Variation in children's cognitive and behavioural adjustment between different types of place in the...

ARTICLE IN PRESS

0277-9536/$ - se

doi:10.1016/j.so

E-mail addr

Social Science & Medicine 62 (2006) 1865–1879

www.elsevier.com/locate/socscimed

Variation in children’s cognitive and behaviouraladjustment between different types of place in the

British National Child Development Study

Andrew McCulloch

2 Napiershall Street, Glasgow G20 6HQ, UK

Available online 3 October 2005

Abstract

Several recent studies suggest that there is a link between children’s development and the types of neighbourhood in

which they live. This paper examines the relationship of a classification of different types of neighbourhood to children’s

cognitive and behavioural outcomes using British data on children who have a parent who is a member of the 1958 Birth

Cohort Study. Substantial variations in children’s outcomes were observed between the neighbourhood types. Part of the

variation between neighbourhood types was accounted for by various family risk factors but children living in

neighbourhoods classified as Deprived City Areas continued to have higher levels of behaviour problems and lower

cognitive test scores than average after adjustment for these factors. Significant adjusted associations were also found

between living in neighbourhoods described as Middling Britain and lower cognitive test scores and neighbourhoods

described as Prosperous Areas and better cognitive test scores. For children’s cognitive outcomes, both the size and

statistical significance of coefficients on the neighbourhood categories were smaller than those of family-level measures. In

contrast, residence in a Deprived City Area was as significant as the family factors in predicting higher levels of behaviour

problems. The association between behaviour problems and family risk factors was mediated by the home environment

and cohort member parent’s psychological health. Mediated effects were stronger for the family risk factors behavioural

problems association than for the neighbourhood classification.

r 2005 Elsevier Ltd. All rights reserved.

Keywords: UK; Children’s outcomes; Office for National Statistics (ONS) ward classification; Neighbourhood

Introduction

There is a growing body of research whichexamines the characteristics of neighbourhoodsand communities that affect children, the natureof those effects and the mechanisms and mediatingprocesses at the community and family levelsthrough which the effects operate. Recent research

e front matter r 2005 Elsevier Ltd. All rights reserved

cscimed.2005.08.048

ess: [email protected].

in the USA has demonstrated a significant relation-ship between neighbourhood characteristics and thewell-being of the children who live in them. Evenafter controlling for child and family backgroundcharacteristics, residence in a low-income neigh-bourhood has negative effects on children’s schoolreadiness and academic achievement and is asso-ciated with higher rates of externalising behaviourproblems, dropout from high school and teenageparenthood (Brooks-Gunn, Duncan, & Aber, 1997).

.

www.elsevier.com/locate/socscimed

ARTICLE IN PRESSA. McCulloch / Social Science & Medicine 62 (2006) 1865–18791866

In addition, neighbourhood income effects havebeen found to be smaller than the effects of familyincome, maternal education or maternal maritalstatus (Klebanov, Brooks-Gunn, McCarton, &McCormick, 1998).

A variety of perspectives have been used toexplain the association between residence in a low-income neighbourhood and children’s outcomes.Most of these perspectives can be characterised asemphasising either neighbourhood economic re-sources or processes of socialisation and socialcontrol in neighbourhoods. As the name implies, thefirst of these perspectives focuses on the resourcesavailable to children in different types of neighbour-hoods as a central mechanism generating differen-tial outcomes. While the availability and qualityof resources may be affected directly by Govern-ment policy, they are also likely to be relatedto neighbourhood socioeconomic characteristics(Aber, Bennett, Conley, & Li, 1997; Jencks &Mayer, 1990). Children living in low-income neigh-bourhoods may have access to fewer neighbour-hood resources such as good schools, parks andlibraries, and the lack of these resources contributesto the negative outcomes associated with living in alow-income neighbourhood.

More attention has recently been paid, however,to the socialisation and social control perspectives.Proponents of models of collective socialisationsuggest that neighbourhood adults play an impor-tant role by providing role models. Wilson (1996)offers perhaps the most influential statement of thecollective socialisation perspective, arguing thatneighbourhoods with high levels of poverty, job-lessness, and other forms of disadvantage are morelikely than affluent neighbourhoods to provideopportunities for children to be exposed to negativeadult role models and deviant peers. Social capitalmodels share many of the features of the collectivesocialisation approach but emphasise the role ofdense and overlapping social ties (Coleman, 1988).In the context of the community, social capitalexists in the norms, social networks and interactionsbetween adult members of the community andchildren that serve to facilitate or support children’soutcomes. From the perspective of this approach,children’s outcomes are related to the extent towhich children have access to social networks thatcan provide beneficial resources and information. Incontrast, the social control perspective emphasisesthe link between working trust between neighbour-hood adults and shared expectations for action to

maintain public order. Jencks and Mayer (1990)argue that adults in more affluent neighbourhoodsare more invested in their neighbourhoods and playan important role by acting as potential enforcerswho keep children from engaging in risky beha-viour, a notion echoed by Brooks-Gunn, Duncan,Klebanov, and Sealand (1993) and in literaturerelated to juvenile delinquency and crime (Sampson,Morenoff, & Earls, 1999; Sampson, Raudenbush, &Earls (1997)).

A separate line of research has examined the rolethat families play in determining how neighbour-hoods impact on the lives of children. This researchshows that the characteristics of neighbourhoodsmay directly influence parenting behaviour. Parentsin disadvantaged neighbourhoods are more likely toexhibit more punitive, authoritarian, and coerciveparenting styles which are likely to have detrimentalconsequences for children’s cognitive and beha-vioural development (Jarrett, 1997). Parents andfamilies may also act as mediators between childrenand neighbourhood conditions. For example, qua-litative studies suggest that parents in poor neigh-bourhoods use a variety of strategies both to protecttheir children from negative aspects of the neigh-bourhood and to find the resources their childrenneed (Furstenberg, 1993). Nonetheless, identifyingthe precise mechanisms by which families influencethe relationship between neighbourhoods and childoutcomes remains a challenge for researchers.

In the UK, there is comparatively little researchwhich examines the influence of neighbourhoods onchildren’s development. Rutter, Cox, Tupling,Berger, and Yule (1975), Rutter, Yule et al. (1975)compared the developmental outcomes of childrenaged 10 years from an Inner London borough andthe Isle of Wight. Emotional disorders, conductdisorders and reading retardation were all twice ascommon in the group of children living in InnerLondon (Rutter, Cox et al., 1975). The adversedevelopmental outcomes experienced by the chil-dren living in London could be attributed to thehigher prevalence of family risk factors includingfamily discord, parental deviance and social dis-advantage (Rutter, Yule et al., 1975). In contrast,Thompson et al. (1996), in a study of 1047 familieswith a 3 year-old child in South West England,found few differences between urban and rural areasin the prevalence rates of behaviour problems.Meltzer, Gatward, Goodman, and Ford (2000) usedthe ACORN classification of the electoral ward ofresidence to add a geographical dimension to the

ARTICLE IN PRESSA. McCulloch / Social Science & Medicine 62 (2006) 1865–1879 1867

investigation of the prevalence of mental disordersin a national sample of children in England andWales between the ages of 5 and 15 years. Theprevalence of disorders was related to the approx-imate level of prosperity of the electoral ward ofresidence. The highest proportion of children withany disorder was found among families living inwards classified as striving (13%). This was aboutdouble the proportion in wards classified asexpanding (7%) and two and a half times theproportion in wards in the thriving category (5%).In another study, McCulloch and Joshi (2001)found that the level of deprivation in electoralwards, measured using the Townsend deprivationindex,1 had a significant association with lowercognitive test scores in children aged 4–5 yearswhich, though somewhat attenuated, was indepen-dent of other socioeconomic indicators. Amongchildren aged between 6 and 18 years, levels of warddeprivation were for the most part statisticallyinsignificant predictors of both cognitive test scoresand maternal reports of children’s behaviourproblems.

This paper uses the same data source asMcCulloch and Joshi (2001) to examine therelationship between measures of children’s devel-opment and a classification of the electoral wards inwhich they live. The Townsend index assesses theeffect of a particular summarisation of the socialcomposition of the ward. Although there are clearadvantages to such data reduction, the use of aunidimensional deprivation index does not takeaccount of differences across a number of dimen-sions. The ward classification is based on astatistical cluster analysis of 37 socioeconomic anddemographic characteristics recorded at the 1991census (Wallace, Charlton, & Denham, 1995). Thisresults in a descriptive classification in which wardssharing similar patterns of socioeconomic, demo-graphic, housing and environmental features aregrouped together and with maximum differentiationbetween clusters. Membership of clusters is basednot on geographical contiguity but on similarity interms of the selected census variables. As the wardclassification is the product of a cluster analysis of awide range of variables it may describe neighbour-

1The components of the Townsend index are: the proportion of

the labour force unemployed, the proportion of households with

no car access, the proportion of households with 1 person per

room and over and the proportion of households not owning

their own home.

hoods in a more informative way than the depriva-tion index used by McCulloch and Joshi (2001).Different area classifications have been found toexplain variation between areas in the social andhealth outcomes of adults (Chandola, 2001; Drever& Whitehead, 1995; Ecob & Jones, 1998). It ispossible that a classification of different types ofplace may similarly distinguish between areas interms of children’s outcomes.

Data and statistical methods

The National Child Development Study (NCDS)is a study of over 17,000 people in Britain bornbetween the 3rd and 9th of March in 1958. Follow-up sweeps took place in 1965, 1969, 1974, 1981,1991 and 2000. The 1991 NCDS follow-up obtainedinformation not only from the cohort member, butfrom the children of 1 in 3 cohort members andfrom the mother of these children (Ferri, 1993). Thisanalysis uses two measures of children’s develop-ment from the 1991 sweep. The behaviouraladjustment of children who were aged 7 years orolder was assessed using the Children’s BehaviourQuestionnaire for completion by parents (ChildScale A: Rutter, 1967). The instrument was designedto provide valid and reliable screening measures of achild’s behaviour at home. The version of thequestionnaire contained in the NCDS consists of aseries of 18 behavioural items to which the motherreplies ‘‘does not apply’’ scored 3, ‘‘applies some-what’’ scored 2 and ‘‘certainly applies’’ scored 1(McCulloch, Wiggins, Joshi, & Sachdev, 2000).Each item was recoded to take the value 1 if theitem ‘‘certainly applies’’ or ‘‘applies somewhat’’ tothe child and the value 0 if it ‘‘does not apply’’ and ameasure of externalising behavioural problemscalculated as the unweighted sum of the responsesto the following items: destructiveness, fighting,irritability, not liked, disobedience, lying and bully-ing (Fig. 1). The paper does not report results forchildren’s internalising behavioural problems(McCulloch et al., 2000) because the level of suchproblems showed no relationship to the categoriesof the ward classification (or to the Townsenddeprivation index). In this paper, children’s perfor-mance on the Peabody Picture Vocabulary Test(PPVT; Dunn & Markwardt, 1970) is used as anindicator of cognitive ability. The PPVT is anindividually administered test of hearing vocabu-lary, with 175 test items arranged in order ofincreasing difficulty. Each item presents a multiple

ARTICLE IN PRESS

200

150

100

50

0

Fre

quen

cy

0 1 2 3 4Behaviour Problems

5 6 7

Fig. 1. Histogram of the sum of the number of positive items

from the externalising behaviour problems scale.

400

300

100

200

0

Fre

quen

cy

0.0 0.2 0.4 0.6 0.8 1.0

PPVT % test score

Fig. 2. Histogram of children’s PPVT test scores.

2Previous analyses have used imputation to account for

nonresponse on the family income variable (Joshi et al., 1999;

McCulloch & Joshi, 2002). The latter suggested that car access

and housing tenure were more systematically related to child

outcomes than current family income per person.

A. McCulloch / Social Science & Medicine 62 (2006) 1865–18791868

choice between four illustrations, the aim being toselect the picture best illustrating the meaning of astimulus word, presented orally by the examiner.The PPVT results are available for children aged 4years and older. In the analyses the test scores foreach child are divided by the total number of itemsin the test. The test scores range between aminimum of 0.04 and a maximum of 1 and have astandard deviation of 0.16 (Fig. 2).

In this analysis the definition of neighbourhood isdictated by the geographical units for which dataare available. Wards are the constituency for localgovernment elections. Being based on politicalboundaries, it is arbitrary, but nevertheless repre-sents a useful framework for analysis. Their averagepopulation is around 5500, but this varies consider-ably. In this study, the Office for National Statistics(ONS) classification of wards is used to classifywards on a demographic and socioeconomic basis(Wallace et al., 1995). The classification assignswards into 14 groups, and at a lower level withinthese, 43 clusters. As the level of aggregationincreases, the description of the areas becomesbroader, but the categories are still intended todistinguish between different types of neighbour-hood. In this study, the classification into groupswas used. Ten categories are used here as there wereinsufficient numbers of children living in three typesof ward (Inner City Estates, Metropolitan Profes-sionals, Transient Populations) to allow analysisand one category (Scotland) was not represented inour data. The sample is restricted to childrenresident in England and Wales in 1991.

In examining the association between differenttypes of place and children’s development, theanalysis controls for a range of indicators of

socioeconomic position. The current employmentstatus of the mother interacted with the presence ofa father is used as an indicator of the family’seconomic status. The presence of a father isincluded on the presumption that children thrivewhen they have the attention of more than oneadult. The employment of the mother is interactedwith family structure, because there can be no morethan one earner in a lone parent family. Previousresearch has shown that family poverty has asignificant effect on children’s outcomes because itinfluences how parents interact with their children(Conger, Conger, & Elder, 1997). How povertyinfluences behavioural problems in adolescents hasalso been examined by Gerard and Buehler (1999)who point to links between the educational qualityof the home environment, academic difficulties andincreased problem behaviours. It was not possibleto use family income for the year of interview as ameasure of the family’s economic resources becauseof the nonresponse of approximately one third ofthe cases. As an indicator of family poverty theanalysis uses residence in social housing (i.e. rentedfrom council or housing association). Social hous-ing is commonly used as an indicator of deprivation(Townsend, Phillimore, & Beattie, 1989) and iscertainly related strongly to income (Davies, Joshi,& Clarke, 1997).2 The cohort member parent’shighest level of education is used as a generalindicator of social advantage. Factors other thansocioeconomic position are likely to contribute to


individual differences in children’s outcomes. Thepossibility that other factors such as birth orderwould also be predictive of children’s outcomes wasalso explored.

In this study, the cohort member parent’spsychological health and the quality of the homeenvironment are examined as potential mediators ofthe link between the neighbourhood categories, thesocioeconomic position of the family and children’soutcomes. Klebanov et al. (1998) found that HOMEscores mediated the influence of living in a low-income neighbourhood and family poverty on thedevelopmental outcomes of children aged less than3 years. Neighbourhood conditions appear toinfluence children’s learning environment withinthe home independently of the influence of familycharacteristics. Similarly living in a disadvantagedneighbourhood has been shown to increase the riskof psychiatric morbidity in adults (Ross, 2000) andsymptoms of depression such as irritability and lackof emotional responsiveness may reduce the capa-city of adults to care for children.

The cohort members summed responses to the 24-item malaise inventory of psychological and somaticsymptoms (Rutter, Tizard, & Whitmore, 1970) areused as an indicator of psychological health. Thequality of the child’s home learning environmentwas assessed using the cognitive stimulation sub-scale from the short form of the HOME inventory(Smith, Brooks-Gunn, & Klebanov, 1997; Suglandet al., 1995). The emotional support subscale fromthe HOME inventory, which measures the warmthof mother–child interactions, was not used becauseit has been found not to meet adequate internalconsistency levels (Klebanov, Brooks-Gunn, Chase-Landsdale, & Gordon, 1997). Each item on theHOME inventory is scored in a binary fashion; ascore of 1 indicates the presence of a developmen-tally supportive aspect of the environment, and ascore of 0 otherwise at the time of the assessment.The maximum value of each overall HOME scorevaries since the number of individual items adminis-tered to children of different ages is not the same.An internally normed percentile score created on anage specific basis (i.e., each single year of age wasnormed separately) was therefore used.

In general, the NCDS has achieved very highresponse rates at each sweep (Ferri, 1993). Somesubject attrition and missing data are, however,inevitable in a study of this complexity. Previousanalyses suggest that complete data cases tend tounder-represent subjects from lower social class

backgrounds, and with lower attainment and abilityscores in childhood (Ferri, 1993). In the data used inthis study the selective attrition of adult cohortmembers from disadvantaged backgrounds may,however, be compensated for by the nature of thesample. In the data used here any children aged over13 would have been born to a teenage parent. Thesample design is thus biased towards children withyoung and hence more disadvantaged parents thana full cross-section of children.

The relationship of child and family character-istics and area type to children’s externalisingbehaviour problems was examined using a multi-level binomial regression model. For child i infamily j this is specified as

log itðpijÞ ¼ aþX

X ijbþX

Zareagþ kj, (1)

where b represents the effects of child and familycharacteristics, X, g represent the effects of areatype, Z, and p is the probability of a positiveresponse to each scale item. In this paper, the ONSgroup codes are viewed as fixed effects, eachconsidered as a dummy variable with reference toa default category (taken as the average of thesample as a whole. The number of items contribut-ing to the behaviour problems scale is used as abinomial denominator. The kj represent indepen-dent and identically distributed family specificdifferences in the log-odds of the response. It isassumed that the kj have a N(0,s2) normal distribu-tion. They summarise the degree to which the familyspecific probabilities differ from their averagepopulation values. Stated informally: the unex-plained family effects are governed by mechanismsthat are roughly the same from one family to thenext, and operate independently between thefamilies. The families are said to be exchangeable.This assumption helps to counteract the paucity ofthe data that results from relatively small number ofchildren per family. Since all family effects areassumed to come from the same population, thedata from each family also have a bearing oninference with respect to the other families, namely,through the information it provides about thepopulation of families. The sample contains 912children who come from 658 families. The data hasrelatively few families per ward, which makesmultilevel analysis at the level of the neighbourhoodimpractical.

The analysis of children’s cognitive test scoresuses a multilevel linear regression model. For child i


in family j this is specified as

Y ij ¼ aþX

X ijbþX

Zareagþ kj þ �ij . (2)

The model contains two random effects: kj and eij

which are both assumed to follow a normaldistribution. Each of these indicates a differentsource of unexplained variation. The randomintercept kj indicates unexplained differences be-tween families in the average Y-values (controllingfor the effect of Xij and Zarea). The random residualeij, indicates unexplained variation among thechildren, relative to their families. The samplecontains 1629 children who come from 1087families. The analyses were carried out using theR software package and libraries contributed byD. Bates and J. Lindsey.

Results

Table 1 presents the means of the variables usedin the analyses for the different categories of theONS area classification. Table 1 shows that thereare substantial variations between different types ofplace in maternal reports of children’s behaviouralproblems. Children living in wards classified asDeprived City Areas and Industrial and Manufac-turing Towns had the highest reported levels andchildren living in wards classified as ProsperousAreas had the lowest reported levels. Children’sPPVT test scores also vary across type of area withchildren living in wards classified as Rural Areashaving the highest test scores and children living inwards classified as Established Owner Occupiers thelowest test scores. The variation in PPVT test scoresbetween different types of place is, however, difficultto interpret directly because of differences in theaverage age of children between different types ofarea. In particular, older children are more likelythan younger children to live in Deprived CityAreas and Industrial and Manufacturing Towns.This finding may be due to the nature of the samplewhich imposes a negative relationship between theage of the child in 1991 and the age of the cohortmember parent when the child was born. Olderchildren are more likely to be born to youngermothers with fewer years of education and thus tobe reared in more disadvantaged neighbourhoods.

Table 1 also shows that different aspects of socialand economic deprivation tend to co-exist in ageographically concentrated way. Children living inDeprived City Areas and Industrial and Manufac-

turing Towns are more likely than children living inother neighbourhood types to have a cohortmember parent with low educational qualifications,to have low family incomes or to live in socialhousing. There are also differences in homeenvironment between children living in differentneighbourhood types. Children living in DeprivedCity Areas are more likely than children in othertypes of neighbourhood to have a cohort memberparent with a high malaise score and to live inhomes lacking cognitive stimulation. The level ofdeprivation of the ward as measured by the Town-send index calculated using 1991 Census data(Townsend et al., 1989) also distinguishes thedifferent types of area. The index is standardisedto the UK and the least deprived wards are thosewith the highest negative numbers. Deprived CityAreas and Industrial and Manufacturing Towns arethe most deprived types of neighbourhood andProsperous Areas and Rural Areas are the mostprosperous types of neighbourhood. The low levelsof deprivation in wards classified as Rural Areasmay be due to the inappropriateness of the Town-send index as a measure of deprivation in RuralAreas. For example, ownership of a car is so basicnecessity for households in the most remote areasthat it cannot reasonably be interpreted as anindicator of affluence. Nevertheless, Table 1 showsthat children residing in Prosperous Areas are lesslikely to be from poor families than are children inDeprived City Areas and Industrial and Manufac-turing Towns. Poor children may be disadvantagedin large part because of their communities ofresidence, not just because of their family situations.

The results from estimating regression models fordifferent predictors are presented in Tables 2 and 3for behaviour problems and PPVT test scores,respectively. The analyses were organised as fol-lows. First, analyses were conducted in order tolook at the effect of neighbourhood type inpredicting child outcomes also controlling for childage and gender (Model A). Model terms in age andage-squared were included to allow for nonlinea-rities in the association between age and childoutcomes. Then, regressions with family-level vari-ables added were run, to see whether the effect ofneighbourhood type was independent of thesefamily-level variables (Model B). Finally, in orderto look at whether any of the observed neighbour-hood or family effects operated through the homeenvironment or cohort member’s psychologicalhealth two additional analyses were run. First,

ARTICLE IN PRESS

Table

1

Meansofvariablesusedin

theanalysisbycategories

ofONSareaclassification

Variable

ONSGROUP

Deprived

city

areas

Established

owner

occupiers

Industrialand

manufacturing

towns

Lower

status

owner

occupiers

Mature

populations

Middling

Britain

Prosperous

areas

Ruralareas

Ruralfringe

Suburbia

Girl

0.56

0.53

0.47

0.44

0.52

0.54

0.53

0.48

0.46

0.48

Age(years)

9.34

7.31

9.47

8.81

8.28

8.83

7.35

9.04

8.14

8.24

Birth

order

1.68

1.46

1.56

1.62

1.46

1.61

1.49

1.63

1.67

1.57

PPVT

test

score

0.51

0.47

0.54

0.52

0.52

0.52

0.50

0.56

0.50

0.50

Externalisingbehaviour

problemsscore

2.93

1.97

2.40

1.99

1.86

2.34

1.79

2.21

2.05

2.42

Cohort

mem

ber’sage

firstbirtho20years

0.17

0.06

0.24

0.15

0.09

0.13

0.00

0.00

0.10

0.13

Intact

familyand

nonworkingmother

0.22

0.22

0.23

0.17

0.17

0.23

0.15

0.07

0.29

0.16

Intact

familyand

workingmother

0.51

0.61

0.51

0.68

0.74

0.64

0.78

0.55

0.61

0.68

Stepfamilyand

nonworkingmother

0.07

0.06

0.06

0.04

0.02

0.02

0.00

0.07

0.02

0.02

Stepfamilyandworking

mother

0.05

0.07

0.08

0.07

0.03

0.04

0.05

0.20

0.03

0.07

Lonemother

and

nonworking

0.09

0.01

0.08

0.01

0.00

0.04

0.00

0.11

0.00

0.03

Lonemother

and

working

0.07

0.02

0.04

0.03

0.03

0.03

0.02

0.00

0.05

0.04

Cohort

mem

bershighest

qualification

1.74

2.54

1.86

2.06

2.39

2.00

2.74

1.90

2.41

1.98

Socialhousing

0.50

0.14

0.42

0.13

0.15

0.23

0.10

0.30

0.16

0.24

Net

familyincome(per

head)

56.95

80.21

63.38

65.21

68.39

67.97

119.27

71.48

73.50

87.28

Cohort

mem

ber’s

malaisescore

3.75

2.83

3.24

3.40

2.32

2.78

1.64

1.34

2.22

2.63

HOME:cognitive

stim

ulation

47.53

65.05

53.05

55.14

65.77

52.92

65.41

62.91

60.01

52.54

Ward

Townsend

DeprivationIndex

7.83

�2.90

3.73

2.22

�0.82

0.71

�3.31

�2.00

�1.97

�1.34

Number

ofcases

147

192

223

139

92

303

59

44

140

290

No

tes:Birth

order:1¼

firstborn

to3¼

thirdborn

orlater.

Cohort

mem

ber’shighesteducationalqualification:1¼

Noneto

4¼

Degreeorequivalent.

Number

ofchildrenaged

7years

andolder

forwhichexternalizingbehaviourproblemsoutcomeisavailable

are,respectively,95,74,134,88,51,179,33,29,73and156.

Number

ofcasesforwhichincomevariable

isavailable

are,respectively,106,125,163,106,54,220,34,38,103and98.

A. McCulloch / Social Science & Medicine 62 (2006) 1865–1879 1871

ARTICLE IN PRESS

Table 2

Regression coefficients on a log it scale and t-statistics of factors predicting child externalising behaviour problems (children age 7 years

and older)

Model A Model B Model C

b t b t b t

Constant �0.46 �0.78 �0.70 �1.11 �0.70 �1.10

Girl �0.32 �5.39 �0.31 �5.13 �0.28 �4.58

Age (years) �0.05 �0.45 �0.02 �0.18 0.03 0.24

Age-squared/100 0.12 0.24 0.02 0.04 �0.28 �0.49

Area classification

Deprived city areas 0.56 6.51 0.39 4.43 0.34 3.80

Established owner occupiers �0.25 �2.38 �0.12 �1.16 �0.20 �1.84

Industrial and manufacturing towns 0.17 2.10 �0.05 �0.62 �0.03 �0.38

Lower status owner occupiers �0.15 �1.58 �0.08 �0.87 �0.18 �1.87

Mature populations �0.20 �1.60 �0.08 �0.66 �0.07 �0.54

Middling Britain 0.17 2.39 0.15 2.12 0.11 1.51

Prosperous areas �0.23 �1.56 �0.05 �0.33 �0.03 �0.21

Rural areas �0.02 �0.11 �0.17 �1.06 0.05 0.33

Rural fringe �0.14 �1.34 �0.09 �0.89 �0.08 �0.76

Suburbia 0.10 1.30 0.11 1.42 0.09 1.21

No qualifications — — — —

CSE grade C or equivalent �0.39 �4.82 �0.20 �2.40

A level or equivalent �0.22 �2.64 �0.06 �0.76

Degree or equivalent �0.29 �2.84 �0.04 �0.35

Intact family and nonworking mother — — — —

Intact family and working mother 0.08 1.03 0.11 1.40

Step family and nonworking mother 0.24 1.62 0.19 1.30

Step family and working mother 0.12 0.93 0.08 0.65

Lone mother and nonworking 0.87 5.10 0.70 4.07

Lone mother and working �0.08 �0.51 �0.10 �0.63

Birth order I — — — —

Birth order II 0.07 0.93 �0.01 �0.12

Birth order III or later 0.45 4.08 0.30 2.66

Social housing 0.29 4.07 0.18 2.46

Cohoirt member’s age first birtho20 years 0.25 2.13 0.20 1.71

Cohort member’s malaise score 0.05 5.81

HOME: cognitive stimulation �0.71 �6.20

s2j 1.32 31.04 1.27 30.06 1.21 29.67

w2 ONS area classificationa 15.13 (0.09) 7.47 (0.59) 6.69 (0.67)

aLog-likelihood ratio significance test of ONS area classification term (9 degrees of freedom, p-value in parentheses).


regressions with the home environment and cohortmember’s malaise added were run (Model C).Second, regressions were run to examine theassociation of the independent neighbourhood,family and child variables with both the homeenvironment and cohort member’s malaise score(Model D). To establish mediation, the followingconditions must hold (Baron & Kenny, 1986;Mirowsky, 1999): First, the malaise score and homeenvironment must influence child outcomes; second,there should be a reduction in the magnitude of theassociation between neighbourhood and familyfactors and child outcomes when the malaise score

and home environment are included (perfect media-tion occurs if either neighbourhood or familyfactors have no effect once the malaise score andhome environment are controlled); and third,neighbourhood and family factors should be asso-ciated with the malaise score and home environ-ment.

The results in Table 2 (Model A) show that con-trolling for child age and gender, the area classifica-tion term has only limited explanatory power(w2 ¼ 15:13, p-value ¼ 0:09). The estimated log it

of behaviour problems in relation to the sample as awhole is, however, statistically significant at the 5%


level for the following neighbourhood types: De-prived City Areas, Established Owner Occupiers,Industrial and Manufacturing Towns and MiddlingBritain. Children with higher levels of externalisingbehaviour problems were more likely to be residentin Deprived City Areas, Industrial and Manufactur-ing Towns and Middling Britain than children withlower levels of such problems, whereas living in award in the Established Owner Occupier categorywas associated with lower levels of behaviourproblems. The coefficients are of a significant orderof magnitude: the odds of externalising behaviourproblems among children living in Deprived CityAreas was around 1.7 times that of the averagechild. In comparison to boys, girls are reported asshowing lower levels of externalising behaviourproblems. In a multilevel analysis the influence offamily context can be detected from the randomfamily-level terms included to capture unobservedcharacteristics of families (kj). The statistical sig-nificance of the kj suggest that there is substantialheterogeneity in the log-odds of the outcome at thefamily level. That is to say, there is substantialvariation between families in the log-odds ofbehaviour problems controlling for child ageand gender.

The introduction of terms describing the structureof the family, the mother’s employment, the cohortmember’s education, the child’s birth order andfamily poverty removes the significant associationbetween the neighbourhood categories EstablishedOwner Occupiers and Industrial and ManufacturingTowns and behaviour problems and attenuates theassociation of Deprived City Areas and behaviourproblems (Table 2; Model B). The odds ofexternalising behaviour problems among childrenliving in Deprived City Areas remains substantialhowever at around 1.5 times that of the averagechild. The area classification term contributesnegligibly to the fit of the model and is highlystatistically insignificant (w2 ¼ 7:47, p-value ¼ 0:59).The model terms for the structure and number ofearners in the family do not predict much systematicadvantage to having two natural parents, nor tohaving mother in employment. In comparison tochildren from intact families with a nonworkingmother, children in lone mother families have ahigher odds of externalising behaviour problems ifthe mother does not work. Behaviour problemswere related, however, to adversity in the child’sfamily background with children who live in socialhousing or whose cohort member parent became a

parent while still a teenager showing higher reportedlevels. Higher levels of parental education arestrongly associated with lower levels of childbehaviour problems. The number of older siblings(birth order) has the opposite association withbehaviour problems. The term for child genderremains statistically significant.

The addition of the cognitive stimulation sub-scale from the HOME inventory and the cohortmember’s malaise score to the model removes thestatistically significant association between resi-dence in a ward described as Middling Britain andbehaviour problems (Table 2; Model C). Otherwisethere is little change in sign or statistical significancein the coefficients of the area type categories incomparison to the previous model. The term forresidence in a Deprived City Area remains statisti-cally significant and as important as most of thefamily factors in predicting higher levels of beha-viour problems. The cohort member’s malaise scoreis positively related to the odds of behaviouralproblems. Children’s chances of developing beha-viour problems depend in a significant way on thepsychological health of their parents. The quality ofthe home environment, its opportunities for learn-ing and the physical condition of the home, isnegatively associated with children’s behaviouralproblems. Children who are growing up in such acognitively deprived family are disadvantaged intheir behavioural development at all levels ofmaterial disadvantage and socioeconomic status.

The results for children’s cognitive test scoresin Table 3 show that the area classificationterm is highly statistically significant (w2 ¼ 72:41,p-value ¼o0:01). Children living in neighbour-hoods classified as Deprived City Areas, Industrialand Manufacturing Towns and Middling Britainhave lower PPVT test scores than average control-ling for child age and gender and children living inneighbourhoods classified as Established OwnerOccupiers and Prosperous Areas have higher testscores than average (Table 3; Model A). Thecoefficients of the neighbourhood classification areof a modest order of magnitude: children living inDeprived City Areas have PPVT scores 4.1 percen-tages points lower in comparison to the averagechild. This is not large in comparison with the age-adjusted standard deviation of PPVT at 7.9percentage points. Child age is as expected posi-tively related to test scores. The estimates of therandom part of the model suggests that families dodiffer in their average PPVT scores and that there is

ARTICLE IN PRESS

Table 3

Regression coefficients and t-statistics of factors predicting child PPVT test scores (children age 4 years and older)

Model A Model B Model C

ba t b t b t

Constant �1.45 �9.65 �1.58 �9.68 �1.80 �10.90

Girl �0.02 �0.67 �0.02 �0.56 �0.05 �1.39

Age (years) 1.11 32.31 1.11 31.50 1.10 31.46

Age-squared/100 �3.43 �18.75 �3.48 �17.52 �3.41 �17.35

Area classification

Deprived city areas �0.41 �5.94 �0.28 �4.18 �0.25 �3.85

Established owner occupiers 0.18 3.00 0.09 1.60 0.08 1.36

Industrial and manufacturing towns �0.19 �3.33 �0.10 �1.82 �0.10 �1.88

Lower status owner occupiers �0.13 �1.82 �0.14 �2.02 �0.12 �1.88

Mature populations 0.12 1.47 0.05 0.62 0.03 0.37

Middling Britain �0.15 �2.88 �0.13 �2.64 �0.11 �2.23

Prosperous areas 0.39 3.83 0.26 2.65 0.25 2.61

Rural areas 0.12 0.97 0.20 1.71 0.17 1.46

Rural fringe 0.03 0.45 0.00 0.03 �0.00 �0.04

Suburbia 0.03 0.67 0.05 0.92 0.07 1.37


CSE grade C or equivalent 0.22 4.06 0.17 3.11

A level or equivalent 0.23 4.07 0.18 3.31

Degree or equivalent 0.43 6.71 0.35 5.52


Intact family and working mother 0.09 1.65 0.08 1.62


Step family and working mother �0.12 �1.35 �0.10 �1.14

Lone mother and nonworking 0.01 0.05 0.05 0.41

Lone mother and working 0.22 1.95 0.26 2.31


Birth order II �0.11 �2.68 �0.09 �2.19

Birth order III or later �0.31 �5.29 �0.27 �4.62

Social housing �0.25 �4.64 �0.20 �3.73

Cohort member’s age first birtho20 years 0.03 0.32 0.05 0.65

Cohort member’s malaise score �0.00 �0.15

HOME: cognitive stimulation 0.01 7.33

s2j 0.48 18.53 0.43 16.06 0.40 14.96

s2ij 0.61 34.83 0.60 34.91 0.60 35.18

w2 ONS area classificationb 72.41 (o0.01) 37.63 (o0.01) 33.20 (o0.01)

aAll coefficients are � 10.bLog-likelihood ratio significance test of ONS area classification term (9 degrees of freedom, p-value in parentheses).


even more variation among children within families.The variance component within families (eij) inModel A is around twice the magnitude of thevariance component between families (kj).

The introduction of the family-level characteris-tics to the model describing children’s PPVT testscores removes the statistically significant associa-tion between the neighbourhood categories Estab-lished Owner Occupiers and Industrial andManufacturing Towns and test scores (Table 3;Model B). The coefficients of the other significantneighbourhood types are somewhat attenuated but

the likelihood-ratio test shows that the term remainshighly statistically significant. Higher levels ofparental education are strongly associated withhigher test scores. The number of older siblings(birth order) has the opposite association with testscores. The social housing coefficient shows thatchildren from deprived circumstances had lower testscores compared to children from more affluentbackgrounds. In comparison to children in intactfamilies with a nonworking mother children in lone-mother families had higher test scores if the motheris in employment.

ARTICLE IN PRESS

Table 4

Regression coefficients and t-statistics of factors predicting HOME cognitive stimulation scores and cohort member’s malaise score

(N ¼ 1629)

Model D

HOME: cognitive stimulation Cohort member’s malaise score

b t b t

Constant 46.30 7.32 2.89 3.99

Age (years) 2.53 1.78 �0.10 �0.60

Age-squared/100 �14.06 �1.75 1.12 1.22

Area classification

Deprived city areas �3.95 �1.76 0.52 2.02

Established owner occupiers 3.36 1.68 0.57 2.49

Industrial and manufacturing towns �0.17 �0.09 0.09 0.39

Lower status owner occupiers �3.41 �1.51 0.75 2.90

Mature populations 4.60 1.70 0.02 0.06

Middling Britain �4.40 �2.68 0.09 0.50

Prosperous areas 1.18 0.35 �0.26 �0.67

Rural areas 6.74 1.75 �1.59 �3.59

Rural fringe 0.72 0.32 �0.21 �0.82

Suburbia �4.69 �2.80 0.01 0.08


CSE grade C or equivalent 10.41 5.69 �0.80 �3.81

A level or equivalent 9.04 4.76 �0.73 �3.33

Degree or equivalent 14.87 6.72 �1.26 �4.97


Intact family and working mother 1.41 0.81 �0.39 �1.92


Step family and working mother �2.70 �0.84 0.24 0.66

Lone mother and nonworking �8.38 �2.10 1.49 3.25

Lone mother and working �8.20 �2.12 �0.09 �0.20


Birth order II �4.07 �2.50 0.16 0.87

Birth order III or later �9.39 �4.28 0.76 3.03

Social housing �10.00 �5.61 0.30 1.45

Cohort member’s age first birth o20 years �5.15 �1.80 0.74 2.25

R-squared 0.16 0.13

w2 ONS area classificationa 24.77 (o0.01) 27.42 (o0.01)

aLog-likelihood ratio significance test of ONS area classification term (9 degrees of freedom, p-value in parentheses).

A. McCulloch / Social Science & Medicine 62 (2006) 1865–1879 1875

There is little change in sign or statisticalsignificance in the coefficients of the neighbourhoodtype categories in comparison to the previous modelwhen the cognitive stimulation sub-scale from theHOME inventory and the cohort member’s malaisescore are added to the model (Model C). The qualityof the home environment is positively associatedwith PPVT test scores. Children’s PPVT test scoreswere, however, independent of the psychologicalhealth of the cohort member parent given thecontrols for material disadvantage and socioeco-nomic status.

To test the hypothesis that area characteristicsand family socioeconomic status operate through

psychological as well as family interaction dimen-sions, the HOME cognitive stimulation sub-scaleand the cohort member parent’s malaise score wereregressed on neighbourhood and child and familyvariables. Table 4 presents the results. The percen-tage of variance accounted for by the models was16% for the HOME score and 13% for the malaisescore, respectively. Consistent with expectations,parents who were more highly educated providedbetter quality home environments and had lowermalaise scores. Moreover, as shown in Table 2 theregression controlling for the home environ-ment and malaise score resulted in a reduction ofthe effect of highest educational qualification on


children’s behaviour problems. A likelihood-ratiotest of the education term in Model B gives a w2

of 7.88 (p-value 0.05), compared to a w2 of 1.88(p-value 0.60) for the education term in Model C.Both the home environment and malaise score weresignificant predictors of children’s behaviour pro-blems and the results therefore suggest that thehome environment and malaise score may act asmediators of the association between highesteducational qualification and behaviour problems.Residence in different types of neighbourhood wasstrongly associated with variations in the cohortmember’s malaise score. Neighbourhood type wasassociated with higher malaise scores net of all othereffects in neighbourhoods classified as DeprivedCity Areas, Established Owner Occupiers andLower Status Owner Occupiers and with lowermalaise scores in neighbourhoods classified as RuralAreas. Children’s home environments were, how-ever, only weekly associated with neighbourhoodtype. Neighbourhood type was only associated withless cognitively stimulating homes net of all othereffects in neighbourhoods classified as Lower StatusOwner Occupiers and more cognitively stimulatinghomes in neighbourhoods classified as MaturePopulations. Moreover, as shown by the like-lihood-ratio tests in Tables 2 and 3, the introductionof the home environment and malaise score hadlittle effect on the significance of the neighbourhoodcategories in explaining children’s outcomes. Ittherefore appears that the parent’s psychologicalhealth and home environment only play a limitedrole in mediating the impact of neighbourhoodcharacteristics on children’s outcomes.

The remaining factors that contributed uniquelyto HOME scores were residence in social housingand living in a lone mother family. Children wholived in poor quality accommodation or in a lone-mother family had less supportive home environ-ments. The birth order of the child made a modestunique contribution to predicting HOME scores.Children with lower birth orders children appear tohave more cognitively stimulating home environ-ments than do children with higher birth orders.The remaining factors that contributed uniquely tomalaise scores were living in either a lone-motherfamily or a step-family where the mother does notwork or in an intact family where the mother works.Mothers who go out to work had lower malaisescores. Cohort members who had their first childwhile still a teenager reported higher malaise scores.Birth order also shows a relationship to malaise

scores with children of birth order 3 or higher,indicating a large family size, more likely to have aparent with a high malaise score.

Discussion

This paper has examined the relationships be-tween children’s externalising behavioural pro-blems, scores on a cognitive test and aclassification of the types of neighbourhood inwhich they live using data collected from childrenborn to a cohort of young parents in the UK. Afteradjustment for child and family characteristicsdifferences in children’s behavioural adjustmentbetween different types of place was restricted tohigher levels of behaviour problems in neighbour-hoods described as Deprived City Areas. The higherrisk of behaviour problems among children living inDeprived City Areas can be judged relativelysignificant: the odds of externalising behaviourproblems among children living in Deprived CityAreas was 1.6 times that of the average child.Children living in a range of neighbourhood types(Deprived City Areas, Lower Status Owner Occu-pier, Middling Britain) had lower PPVT test scoresand children in Prosperous Areas had higher PPVTtest scores. Neighbourhood type was of lessimportance in predicting children’s test scores thanindicators of material disadvantage such as socialhousing. Both the size and statistical significance ofcoefficients on the neighbourhood categories wereusually smaller than those of family-level measures.

In contrast to McCulloch and Joshi’s (2001)findings, this analysis finds some evidence fordifferences in children’s behavioural outcomesbetween neighbourhoods, when adjusted for differ-ential socioeconomic composition at the child andfamily level. This is likely to be due mainly to thegreater informativeness of the ONS classificationthan the composite one-dimensional deprivationindex used by McCulloch and Joshi (2001). Theadvantages of the ward classification include abroader view of social circumstances than simplydeprivation and the ability to identify characteristictypes of neighbourhoods. It has been argued thatconceptualising deprivation as a unidimensionalvariable neglects the multidimensional nature ofthe problem both for individuals and between areas.Indeed, there are a range of potential factors thatmight be used to describe a local neighbourhood,with certain factors having more significance forsome residents than others. Whichever factors are


chosen no two neighbourhoods are likely to haveexactly the same characteristics. It would also seemreasonable to suggest that every deprivation mea-sure has a different meaning in different contexts.The association between children’s outcomes andsocioeconomic disadvantage measured at the arealevel might vary across neighbourhoods. This wouldseem reasonable given that the meaning of theconstituent elements of an index may be expected tovary across different types of neighbourhood, aswith the car ownership component of the Townsendindex.

In our analyses, both HOME scores and theparent’s malaise score mediated the effects of cohortmember educational qualifications on children’sbehavioural problems. Families with children withmore educated cohort member parents were asso-ciated with better home environments and lowermalaise scores, which in turn are associated withlower levels of behavioural problems. The media-tion effect is not due to either the HOME scores ormalaise score being correlated with other familyfactors, such as maternal education or teenageparenthood. Education may provide parents withthe knowledge, skills and self-confidence necessaryfor happy marital relationships and successful childrearing. On the other hand lack of educationalqualifications may also be a more general indicatorof social disadvantage. The quality of the homeenvironment and the parent’s psychological healthdid not mediate the association between the type ofneighbourhood and children’s outcomes. Althoughthe analyses provide some insight into the mechan-isms through which parental education mightinfluence children’s outcomes they give little indica-tion of the mechanisms through which the neigh-bourhood influences children’s outcomes.

The results point to residence in a Deprived CityArea as having a particularly negative impact onchildren’s cognitive and behavioural development.The study provides little insight into the processesunderlying this association. It is likely, however,that the increased prevalence of behaviour problemsis due to environmental factors in urban environ-ments which cause or prolong behaviour problems(McArdle, O’Brien, & Kolvin, 1995). This effect is,at least in part, specific for externalising behaviourproblems because associations with other outcomeswere much weaker. For this reason, a generic effectof stress is unlikely to explain the entire associationbetween externalising behaviour problems andresidence in deprived urban areas. Stress due to

environmental pollutants such as noise have beensuggested (Haines, Stansfeld, Job, Berglund, &Head, 2001), but most attention has been paid tosocial factors. Qualitative and quantitative studiesof urban communities have revealed that safety is anoverriding concern. It factors strongly into adults’decisions about the breadth and frequency of theirown social interactions and about whether to permityouth to spend time in educational activities outsideof school or during evening hours. Not surprisingly,children’s exposure to community violence isassociated with children’s developmental outcomes.Furthermore, there is evidence that children whobehave aggressively with peers often have beenvictims of some form of violence themselves(Margolin & Gordis, 2000). It is likely that byobserving adults using coercive strategies to solvetheir problems, children develop the view thataggression is an acceptable approach to solvingdisagreements. Association with deviant peers andconflict in the parent–child relationship arising, forexample, from parental attempts to restrict chil-dren’s activities in neighbourhoods characterised bycrime and disorder are other examples of pathwaysby which residence in a Deprived City Areamay influence children’s externalising behaviourproblems.

This study has several limitations. Most signifi-cantly we are unable to distinguish nonrandomselection processes from genuine neighbourhoodeffects. Sociologists have argued that the family-level variables often used in neighbourhood studiesdo not control for unmeasured differences betweenfamilies who live in one neighbourhood versusanother thus neighbourhood effects may simply be areflection of the characteristics of the families wholive in certain neighbourhoods (Duncan, Connell, &Klebanov, 1997; Tienda, 1991). This view issupported by studies of population trends whichsuggest that many people find urban areas unat-tractive places to live (Atkins, Champion, Coombes,Dorling, & Woodward, 1996). Conversely, it hasrecently been suggested that the impact of familycharacteristics such as living in a single-parentfamily on children’s outcomes might be capturingaspects of the poor neighbourhood environments towhich children from such families are exposed(Crowder & Teachman, 2004). Further researchwith longitudinal designs is clearly required todetermine the extent to which the effects of familycharacteristics and neighbourhood characteristicsare confounded. It also needs to be emphasised that


these findings are based upon the mother’s percep-tions of children’s behaviour. Rutter et al. (1970)have reported a relatively low correlation betweenteachers’ and parents’ ratings of the same child(r ¼ 0:5). McGee, Silva, and Williams (1983) andCundall (1987) found that greater psychologicalsymptoms among mothers were associated withidentification of disturbance among children basedon mothers’ but not teachers’ ratings. The relation-ship between increasing behavioural ratings andparental psychological distress in the NCDS musttherefore be interpreted with some caution.

Another limitation of the analyses is that theneighbourhood classification only indirectly cap-tures neighbourhood characteristics that are con-sidered most relevant to neighbourhood effects onchildren. The potential remains for examiningadditional area characteristics more pertinent tochildren, such as quality of health services or schoolquality. A major disadvantage lies in providinglimited information on the depth and quality ofsocial relationships. Differences in motivation,emotional health, self-efficacy, literacy and so oncould account for any observed neighbourhoodeffects (Duncan et al., 1997). The boundaries oflocal communities also do not necessarily coincidewith those of wards. As far as teenagers areconcerned the electoral ward may be a poorsubstitute for measurements on school and peergroups. Given these concerns, we must be cautiousabout drawing firm conclusions regarding the sizeand nature of neighbourhood effects. However,despite these limitations the study provides usefulinformation on the variation in children’s outcomesacross different types of neighbourhood whichearlier studies with smaller or less geographicallyrepresentative samples could not provide.

In summary, this paper is one of the relatively fewto combine area statistics from the 1991 census withindividual-level information, and is also one of thefew data sources containing detailed nationalinformation on children’s cognitive and behaviouraloutcomes. Both the kinds of neighbourhoods inwhich children live and the characteristics of thefamily were found to have important influences onchildren’s outcomes. For children’s cognitive out-comes, both the size and statistical significance ofcoefficients on the neighbourhood measures wereusually smaller than those of family-level measures.In contrast, residence in a Deprived City Area wasas significant as the family factors in predictinghigher levels of behaviour problems. Since the data

used in this article was collected a range of policiesdesigned to improve the circumstances of the leastwell-off families and children have been introduced.These include the introduction of Sure Start, anarea-based initiative which aims to improve thehealth and well-being of families and children frombefore birth to 4 years of age, so that children areready to learn when they go to school. The results inthis paper should help future studies assess theimpact that such policies have had on the relation-ship between neighbourhood factors and children’soutcomes, and whether this relationship has chan-ged over time more generally.

References

Aber, J. L., Bennett, N. G., Conley, D. C., & Li, J. (1997). The

effect of poverty on child health and development. Annual

Review of Public Health, 18, 463–483.

Atkins, D., Champion, T., Coombes, M. G., Dorling, D., &

Woodward, R. (Eds.). (1996). Urban trends in England: Latest

evidence from the 1991 census. London: HMSO.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator

variable distinction in social psychological research: Con-

ceptual, strategic and statistical considerations. Journal of

Personality and Social Psychology, 51, 1173–1182.

Brooks-Gunn, J., Duncan, G., & Aber, J. L. (Eds.). (1997).

Neighbourhood poverty: Context and consequences for devel-

opment. New York: Russell Sage Foundation.

Brooks-Gunn, J., Duncan, G. J., Klebanov, P. K., & Sealand, N.

(1993). Do neighbourhoods affect child and adolescent

development? American Journal of Sociology, 99, 353–395.

Chandola, T. (2001). The fear of crime and area differences in

health. Health and Place, 7, 105–116.

Coleman, J. (1988). Social capital in the creation of human

capital. American Journal of Sociology, 94(Suppl.), S95–S120.

Conger, R., Conger, K., & Elder, G. (1997). Family economic

hardship and adolescent adjustment: Mediating and moder-

ating processes. In G. Duncan, & J. Brooks-Gunn (Eds.),

Consequences of growing up poor (pp. 288–310). New York:

Russell Sage Foundation.

Crowder, K., & Teachman, J. (2004). Do residential conditions

explain the relationship between living arrangements and

adolescent behavior. Journal of Marriage and the Family, 66,

721–738.

Cundall, D. B. (1987). Children and mothers at clinics: Who is

disturbed? Archives of Disease in Childhood, 62, 820–824.

Davies, H., Joshi, H., & Clarke, L. (1997). Is it cash the deprived

are short of ? Journal of the Royal Statistical Society, Series A,

160, 107–126.

Drever, F., & Whitehead, M. (1995). Mortality in regions and

local authority districts in the 1990’s: Exploring the relation-

ship with deprivation. Population Trends, 82, 19–26.

Duncan, G. J., Connell, J. P., & Klebanov, P. K. (1997).

Conceptual and methodological issues in estimating causal

effects of neighbourhoods and family conditions on individual

development. In J. Brooks-Gunn, G. Duncan, & J. L. Aber

(Eds.), Neighbourhood poverty: Context and consequences for


development (pp. 219–250). New York: Russell Sage Founda-

tion.

Dunn, L., & Markwardt, C. (1970). Peabody individual

achievement test manual. Circle Pine, Minnesota: American

Guideance Service.

Ecob, R., & Jones, K. (1998). Mortality variations in England

and Wales between types of place: An analysis of the

ONS Longitudinal study. Social Science & Medicine, 47,

2055–2066.

Ferri, E. (1993). Life at 33: The fifth follow-up of the National

Child Development Study. London: National Childrens

Bureau.

Furstenberg, F. (1993). How families manage risk and opportu-

nity in dangerous neighborhoods. In W. J. Wilson (Ed.),

Sociology and the public agenda (pp. 231–258). Newbury Park,

CA: Sage Publications.

Gerard, J. M., & Buehler, C. (1999). Multiple risk factors in the

family environment and youth problem behaviors. Journal of

Marriage and the Family, 61, 343–361.

Haines, M. M., Stansfeld, S. A., Job, R. F. S., Berglund, B., &

Head, J. (2001). Chronic aircraft noise exposure, stress

responses, mental health and cognitive performance in school

children. Psychological Medicine, 31, 265–277.

Jarrett, R. L. (1997). African American family and parenting

strategies in impoverished neighborhoods. Qualitative Sociol-

ogy, 20, 275–287.

Jencks, C., & Mayer, S. (1990). The social consequences of

growing up in a poor neighbourhood. In L. E. Lynn Jr., & M.

G. McCeary (Eds.), Inner-city poverty in the United States

(pp. 111–186). Washington, DC: National Academy Press.

Joshi, H., Cooksey, E., Wiggins, R. D., McCulloch, A.,

Verropoulou, G., & Clarke, L. (1999). Diverse family living

situations and child development: A multi-level analysis

comparing longitudinal evidence from Britain and the United

States. International Journal of Law, Policy and the Family, 13,

292–314.

Klebanov, P. K., Brooks-Gunn, J., Chase-Landsdale, P. L., &

Gordon, R. A. (1997). Are neighbourhood effects on young

children mediated by features of the home environment? In J.

Brooks-Gunn, G. Duncan, & J. L. Aber (Eds.), Neighbour-

hood poverty: Context and consequences for development

(pp. 119–145). New York: Russell Sage Foundation.

Klebanov, P. K., Brooks-Gunn, J., McCarton, C., & McCor-

mick, M. C. (1998). The contribution of neighbourhood and

family income to developmental test scores over the first three

years of life. Child Development, 69, 1420–1436.

Margolin, G., & Gordis, E. (2000). The effects of family and

community violence on children. Annual Review of Psychol-

ogy, 51, 445–479.

McArdle, P., O’Brien, C., & Kolvin, I. (1995). Hyperactivity,

prevalence and relationship with conduct disorders. Journal of

Child Psychology and Psychiatry, 36, 279–304.

McCulloch, A., & Joshi, H. (2001). Neighbourhood and

family influences on the cognitive ability of children in the

British National Child Development Study. Social Science &

Medicine, 53, 579–591.

McCulloch, A., & Joshi, H. (2002). Child development and

family resources: An exploration of evidence from the second

generation of the 1958 British birth cohort. Journal of

Population Economics, 15, 283–304.

McCulloch, A., Wiggins, R. D., Joshi, H. E., & Sachdev, D.

(2000). Internalising and externalising children’s behaviour

problems in Britain and the US: Relationships to family

resources. Children and Society, 14, 368–383.

McGee, R., Silva, P. A., & Williams, S. (1983). Parents’ and

teachers’ perceptions of behaviour problems in seven year old

children. Exceptional Children, 30, 151–161.

Meltzer, H., Gatward, R., Goodman, R., & Ford, T. (2000).

Mental health of children and adolescents in Great Britain.

London: The Stationary Office.

Mirowsky, J. (1999). Analyzing associations between social

circumstances and mental health. In C. S. Aneshensel, & J.

C. Phelan (Eds.), Handbook of the sociology of mental health

(pp. 105–123). New York: Kluwer.

Ross, C. (2000). Neighborhood disadvantage and adult depres-

sion. Journal of Health and Social Behavior, 41, 177–187.

Rutter, M. (1967). A children’s behaviour questionnaire for

completion by teachers: Preliminary findings. Journal Child

Psychology and Psychiatry, 8, 1–11.

Rutter, M., Cox, A., Tupling, C., Berger, M., & Yule, W. (1975).

Attainment and adjustment in two geographical areas: I—the

prevalence of psychiatric disorder. British Journal of Psychia-

try, 126, 493–509.

Rutter, M., Tizard, J., & Whitmore, K. (1970). Education, health

and behaviour. London: Longmans.

Rutter, M., Yule, B., Quinton, D., Rowlands, O., Yule, W., &

Berger, M. (1975). Attainment and adjustment in two

geographical areas: III—some factors accounting for area

differences. British Journal of Psychiatry, 126, 520–533.

Sampson, R. J., Morenoff, J. D., & Earls, F. (1999). Beyond

social capital: Spatial dynamics of collective efficacy for

children. American Sociological Review, 64, 633–660.

Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997).

Neighbourhoods and violent crime: A multilevel study of

collective efficacy. Science, 277, 918–924.

Smith, J. R., Brooks-Gunn, J., & Klebanov, P. K. (1997).

Consequences of living in poverty for young children’s

cognitive and verbal ability and early school achievement. In

G. Duncan, & J. Brooks-Gunn (Eds.), Consequences of growing

up poor (pp. 132–189). New York: Russell Sage Foundation.

Sugland, B., Zaslow, M., Smith, J., Brooks-Gunn, J., Coates, D.,

Blumenthal, C., et al. (1995). The early childhood HOME

Inventory and HOME Short Form in differing racial/ethnic

groups: Are there differences in underlying structure, internal

consistency of subscales, and patterns of prediction. Journal

of Family Issues, 16, 632–663.

Thompson, M. J., Stevenson, J., Sonuga-Barke, E., Nott, P.,

Bhatti, Z., Price, A., et al. (1996). Mental health of preschool

children and their mothers in a mixed urban/rural population.

I. Prevalence and Ecological Factors, 168, 16–20.

Tienda, M. (1991). Poor people and poor places: Deciphering

neighbourhood effects on poverty outcomes. In J. Huber

(Ed.), Macro-micro linkages in sociology (pp. 244–267).

Thousand Oaks, California: Sage.

Townsend, P., Phillimore, P., & Beattie, A. (1989). Health and

deprivation. London: Routledge.

Wallace, M., Charlton, J., & Denham, C. (1995). OPCS area

classifications. Population Trends, 79, 15–30.

Wilson, W. J. (1996). When work disappears: The world of the new

urban poor. New York: Vintage Books.

Variation in children's cognitive and behavioural adjustment between different types of place in the...

Documents

Transcript of Variation in children's cognitive and behavioural adjustment between different types of place in the...