intervening variables. Given that Model 2 was the most parsimonious in both cohorts, it is described in detail below. Table 3.4 gives the standardized parameter estimates of the measurement model for both cohorts. The standardized regression weights of the indicator variables on their latent variable differed slightly for the two cohorts. Social class had a similar weighting in both cohorts, and was the most important indicator of socioeconomic disadvantage. Material deprivation gained more importance as the NCDS cohort members grew older, whereas for the BCS70 cohort there was a peak at age 10. The most important indicators of a child’s adjustment in the NCDS were generally reading and math test scores, and at age 5 also the Human Figure Drawing and Copy-a-Design tests. The BCS70 cohort had slightly different indicators: at age 5 the Copy-a-Design test and the reading test were the principal indicators, whereas at age 10, the reading and math tests were the principal indicators. For both cohorts, test performance at age 10/11 was more crucial than earlier performance as a determinant of a child’s adjustment. At age 16, the exam score was the key indicator of a child’s adjustment in both cohorts, particularly so in the later-born cohort. Social status for the 33-year-old NCDS cohort members was slightly better identified by the CS than by the RGSC, whereas in the BCS at age 30 years, the RGSC was a better indicator than the CS. Figures 3.2 and 3.3 show the pathways between the latent or unobserved variables, which represent continuities and interactions of social risk and academic adjustment, and give the standardized coefficients for the structural model, fitted separately for both cohorts. Several covariances between the error terms for the
observed variables were included a priori to account for the autocorrelations over time. Parental social class and material deprivation were hypothesized to covary for the measure- ment points at consequent ages, as were the aca- demic attainment scores. The variables shown were all latent or unobserved variables. The hypothesized pathways were supported by the data, with the parameter estimates all being significantly different from 0, p < .005, and in the predicted direction. The effect sizes of the parameter estimates are described as small, r =. 10, medium, r = .30, and large, r =. 50, fol- lowing Cohen’s 1992 power primer. In both cohorts a stark chaining, or continuity of risk factors, was found: parental social class at birth predicted the experience of risk at subsequent ages, and the experience of risk at one time point increased the probability that risk would also be encountered at a later time point. Parental social class also had a moderate influence on academic adjustment. The experience of social risk at birth influenced the level of later academic attainment. Also observed were continuities in academic adjustment level over time. Academic attainment at one time point was a significant predictor of academic attainment at a later time point. Continuities occurred because current adjust- ment encompassed previous adjustment as well as earlier structural and functional change. The detrimental effect of experiencing disadvantage at one measurement point was carried forward into the future via decreased individual adjust- ment levels. The model also shows concurrent associa- tions between social risk and academic attain- ment. In the NCDS, these current associations (coefficients “c”) were of moderate size at age 7 and 16, and were only of small size at age 11. In the BCS70, the concurrent association between social risk and academic attainment at age 5 was of moderate size, whereas at age 10 and 16 the associations were only small. In addition, there were small time-lagged risk effects (paths “d”) which indicated the added negative influence of social risk on subsequent attainment not accounted for by the risk carried
Life-Span Developmental Theory (^) • 55
forward in time. In both cohorts these time-lagged risk effects were greatest at the transition from late childhood to adolescence. The best predictor of attained social status in early adulthood was academic attainment at age 16, which confirmed the crucial role of education in determining adult outcomes. In both cohorts, the direct influence of social risk experienced at
age 16 on adult social status was only small, yet in comparison with the NCDS the effect size had nearly trebled in the BCS70. It is important to remember that the experience of social risk does have additional indirect effects on social status operating through the child’s academic adjust- ment. For the NCDS cohort, the combined effect of the variables in the model explained 37% of
Life-Span Developmental Theory (^) • 57
. .97.
−.
.78. .
−.38 −.21 −. −.19 −.
−.
Birth
Parental Social Class
SOC
ACA
RISK
ACA
RISK
ACA
RISK
Early Childhood Late Childhood Adolescence Adulthood
Figure 3.2 The National Child Development Study: Developmental-contextual model of accumulating risk effects from birth to adulthood. ACA = academic adjustment; Risk = socioeconomic risk; SOC = social position in adulthood. See text for further details.
. .93.
−.
.81. .
−.45 −.15 −. −.07 −.
−.
Birth
Parental Social Class
SOC
ACA
RISK
ACA
RISK
ACA
RISK
Early Childhood Late Childhood Adolescence Adulthood
Figure 3.3 The British Cohort Study: Developmental-contextual model of accumulating risk effects from birth to adulthood. ACA = academic adjustment; Risk = socioeconomic risk; SOC = social position in adulthood. See text for further details.
the variation in adult social status; and for the BCS70 cohort, the model explained 54% of the variation.
Discussion
In the study of human development, laborato- ries are rare. The closest one can get to them is when the opportunity arises to compare and con- trast experiences in different societies or in dif- ferent sociohistorical contexts. The present study used data collected for two national birth cohort studies born 12 years apart. On comparison of the prevalence of risk indicators that were assessed at roughly the same time (e.g., parental social class or housing tenure experienced by 16- year-olds born in 1958 and 5-year-olds born in 1970, assessed in 1974 and 1975, respectively), it appears that the samples represented well the state of affairs over the historical period covered. The data suggest that there was a real-time secular shift in material resources. For cohort members born in 1970, the indicator variables point to improved material circumstances for the cohort as a whole. Furthermore, the parents of cohort members born in 1970 experienced upward social mobility while bringing up their children. Those cohort members born in 1970 who suffered deprivation, however, appear to have been more disadvantaged relative to other children in the same cohort than similarly affected children born earlier, in 1958. A strong continuity of social risk effects was observed, which was of similar strength for both cohorts. The strength of the association can be explained by the fact that in addition to indicators of material disadvantage, indicators of SES, which denote relative position in society, were also used. However, despite improved material condi- tions, and the experience of upward social mobility among parents of the BCS70 cohort, the relative social position remained remarkably stable. Furthermore, the influence of parental social class at birth on academic adjustment was greater for cohort members born in 1970 than for those born in 1958, and the direct influence of social risk experienced in adolescence on adult
attainment was only small, but nearly trebled for the later-born BCS70 cohort. These findings suggest that for cohort members born in 1970, contextual factors, to some extent, became more important than for cohort members born in 1958 in shaping the development of academic adjustment, and in influencing attainments in adulthood. Material conditions improved, yet socioeconomic disadvantage continued to be a barrier for individual achievements. It has been argued that SES at the time of a child’s birth is an indicator of the social context, but also reflects hereditary influences, which are assumed to have some role in determining the level of academic functioning (Plomin & Bergeman, 1991; Plomin & McClearn, 1993; Scarr, 1992). The present study’s data show that the influence of social class at birth on academic adjustment was of moderate size, and differed slightly for the two cohorts. Socioeconomic disadvantage experienced consequently during childhood had an additional detrimental influ- ence on the level of academic adjustment, sug- gesting that contextual factors have a role in shaping the level and maintenance of academic adjustment throughout childhood and adoles- cence. Furthermore, there were smalltime-lagged effects of earlier risk on later academic achieve- ment, indicating that social risk does not always have an immediate impact, and that vulnerabili- ties may emerge only later in life. On the other hand a considerable stability of individual adjustment was also seen. These results suggest that adaptation is a product of both developmen- tal history and current circumstances (Clarke & Clarke, 1981, 2000; Sroufe, Egeland, & Kreuzer, 1990), shaped by the interactions between individual and context (Bronfenbrenner & Ceci, 1994). Generally, the findings of this study concur with a large body of previous research, which has reported consistent correlations between measures of social disadvantage and measures of individual academic achievement and occupa- tional attainment (Blau & Duncan, 1967; Bolger et al., 1995; Bynner, Joshi, & Tsatsas, 2000; Campbell & Ramey, 1994; Duncan et al., 1994;
58 • READINGS IN FAMILY THEORY
