Family and Neighborhood Socioeconomic Status and Temperament Development from Childhood to Adolescence

Abstract

Objective:

Children differ in their temperament and these differences predict consequential outcomes, including mental health, peer relations, substance use, academic performance, and adult personality. Additionally, children’s temperament develops over time in response to environmental factors, such as the socioeconomic status (SES) of their family and the neighborhood in which they are raised. However, there has been lack on research on the relation between neighborhood SES and the development of temperament or personality.

Method:

Using data from two cohorts of the Longitudinal Study of Australian Children (LSAC; N = 9,217) that followed children from 4–15 years old, the present analyses examined whether parent income, parent education, and neighborhood disadvantage were associated with three child temperament traits that are precursors to five-factor model (FFM) adult personality traits.

Results:

Longitudinal hierarchical linear models (HLM) generally found that children with lower neighborhood SES or family SES tended to have lower sociability, higher reactivity, and lower persistence and these associations did not decrease over time.

Conclusions:

This research demonstrates that both neighborhood and family SES in childhood are important for the development of temperament across childhood and adolescence.

Children differ in their general temperament and these differences predict a wide variety of consequential outcomes, such as mental health (e.g., Lewis & Olsson, 2011), peer relations (e.g., Sanson et al., 2004), substance use (e.g., Williams, Sanson, Toumbourou, & Smart, 2000), academic performance (e.g., Schoen & Nagle, 1994), and adult personality (see Prior, Sanson, Smart, & Oberklaid, 2000). Additionally, children’s temperament develops over time in response to both genetic and environmental factors (see Smart & Sanson, 2005). One important environmental factor that may contribute to temperament development is the socioeconomic status (SES) of the family and neighborhood in which children are raised.

Temperament and Personality

Child temperament commonly refers to relatively stable individual differences in children’s attentional, emotional, and behavioral reactivity and self-regulation (Prior et al., 2000; Sanson & Oberklaid, 2013). Child temperament is strongly associated with adult personality but is traditionally thought to be more directly linked to biology and to have a narrower range of attributes than adult personality (De Pauw, 2017; Prior et al., 2000). Both child temperament and adult personality are composed of a vast array of different traits. As an organizing framework for these traits, most research on adult personality uses the five-factor model (FFM; also known as the Big Five), which clusters closely-related traits into five over-arching factors: neuroticism, extraversion, openness, agreeableness, and conscientiousness (McCrae & John, 1992). For child temperament there is less consensus on a single organizing framework, but one set of commonly-investigated temperament traits is sociability, reactivity, and persistence (Sanson & Oberklaid, 2013; Smart & Sanson, 2005).

Sociability is the degree to which a child is comfortable with and approaches new people and new situations as opposed to being shy or reserved (Sanson & Oberklaid, 2013; Smart & Sanson, 2005). Children high in sociability are less likely to experience internalizing disorders like anxiety and depression (Letcher et al., 2004; Lewis & Olsson, 2011) and have better social skills and fewer peer problems (Sanson, Hemphill, & Smart, 2004; Smart & Sanson, 2005), but they tend to be at higher risk for engaging in substance use in adolescence (Prior et al., 2000; Williams et al., 2000). Children’s sociability is associated with adult extraversion (Prior et al., 2000), which encompasses warmth, gregariousness, assertiveness, activity, excitement-seeking, and positive emotions (McCrae & John, 1992). Higher extraversion predicts a wide variety of positive outcomes such as mental health and well-being (Strickhouser, Zell, & Krizan, 2017), life-satisfaction (Duckworth, Weir, Tsukayama, & Kwok, 2012), and relationship satisfaction (Malouff, Thorsteinsson, Schutte, Bhullar, & Rooke, 2010).

Reactivity is the degree to which a child has intense and volatile emotional responses to challenges or frustration as opposed to being flexible and easy-going (Sanson & Oberklaid, 2013; Smart & Sanson, 2005). Children high in reactivity are at greater risk for poor outcomes including externalizing disorders (Sanson & Prior, 1999; Smart & Sanson, 2005), internalizing disorders (Letcher et al., 2004; Lewis & Olsson, 2011), eating disorders (Martin et al., 2000), adolescent substance use (Prior et al., 2000; Williams et al., 2000), as well as poor social skills and peer problems (Sanson et al., 2004; Smart & Sanson, 2005). Children’s reactivity is associated with adult neuroticism (Prior et al., 2000), which encompasses anxiety, hostility, depression, self-consciousness, impulsiveness, and vulnerability (McCrae & John, 1992). Higher neuroticism predicts a wide variety of negative outcomes such as worse mental health and well-being (Strickhouser et al., 2017), less life-satisfaction (Duckworth et al., 2012), and less relationship satisfaction (Malouff et al., 2010).

Persistence is the degree to which a child can maintain attention and stay focused on tasks until completion as opposed to giving up or being distracted (Sanson & Oberklaid, 2013; Smart & Sanson, 2005). Children high in persistence are less likely to experience externalizing disorders (Sanson & Prior, 1999 ; Smart & Sanson, 2005) or eating disorders (Martin et al., 2000), are less likely to engage in substance use in adolescence (Prior et al., 2000; Williams et al., 2000), have better social skills and fewer peer problems (Sanson et al., 2004; Smart & Sanson, 2005), and have better reading and mathematical skills (Schoen & Nagle, 1994). Children’s persistence is associated with adult conscientiousness (Prior et al., 2000), which encompasses competence, order, dutifulness, achievement-striving, self-discipline, and deliberation (McCrae & John, 1992). Higher conscientiousness predicts to a wide variety of positive outcomes such as physical and mental health (Strickhouser et al., 2017) income and life-satisfaction (Duckworth et al., 2012), academic achievement (Schneider & Preckel, 2017), and relationship satisfaction (Malouff et al., 2010).

Although temperament and personality are most often investigated as stable predictors of future outcomes, they develop and change across the lifespan in response to both genetic and environmental factors (see Caspi, Roberts, & Shiner, 2005; Smart & Sanson, 2005). Behavioral genetics research indicates that approximately 50% of the variance in temperament and personality can attributed to variance in genetics, implying that the remaining 50% may be attributed to the environment (e.g., Saudino, 2005; van den Berg et al., 2014). A variety of environmental factors have been shown to contribute to the development of children’s temperament. For instance, maternal smoking during pregnancy is associated with children who have lower sociability, lower persistence, and higher reactivity and these associations remain stable throughout childhood (Sutin, Flynn, & Terracciano, 2018). Also, a warm and nurturing parenting style predicts decreases in shyness over time when compared to other parenting styles (Sanson & Oberklaid, 2013) and interventions to teach a more responsive parenting style help children to become more cooperative 2-3 years later (van den Boom, 1995). Additionally, exposure to more screen time or less participation in sports predicts decreases in children’s sociability and persistence over two years (Allen, Vella, & Laborde, 2015). Other factors, such as the socioeconomic status of the family and neighborhood in which the child is raised, may also contribute to how temperament develops across childhood.

Socioeconomic Status

SES is a broad construct that represents the relative ranking of a person or group’s social and economic resources compared to others in their social system. A wide variety of factors are used as rough indicators of SES, but the most common SES indicators are income, education, and occupational prestige (see Baker, 2014). Researchers sometimes investigate the correlates of each of these indicators individually and other times combine multiple indicators into a single composite index of overall SES (e.g., Blakemore, Strazdins, & Gibbings, 2009).

SES is most often examined at the individual or family level, but the overall SES of the neighborhood that a family lives in may be just as important (see Diez Roux & Mair, 2010; Leventhal & Brooks-Gunn, 2000). Many opportunities and resources available to higher SES children are provided by the neighborhood rather than the family. For example, higher SES neighborhoods may provide a wider variety of material and social benefits through schools, recreational facilities, local shopping, public transportation, and availability of jobs. Children living in lower SES neighborhoods, in contrast, are often exposed to more violence, drug abuse, and other criminal activity (e.g., Chong, Lee, & Victorino, 2015; Karriker-Jaffe, 2013; Odgers et al., 2009). This exposure likely increases children’s chronic stress and anxiety as well as increasing the risk that they will join their peers in those antisocial activities.

Childhood SES is strongly related to a variety of consequential life outcomes. Children in higher SES families, for example, complete more years of education, earn higher income, and hold more prestigious occupations as adults (e.g., Deary et al., 2005; Furnham & Cheng, 2013; Hertz et al. 2007; McGue, Rustichini, & Iacono, 2017). People who grow up in lower SES families are more likely to experience problems with physical health or cognitive functioning in childhood (Currie, 2009) and these increased risks persist into older adulthood (Luo & Waite, 2005). Additionally, children and adolescents from lower SES families report worse self-rated health, mental well-being, and health-related quality of life (Currie et al., 2008; von Reuden, Gosch, Rajmil, Bisegger, & Ravens-Sieberer, 2006).

The SES of the neighborhood that children grow up in also predicts numerous consequential life outcomes, above and beyond family SES. Higher neighborhood SES in childhood predicts more years of education and higher income in adulthood, controlling for family SES (Chetty & Hendren, 2018; Chetty, Hendren, & Katz, 2016). Independent of family SES, children born in lower SES neighborhoods are at greater risk of low birthweight and infant mortality (see Pickett & Pearl, 2001; Leventhal & Brooks-Gunn, 2000), as well as being at greater risk of poor overall physical health and susceptibility to illness (Drukker, Kaplan, Feron, & van Os, 2003) and worse self-rated health (Torsheim et al., 2004). Adolescents in lower SES neighborhoods are also more likely to engage in sexual activity, are less likely to use birth control, and have higher rates of nonmarital childbearing, controlling for family SES (see Leventhal & Brooks-Gunn, 2000).

Family and neighborhood SES in childhood also predict mental health outcomes. Children are more likely to experience internalizing and externalizing mental health problems if they have lower family SES (Reiss, 2013) or lower neighborhood SES (Leventhal & Brooks-Gunn, 2000). Furthermore, when family and neighborhood SES are examined simultaneously, they each have independent associations with better mental health in children (e.g., Priest, Baxter, & Hayes, 2012), and longitudinal studies find that these independent associations persist over time (e.g., Christensen, Fahey, Giallo, & Hancock, 2017). These associations cannot be explained solely by shared genetics, as a twin study showed that genetic factors only explained 55% of the variance in children’s mental health outcomes, while shared environment factors accounted for 20% of the variance and neighborhood deprivation accounted for 5% of that variance (Caspi, Taylor, Moffitt, & Plomin, 2000). Finally, an experiment that randomly assigned some families to move to better neighborhoods demonstrated that higher neighborhood SES had a causal impact leading to better mental health outcomes for children (Leventhal & Brooks-Gunn, 2003).

Childhood SES and Temperament/Personality

Lower family SES in childhood is associated with the development of temperament in children. For instance, lower parent income in adolescence predicts less positive emotionality, more negative emotionality, and less constraint (Conger et al., 2012). Furthermore, these traits go on to predict lower income in adulthood, which subsequently predicts a similar pattern of trait development in the next generation (Conger et al., 2012). Lower maternal education also predicts more irritability in children and lower parent income predicts more irritability, more fear, and less effortful control, and these relations remain stable over time (Lengua, 2006). Additionally, children in families with lower income are less likely to have a resilient temperament, children whose mothers had less education are more likely to have an under-controlled temperament, and these associations remain stable over time (Hart, Eisenberg, & Valiente, 2007). Finally, a recent meta-analysis (Ayoub et al., 2018) that combined data from at least nine different studies and 4,400 children found that lower parent SES was associated with higher negative emotionality and lower effortful control, but was unrelated to positive emotionality.

Two meta-analyses have also investigated the relation between family SES in childhood and the development of adult personality traits. The first meta-analysis used data from seven large-scale longitudinal samples to examine the relation between parent education and adult FFM traits controlling for a variety of demographic factors (Sutin, Luchetti, Stephan, Robins, & Terracciano, 2017). Lower parent education predicted lower extraversion, lower openness, and higher neuroticism in adulthood, but was unrelated to agreeableness or conscientiousness. Furthermore, there were also significant longitudinal trends indicating that the association with parent education became larger over time, such that lower parent education predicted relative declines in extraversion and openness as well as relative increases in neuroticism over time. The same associations were found when analyzing a subsample of adopted children, which suggests that the relation between parent education and personality cannot be explained by just shared genetics.

The second meta-analysis used data from at least 16 different studies to examine bivariate correlations between broadly-defined parent SES and adult FFM traits (Ayoub et al., 2018). As in the first meta-analysis, lower parent SES predicted lower extraversion, lower openness, and higher neuroticism and was unrelated to agreeableness. In contrast to the first meta-analysis, however, lower parent SES also significantly predicted lower conscientiousness. Additionally, the relation between parent SES and FFM traits was not significantly different between younger and older samples, which is not entirely consistent with the significant increase in association over time found in the first meta-analysis. Still, the results of both meta-analyses agree that the associations do not significantly decrease over time, suggesting that family SES in childhood has an association with personality that persists across the lifespan.

In contrast to the numerous studies investigating the relation between family SES in childhood and temperament or personality, at present there has been a lack of research on the relations with neighborhood SES. To our knowledge, only one study has investigated the relation between neighborhood SES and temperament or personality (Hart, Atkins, & Matsuba, 2008). This study found that children 3 or 4 years old who lived in neighborhoods with a high percentage of households below the federal poverty level were less likely to have a resilient temperament. Furthermore, there was a significant longitudinal association such that the risk of having a non-resilient temperament increased over time for children in higher poverty neighborhoods, controlling for family SES.

More research is needed to investigate the ways in which neighborhood SES in childhood is associated with temperament. In particular, it is important to examine the relation between neighborhood SES and the development of children’s sociability, reactivity, and persistence, because these child temperament traits are linked to the adult FFM personality traits of extraversion, neuroticism, and conscientiousness, respectively (Prior et al., 2000). Thus, any relation between neighborhood SES and sociability, reactivity, or persistence may have life-long consequences for adult personality.

Present Research

The present research examined whether SES in childhood is associated with temperament and its development over time. We investigated longitudinal changes from 4 to 15 years old in three aspects of child temperament related to adult FFM personality traits: sociability, reactivity, and persistence. We tested whether three dimensions of SES were associated with temperament development: parent income, parent education, and neighborhood disadvantage. We predicted that lower SES and decreases in SES over time would be associated with more negative temperament traits and that the association would remain stable or become stronger over time. Specifically, we predicted that children whose parents had lower income, less education, or who lived in more disadvantaged neighborhoods would be rated lower on sociability, higher on reactivity, and lower on persistence and that this discrepancy would not significantly decrease over time.

Method

Participants and Procedure

Participants were drawn from the Longitudinal Study of Australian Children (LSAC; Australian Institute of Family Studies, 2015). The LSAC is a nationally representative study of two Australian age cohorts: an older cohort (labeled K) born March 1999 – February 2000 (N = 4,983) and a younger cohort (labeled B) born March 2003 – February 2004 (N = 5,107). The first wave of data collection for both cohorts was conducted March 2004 – January 2005 and a follow-up wave has occurred every 2 years. The study was approved by the Australian Institute of Family Studies Ethics Committee and each family provided written informed consent before participating. LSAC data can be requested from the Australian Institute of Family Studies (http://growingupinaustralia.gov.au/data/dataaccessmenu.html).

The present analyses make use of the LSAC data collected when children were 4–15 years old. For the older cohort K, data from waves 1–6 was included (collected March 2004 – February 2015). For the younger cohort B, data from waves 3–7 was included (collected April 2008 – June 2017). There were 553 children in cohort B who dropped out before age 4 and are not included in these analyses. Additionally, we excluded 262 children who were missing relevant temperament data at all waves and 58 children who were missing relevant parent income (n = 5) or parent education (n=54) data at all waves. Otherwise, there were no exclusion criteria and all participants with relevant data were included in the analyzed sample. There were a total of 9,217 children included in the present analyses: 4,759 from Cohort K and 4,458 from Cohort B. Descriptive statistics for the included children are provided in Table 1.

Table 1.

Descriptive Statistics for Included Children by Cohort

	Both Cohorts	Cohort K	Cohort B
Number of children	9217	4759	4458
Female children	48.90%	48.90%	48.90%
Indigenous children	3.60%	3.60%	3.60%
At Baseline
Log10 weekly Parent income	3.00(0.20)	2.94(0.18)	3.05(0.19)
Years of parent education	15.54(2.35)	15.34(2.41)	15.75(2.28)
SEIFA-IRSD, Neighborhood disadvantage	−1012(52.7)	−1011(50.5)	−1013(54.9)
Linear Change per Year
Log10 weekly Parent income	0.016(0.009)	0.019(0.008)	0.013(0.007)
Years of parent education	0.031(0.115)	0.037(0.115)	0.025(0.115)
SEIFA-IRSD, Neighborhood disadvantage	0.078(4.174)	0.112(4.222)	0.042(4.122)

Note. Values represent a count, percent, or mean (standard deviation).

Measures

Parent income.

Parents were asked to report the usual weekly income for themselves and their partner (if applicable). In cases of non-response, attempts were made to impute income based on other demographic variables and this imputed income is provided in the LSAC data set (for method see Mullan, Daraganova, & Baker, 2015). For one-parent households we took the parent’s imputed weekly income as the value for parent income, while for two-parent households we took the highest of either parents’ imputed weekly income as the value for parent income. In two-parent households where one or both parents’ weekly income was not reported and could not be imputed, we set parent income to missing. To account for the logarithmic utility of income we transformed parent income using the function LogIncome = log₁₀(Income + 1).

Parent education.

Parents reported 2 separate education variables for themselves as well as their partner (if applicable): highest year of primary or secondary schooling completed and highest post-secondary qualification. We transformed these 2 variables into a single estimate of years of education for each parent, utilizing the same method as previous LSAC research (see Blakemore et al., 2009). For those with qualifications, our estimate of years of education was: postgraduate degree = 20, graduate diploma or bachelor degree = 17, advanced diploma or diploma = 16, and certificate or other qualification = 14. For those without qualifications, our estimate was: Year 12 completed or still in school = 13, Year 11 completed = 12, Year 10 completed = 11, Year 9 completed = 10, Year 8 or below completed = 9, Never attended school = 0. For one-parent households, we took the parent’s estimated years of education as the value for parent education; for two-parent households, we took the highest estimated years of education for either parent as the value for parent education. In two-parent households where one or both parents’ education was not reported, we set parent education to missing.

Neighborhood disadvantage.

The families’ addresses were geocoded and linked to a statistical local area or, when that was unavailable, to their postcode area. These areas were then linked to the Australian Socio-Economic Indexes for Areas (SEIFA; Pink, 2006). There are 4 different SEIFA indexes assessing various socioeconomic aspects of areas. In the present analyses, we chose the Index of Relative Socio-economic Disadvantage (IRSD), as it encompasses a broad array of indicators for disadvantaged neighborhoods.¹ These indicators include having a high percentage of residents with low income, no educational qualifications, who are unemployed or in a low-skill occupation, and living in overcrowded housing, as well as a low percentage of car ownership and home internet access. The IRSD is scored such that lower values indicate more disadvantaged areas, but for ease of interpretation in our analyses we reversed the scale so that higher neighborhood disadvantage scores indicated more disadvantaged areas.

Child temperament.

At each wave of data collection, parents completed a 12-item age-appropriate measure of their child’s sociability, reactivity, and persistence, with four Likert-type items for each trait. The items administered at 4 to 7 years old were drawn from the Short Temperament Scale for Children (STSC; Sanson, Smart, Prior, Oberklaid, & Pedlow, 1994), while the items administered at 8 years old and up were drawn from the School-Age Temperament Inventory (SATI; McClowery, 1995). For sociability, the items were selected from the STSC sociability subscale (e.g., “This child will go up to strange children and join in their play”) or the SATI approach/withdraw subscale (e.g., “Approaches children his/her age even when he/she doesn’t know them”). For reactivity, the items were selected from the STSC inflexibility subscale (e.g., “When this child is angry about something, it is difficult to sidetrack him/her”) or the SATI negative reactivity subscale (e.g., “When angry, yells or snaps at others”). For persistence, the items were selected from the STSC persistence subscale (e.g., “This child stays with an activity for a long time”) or the SATI task persistence subscale (e.g., “Goes back to the task at hand after an interruption”). All items from the STSC had six response options (1 = “Almost never”, 2 = “Not often”, 3 = “Variable, usually does not”, 4 = “Variable, usually does”, 5 = “Frequently”, 6 = “Almost always”), while all items from the SATI had five response options (1 = “Never”, 2 = “Rarely”, 3 = “Half the time”, 4 = “Frequently”, 5 = “Always”).

Covariates.

We controlled for three additional demographic variables that might be expected to relate to child temperament, parent income, parent education, or neighborhood disadvantage: child gender, child indigenous status, and study cohort.

Statistical Approach

First, for each of the three SES variables, we obtained an estimate of the linear change over time as well as an estimate of the value at baseline. We thus computed three separate Hierarchical Linear Models (HLM; Raudenbush & Bryk, 2002) with SES at each timepoint predicted at level-1 by child age (centered at 5.0 years old) and at level-2 by the child’s date of birth, cohort, and their interaction. We then extracted the estimated regression intercept and slope over time for each child to serve as the baseline value and linear change over time in our primary analyses.

In our primary analyses, we analyzed the degree to which baseline parent income, parent education, and neighborhood disadvantage and their linear change over time predict temperament development. Temperament was assessed with different measures at different ages (the STSC from ages 4–7 and the SATI from ages 8 and up) so the dataset was split according to measure and separate HLMs were calculated for each group. At level-1, a linear model was fit for each child predicting their temperament at each assessment from their age at the time. The age variable was centered at the mean for that subset of the data, 5.80 years for the STSC HLM and 11.36 years for the SATI HLM. At level-2, the slope and intercept of the level-1 models for each child were predicted from one or more SES predictors’ baseline value and linear change over time, controlling for the demographic covariates. All level-2 predictors were centered on their grand mean and standardized. Temperament outcomes were centered at the scale midpoint and were not standardized. We conducted separate longitudinal HLM analyses predicting each of the three temperament aspects. Additionally, for each temperament aspect, our primary analyses tested four different predictor models: 1) intercept and slope only, 2) parent income controlling for covariates, 3) parent education controlling for covariates, and 4) neighborhood disadvantage, controlling for parent income, parent education, and covariates. Thus, there were three temperament aspects assessed by two measures and predicted by four models for a total of 24 longitudinal HLMs.

Results

In this sample of 9,217 Australian children, there was an even distribution of sexes (49% female, 51% male) and 3.6% of the children were indigenous Aboriginal or Torres Strait islander, which is representative of the overall Australian population. At baseline (5.0 years old), on average the children’s highest income parent earned AU$1,000 per week, their most educated parent had 15.54 years of education, and they lived in a neighborhood with a SEIFA-IRSD disadvantage score of −1012. Over time, parent income increased on average by 3.8% per year, parent education increased by 0.03 years of education per year, and neighborhood disadvantage increased by .08 points per year. Comparing these changes over time to the standard deviations at baseline, parent income increased by 0.08 SD per year, parent education increased by 0.01 SD per year, and neighborhood disadvantage increased by 0.001 SD per year. These demographic characteristics were similar in both the younger cohort B and the older cohort K (see Table 1).

Model 1: Intercept and Slope (Table 2)

Table 2.

Model 1: Results for Intercept and Slope

		STSC, 4-7yo			SATI, 8-15yo
Outcome	Predictor	B[95% CI]	p	rp	B[95% CI]	p	rp
Sociability	Intercept	.452[.430, .474]	<.001	--	.381[.367, .396]	<.001	--
	Age	.121[.110, .131]	<.001	.265	−.017[−.020, −.014]	<.001	−.079
Reactivity	Intercept	−.983[−1.00, −.966]	<.001	--	−.586[−.601, −.571]	<.001	--
	Age	−.124[−.133, −.115]	<.001	−.307	−.027[−.030, −.023]	<.001	−.121
Persistence	Intercept	.458[.440, .475]	<.001	--	.483[.467, .499]	<.001	--
	Age	.060[.050, .069]	<.001	.149	.049[.045, .053]	<.001	.191

Note. B[95% CI] = unstandardized coefficient [95% confidence interval], r_p = partial correlation. Bolded values are statistically significant α = .05. Outcomes were centered at the scale mid-point. Age was centered at the population average.

For sociability, parents tended to rate their children significantly above the midpoint on both the younger STSC measure and the older SATI measure. On the STSC, sociability significantly increased from 4 to 7 years old, but on the SATI sociability significantly decreased from 8 to 15 years old. This may indicate that sociability follows a somewhat curvilinear trend, increasing in younger children, but decreasing in middle childhood and adolescence.

For reactivity, parents tended to rate their children significantly below the midpoint on both the younger STSC measure and the older SATI measure. Additionally, reactivity significantly declined over time on both the STSC and the SATI. Thus, parents tended to perceive their children as consistently decreasing in reactivity from 4 to 15 years old.

For persistence, parents tended to rate their children significantly above the midpoint on both the younger STSC measure and the older SATI measure. Additionally, persistence significantly increased over time on both the STSC and the SATI. Thus, parents tended to perceive their children as consistently increasing in persistence from 4 to 15 years old.

Model 2: Parent Income (Table 3)

Table 3.

Model 2: Results for Parent Income

			STSC, 4-7yo			SATI, 8-15yo
Outcome	Predictor	Effect	B[95% CI]	p	rp	B[95% CI]	p	rp
Sociability	SES at baseline	Main Effect	.026[−.006, .059]	.115	.017	.016[−.006, .037]	.151	.011
		Age Interaction	.003[−.012, .018]	.697	.005	−.001[−.006, .003]	.549	−.005
	Linear change in SES per year	Main Effect	.001[−.033, .035]	.944	.001	.020[−.002, .042]	.068	.014
		Age Interaction	−.004[−.020, .012]	.630	−.006	.002[−.003, .007]	.438	.006
Reactivity	SES at baseline	Main Effect	−.078[−.103, −.054]	<.001	−.066	−.015[−.037, .006]	.161	−.011
		Age Interaction	.007[−.007, .020]	.315	.012	−.006[−.011, −.002]	.009	−.020
	Linear change in SES per year	Main Effect	−.004[−.030, .021]	.754	−.003	−.029[−.051, −.007]	.010	−.020
		Age Interaction	−.007[−.021, .006]	.289	−.013	.003[−.001, .008]	.180	.010
Persistence	SES at baseline	Main Effect	.076[.051, .102]	<.001	.062	.089[.066, .111]	<.001	.059
		Age Interaction	−.007[−.021, .006]	.291	−.013	.017[.012, .023]	<.001	.047
	Linear change in SES per year	Main Effect	−.002[−.029, .025]	.885	−.002	.015[−.008, .038]	.196	.010
		Age Interaction	.009[−.005, .023]	.205	.015	−.003[−.009, .002]	.249	−.009

Note. Models control for grand-mean-centered demographic covariates. B[95% CI] = unstandardized coefficient [95% confidence interval], r_p = partial correlation. Bolded values are statistically significant α = .05. Intercept and slopes for these models are not presented as they are essentially identical to Model 1 (see Table 2)

Parent income at baseline was not significantly related to child sociability on either the younger STSC measure or the older SATI measure and this relation did not significantly vary by age. Similarly, linear change in parent income over time was not related to child sociability on either measure and the relation did not vary by age.

Lower parent income at baseline was significantly related to higher reactivity on the younger STSC measure and this relation was stable from 4–7 years old. In contrast, on the older SATI measure there was no significant main effect for parent income, but there was a significant age interaction such that lower parent income at baseline was associated with less decrease in reactivity from 8–15 years old. Linear change in parent income over time was not related to child reactivity on the STSC and the relation did not vary by age. However, on the SATI, relative decreases in parent income over time were associated with higher reactivity and this relation was stable from 8–15 years old.

Lower parent income at baseline was significantly related to lower persistence on both the younger STSC measure and the older SATI measure. On the STSC this relation was stable from 4–7 years old, while on the SATI there was a significant age interaction such that lower parent income at baseline was associated with less increase in persistence from 8–15 years old. Linear change in parent income over time was not related to child persistence on either measure and the relation did not vary by age.

Model 3: Parent Education (Table 4)

Table 4.

Model 3: Results for Parent Education

			STSC, 4-7yo			SATI, 8-15yo
Outcome	Predictor	Effect	B[95% CI]	p	rp	B[95% CI]	p	rp
Sociability	SES at baseline	Main Effect	.008[−.015, .031]	.500	.007	.017[.002, .032]	.024	.017
		Age Interaction	−.015[−.026, −.004]	.009	−.032	.000[−.003, .004]	.826	.002
	Linear change in SES per year	Main Effect	−.008[−.031, .015]	.509	−.007	.003[−.012, .017]	.727	.003
		Age Interaction	.001[−.010, .012]	.922	.001	−.001[−.005, .002]	.498	−.005
Reactivity	SES at baseline	Main Effect	−.081[−.099, −.063]	<.001	−.095	−.032[−.048, −.017]	<.001	−.032
		Age Interaction	.007[−.002, .017]	.134	.018	−.006[−.009, −.003]	.001	−.026
	Linear change in SES per year	Main Effect	−.020[−.037, −.002]	.027	−.023	.003[−.012, .018]	.716	.003
		Age Interaction	−.002[−.011, .008]	.745	−.004	−.000[−.004, .003]	.888	−.001
Persistence	SES at baseline	Main Effect	.111[.092, .129]	<.001	.125	.101[.085, .117]	<.001	.095
		Age Interaction	−.003[−.013, .007]	.517	−.008	.016[.012, .020]	<.001	.062
	Linear change in SES per year	Main Effect	.009[−.009, .027]	.315	.011	.009[−.007, .025]	.259	.009
		Age Interaction	.003[−.006, .013]	.500	.008	−.000[−.004, .004]	.989	.000

Note. Models control for grand-mean-centered demographic covariates. B[95% CI] = unstandardized coefficient [95% confidence interval], r_p = partial correlation. Bolded values are statistically significant α = .05. Intercept and slopes for these models are not presented as they are essentially identical to Model 1 (see Table 2)

For parent education at baseline there was no significant main effect with child sociability on the younger STSC measure, but there was a significant age interaction such that lower parent education at baseline was associated with more increase in sociability from 4–7 years old. This interaction is contrary to our hypotheses that lower SES would be associated with relatively less increase in sociability over time or would not be associated with changes in sociability. However, lower parent education at baseline was significantly related to lower sociability on the older SATI measure and this relation was stable over time. Linear change in parent education over time was not related to child sociability on either measure and the relations did not vary by age.

Lower parent education at baseline was significantly related to higher reactivity on both the younger STSC measure and the older SATI measure. On the STSC this relation was stable from 4–7 years old, while on the SATI there was a significant age interaction such that lower parent education at baseline was associated with less decrease in reactivity from 8–15 years old. On the SATI, linear change in parent education over time was not related to child reactivity and the relation did not vary by age. However, on the STSC less increase in parent education over time was associated with higher child reactivity and this association was stable from 4–7 years old.

Lower parent education at baseline was significantly related to lower persistence on both the younger STSC measure and the older SATI measure. On the STSC this relation was stable from 4–7 years old, while on the SATI there was a significant age interaction such that lower parent income at baseline was associated with less increase in persistence from 8–15 years old. Linear change in parent income over time was not related to child persistence on either measure and the relation did not vary by age.

Model 4: Neighborhood Disadvantage (Table 5)

Table 5.

Model 4: Results for Neighborhood Disadvantage

			STSC, 4-7yo			SATI, 8-15yo
Outcome	Predictor	Effect	B[95% CI]	p	rp	B[95% CI]	p	rp
Sociability	SES at baseline	Main Effect	−.049[−.074, −.024]	<.001	−.042	−.033[−.049, −.017]	<.001	−.031
		Age Interaction	.000[−.011, .012]	.958	.001	−.000[−.004, .004]	.961	.000
	Linear change in SES per year	Main Effect	−.007[−.030, .016]	.545	−.006	−.017[−.032, −.003]	.021	−.018
		Age Interaction	.011[−.000, .022]	.053	.024	−.001[−.005, .002]	.395	−.006
Reactivity	SES at baseline	Main Effect	.020[.001, .039]	.036	.022	.026[.010, .042]	.002	.024
		Age Interaction	−.002[−.012, .008]	.698	−.005	−.000[−.004, .003]	.879	−.001
	Linear change in SES per year	Main Effect	.035[.018, .052]	<.001	.043	.026[.011, .041]	.001	.026
		Age Interaction	−.006[−.016, .003]	.181	−.016	−.001[−.004, .003]	.645	−.004
Persistence	SES at baseline	Main Effect	−.016[−.035, .003]	.103	−.017	−.001[−.018, .016]	.904	−.001
		Age Interaction	−.007[−.017, .004]	.216	−.015	−.009[−.014, −.005]	<.001	−.033
	Linear change in SES per year	Main Effect	−.009[−.027, .009]	.330	−.010	−.010[−.026, .005]	.189	−.010
		Age Interaction	.005[−.005, .014]	.356	.011	.001[−.003, .005]	.540	.005

Note. Models control for parent income, parent education and grand-mean-centered demographic covariates. B[95% CI] = unstandardized coefficient [95% confidence interval], r_p = partial correlation. Bolded values are statistically significant α = .05. Intercept and slopes for these models are not presented as they are essentially identical to Model 1 (see Table 2).

After controlling for parent income and parent education, greater neighborhood disadvantage at baseline was significantly associated with less sociability on both the younger STSC measure and the older SATI measure and these relations were stable over time from 4–15 years old. Linear change in neighborhood disadvantage over time was not related to child sociability on the STSC and this relation did not vary by age from 4–7 years old. However, linear increases in neighborhood disadvantage over time were related to lower sociability on the SATI and this relation was stable over time from 8–15 years old.

Greater neighborhood disadvantage at baseline, controlling for parent income and parent education, was associated with higher reactivity on both the younger STSC measure and the older SATI measure and these relations were stable over time from 4–15 years old. Additionally, linear increases in neighborhood disadvantage over time were associated with higher reactivity on both the STSC and SATI and these associations were stable over time from 4–15 years old.

Neighborhood disadvantage at baseline, controlling for parent income and parent education, was not associated with persistence on the younger STSC measure and this relation did not vary by age. In contrast, for the older SATI measure, while there was not a significant main effect, there was a significant interaction such that greater neighborhood disadvantage at baseline was associated with significantly less increase in persistence over time from 8–15 years old. Linear change in neighborhood disadvantage over time was not related to child persistence on either measure and the relation did not vary by age.

Discussion

Supporting our hypotheses, there were significant associations between SES in childhood and child temperament development and these associations tended to be stable over time. With a few exceptions, lower parent income, less parent education, and more neighborhood disadvantage were associated with lower sociability, higher reactivity, and lower persistence on both the STSC temperament measure used when the children were younger and the SATI measure used when the children were older. Additionally, these associations were typically stable or increased over time. We also investigated the association of linear change in SES over time with child temperament development. Decreases in SES over time were typically associated with increased reactivity, but changes in SES typically did not have a significant relation with sociability or persistence. Thus, these results are generally consistent with previous research on family SES and child temperament (e.g., Hart et al., 2007; Lengua. 2006) and these results go beyond previous research by demonstrating significant relations between neighborhood SES and child temperament controlling for family SES.

A variety of mechanisms may underlie the observed relations between neighborhood SES and child temperament. The association between neighborhood SES and both sociability and reactivity, controlling for family SES, was generally significant in the same direction for both neighborhood SES at 5.0 years old and for change in neighborhood SES over time. These results suggest that there was not a critical period for the development of sociability and reactivity around 5 years old, but rather a dynamic association that to some degree responds to increases or decreases in neighborhood SES across childhood and adolescence. Thus, the underlying mechanisms are more likely to be broad characteristics that differ between neighborhoods with different SES levels and that impact many areas of life for people of all ages. For example, higher SES neighborhoods are more likely to be able to provide a wide array of material and social benefits through more funding for schools, recreational facilities, public transportation, local shopping, and job opportunities. In contrast, people living in lower SES neighborhoods are more likely to be exposed to violence, drug abuse, and other criminal activity (e.g., Chong, Lee, & Victorino, 2015; Karriker-Jaffe, 2013; Odgers et al., 2009). This exposure likely increases chronic stress and anxiety which may directly impact temperament development via biological pathways, such as elevated stress hormones or chronic inflammation (e.g., Luchetti, Barkley, Stephan, Terracciano, & Sutin, 2014). In addition, children growing up in low SES neighborhoods are more to have higher lead exposure than children in high SES neighborhoods (Berg, Kuhn, &Van Dyke, 2017). Such exposure in childhood has been associated with worse psychological outcomes in adolescence (Winter & Sampson, 2017), an association that continues into adulthood and that extends to personality (Reuben et al., 2019). Chronic exposure to more dangerous or unpredictable environments may also result in long-lasting behavioral adaptations that maximize immediate safety, such as avoiding strangers or quickly reacting to perceived threats, which may become general tendencies associated with lower sociability and higher reactivity (e.g., Chen, Shi, & Sun, 2017).

In contrast to sociability and reactivity, the only significant association for persistence was an age interaction for neighborhood SES at 5.0 years old on the older SATI measure, such that children in lower SES neighborhoods at 5.0 years old tended to increase less in persistence from 8–15 years old. There was no significant association with linear change in neighborhood SES over time, which suggests that early childhood may be a critical period for the association between neighborhood SES and persistence and that later increases or decreases in neighborhood SES may be unimportant. There are many potential mechanisms that could result in such a critical period. One possible mechanism is the differences in early educational environment available in higher and lower SES neighborhoods. Most children start school around 5 years old and if their school lacks the funding to provide high quality education in those first years it may permanently set the children back in many developmental trajectories, including persistence, in ways that cannot be erased even if they later transfer to a better school. For instance, perhaps having the funding to provide a low student to teacher ratio in the younger grades is particularly important for the development of persistence, as lessons can more easily be tailored to provide an appropriate level of challenge for each student (e.g., Blatchford, Bassett, & Brown, 2011).

To our knowledge, this is the first research to examine the association of neighborhood SES and the development of child temperament traits that are associated with adult FFM personality traits, controlling for family SES. The only previous research that investigated neighborhood SES and child temperament focused on a typological model of resilient, overcontrolled, and undercontrolled child temperament, which is not directly linked to the adult FFM (Hart et al., 2008). Lower neighborhood SES in childhood has been repeatedly associated with a wide variety of negative outcomes in adulthood, including lower future education and less income (e.g., Chetty & Hendren, 2018), worse physical health (e.g., Drukker et al., 2003), and worse mental health (e.g., Christensen et al., 2017). The present research expands and supports these prior findings by showing that lower neighborhood SES in childhood is also associated with less positive child temperament and that these associations persist over time. Together, this body of research highlights the substantial importance of neighborhood SES in childhood and throughout the lifespan, above and beyond the association with family SES.

Furthermore, previous research has also shown that child temperament is associated with a wide variety of consequential outcomes, such as mental health (e.g., Lewis & Olsson, 2011), peer relations (e.g., Sanson et al., 2004), substance use (e.g., Williams, Sanson, Toumbourou, & Smart, 2000), and academic performance (e.g., Schoen & Nagle, 1994). The results of the present analyses suggest the possibility that child temperament could be a mediator that contributes to the relation between SES and consequential life outcomes. Children who are raised in lower SES environments tend to develop lower sociability, higher reactivity, and lower persistence and these temperament traits subsequently predict more negative life outcomes, in general. Further research is needed to directly test whether child temperament mediates the relation between childhood SES and adult outcomes.

Strengths, Limitations, and Future Directions

These analyses included a relatively large number of participants (N = 9,217), originally selected as a nationally representative sample for Australia, and we were able to include up to six longitudinal time points per participant, from 4–15 years old. Furthermore, neighborhood SES was measured objectively using an index specifically designed for use in Australia. Thus, these results give a very clear picture of the association between neighborhood SES and the development of temperament from childhood to adolescence in Australia.

Nevertheless, as this research is the first to investigate the association between neighborhood SES and the development of childhood temperament traits, it remains unclear to what extent these associations generalize to other cultures. Future research should examine whether these associations vary by population. For example, differences in government-sponsored social assistance programs across countries might accentuate or ameliorate the impact of neighborhood SES on child temperament. Additionally, neighborhood SES might be more strongly associated with child temperament in cultures higher in collectivism, as neighbors may be more involved in each other’s lives.

Future research would also benefit from more objective measures of key variables. Although the assessment of neighborhood disadvantage was based on publicly available neighborhood statistics, assessments of parent income and parent education were based on parent self-report and may not be entirely accurate. Furthermore, the measures of child temperament were solely based on parent-report, which may have resulted in a degree of positive bias as well as contrast or assimilation effects of parent expectations for their child.

Finally, because these analyses were based on correlational rather than experimental data it is unclear to what degree the observed associations may be the result of unmeasured variables. For example, personality traits or other individual differences in the parents might independently contribute to both SES and child temperament. In the present research, because the LSAC was primarily focused on the children, information on parents was more limited. In addition, more information on the dynamics of neighborhoods, particularly the systematic flow into and out of neighborhoods, would be useful to better differentiate family choices from neighborhood characteristics. Future research is needed to further explore these issues.

Conclusion

The present analyses expand research on the substantial impact of childhood SES by demonstrating associations with temperament development that persist at least into adolescence. Children with lower family or neighborhood SES displayed less positive temperament and the associations tended to remain stable over time from 4 to 15 years old. The temperament traits that were investigated are linked to adult FFM personality traits, so it seems likely that these associations with child temperament may also be associated with adult personality and may persist across the lifespan. Both family and neighborhood SES in childhood are known to be important factors that are associated with a wide variety of consequential life outcomes, including income, education, physical health, and mental health (e.g., Chetty & Hendren, 2018; Christensen et al., 2017; Drukker et al., 2003). The present analyses go further and demonstrate the association between family and neighborhood SES and the development of children’s temperament and personality.