Relations between adolescent perceptions of household chaos and externalizing and internalizing behaviors in low‐ and middle‐income families

Abstract

A large body of literature has established that chaos in the home environment, characterized by high levels of disorganization, lack of household routine, crowding, noise, and unpredictability, undermines social–emotional and behavioral development in early childhood. It is less clear whether household chaos is linked to elevated risk for behavior problems in adolescence. The aims of this study were 3fold: (1) characterize the variability of adolescent and caregiver reports of household chaos over time; (2) examine associations among caregiver and adolescent reports of chaos over a 9‐month period; (3) consider how between‐ and within‐ individual variability in household chaos predicts adolescent externalizing and internalizing problems. This study drew data from the Family Income Dynamics study, a 9‐month longitudinal study. Participants included 104 adolescents between 14 and 16 years old (55% female; M age = 14.85) and their caregiver (92% female) from low‐ and middle‐income families. Results showed that adolescent‐reports of household chaos were more variable over time compared to caregivers' reports. Adolescent‐reports of household chaos had positive within‐ and between‐level associations with externalizing problems and between‐level associations with internalizing, while caregiver‐reports of chaos had no links to behavior. This work highlights the importance of adolescents' own perceptions of household chaos when considering its links to adolescent development.

INTRODUCTION

Mental health problems among adolescents have reached alarming levels (Office of the Surgeon General, 2021). Over the past decade, rates of depressive symptomatology have increased by more than 40% such that in 2019, more than 1/3 of youth reported persistent internalizing issues (e.g., sadness or anxiety; Office of the Surgeon General, 2021). More than 20% of adolescents exhibit externalizing problems (e.g., bullying, rule‐breaking, and aggression). The rise in adolescent mental health issues has garnered much attention from researchers, practitioners, and policymakers and has led to calls for a better understanding of how children's ecological contexts play a role in their behavioral development (Office of the Surgeon General, 2021).

The home environment is one such critical ecological context that shapes children's social and behavioral development (Bronfenbrenner & Evans, 2000). In particular, household chaos, defined as disruptions and/or uncertainty in the home arising from things such as ambient noise, overcrowding, and a lack of daily structure and family routines, has been linked to increased internalizing and externalizing problems in children (Andrews et al., 2021; Evans et al., 2005; Marsh et al., 2020). Household chaos is one proximal process through which more distal family risk factors, including residential instability, disruption in family structure, and neighborhood and socioeconomic disadvantage, can influence children's development.

It is well established that household chaos is associated with worse behavioral development in young children (Marsh et al., 2020). However, less attention is paid to examining how chaos relates to adolescent internalizing and externalizing problems despite adolescence being a time of high vulnerability for behavioral problems, including mood disorders and emotional problems (Costello et al., 2011; Kessler et al., 2005). The little work that exists tying chaos to behavior problems in adolescence fails to assess adolescents' own perceptions of chaos, which are likely important drivers of their behavior (Miller et al., 2024), instead focusing on parents' perceptions. Therefore, research that examines the role of household chaos, and in particular youth perceptions of chaos, in shaping adolescent behavioral development is needed to inform research and policy and prevention efforts aimed at combating this emerging youth mental health crisis.

Theoretical framework

This study is grounded in Bioecological Theory, which argues that children are embedded within multiple contexts that drive key developmental processes that shape their social and behavioral trajectories (Bronfenbrenner, 2005; Bronfenbrenner & Morris, 2006). These contexts range from more proximal environments, such as the home and family contexts, to more distal environments, including schools, neighborhoods, and even larger societal factors like culture, social inequalities, and government policies (Bronfenbrenner & Ceci, 1994). Chaos in the home—and the disorder and uncertainty that it causes—can generate environments and interactions that are dysfunctional, which may disrupt the developmental processes that build behavioral competence (Bronfenbrenner & Evans, 2000; Wachs & Evans, 2010). For example, studies show that chaos negatively affects parental well‐being, and in turn, reduces the quality of parenting, including less parental responsiveness, involvement, and supervision and more harsh parenting (e.g. Coldwell et al., 2006; Corapci & Wachs, 2002; Wachs & Evans, 2010). These aspects of parenting are linked to behavioral functioning such that decreases in responsive parenting and parent involvement and increases in harsh parenting are associated with more internalizing and externalizing behavior problems in children (Gorostiaga et al., 2019; Pinquart, 2017). Chaos in the home may also directly impact child behavior; chaotic households are linked to deficits in self‐regulatory skills and feelings of mastery, which may lead to increased internalizing and externalizing problems (Evans et al., 2005; Evans & Stecker, 2004).

Another important facet of Bioecological Theory is consideration of how time is an important factor for human development (Bronfenbrenner & Morris, 1998). Researchers must account for how similar contexts and processes may have differing effects on children as a function of age or developmental stage (Bronfenbrenner & Morris, 1998). Layering this on the extant research and theory of household chaos, we surmise that adolescence may be a critical developmental stage where household chaos exerts significant impacts on behavior. During adolescence, youth undergo accelerated changes in physical, cognitive, emotional, and interpersonal development (Kerig et al., 2012 ). When it comes to cognition, youth are shifting from concrete to abstract thought with sizeable improvements in reasoning, perspective taking, and meta‐cognition (Steinberg, 2005). This can support youth in more complex goal directed behavior. Moreover, adolescents experience marked changes in arousal, motivation, socioemotional contexts, and brain regions involved with self‐regulation (Casey et al., 2008; Dahl & Gunnar, 2009; Monahan et al., 2016 ; Steinberg, 2008). Several of these features that undergo change during adolescence directly impact adolescents' ability to manage emotions and behaviors.

Chaos in the home may tax an adolescent's ability to self‐regulate, sustain attention and motivation, or manage arousal (Brieant et al., 2017; Hardaway et al., 2012). Disordered, noisy, or crowded homes may hamper youth's ability to down‐regulate emotions or find the space necessary to reset or rest. Unpredictable home contexts may make it difficult for youth to understand connections between actions and outcomes, lead to feelings of helplessness and frustration, and the inability to internalize social rules for proper behavior (Wachs, 1986). Moreover, the requirements for a promotive physical environment may differ for adolescents compared to younger children in the home or caregivers. For example, adolescents may have an amplified need for privacy (e.g., a bedroom) or a quiet, designated space to do schoolwork, both of which are likely less germane for younger children or adults. Accordingly, chaotic home environments could undermine adolescent behavioral functioning by taxing adolescents' systems, confusing youth or preventing them from understanding how behaviors are associated with desired outcomes. Conversely, living in a consistent and stable home could be particularly supportive for youth as they develop a sense of competence from predictability in the environment (Murray & Rosanbalm, 2017; Simmons et al., 1987). For these reasons, in terms of behavioral development, adolescence may be a period where the effects of environmental stimuli, like chaos in the home, are particularly impactful (Blakemore & Mills, 2014; Fuhrmann et al., 2015).

Extant literature

When considering behavioral outcomes, several studies have examined relations between household chaos and externalizing and internalizing problems during adolescence. Cross‐sectional studies illustrate that household chaos in adolescence may be related to more externalizing issues, such as aggression and rule‐breaking (Chatterjee et al., 2020; Delker et al., 2020; Joo & Lee, 2020). Similarly, a longitudinal study found links between higher household chaos at 9 years and increased externalizing problems 3 years later (Jaffee et al., 2012). Empirical support for association between household chaos and adolescent internalizing issues is more mixed. Some studies find no association between chaos and internalizing problems, while others find household chaos is associated with more internalizing problems. For example, a cross‐sectional study by Lobo et al. (2021) finds no direct link between household chaos and adolescent depressive symptoms. On the other hand, a two‐year longitudinal study by Human et al. (2016) found that adolescent reports of household chaos in their home environment positively predicted their depressive symptoms concurrently and over time. Differences in these study findings may be related to the age of participants sampled. The Human et al. study included slightly older adolescents (M = 14.5) compared to the Lobo et al. study (M = 12.8), and internalizing problems tend to be low in middle childhood and increase during adolescence (Lewis & Rudolph, 2014). In sum, there is some evidence of consistent links between household chaos and externalizing behaviors in adolescence, but links between chaos and internalizing problems are less clear.

Limitations of existing literature

There are several limitations of the existing literature examining connections between household chaos and adolescent behavior problems that constrain our understanding of the associations. First, most studies presume the stability of household chaos. Many assessed chaos at one point in time and used it to predict behavior problems measured several years later. For instance, it is common in the current literature for assessments of household chaos to predict problem behaviors measured 1–2 years later or even as much as six or more years later (Deater‐Deckard et al., 2009; Shapero & Steinberg, 2013; Tucker et al., 2017). Even when household chaos is measured repeatedly, there is an assumption of stability of chaos within a year. For instance, Deater‐Deckard et al. (2009) assessed caregiver reports of chaos annually over 3 years, while Jaffee et al. (2012) measured chaos at age nine and then again 3 years later. These studies assumed stability of household chaos from year to year; yet household chaos may vary considerably within a year, as it can be influenced by shifts in caregiving responsibilities, job loss, just‐in‐time work schedules, changes in cohabitation with household members moving in and out of the home, or normative transitions like the start of summer or the beginning of the school year. Failing to assess household chaos at multiple time points during the year may obscure meaningful variability in caregivers' or youths' perceptions of chaos within the household and their concurrent effects on behavior.

Not only does this assumption of stability overlook important variability in household chaos, but it also may downwardly bias links between chaos and behavior problems. Adolescent functioning is characterized by frequent fluctuations in thinking, emotion, and behavior, and accordingly their problem behavior may be particularly sensitive to more temporally proximal experiences of household chaos (Blakemore & Mills, 2014; Buchanan & Hughes, 2011; Fuhrmann et al., 2015). Thus, assessments of chaos taken a year or more prior to the behavioral outcome may be too distal to meaningfully capture children's experiences during this developmental period. Rather, shorter temporal lags (e.g. every few months) between assessment of household chaos and problem behavior may be more relevant when explaining individual differences in adolescent behavior. Thus, research examining intra‐year fluctuations in both household chaos and adolescent problem behaviors on more proximal timescales is necessary to test these associations.

Second, studies have primarily used caregiver reports of household chaos to predict adolescent behavior (e.g., Deater‐Deckard et al., 2009; Dumas et al., 2005; Shamama‐tus‐Sabah et al., 2011; Taylor & Hart, 2014). These adult‐centered perspectives fail to consider how adolescents perceive their home environments and how youth perceptions of chaos relate to their behavior despite the possibility that these perceptions may shape their behavior above and beyond their caregivers' perceptions (Delgado et al., 2013; Mistry et al., 2009; Wadsworth & Berger, 2006). Although some studies asked adolescents to report on the chaos in their home environments (see Chatterjee et al., 2020; Jaffee et al., 2012), far fewer have examined both adolescent and caregiver reports. It is therefore unclear whether these reports are consistent and not clear whether these are similarly predictive of adolescent problem behaviors. In a notable exception, Jaffee et al. (2012) assessed youth and parent reports of household chaos and find contemporaneous correlations of r = .53 and r = .55 across two‐time points. In another study, Human et al. (2016) reported a correlation of r = .36 between adolescent and caregiver reports of chaos. Given these moderate relations between adolescent and caregiver reports of chaos, caregivers and youth seem to perceive their home environments differently. Therefore, studies should consider both caregiver and adolescent perceptions of chaos when predicting adolescent problem behaviors.

Third, studies that examined externalizing and internalizing problems tend to use questionnaire measures that capture more severe problems or are limited in scope. For example, several studies on chaos and problem behaviors captured highly deviant externalizing behaviors, such as carrying a weapon or physical fighting (Chatterjee et al., 2020; Delker et al., 2020; Joo & Lee, 2020), as opposed to attention problems. For internalizing problems, studies used the Center for Epidemiological Studies‐Depression Scale (Human et al. (2016); Lobo et al., 2021), which captures depression but not anxiety or social problems. It is unclear whether these links would replicate when predicting broadband measures of internalizing and externalizing problems, which are more frequently expressed in teens. Therefore, studies that examine a broader set of behaviors, both separately and together, could be beneficial in clarifying associations between household chaos and behavior problems.

Finally, the current literature on household chaos and adolescent behavior relies on between‐individual variability in perceptions of household chaos as an identification strategy, which raises concerns about endogeneity or omitted variable bias (Chamberlain, 1978; Kim & Frees, 2006). In other words, households with high levels of chaos likely differ from those with lower levels of chaos in other ways that may influence youth behavior problems. Failure to control for these differences in statistical models may inflate relations between household chaos and adolescent development. It will be critical to deploy more advanced statistical models that isolate variability between and within participants to overcome this issue.

Study aims

To address these gaps in the literature, the current study will examine whether between‐ and within‐individual variability in youth and caregiver perceptions of household chaos are related to youth externalizing and internalizing problems using data from more than 100 teen‐caregiver dyads collected monthly over 9‐months. The specific aims of this study are 3‐fold. First, it will characterize the stability of adolescent and caregiver reports of household chaos over a nine‐month period. In contrast to current assumptions of year‐to‐year stability, we anticipate that caregiver and adolescent perceptions of chaos will fluctuate over time. Second, this investigation will examine whether caregiver and adolescent reports of chaos are correlated. We anticipate a significant positive association between these reports, but we also hypothesize that there will be unique variability in caregiver and adolescent perceptions of chaos. Third, this study will consider relations between household chaos, as reported by both adolescents and caregivers, and youth problem behaviors, including broad composites of externalizing and internalizing problems as well as well as subdomains of these behaviors: conduct problems, attention problems, emotional problems, and peer problems. We predict that increases in chaos reported by caregivers and adolescents will be associated with elevated externalizing and internalizing problems, and links will be stronger when looking at adolescent's perception of household chaos predicting behavior. Without the existence of empirical evidence to the contrary, we hypothesize that links between chaos and the problem behaviors subdomains will be similar to those observed for the composite measures. Lastly, supplemental analyses will consider whether these relations are robust to the influences of other correlated characteristics of caregivers with household chaos. We expect links between household chaos and behavior problems will still be present and significant even after the inclusion of additional caregiver characteristics in the models.

METHODS

Participants

This study drew data from the Family Income Dynamics Study (FInD), a 9‐month longitudinal study of a large, racially, and socioeconomically mixed cohort of youth and their caregivers in Pittsburgh, PA (Miller et al., 2024). Participants included 104 dyads (55% female) adolescents and (92% female) caregivers. Youths were 14–16 at the first assessment (M = 14.85, SD = 0.83). The data set used stratified sampling to select a sample that equally represented youth and caregivers from low‐income (<2× the federal poverty threshold) and middle‐income (between 2× and 5× the federal poverty threshold) households adjusted according to the number of adults and children in the home. This inclusion criteria allowed us to capture the vast majority of US families; for example, using thresholds based on a 4‐person family we capture 79% of US households (U.S. Census Bureau ACS Data, 2023). Families participated in a phone screen before enrolling in the study to be evaluated on these criteria. Caregivers varied when it came to their educational backgrounds. More specifically, when it came to the highest degree attained by the caregiver who completed the survey, 23% had obtained a high school diploma or less, 23% had some college or associate degree, and 54% had a bachelor's degree or more. Thirty‐seven percent of youth identified as Black, 43% as White, and 20% as multi‐racial or another racial group. Thirty‐three percent of caregivers identified as Black, 61% as White, and 6% multi‐racial or another racial group. Most caregivers were biological mothers (85%), with an additional 7% biological fathers, and the remaining 8% other caregivers (i.e., Adoptive, step‐, or grandparents).

Procedures

Beginning in November 2019, FInD recruited participants on a rolling basis (adolescents and caregivers) via community sampling, using referrals, research registries (e.g., Pitt+Me, Facebook, Craigslist, and flyer postings across Pittsburgh, PA). Study staff screened caregivers to collect information about race, income level, English proficiency, and adolescent disability status. Adolescents with severe or pervasive conditions that limited their ability to complete surveys independently (e.g., Pervasive Developmental Disorder (PDD), diagnosis of blindness, or severely impaired hearing) were excluded from the study. Once screened, staff scheduled caregivers to come into the lab to complete their baseline visit, which included the consent and completion of their first survey. Dyads were given a paper copy of the consent/assent, and the research study staff explained the study details and answered participant questions. Caregivers and adolescents were in separate spaces to complete their baseline survey using an electronic tablet. After COVID‐19 pandemic protocols prohibited in‐person visits, we modified the baseline visit so that dyads consented/assented via conference call and an online consent form. Participants then received their consent/assent forms and baseline surveys via email and were given 24 h to complete the surveys.

After the baseline visit there were no differences in follow‐up procedures for participants entering the study before or after the pandemic. Staff sent monthly follow‐up surveys to all participants through Qualtrics, an online survey platform. Caregivers and youth completed one follow‐up survey each month for eight consecutive months. Participants were given from the 1st through the 5th of each month to complete their survey. Study staff sent customized reminder emails, texts, and phone calls based on participants' contact preferences a few days before surveys launched and during the collection period. Participants were contacted promptly after completion, thanked for participation, and were provided payment details. When participants did not complete their survey by the 5th of each month, extensions were granted on a case‐by‐case basis, giving participants an additional 1–3 days to respond. Recruitment of caregivers and adolescents ended in December 2020, with final surveys completed 8 months later in August 2021. Eighty‐three percent of the total possible surveys were collected after March 2020, the onset of the pandemic in the U.S. Using these methods, retention rates for this sample over time were 98% for caregivers and youth. To reduce participant burden, we collected different measures at different time‐points depending on the construct. Staff collected questionnaires in two ways (1) at every wave, i.e., baseline, 2, 3, 4, 5, 6, 7, 8, and 9; or (2) at baseline, waves 2, 5, and 8. Measures collected at every wave were those expected to show meaningful month‐to‐month variability, whereas those collected less frequently were measures that were hypothesized to exhibit change on a relatively longer timescale.

Measures

Household chaos

The Confusion Hubbub and Order Scale was used to assess household chaos. It was a well‐validated measure capturing the most proximal manifestations of chaos, including noise, overcrowding, and lack of household routines (Matheny Jr et al., 1995) It has been used with adolescents (see Chatterjee et al., 2020; Human et al., 2016; Jaffee et al., 2012). The scale includes 15 items which capture noise (e.g., “It's a real zoo in our home”), disorder (e.g., “We can usually find things when we need them”), and unpredictability (e.g., “No matter what our family plans, it usually doesn't seem to work out”). Youth and caregivers were asked to indicate how much each statement described their home environment using a Likert scale (0 = very much like your own home; 1 = somewhat like your own home; 2 = a little bit like your own home; 3 = not at all like your own home). On many items, higher scores indicated more chaos, and 8 items were reverse scored. Youth and caregiver reports of chaos exhibited strong reliability over time, with youth reports ranging between α = .84 and α = .85, with an average $\overline{\alpha }=.85$. The reliability of caregiver reports varied between α = .85 and α = .89, with an average of $\overline{\alpha }=.87$. Caregiver and youth reports of chaos were collected at baseline and waves 2, 5, and 8.

Problem behavior – Externalizing and internalizing

The Strengths and Difficulties Questionnaire (SDQ; Goodman et al., 2010) was used to measure problem behaviors. The SDQ is a well‐validated measure of prosocial behavior and psychopathology. It is suitable for use with children aged 3–16 years and can be completed by various reporters, including the adolescents themselves (Goodman et al., 2010). Research has established that adolescents can provide reliable reports of behavior problems in a clinical sample, and, moreover, adolescents are often the best reporters of their internalizing problems, which are mostly private feelings/states that can be difficult for caregivers assess (Aebi et al., 2017; Becker et al., 2004). The measure included 25 items related to recent experiences of positive and negative statements assessing five behavioral domains: emotion problems, conduct problems, hyperactivity‐inattention, peer problems, and prosocial behavior. For each item, participants answered on a 0–2 scale (0 = not at all true; 1 = somewhat true; 2 = certainly true). Higher scores indicated more problem behaviors (except for the prosocial subscale which we did not analyze). We took the average of the conduct problems (e.g., “I fight a lot. I can make other people do what I want”) and hyperactivity‐inattention (e.g., “I think before I do things”) subscales to create an externalizing problem composite variable. The internalizing variable was an average of the emotional problems (e.g., “I am often unhappy, downhearted or tearful”) and peer problems (e.g., “Other children or young people pick on me or bully me”) SDQ subscales. Prior work supported combining these subscales into internalizing and externalizing problems (Bevilacqua et al., 2021; Goodman et al., 2010). Reliability for the composite scales was good (youth α = .72–.81; caregiver α = .81–.85). Youth completed the SDQ at all waves.

Covariates

All analyses controlled for factors that may relate to both household chaos and adolescents' externalizing or internalizing problems. Time invariant covariates included adolescent age in years measured at baseline, binary sex assigned at birth, with female as the reference group, and race (Black, white, or Multiracial/other racial group coded as dummy variables, Black participants served as the reference group). Family household size (M = 4.19, SD = 1.50) was defined as the sum of adults and children living in the home as reported by the caregiver at baseline. Monthly income was included as a time‐varying covariate and was based on caregiver reports of total household income (Median = 3411.00), which were naturally log‐transformed to address the highly skewed nature of the income variable. Lastly, we included a time indicator in models to account for the 9 waves of data (coded 0 to 8); the baseline assessment was considered time zero.

Caregiver characteristics

Perceived stress

Caregiver stress was measured with the Perceived Stress Scale, a widely used questionnaire that assesses stress in daily life (Cohen et al., 1995). Caregivers completed all 10 items to evaluate how often they thought or felt about experiences over the past month (e.g., “In the past month, how often have you been upset because of something that happened unexpectedly”). Participants then responded on a 5‐point Likert scale (0 = never; 1 = almost never; 2 = sometimes; 3 = fairly often; 4 = very often). Positively worded items were reverse scored (e.g., “In the past month, how often have you felt you were on top of things”) so that higher scores indicated more stress. The reliability for this measure at each wave ranged from α = .84–.91 for caregivers. Caregiver stress was measured at all waves.

Caregiver depression

Depression was measured with the Center for Epidemiological Studies‐Depression (CES‐D), a commonly used questionnaire that assesses depressive symptoms in the general population (Radloff, 1977). The set of 20 items asks participants to rank how they felt or behaved in the past week (e.g., “I was bothered by things that don't usually bother me”) and (e.g., “I could not get going”). Participants were asked to rate how frequently they felt this way (0 = rare or none of the time, less than 1 day; 1 = some or a little of the time, 1–2 days; 2 = occasionally or a moderate amount of time, 3–4 days; 3 = most or all of the time, 5–7 days). Positively worded items (e.g., “I felt just as good as other people” and “I enjoyed life”) were reverse scored so that higher scores indicated more depressive symptoms. The reliability of this scale at each wave ranged from α = .91–.92. Caregivers were given this questionnaire at the baseline visit and waves 2, 5, and 8.

Caregiver monitoring

Caregiver monitoring was assessed using items adapted from several measures aimed at capturing parenting behaviors associated with expectations and rules related to adolescent behavior. Our measure consisted of 7 items. Three items were adapted from the Alabama Parenting Questionnaire (Elgar et al., 2007). These items asked caregivers to answer statements (e.g., “Your child fails to let you know where he/she is going” and “Your child stays out past the time he/she is supposed to be home”) using a 5‐point scale (1 = never; 2 = almost never; 3 = sometimes; 4 = often; 5 = always). The last four items were developed based on existing work related to monitoring by Conger et al. (1994) and Steinberg et al. (1992). Items asked caregivers to rate “How much do you try to know who your child spends time with” and “How much do you try to know how your child spends his/her free time” and “How often do you set rules or limits on who your child spends time with” and “How often do you set rules or limits on how your child spends their free time?” The response option for these items varied so item‐level scores were converted to z‐scores and then averaged to create a composite score. The reliability for the composite measure ranged from α = .54–.73 across waves. Caregivers were given this questionnaire at all waves.

Adolescent‐caregiver relationship quality

To assess caregivers' perception of warmth and openness in their relationship with their adolescent, they were given six items from the Child–Parent Relationship Scale: Closeness Subscale (Driscoll & Pianta, 2011; Pianta, 1992). Sample items included statements such as, “I share an affectionate, warm relationship with my child” and “It is easy to be in tune with what my child is feeling.” Participants then rated each statement on a 1–5 scale (1 = definitely does not apply; 2 = not really; 3 = neutral, not sure; 4 = applies somewhat; 5 = definitely applies). The reliability of this measure ranged from α = .70–.84. Adolescents completed the same closeness questionnaire for their primary caregiver with language modified to reference the caregiver (“If upset, I will seek comfort from my mother/father/grandmother”). Across waves, reliability for adolescent reports of this measure ranged from α = .88–.92. Caregivers and adolescents completed this questionnaire in all waves.

ANALYTIC STRATEGY

Study analyses were conducted in Stata/SE 17.0 (StataCorp, 2021). We calculated Intraclass Correlations (ICCs) for caregiver and youth report to address our first aim, which was to describe within‐individual change in household chaos. ICCs were used to assess the strength of the correlation of a measure within an individual over time, where a score of “1” indicates no variability or change. If the ICCs for youth or caregivers were close to 1, this would support existing assumptions of within‐person stability in reports of household chaos over the study period. Therefore, ICCs provided a descriptive measure of intra‐individual variability in caregiver and youth reports of chaos over a 9‐month timeframe.

To address our second aim, which considered whether caregiver and adolescent reports of household chaos were related over time, the present study examined within‐ and between‐ individual associations (Curran & Bauer, 2011) among caregiver reports of chaos (CC) and adolescent reports of chaos (AC) using two‐level mixed effects models. The models were estimated using Full Information Maximum Likelihood in Stata with the mixed command (StataCorp, 2021). Level 1 contained repeated measures of caregiver chaos over time (t) (i.e., baseline, wave 2, wave 3, …), which were nested within an individual (i) at Level 2. The Level‐1 model is shown in Equation 1.

$$\text{Level}1{\mathrm{CC}}_{\mathit{ti}}={\mathrm{B}}_{0i}+{\mathrm{B}}_{1i}{\mathrm{AC}}_{\mathit{ti}}+{\mathrm{B}}_{2i}{\mathrm{X}}_{\mathit{ti}}+{\mathrm{B}}_{3i}{\mathrm{W}}_{\mathit{ti}}+{\mathrm{r}}_{\mathit{ti}}$$ (1)

Caregiver reports of chaos for individual i at time t (CC _ti ) were modeled as a function of time‐varying measures of adolescent chaos (AC _ti ), time‐varying covariates (X _ti ), time‐invariant covariates and a wave indicator (W _ti ). All of the time‐varying predictors at Level 1 were group‐mean centered, also known as within‐individual centering, which reduces bias from unobserved heterogeneity or unmeasured factors that vary across individuals over time (Raudenbush & Bryk, 2002; Singer et al., 2003). Between‐individual associations of average caregiver chaos and average adolescent reports of chaos were estimated at Level 2. Here variability in the Level 1 intercept was modeled using equation 2 through equation 5 below:

$$\text{Level}2{\mathrm{B}}_{0i}={\mathrm{Y}}_{00}+{\mathrm{Y}}_{01}{\mathrm{AC}}_i+{\mathrm{Y}}_{02}{\mathrm{X}}_i+{\mathrm{Y}}_{03}{\mathrm{W}}_i+{\mathrm{u}}_{0i}$$ (2)

$${\mathrm{B}}_{1i}={\mathrm{Y}}_{10}$$ (3)

$${\mathrm{B}}_{2i}={\mathrm{Y}}_{20}$$ (4)

$${\mathrm{B}}_{3i}={\mathrm{Y}}_{30}$$ (5)

Variability in mean levels of caregiver chaos (B_0i ) was explained with individual‐level averages of adolescent chaos (AC _i ) along with average levels of time‐varying covariates (X _i ), and time‐invariant covariates (W _i ). A random effect for the intercept was estimated at Level 2 (u_0i ), and all other Level 1 coefficients were estimated as fixed at Level 2. Time‐varying predictors were grand‐mean centered in the Level 2 equations.

The mixed effects modeling framework captured both between‐ and within‐ adolescent variability. The Level‐1 coefficient of adolescent chaos considered whether within‐individual changes in adolescent reports of chaos were related to changes in caregiver chaos. The Level 2 coefficients on adolescent chaos reflected whether adolescents whose caregivers reported higher than their average level also tended to report more chaos in their home environment. We examined how within‐individual variability in household chaos over time predicted behavior problems to help address endogeneity concerns. Analysis of within‐individual variability allowed us to estimate levels of behavior problems when reported household chaos is higher or lower than average, with adolescents serving as their own comparison over time. This approach helped to reduce the threat of bias posed by time‐invariant characteristics of children and families that may give rise to spurious links between household chaos and adolescent development (Hofmann, 1997; Miller et al., 2016; Osborne, 2000).

The third aim considered whether adolescent and caregiver reports of household chaos predicted adolescent reports of problem behaviors using this same two‐level modeling framework. Externalizing and internalizing behavior problems were estimated in separate models. To investigate whether certain subdomains of externalizing or internalizing behaviors drove the association between household chaos and outcomes, models were estimated separately for behavior problems subscales, including conduct problems, hyperactivity‐inattention, emotional problems, and peer problems. In Equation 6, behavior problems for individual i at time t (BP _ti ) were estimated as a function of time‐varying measures of adolescent chaos (AC _ti ), caregiver chaos (CC _ti ), time‐varying covariates (X _ti ), time‐invariant covariates and a wave indicator (W _ti ). Like the prior model, all time‐varying predictors at Level 1 were group‐mean centered (Raudenbush & Bryk, 2002; Singer et al., 2003).

$$\text{Level}1{\mathrm{BP}}_{\mathit{ti}}={\mathrm{B}}_{0i}+{\mathrm{B}}_{1i}{\mathrm{AC}}_{\mathit{ti}}+{\mathrm{B}}_{2i}{\mathrm{CC}}_{\mathit{ti}}+{\mathrm{B}}_{3i}{\mathrm{X}}_{\mathit{ti}}+{\mathrm{B}}_{4i}{\mathrm{W}}_{\mathit{ti}}+{\mathrm{r}}_{\mathit{ti}}$$ (6)

Between‐individual associations of average behavior problems and average adolescent and caregiver reports of chaos were estimated at Level 2. Variability in the level 1 intercept was modeled using equation 7:

$$\text{Level}2{\mathrm{B}}_{0i}={\mathrm{Y}}_{00}+{\mathrm{Y}}_{01}{\mathrm{AC}}_i+{\mathrm{Y}}_{02}{\mathrm{CC}}_i+{\mathrm{Y}}_{03}{\mathrm{X}}_i+{\mathrm{Y}}_{03}{\mathrm{W}}_i+{\mathrm{u}}_{0i}$$ (7)

Changes in the mean level of behavior problem (B_0i ) were explained with individual‐level averages of adolescent chaos (AC _i ), caregiver chaos (CC _i ), time‐varying covariates (Xi), time‐invariant covariates and a wave indicator (W _i ). A random effect for the intercept was estimated at Level 2 (u_0i ) and all other Level 1 coefficients were estimated as fixed at Level 2. Time‐varying predictors were grand‐mean centered in the Level 2 equations.

Sensitivity analysis

Sensitivity analyses tested the robustness of the results to the inclusion of several constructs that tend to be related to both household chaos and adolescent behavioral outcomes, including caregiver depression and stress (Allen et al., 2010; Goodman et al., 2010; Henry et al., 2020), caregiver‐child relationship quality (Tucker et al., 2017), and monitoring (i.e., knowing the location of the adolescent, who they are with, or activities they are engaged in; Barnes et al., 2006; Ahmadi et al., 2013). Therefore, additional models, which included caregiver stress, depression, monitoring, and adolescent‐caregiver ratings of relationship quality were included in our supplemental materials. Controlling for these other caregiver characteristics helped to strengthen discriminant validity of our findings, address omitted variable bias, and strengthen causal inference. These sensitivity analyses clarify that the observed associations between chaos and problem behavior are not an artifact of correlated caregiver characteristics, such as caregiver stress and depression, and other dimensions of parenting, including monitoring adolescent behavior and the quality of the caregiver‐adolescent relationship, that may be correlated with household chaos and problem behavior. In other words, we had greater certainty that household chaos, and not another closely related construct, influences adolescents' externalizing and internalizing problems.

RESULTS

Stability of household chaos

Analyses first examined descriptive statistics for the study sample and correlations (see Table 1 for additional descriptive statistics and Table 2 for Level 1 and Level 2 correlations of all key variables). To address aim one, the study estimated the ICC of individual reports of household chaos over four time points using 95% confidence intervals in Stata/SE 17.0. Estimates were based on a one‐way random effects model. For the adolescent report, the ICC was 0.69 with a 95% CI [0.61–0.76]. For the caregiver report, the ICC was 0.83 with a 95% CI [0.78–0.87]. Thus, adolescents' perceptions of household chaos were more variable over 9 months.

TABLE 1.

Descriptive statistics.

Observation N (level 1) = 934
Dyad N (level 2) = 104
	Mean	SD	ICC	Min	Max
Adolescent chaos	0.95	0.52	0.69	0	2.73
Caregiver chaos	0.95	0.55	0.83	0	2.87
Adolescent baseline age	14.85	0.83		14	16
Adolescent sex (% female)	55%
Adolescent race
Asian	4%
Black	37%
Latinx	6%
Multi‐ or another race	10%
White	43%
Caregiver education
High school	23%
Associates or some college	23%
Bachelor's or higher	54%
Caregiver marital status
Married	46%
Monthly income	$4214	$3106	0.44	$0	$20,192
Household size	4.19	1.5		1	10

TABLE 2.

Between and within correlation matrix including sensitivity constructs.

	Between
	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
1. Adolescent chaos	1	0.04	–	0.02	–	0.28	0.25	0.19	0.18	0.15	0.12	−0.01	0.07	−0.08	−0.13	Within
2. Caregiver chaos	0.52	1	–	−0.05	–	0.05	0.01	0.07	0.02	0.04	−0.02	−0.02	−0.23	0.26	−0.03
3. Age	0.02	−0.05	1	–	–	–	–	–	–	–	–	–	–	–	–
4. Monthly income	0.00	−0.13	−0.15	1	–	−0.04	−0.01	−0.05	0.00	0.00	0.00	0.02	−0.01	−0.07	−0.04
5. Household size	0.20	0.46	−0.09	0.13	1	–	–	–	–	–	–	–	–	–	–
6. Externalizing	0.59	0.29	−0.03	0.01	0.04	1	0.71	0.82	0.36	0.36	0.17	−0.05	−0.12	0.00	−0.14
7. Conduct	0.54	0.33	−0.07	−0.08	0.09	0.83	1	0.18	0.36	0.31	0.23	−0.02	−0.07	−0.01	−0.08
8. Hyperactive inattention	0.52	0.21	0.01	0.07	0.00	0.93	0.57	1	0.21	0.25	0.05	−0.05	−0.11	0.00	−0.13
9. Internalizing	0.46	0.24	0.11	−0.05	0.00	0.59	0.44	0.58	1	0.81	0.72	−0.05	−0.04	−0.10	−0.08
10. Emotion	0.46	0.24	0.12	0.09	0.13	0.54	0.38	0.54	0.89	1	0.18	−0.05	−0.08	−0.04	−0.05
11. Peer	0.27	0.14	0.04	−0.23	−0.19	0.42	0.34	0.40	0.75	0.38	1	−0.02	0.02	−0.13	−0.08
12. Caregiver monitoring	−0.23	−0.05	0.02	−0.19	0.12	−0.11	−0.08	−0.12	−0.04	−0.08	0.02	1	0.16	−0.03	−0.05
13. Caregiver relation‐qual	−0.23	−0.17	0.12	0.03	0.05	−0.31	−0.30	−0.26	−0.18	−0.12	−0.20	0.42	1	−0.16	0.12
14. Caregiver depression	0.24	0.50	−0.01	−0.32	0.04	0.24	0.23	0.20	0.25	0.18	0.24	−0.03	−0.16	1	−0.02
15. Adolescent relation‐qual	−0.40	−0.02	−0.05	0.02	0.06	−0.41	−0.34	−0.38	−0.31	−0.27	−0.26	0.14	0.46	−0.19	1

Note: Items shaded in blue are part of the supplemental analysis.

Adolescent chaos predicting caregiver chaos

Mixed effects models were used to examine how adolescent reports of household chaos predicted caregiver reports. The between‐dyad analysis explored whether adolescents who reported higher than average chaos also had caregivers who reported higher chaos on average compared to other dyads. In comparison, within‐person analyses indicated if deviations above or below the mean in adolescent's reports of chaos were associated with changes in caregiver reports of chaos. The results of the mixed effects models examining inner‐year variability in adolescent and caregiver reports of household chaos are found in Table 3. There was a significant positive association between adolescent reports of household chaos and caregiver reports of chaos at the between‐dyad level. An increase in adolescent report of chaos was associated with an increase in caregiver reports of chaos (β = .51, p < .001). Several control variables were significant as well. Family income was negatively related to household chaos at the between‐level (β = −.30, p = .015). Larger households tended to be more chaotic (β = .12, p < .001). Individual variability or a within‐level association of adolescent reports of chaos was unrelated to caregiver reports of chaos.

TABLE 3.

Adolescent chaos predicting caregiver chaos.

Predictors	Unadjusted			Adjusted
	Estimates	SE	p	Estimates	SE	p
Between
Adolescent chaos	0.59	0.10	<.001	0.51	0.09	<.001
Adolescent age				−0.01	0.05	.809
Adolescent sex				0.14	0.08	.069
Adolescent race
Multi or another race				0.10	0.11	.359
White				0.05	0.09	.562
Monthly income				−0.30	0.12	.015
Household size				0.12	0.03	<.001
Time				−0.01	0.00	.093
Within
Adolescent chaos	0.03	0.05	.483	0.04	0.05	.393
Monthly income				−0.02	0.03	.481
Intercept	0.95	0.04	<.001	0.50	0.72	.489
N	104famid			103famid
Observations	412			404

Note: Bold indicates P < .05.

Household chaos predicting externalizing problems

Mixed effects models were estimated to address the third aim, which considers the relations between chaos and externalizing problems (Table 4). At the between level, a one‐unit increase in the average level of adolescent‐reported household chaos was associated with an increase in externalizing problems (β = .37, p < .001). Additionally, the within‐level associations showed that in waves where adolescents report higher than their average level of chaos there was an increase in externalizing problems (β = .14, p < .001). However, between‐ or within‐level associations among caregiver reports of chaos and adolescent externalizing problems were not significant. For covariates, only between‐level associations for adolescent race significantly related to externalizing problems. In particular, Black adolescents reported more externalizing problems than White (β = −.13, p = .032) and multi‐racial or another‐race (β = −.14, p = .054) adolescents. Results of the models focused on the conduct problems and hyperactivity‐inattention subscales (see Table 4) have comparable coefficients to the main externalizing models.

TABLE 4.

Chaos predicting externalizing problems and subscales.

Predictors	Externalizing			Conduct			Hyperactive inattention
	Estimates	SE	p	Estimates	SE	p	Estimates	SE	p
Between
Adolescent chaos	0.37	0.07	<.001	0.30	0.06	<.001	0.45	0.10	<.001
Caregiver chaos	0.00	0.07	.977	0.04	0.06	.570	−0.03	0.10	.748
Adolescent age	−0.03	0.03	.293	−0.05	0.03	.111	−0.01	0.04	.820
Adolescent sex	0.04	0.05	.410	0.04	0.05	.431	0.05	0.07	.497
Adolescent race
Multi or another race	−0.14	0.07	.054	−0.05	0.07	.440	−0.23	0.10	.026
White	−0.13	0.06	.032	−0.07	0.05	.176	−0.18	0.08	.030
Monthly income	0.04	0.08	.646	−0.08	0.08	.264	0.17	0.12	.155
Household size	−0.01	0.02	.635	0.00	0.02	.891	−0.02	0.03	.444
Time	0.00	0.00	.340	0.00	0.00	.532	0.00	0.00	.432
Within
Adolescent chaos	0.14	0.03	<.001	0.13	0.04	.001	0.16	0.05	.002
Caregiver chaos	0.04	0.04	.307	0.01	0.04	.796	0.07	0.06	.275
Monthly income	−0.03	0.02	.145	−0.03	0.03	.220	−0.04	0.03	.284
Intercept	1.14	0.48	.017	1.01	0.44	.024	1.14	0.68	.093
N	103famid			103famid			103famid
Observations	404			404			404

Note: Bold indicates P < .05.

Household chaos predicting internalizing problems

Results of mixed effects models examining household chaos predicting internalizing problems as the outcome are shown in Table 5. At the between‐level, adolescent‐reported chaos positively predicted the internalizing problems composite measure (β = .29, p < .001). However, there was no within‐person association between adolescent chaos and internalizing problems (β = .06, p = .100). There were no significant between‐ and within‐associations among caregiver reports of chaos and adolescent internalizing problems. None of the covariates included in the model were significantly predictive of internalizing problems.

TABLE 5.

Chaos predicting internalizing problems and subscales.

Predictors	Internalizing			Emotion			Peer
	Estimates	SE	p	Estimates	SE	p	Estimates	SE	p
Between
Adolescent chaos	0.29	0.07	<.001	0.42	0.10	<.001	0.16	0.07	.019
Caregiver chaos	0.01	0.07	.855	0.00	0.10	.964	0.03	0.07	.643
Adolescent age	0.01	0.03	.816	0.03	0.04	.483	−0.02	0.03	.576
Adolescent sex	−0.09	0.05	.096	−0.15	0.08	.050	−0.03	0.05	.558
Adolescent race
Multi or another race	0.00	0.07	.958	0.03	0.11	.758	−0.03	0.07	.732
White	0.05	0.06	.403	0.13	0.08	.124	−0.03	0.06	.602
Monthly income	−0.03	0.08	.738	0.09	0.12	.450	−0.14	0.08	.085
Household size	−0.01	0.02	.602	0.02	0.03	.449	−0.04	0.02	.028
Time	0.01	0.00	.029	0.00	0.00	.418	0.01	0.00	.011
Within
Adolescent chaos	0.06	0.04	.100	0.06	0.05	.234	0.06	0.04	.207
Caregiver chaos	0.02	0.04	.668	0.05	0.06	.358	−0.02	0.05	.714
Monthly income	−0.02	0.02	.416	−0.02	0.03	.632	−0.02	0.03	.429
(Intercept)	0.53	0.49	.273	0.09	0.69	.892	1.00	0.49	.041
N	103famid			103famid			103famid
Observations	404			404			404

Note: Bold indicates P < .05.

The models examining the internalizing subscales had similar results (see Table 5). More specifically, there were significant positive between‐level associations among adolescent‐reported chaos and emotional problems (β = .42, p < .001) and peer problems (β = .16, p = .019). Within‐person associations between adolescent chaos and the subscales were not significant. Nor were there associations between caregiver reports of chaos and emotional or peer problems at either the between‐ or within‐person level. Contrary to the main internalizing model, males scored lower on emotional problems compared to females on average (β = −.15, p = .050). Household size was negatively related to peer problems (β = −.04, p = .028) but not emotional problems at the between‐level.

Sensitivity analyses

For the sensitivity analyses, the study examined relations among adolescent and caregiver perceptions of household chaos relation and both externalizing and internalizing problems, while controlling for variables closely associated with household chaos and problem behavior (see Table S1 for descriptives). For models controlling for caregiver monitoring, depression, stress and caregiver and adolescent reports of relationship quality the pattern of results was consistent (see Tables S2–S6) with our main models. Adolescent reports of household chaos were significant at the between level for externalizing and internalizing problems. Adolescent reports of household chaos were significant at the within level for externalizing problems but not internalizing problems. Caregiver reports of household chaos was not significant at the between or within level. Inclusion of caregiver monitoring, depression, stress and caregiver relationship quality were not significantly associated with externalizing or internalizing problems. However, the model controlling for adolescent reports of relationship quality (see Table S6) was significantly related to externalizing problems.

DISCUSSION

Taking advantage of an intensive 9‐month longitudinal design that drew data from adolescents and their caregivers, this study explored how household chaos is related to adolescent externalizing and internalizing problems. The results of this study challenge prevailing assumptions of stability in household chaos. Both adolescents and their caregivers reported variability in household chaos over 9 months. Compared to adolescents (ICC: 0.69), caregivers reported (ICC: 0.83) more stability in household chaos, which is consistent with past findings in the existing literature (Cassinat et al., 2021). Further, adolescent reports of chaos predicted parent reports, but surprisingly, only when looking between dyads. In other words, looking within‐dyads, when a teen reported higher‐than‐average levels of chaos in the home, it did not relate to changes in the caregiver's report of chaos. Lastly, household chaos was related to youth behavior problems. Adolescent reports of chaos predicted externalizing problems at both the within‐ and between‐person levels. For internalizing, this association was evident only at the between‐person level. Of note, caregiver reports of household chaos did not relate to either type of problem behaviors at the within‐ or between‐person levels.

Adolescents report more variability in household chaos compared to adult caregivers

The first aim of this study was to characterize the stability of household chaos over a 9‐month period. Most existing research treats household chaos as stable year‐to‐year or over several years (e.g., Deater‐Deckard et al., 2009; Human et al., 2016; Jaffee et al., 2012). The relatively large intraclass correlation for caregivers‐reports of chaos gives some support to the assumption that household chaos has intra‐year stability when considering caregiver perceptions of chaos. However, comparing adolescent ratings to caregiver ratings, adolescent perceptions of chaos are less stable, raising questions about the validity of assumptions of stability when considering adolescents' experiences. A potential explanation for these findings centers on adolescents' emotional variability. Adolescents experience more emotional variability than adults, meaning they feel more intense highs and lows and experience more swings in emotions, compared to adults (Bailen et al., 2019; Maciejewski et al., 2015; McKone & Silk, 2022). Insomuch that emotionality may relate to reports of household chaos (i.e., reports of chaos are colored by “high” and “low” feelings), adolescents may be more likely to report variation in household chaos depending on their mood. Moreover, adolescents have heightened sensitivity to environmental stimuli, which could also influence their perceptions of chaos (Somerville et al., 2010). The smoothing of emotional volatility and desensitization to environmental stimuli that develops in adulthood may explain reduced variability in adults' perceptions of chaos (Bailen et al., 2019; Maciejewski et al., 2015; McKone & Silk, 2022). An interesting future extension of this study could directly test whether variability in perceptions of household chaos changes throughout development and stabilizes in adulthood.

Adolescent and caregiver perceptions of household chaos are moderately correlated

Adolescent and caregiver reports of chaos were significantly associated, and this finding is consistent with prior work (Human et al., 2016; Jaffee et al., 2012). In other words, in families where caregivers report higher levels of household chaos, their adolescents tend to report higher levels of chaos. It is important to note, however, that while adolescent and caregiver reports were linked, the correlation was only moderate (about 0.5). This illustrates that adolescents have perceptions of chaos that are unique from their caregivers.

Interestingly, the association between adolescent and caregiver reports of chaos was non‐significant when examined within the dyad. Stated differently, while levels of adolescent‐reported chaos predicted caregiver‐reported chaos, fluctuations in an adolescent's perceptions chaos from their overall mean over time did not correlate with fluctuations in their caregiver's reports of chaos. This difference suggests that caregivers are not perceiving the same changes in their home environment that their adolescents are observing. These findings highlight the importance of measuring adolescents' perceptions of their environments in studies aimed at understanding their well‐being and development. As children age and enter adolescence, their social cognitive processes, and abilities to understand complex and intimate relationships improve (Blakemore & Mills, 2014). To the extent that social cognition acts as a pathway through which adolescents' environments affect their behavior (e.g., Bradshaw & Garbarino, 2004), understanding adolescents' perceptions of the home environment and their cognitions related to their environments is vital to fully understanding how chaos influences behavioral functioning. This study is an initial step in such work. Future studies should directly measure social cognition to explore their role as a mediator of links between chaos and behavior problems.

Household chaos has stronger links to externalizing as compared to internalizing

Adolescent reports of household chaos related to externalizing problems in both the between‐ and within‐parts of the models, while adolescent reports were only significantly associated with internalizing problems when examining between dyads. Additionally, the lack of a within‐person association for internalizing was consistent with prior studies, which have shown more stable relations between children's family contexts and externalizing versus internalizing problems (see Costello et al., 2003; Dearing et al., 2006 examining links between family income and behavior).

There are a few plausible explanations for the more consistent associations between household chaos and externalizing versus internalizing problems. First, household chaos, through heightened unpredictable contexts, unpredictability and higher stress responses, may undermine executive functions; these are the higher order cognitive processes that allow individuals to plan and execute goal‐directed control over their behaviors, thoughts and emotions (Dickerson & Kemeny, 2004). A recent meta‐analysis of 167 prospective longitudinal studies show that executive functions tend to be more consistently and strongly related to externalizing problems when compared to internalizing problems (Yang et al., 2022). When it came to links between executive functions and broad externalizing problems, the combined effect size was r = −.11 whereas the association with broad internalizing problems was r = −.07. Additionally, executive functions were linked to all 4 subtypes of externalizing problems under investigation, including substance use, oppositional defiant disorder symptoms, conduct problems, and attention‐deficit/hyperactivity disorder symptoms. Second, chaotic environments may lead some adolescents to act out more through rule‐breaking to exert some control or influence over their environment. A lack of predictability, structure, and routine in the environment can undermine one's sense of control and autonomy (Mittal & Griskevicius, 2014). Connected to this, some have similarly suggested that externalizing behaviors may be a way to communicate unmet needs, distress, and the need for parental and other support (Marcynyszyn et al., 2008). As such, externalizing behaviors may elicit desired attention from others and provide a sense of outlet or relief. Notably, recent work has found that links between stress exposure and negative health outcomes are attenuated for adolescents who scored high on externalizing behaviors (Dich et al., 2017; Doan et al., 2019). This connects to recent theories about developmental adaptations to environmental challenges that posit some behavioral responses to stress (or chaos) may be adaptive in the short‐term, but eventually costly in the long run (Ellis & Del Giudice, 2019). While more research is warranted, the current evidence suggests household chaos differentially impacts adolescent development in ways that manifest more strongly in externalizing behaviors rather than internalizing issues due to disruptions in executive functions and other potential developmental pathways.

Strengthening internal validity of connections between household chaos and problem behaviors

Compared to prior studies, the current investigation strengthens the internal validity of associations between household chaos and youth problem behaviors by reducing omitted variable bias in two different ways. First, by examining connections among within‐variability in household chaos and problem behaviors this study helps to reduce bias from time‐invariant characteristics of children and families that may be correlated with chaos and problem behaviors. While there is no way to be sure these connections are causal, based on correlational data, this analytic approach provides a more methodologically rigorous approach to testing these connections. In doing so it moves literature forward (Miller et al., 2016). Second, the sensitivity analysis directly tests whether linkages between household chaos and problem behaviors are robust to the inclusion of several other caregiver characteristics that were measured in the current study, but not the primary focus for this investigation, including caregiver monitoring, stress, depression, and relationship quality (see Tables S1–S6). Prior studies have generally not performed this type of sensitivity analysis. The results of these models were similar to those that excluded these characteristics thereby strengthening the conclusion that the underlying associations are not simply an artifact of other dimensions of caregiving that may be driving be simultaneously driving chaos and problem behaviors.

Adding specificity to bioecological models of the effects of chaos on children

Finally, it is important to tie the findings of this study to theoretical frameworks of the impacts of chaos on children and families. Bioecological Theory stresses the importance of examining how time impacts the contexts and proximal processes occurring therein that drive development (Bronfenbrenner & Morris, 1998). While most literature on chaos is focused on young children, this study reinforces the notion that chaos in the home environment is an important contextual predictor of well‐being (in this case behavioral well‐being) into adolescence as well.

Our findings also show that time is an important consideration when thinking about links between chaos and behavioral development in two major ways. First, while chaos in the home still appears to shape youth behavioral development, our results suggest that by adolescence, it is the teens' own perceptions of chaos in the home that relate to behaviors problems—not caregiver perceptions. Adolescence is a key developmental stage where teens begin to reorient towards the external world and gain acute awareness of social structures and institutions and social roles and norms (Blakemore & Mills, 2014). As a result, teens more frequently engage in social comparisons and deeply internalize the views of others (Albert et al., 2013; Jacobs, 2003; O'Brien & Bierman, 1988; Rivenbark et al., 2020; Wrzus et al., 2013). Contrary to young children who have little frame of reference to make social comparisons, this may lead adolescents to view their home context differently than their caregivers and their perceptions are stronger drivers of their behavioral functioning. This illustrates why Bioecological Theory stresses the importance of considering developmental timing. Second, these results show that the proximal context of chaos in the home environment is not stable over time. Rather, in just 9 months, we observed variation in chaos levels. Accordingly, this bolsters Bioecological Theory's focus on time as children's contexts do not remain stable over time. We continue to need rigorous research that follows children over time to measure changes in these contexts and relates contextual change with developmental change (Bronfenbrenner & Morris, 1998).

Limitations and future directions

While the 9‐month longitudinal nature of this study allowed us to use an analytic strategy that strengthens the internal validity of results (i.e., utilizing within‐person change allows individual reporters to be their own comparison group), there are a few limitations that must be discussed. First, although the study used repeated measurements of key variables of interest, this study was not experimental, so we cannot draw causal inferences. Second, although this study was interested in changes in household chaos during a year, we were only able to capture household chaos during 4 of the 9 waves. Future work could consider assessing household chaos monthly to better capture the relations with adolescent behavior. Third, the global pandemic caused disruptions to household routines and levels of disorder in the home, including stress on schedules and family relationships, so our study may be capturing levels of chaos that are out of the ordinary (Cassinat et al., 2021). The pandemic upended adolescents' social lives. Our observed associations between household chaos and internalizing problems may be underestimated since adolescents may have engaged in fewer peer interactions or engaged differently due to increased online exchanges (Larivière‐Bastien et al., 2022). Due to the timing of data collection, the study was not sufficiently powered to probe pre‐ and post‐pandemic associations. Lastly, the study did not capture processes driving adolescents' changing perceptions of household chaos. For example, the selected analytic approach did not test moderation/mediation effects with important variables such as emotional reactivity or executive functioning.

Although there are limitations, this study highlights the importance of considering adolescent perceptions of their contexts, as caregiver reports may not capture the whole lived experience of adolescents. Additionally, these findings show that household chaos may not be stable during adolescence and within‐year variability in chaos is important to measure, particularly when understanding externalizing problems. This study emphasizes that adolescents have unique perceptions from their caregivers. Therefore, future work should do more to identify factors which contribute to differences in perceptions. This investigation also shows that adolescent problem behaviors, especially externalizing problems, respond to changes in the home environment assessed over a shorter time intervals. Therefore, reducing chaos in the home environment would help diminish adolescents' problem behaviors, particularly externalizing issues. Policymakers could support families in increasing structure and routines by addressing structural factors that contribute to household chaos, such as residential instability and financial precarity through access to affordable housing and to income supports that create greater financial stability for households. However, suggesting specific solutions at the individual level is challenging, as the causes and impacts of household chaos vary widely. Addressing housing issues could reduce crowding or noise, and learning skills to create schedules could increase routines. Therefore, interventions should be adaptable to the unique challenges each family faces. To advance our understanding of how household chaos shapes adolescent behavior problems, future studies should explore alternative models that capture the dynamic nature of household chaos over time. This could involve implementing more frequent measurements of chaos, such as monthly, to capture its fluctuations or patterns of changes. Additionally, it would be useful to investigate potential mediators, such as executive functioning or emotional variability, which may play a role in the relation between household chaos and adolescent behavior problems. Furthermore, development of experimental studies that identify interventions to target chaos could reduce externalizing behavior; this would further bolster our knowledge of the role of household chaos in influencing adolescent behavior problems.

FUNDING INFORMATION

Work was funded by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01HD110423) and also an internal grant from the Learning Research and Development Center at the University of Pittsburgh awarded to Drs. Miller, Hanson and Votruba‐Drzal.

CONFLICT OF INTEREST STATEMENT

We have no known conflicts of interest disclose.

STATEMENT OF INFORMED CONSENT

Informed consent was obtained from participants for their anonymized information to be published in this article.

Supporting information

Tables S1‐S6.

Hunter‐Rue, D. S. , Miller, P. , Hanson, J. L. , & Votruba‐Drzal, E. (2024). Relations between adolescent perceptions of household chaos and externalizing and internalizing behaviors in low‐ and middle‐income families. Journal of Research on Adolescence, 34, 1500–1516. 10.1111/jora.13016

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.