Latent classes and longitudinal patterns of material hardship as predictors of child well-being

Abstract

This study examined the associations of multifaceted material hardship measured cross-sectionally and longitudinally with children’s wellbeing in the United States. Results from linear regression and child fixed effects models indicated that more intense material hardship had consistent, detrimental associations with child health status and internalizing and externalizing behaviors. More intense longitudinal patterns of material hardship were consistently associated with behaviors only. These findings examine new, multifaceted measures of material hardship and suggest associations between child wellbeing, particularly behavior challenges, and exposure both to multiple forms of material hardship and to more intense long-term patterns of hardship.

Introduction

Recent evidence suggests that in the United States (US), more than 1 in 3 children experiences material hardship in a given year, a substantial majority of whom live in households with income above the federal poverty threshold (Rodems & Shaefer, 2020). Children may face lifelong consequences from childhood deprivation (Ciula & Skinner, 2015; Duncan et al., 2010), and, importantly, deprivation may impact children differently over developmental stages and when experienced chronically (American Academy of Pediatrics, 2016; Brooks-Gunn & Duncan, 1997; Duncan et al., 1994, 1998; Ratcliffe & Mckernan, 2010). Despite this knowledge, no prior work has examined trajectories of material hardship – the experience of unmet basic needs – as they affect children over time. This study assessed how different classes and longitudinal patterns of material hardship exposure were associated with children’s wellbeing, including general health status and internalizing and externalizing behavior challenges. An important feature of this study is the inclusion of families across the income distribution in an effort to understand experiences of material hardship for children broadly, recognizing the widespread deprivation among families who do not experience income poverty (Rodems & Shaefer, 2020). These analyses provide new and compelling empirical evidence about the associations between material hardship and critical measures of child wellbeing.

Background

Deprivation and Child Wellbeing: Evidence from the Study of Income Poverty

Core theories conceptualizing the mechanisms through which income poverty leads to worse child outcomes have emphasized two complementary pathways. First, the resource and investment model has hypothesized that income poverty limits family resources, resulting in a range of reduced investments in goods and services from essential needs like food and housing to more complex investments like enrichment activities (Chaudry & Wimer, 2016; Magnuson & Votruba-Drzal, 2009). Second, the family stress model has proposed that income poverty causes parental stress about monetary concerns and results in personal, relational, and parenting stresses, including increased parental depression, marital conflict, and increased negative or detached parenting behaviors (Chaudry & Wimer, 2016; Elder, 1974; Magnuson & Votruba-Drzal, 2009). Very little extant research has examined material hardship as an independent variable, a gap the present study sought to address. Recognizing the dearth of prior research, this study drew on empirical evidence from the income poverty literature to inform study expectations and analytic design, acknowledging the close affinity between the resource and investment and family stress models and the conceptualization of material hardship as a direct measure of deprivation.

Research linking family income poverty and children’s health, mental health, behavior, and academic outcomes has been clear and consistent: overall, poor children face worse outcomes than non-poor children (Brooks-Gunn & Duncan, 1997; Chaudry & Wimer, 2016). More specifically, income-poor children have higher rates of chronic health issues, worse general health, weaker academic performance, higher rates of learning disability, and more emotional and behavior problems (American Academy of Pediatrics, 2016; Brooks-Gunn & Duncan, 1997; Chaudry & Wimer, 2016). Empirical evidence has also demonstrated the heightened negative impact of early childhood and persistent poverty (American Academy of Pediatrics, 2016; Brooks-Gunn & Duncan, 1997; Duncan et al., 1994, 1998; Ratcliffe & Mckernan, 2010).

Material Hardship and Child Wellbeing

Understanding material hardship in addition to income poverty may improve our ability to predict, ameliorate, and prevent harm to children that results from their deprivation in terms of essential needs such as hunger, unmet medical needs, homelessness, instability, and stress (Heflin, 2017). Given the abundant literature which has demonstrated clear and consistent relationships between family income poverty and child health (American Academy of Pediatrics, 2016; Brooks-Gunn & Duncan, 1997; Magnuson & Votruba-Drzal, 2009), school achievement (Brooks-Gunn & Duncan, 1997; Chaudry & Wimer, 2016; Chmielewski & Reardon, 2016; Duncan et al., 1998), and behavior outcomes (Driscoll et al., 2002; Magnuson & Votruba-Drzal, 2009; Ratcliffe & Mckernan, 2012), and the complementary nature of income poverty and material hardship, material hardship is likely to negatively impact child wellbeing.

Consistent with the resource and investment and family stress models, which have posited that income poverty impacts children’s outcomes through diminished access to resources and heightened family stress, the present study hypothesized that material hardship would be associated with children’s health and behavior outcomes in broadly similar ways as is income poverty. That is, material hardship may limit children’s access to resources (such as food, parental time, or safe housing) and increase children’s exposure to family stress (such as harsher parenting behavior or increased anxiety), which are likely to lead to poorer child wellbeing. At the most general level, this study hypothesized that increased material hardship would be associated with worse outcomes, including health status and internalizing and externalizing behaviors. Further, recognizing consistent findings in the income poverty literature, this study anticipated that early childhood material hardship (Brooks-Gunn & Duncan, 1997; Duncan et al., 1998, 2010; Ratcliffe & Mckernan, 2012), persistent material hardship (Duncan et al., 1994; Ratcliffe & Mckernan, 2010), and the experience of multiple concurrent types of hardship (American Academy of Pediatrics, 2016; Magnuson & Votruba-Drzal, 2009; Ratcliffe & Mckernan, 2012) would predict worse child outcomes than would later, less persistent, or more limited experiences of hardship.

The limited body of available research examining material hardship and child outcomes provides general support for theses hypotheses, pointing to negative implications of material hardship exposure for several dimensions of child wellbeing, including physical and mental health, cognition, behavior challenges, and educational attainment (Chaudry & Wimer, 2016; Neckerman et al., 2016; Yoo et al., 2009; Zilanawala & Pilkauskas, 2012). For instance, Zilanawala and Pilkauskas (2012) found material hardship was positively associated with increased internalizing and externalizing behavior problems among young children. Yoo and colleagues (2009) found that worse child health status was associated with material hardship, particularly food hardship, and moreover was better predicted by material hardship exposure than by family socioeconomic status. Thomas and Waldfogel (2022) also found material hardship to be a stronger predictor than income poverty of children’s involvement with child protective services.

Material Hardship as an Independent Variable

This study was motivated by the evidence that material hardship captures distinct and meaningful experiences of deprivation that are likely to impact children’s outcomes. The limited extant research on the effects of material hardship and its modest correlation with income poverty raise the real possibility that this form of deprivation may have different impacts on children than does income poverty. Material hardship measures actual experiences of unmet need, and evidence about the impacts of some discrete forms of hardship suggests that, independent of income, experiencing deprivation can be harmful to children’s wellbeing (Thomas et al., 2019). Additionally, material hardship affects different groups of children than does income poverty (Thomas, 2022; Rodems & Shaefer, 2020), perhaps posing differing risks due to these families’ varied circumstances. The very limited current research on multidimensional measures of deprivation in the US. further highlights our insufficient knowledge about the impacts of material hardship on wellbeing, a necessary prerequisite to clarifying the nuanced interplay of different sources of deprivation in shaping children’s development and wellbeing (Dhongde & Haveman, 2015; Neckerman et al., 2016).

Because research on material hardship is nascent, particularly when material hardship is considered as an independent variable, theoretical and empirical work on income poverty provides the clearest set of expectations about how material hardship may impact child wellbeing. Prior material hardship work has highlighted meaningful differences between income poverty and material hardship, which motivated the current study’s effort to improve our understanding of material hardship while building on what knowledge we have about the role of deprivation in children’s development and wellbeing.

Current Study

Drawing on prior work establishing data-driven classes and patterns of material hardship experience (Thomas, 2022), this study employed multivariate regression techniques to model the associations between material hardship classes and patterns and children’s health and behavioral outcomes. These multifaceted measures of material hardship allow for analysis of the associations of commonly co-occurring material hardship experiences with child outcomes, rather than treating individual forms of material hardship as distinct. The longitudinal patterns allow for analysis of differences in the associations between these multifaceted material hardship categories and child wellbeing among children experiencing different trajectories of hardship over a 15-year period.

This study addressed two specific research questions:

1. Are material hardship classes associated with child health status, internalizing behaviors, and externalizing behaviors?

2. Are longitudinal material hardship patterns associated with child health status, internalizing behaviors, and externalizing behaviors?

Methods

Data

This study used data from the Fragile Families and Child Well-Being Study (FFCWS). The FFCWS is an ongoing longitudinal birth cohort study, which gathers information about a baseline sample of approximately 4,900 births between 1998 and 2000 in large US cities (Reichman et al., 2001). Following baseline interviews conducted in hospitals at the focal children’s births, data have been collected when the focal child was approximately 1 year, 3 years, 5 years, 9 years, and 15 years old. At each wave, data were collected about the focal child, biological mother, and biological father, wherever possible, except in the 15-year wave, when only a single primary caregiver survey was administered. The study retained approximately 3,600 children in the 15-year wave (Princeton University, 2017).

The FFCWS intentionally oversampled families who were at higher risk of challenging circumstances, using the proxy of unmarried parent status (75% of the sample mothers were unmarried at the child’s birth) to recruit a sample which is disproportionately exposed to social, economic, racial, and material marginalization (Reichman et al., 2001; Waller, 1999). These families represent a population often harder to reach and thus not included in some data sources and as such offer valuable information about the experiences of lower-resourced, more frequently marginalized, US families.

In addition to the value of the sampling frame, FFCWS is the sole large-scale, US dataset with comprehensive information on material hardship collected at more than two time points in a national sample. No other data source provides consistent, repeated measures of numerous types of material hardship, allowing for the longitudinal analyses undertaken in this study. As a form of essential resource deprivation, material hardship is likely to have negative impacts on children’s wellbeing which may differ meaningfully over life course stages (Brooks-Gunn & Duncan, 1997; Duncan et al., 2012; Ratcliffe & Mckernan, 2012). Thus, the availability in FFCWS of material hardship measures across early and middle childhood and into adolescence provided not just repeated measures but developmentally distinct measures of material hardship experience, providing a unique and valuable tool for examining the impacts of material hardship on child wellbeing.

In addition, the material hardship measures in FFCWS are consistent with those used by the US Census Bureau in the Survey of Income and Program Participation (SIPP), which gathers cross-sectional national data on material hardship (Ouellette et al., 2004). Further, FFCWS includes comprehensive baseline information about family members’ personal identities and relationships, economic resources, and social positions and experiences as well as detailed information about children’s health and behavior. This combination of family characteristic, health, and wellbeing data are not available in conjunction with repeated material hardship measures in any other US dataset.

Sample

This study drew on a sample of 2,772 children representing all those families in which the focal child resided primarily with their biological mother and/or father and in which the custodial parent had complete material hardship data at all post-baseline study waves (1-year, 3-year, 5-year, 9-year, and 15-year). For missing family characteristics, I relied on a complete case sample, opting to avoid the biases which accompany all approaches to imputing missing while recognizing this likely resulted in a more advantaged sample and produced estimates of the effects of material hardship on child wellbeing which may not generalize to all families, particularly the most disadvantaged (see Thomas, 2022 for detailed discussion of approach to missing material hardship data). Because material hardship occurs among families with incomes substantially higher than US federal income poverty thresholds (Thomas, 2022; Karpman et al., 2018; Rodems & Shaefer, 2020), the sample was not restricted by income. Table 1 provides descriptive statistics about the sample children and their families as well comparisons to those families excluded from the analytic sample.

Table 1.

Characteristics of the sample (n=2,772) and comparison of included and excluded family characteristics

	Analytic Sample		Excluded Sample		Significant difference between samples
	(n=2,772)		(n=2,126)
Child Characteristics	Mean	SD	Mean	SD
Child’s sex (proportion female)	0.48	0.50	0.47	0.50
Child’s age, 1-year wave (months)	14.92	3.34	15.19	3.68	*
Child born at low birth weight	0.09	0.29	0.11	0.32	*
Child very good/excellent health (indicator)
1-year wave	0.87	0.33	0.85	0.36	*
3-year wave	0.89	0.31	0.86	0.35	***
5-year wave	0.89	0.31	0.88	0.33
9-year wave	0.85	0.36	0.81	0.39	**
15-year wave	0.84	0.36	0.78	0.42	***
Child internalizing behavior (std. score)
1-year wave	−0.02	0.69	0.03	0.70	*
3-year wave	−0.01	0.60	0.04	0.66	*
5-year wave	0.00	0.61	−0.01	0.62
9-year wave	0.00	0.65	0.01	0.66
15-year wave	0.00	0.66	0.01	0.71
Child externalizing behavior (std. score)
1-year wave	−0.01	0.74	0.02	0.76
3-year wave	0.01	0.59	−0.01	0.62
5-year wave	0.01	0.58	−0.03	0.58
9-year wave	0.00	0.63	−0.01	0.63
15-year wave	0.01	0.68	−0.03	0.69
Mother Characteristics
Mother’s age (years)	25.21	6.04	25.36	6.04
Mother’s race-ethnicity					***
white non-Latinx	0.23		0.18
Black non-Latinx	0.50		0.44
Latinx	0.23		0.33
Other	0.03		0.05
Mother’s education level					***
Less than high school	0.30		0.41
High school or equivalent	0.32		0.28
Some college or technical school	0.26		0.22
College degree or more	0.12		0.09
Mother’s housing tenure					***
own	0.37		0.30
rent, no gov. assistance	0.46		0.54
rent, with gov. assistance	0.05		0.07
public housing	0.11		0.10
other	0.00		0.00
Mother’s relationship status to bio father
Married	0.25		0.23
Cohabitating	0.36		0.38
Other	0.39		0.39
Mother US-born	0.88	0.33	0.77	0.42	***
Mother total number of children	2.10	1.25	2.26	1.51	***
Mother good/fair/poor health (vs. very good/excellent)	0.67	0.47	0.63	0.48	***
Mother meets criteria for depression, 1-year wave	0.16	0.36	0.15	0.36
Mother worked in year prior to child’s birth	0.15	0.36	0.22	0.42	***
Mother lived with both own parents at age 15	0.42	0.49	0.45	0.50	*
Household Characteristics
Income to poverty ratio (top coded)	2.29	2.29	2.01	2.12	***
Income to poverty ratio, categorical					***
0–49% FPL	0.17		0.22
50–99% FPL	0.17		0.18
100–199% FPL	0.26		0.25
200–299% FPL	0.16		0.15
300%+ FPL	0.25		0.20
TANF participation
1-year wave	0.25	0.43	0.23	0.42
3-year wave	0.21	0.41	0.22	0.41
5-year wave	0.18	0.38	0.19	0.39
9-year wave	0.12	0.33	0.13	0.34
15-year wave	0.10	0.30	0.13	0.34	*
Ever participated	0.42	0.49	0.96	0.19	***
No. of waves participated	0.86	1.27	0.61	0.96	***
SNAP participation
1-year wave	0.37	0.48	0.36	0.48
3-year wave	0.38	0.49	0.37	0.48
5-year wave	0.41	0.49	0.39	0.49
9-year wave	0.44	0.50	0.46	0.50
15-year wave	0.43	0.49	0.43	0.49
Ever participated	0.67	0.47	0.99	0.11	***
No. of waves participated	2.03	1.86	1.23	1.32	***
Medicaid participation
1-year wave	0.57	0.49	0.56	0.50
3-year wave	0.59	0.49	0.58	0.49
5-year wave	0.55	0.50	0.59	0.49	*
9-year wave	0.60	0.49	0.67	0.47	***
15-year wave	0.59	0.49	0.66	0.48	***
Ever participated	0.81	0.39	0.99	0.10	***
No. of waves participated	2.90	1.89	1.92	1.47	***

^***<.001;

^**<.01;

^*<.05

Some statistically significant but primarily clinically unimportant differences were evident in comparing families in the study sample with those who were excluded. For instance, children in the analytic sample had slightly higher rates of very good/excellent health, by 2% to 6% in a given wave. Similarly, mothers’ income-to-poverty ratio was slightly higher on average in the analytic sample (2.29) than among those excluded from the sample (2.01). Most differences were small in magnitude; however, sizable differences emerged in mothers’ race-ethnicity and in mothers’ educational attainment. Mothers in the analytic sample were more likely to be White or Black and less likely to be of another race or of Latinx ethnicity than those excluded from the sample. Mothers in the sample were also more likely to have higher levels of educational attainment than those excluded from the sample. Additionally, families in the excluded sample were more likely to have ever participated in income-restricted public assistance programs, including each of Temporary Assistance for Needy Families (TANF), the Supplemental Nutrition Assistance Program (SNAP), and Medicaid, than were families in the analytic sample.

Overall, these differences indicate the analytic sample included a disparate number of non-Latinx-identified families and represented families with more socioeconomic advantage relative to the excluded sample. This comparative social and economic advantage is typical of survey data, as respondents facing the greatest barriers to participation, such as economic and social instability, may be more difficult to contact for follow up surveys (Hardy, 2014; Pilkauskas et al., 2012; Stone & Rose, 2011). While expected, the analytic sample’s relative socioeconomic advantage means that the results of this study may be biased downward, while the underrepresentation of Latinx families suggests particular caution in generalizing findings to Latinx children.

Measures

Material hardship.

Material hardship is measured in every wave of the FFCWS data except for the baseline. Questions ask parent respondents to report on material hardships experienced in the past 12 months. Across all waves, nine material hardship questions are asked consistently and form the basis of the measures of material hardship used in this study, consistent with prior research (Heflin et al., 2009; Heflin & Iceland, 2009; Nepomnyaschy & Garfinkel, 2011; Pilkauskas et al., 2012; Zilanawala & Pilkauskas, 2012). This study used measures of five discrete hardships in the areas of food, housing, medical care, utilities, and essential bill-paying, in keeping with prior research that has demonstrated meaningful differences among these specific hardship types (Heflin et al., 2009). These measures are consistent with prior research on material hardship using the FFCWS and are summarized in Table 2 (Thomas, 2022; Pilkauskas et al., 2012; Zilanawala & Pilkauskas, 2012).

Table 2.

Material Hardship Measures

Domain	Measures	Citations
Food Hardship	Indicator for yes response: 1) In the past twelve months, did you receive free food or meals?	Heflin and Iceland, 2009; Pilkauskas et al., 2012; Zilanawala & Pilkauskas, 2012.
Housing Hardship	Indicator for any one or more yes responses: 1) In the past 12 months, were you evicted from your home or apartment for not paying the rent or mortgage? 2) In the past 12 months, did you move in with other people even for a little while because of financial problems? 3) In the past 12 months, did you stay at a shelter, in an abandoned building, an automobile or any other place not meant for regular housing even for one night?	Heflin and Iceland, 2009; Pilkauskas et al., 2012; Zilanawala & Pilkauskas, 2012.
Medical Hardship	Indicator for yes response: 1) In the past 12 months, was there anyone in your household who needed to see a doctor or go to the hospital but couldn’t go because of the cost?	Heflin and Iceland, 2009; Pilkauskas et al., 2012; Zilanawala & Pilkauskas, 2012.
Utility Hardship	Indicator for any one or more yes responses: 1) In the past 12 months, was service turned off by the gas or electric company, or did the oil company not deliver oil? 2) In the past 12 months, was service disconnected by the telephone company because payments were not made?	Pilkauskas et al., 2012; Zilanawala & Pilkauskas, 2012.
Bill-Paying Hardship	Indicator for any one or more yes responses: 1) In the past 12 months, did you not pay the full amount of rent or mortgage payments because there was not enough money? 2) In the past 12 months, did you not pay the full amount of a gas, oil or electricity bill because there was not enough money?	Pilkauskas et al., 2012; Zilanawala & Pilkauskas, 2012.

The present study operationalized material hardship as a latent construct reflecting classes made up of distinct groups of specific hardships, identified using latent class analysis (LCA). In addition to cross-sectional classes of material hardship, the study also examined longitudinal patterns of material hardship, reflecting patterns of transition and stability between latent class groups over the five survey time points in the FFCWS. These patterns were identified using latent transition analysis (LTA). Prior work described the development of these classes and patterns in extensive detail (Thomas, 2022); however, in brief, LCA and LTA methods examine measured variables which are hypothesized, when considered together, to approximate an unmeasured, or latent, construct. In this case, the LCA analysis examined measured material hardship indicators and derived latent, probability-based groupings (classes) of those indicators which constituted the values of a single, unmeasured material hardship class construct. Following similar logic, the LTA analysis described patterns of membership in these material hardship classes across five time points, ultimately identifying the six most statistically probable patterns of class membership over the study period.¹ That study was the first to identify latent material hardship constructs and to measure longitudinal patterns of material hardship. Specifically, the study identified cross-sectional material hardship classes, with three values at each time point, and longitudinal material hardship patterns, with six values across the study period.

The cross-sectional class values, labeled limited, moderate, and severe material hardship, were distinguished by the presence of specific forms of hardship (e.g., bill-paying hardship) and the number of hardships (e.g., 3–4 unique forms of hardship) common to each class. Figure 1 illustrates the make-up of the material hardship classes (Thomas, 2022). The prevalence of each material hardship class differed somewhat by survey wave, though the limited class was consistently most common while the severe class was consistently least common. The proportion of the sample in the limited class ranged from 44% at the 9-year wave to 61% at the 15-year wave; the moderate class accounted for between 25% of the sample at the 15-year wave and 42% at the 3-year and 9-year waves; finally, the severe class included 7% at the 1-year wave and 14% at the 15-year wave (see Thomas, 2022, for additional details).

Figure 1. Material Hardship Item Probability Parameters by Class (as published in Thomas, 2022).

The longitudinal patterns of material hardship reflect transitions over time between the three material hardship classes. Three of the six patterns indicate relative stability in a given hardship class (the mostly limited, mostly moderate, and mostly severe patterns); two of the six patterns indicate changing intensity of hardship class over time (the improving and worsening patterns); and the final, inconsistent pattern includes a small proportion of the sample (7%) with no clear pattern across time. Figure 2 demonstrates the prevalence of each pattern within the sample (Thomas, 2022).

Figure 2. Prevalence of Material Hardship Patterns (as published in Thomas, 2022).

The present study uses these measures of material hardship class (with three values, limited, moderate, and severe) and material hardship pattern (with six values, mostly limited, mostly moderate, mostly severe, improving, worsening, and inconsistent) as the primary analytic measures of material hardship throughout the analyses.

Child wellbeing.

Child wellbeing measures were derived from primary caregiver data. Using data from each of the 1-year to 15-year waves, child health status was measured as an indicator for very good or excellent health versus good, fair, or poor health, as indicated by primary caregiver report.

Internalizing and externalizing behavior were measured using primary caregiver responses to two different, established child behavior scales. At the 1-year wave, primary caregivers responded to questions from the emotionality and shyness scales of the Emotionality, Activity, and Sociability (EAS) Temperament Survey for Children (Mathiesen & Tambs, 1999). Internalizing behavior was measured as a standardized index of three questions from the shyness scale, while externalizing behavior was measured as a standardized index of three measures from the emotionality scale (Table 3). In the 3-year through 15-year waves, primary caregivers responded to items from the Child Behavior Checklist (CBCL)/6–18 (Achenbach & Rescorla, 2001). Specifically, internalizing behavior was measured as a standardized index of five questions from the CBCL/6–18 anxious/depressed and withdrawn subscales (Appendix C). Externalizing behavior was measured as a standardized index of nine questions from the CBCL/6–18 aggressive and rule-breaking subscales (Table 4).

Table 3.

Emotionality, Activity, and Sociability (EAS) Temperament Survey for Children subscale items from Fragile Families and Child Wellbeing Study 1-Year Data User’s Guide

	Item in FFCWS	Source Item
Shyness
	m2b17a, f2b16a, m2b43a, f2b37a	Tends to be shy
		Makes friends easily (R)
	m2b17c, f2b16c, m2b43c, f2b37c	Is very sociable (R)
		Takes a long time to warm up to strangers
	m2b17f, f2b16f, m2b43f, f2b37f	Is very friendly with strangers (R)
Emotionality		Cries easily
		Tends to be somewhat emotional
	m2b17b, f2b16b, m2b43b, f2b37b	Often fusses and cries
	m2b17d, f2b16d, m2b43d, f2b37d	Gets upset easily
	m2b17e, f2b16e, m2b43e, f2b37e	Reacts intensely when upset

Table 4.

Child Behavior Checklist (CBCL)/6–18 subscale items from Fragile Families and Child Wellbeing Study 15-Year Data User’s Guide

CBCL Subscale	FFY15 PCG Survey Item
Aggressive behavior	1. Child is cruel, bullies, or shows meanness to others
	2. Child destroys things belonging to the family or others
	3. Child is disobedient at home
	4. Child is disobedient at school
	5. Child gets in many fights
	6. Child physically attacks people
	7. Child is stubborn, sullen, or irritable
	8. Child has temper tantrums or a hot temper
	9. Child threatens people
	10. Child is unusually loud
	11. Child argues a lot
Anxious/depressed behavior	1. Child cries a lot
	2. Child feels worthless or inferior
	3. Child is nervous, highstrung, or tense
	4. Child is too fearful or anxious
	5. Child feels too guilty
	6. Child worries
Rule-breaking behavior	1. Child doesn’t seem to feel guilty after misbehaving
	2. Child hangs around with others who get in trouble
	3. Child lies or cheats
	4. Child runs away from home
	5. Child sets fires
	6. Child steals at home
	7. Child steals outside the home
	8. Child swears or uses obscene language
	9. Child vandalizes
Withdrawn	1. Child is underactive, slow moving, or lacks energy
	2. Child is unhappy, sad or depressed
Internalizing Behaviors	All variables from anxious/depressed and withdrawn subscales
Externalizing Behaviors	All variables from aggressive and rule-breaking subscales

Family characteristics.

A comprehensive set of variables reflecting families’ social positioning characteristics were included in this study. These characteristics were drawn from mothers’ reports at the baseline interviews and specifically reflected prior research identifying important demographic correlates of material hardship (Heflin, 2017; Karpman et al., 2018; Wimer et al., 2014). Baseline measures were selected for several reasons: (1) they represented the family’s situation at the beginning point of the longitudinal trajectories this study examined; (2) they captured family circumstances prior to the focal child’s exposure to any material hardship; and (3) they were indicative of the family’s situation in the critical early childhood developmental period.

Child measures included: sex at birth, whether child was born at low birth weight, and child’s age at the 1-year interview. This child age measure was used because all children were within days of birth at the baseline interview, but 1-year interviews captured children ranging in age from 9 to 30 months.

Mother measures included: age (in years), race (white, Black, and other race(s)) and Latinx ethnicity (yes or no, in combination with any selected race), education level (less than high school to college degree or more), housing tenure (own, rent with or without government assistance, public housing, other), relationship status to child’s biological father (married, cohabitating, other), whether mother was born in the US, how many children mother had, an indicator measure of health status (good/fair/poor versus very good/excellent), whether mother met criteria for depression (per Composite International Diagnostic Interview - Short Form (CIDI-SF); 1-year data used because mental health was not measured at baseline), whether mother worked during the year prior to the focal child’s birth, and whether mother lived with both of her own parents at age 15.

Household measures included: income-to-poverty ratio (continuous measure top-coded at 95% to account for outliers and categorical measure for ease of interpretation) and public assistance receipt. The FFCWS data contain sufficient information to create family-level measures of public assistance receipt for key public programs. This study used indicator measures drawn from primary caregiver questionnaire responses at each post-baseline wave. These measures indicated whether the study focal child or caregiver participated in TANF, SNAP, or Medicaid in a given wave. Additionally, two cumulative measures were constructed: an indicator variable for families having ever participated in each of TANF, SNAP, and Medicaid over the 1-year to 15-year timeframe and a variable summing the total number of waves in which families participated in each of the programs.

Analyses

I conducted two sets of analyses to examine the two different material hardship measures described in the Methods section. First, using the cross-sectional material hardship class measure, I examined associations between wave-specific, time-varying material hardship class and wave-specific child wellbeing. Second, I assessed associations between the time-invariant longitudinal material hardship pattern measure, which captures the trajectory of hardship experience across the study’s 15-year timeframe, and child wellbeing measured at the 15-year wave.

To examine the associations between cross-sectional material hardship class and child wellbeing outcomes I first used linear regression to predict pooled outcomes from material hardship class, controlling for baseline family background characteristics. As the first study examining material hardship class as a predictive variable, these analyses are a novel contribution. However, there is likely substantial endogeneity between material hardship experiences and child outcomes. So, while this analysis controlled for a number of measured covariates, the results should still be considered non-causal associations.

Second, because of this likely endogeneity in the prediction of outcomes from material hardship class, I capitalized on the robust analytic techniques available for use with repeated measures data to get closer to an unbiased estimate of the impact of material hardship class on child wellbeing. Specifically, I used child fixed effects linear regression models to predict within-child differences in health and behavior outcomes from material hardship class. Fixed effects models evaluate change in the outcome over time while omitting the effects of all time-invariant predictors, both measured and unmeasured (Allison, 2009). This capacity is particularly useful for eliminating bias from unmeasured but likely time-invariant characteristics that might predict both material hardship and child wellbeing, such as personality traits or family norms (Firebaugh et al., 2013).

To assess the association between longitudinal material hardship patterns and wellbeing outcomes, I used linear regression to predict each 15-year outcome from longitudinal material hardship pattern, controlling for baseline family characteristics. In these analyses, the mostly limited material hardship pattern, representing the least intensity of material hardship, served as the reference group. Thus, the models identified whether children in families experiencing each of the five other material hardship patterns were more or less likely to have better health status, internalizing behavior, and externalizing behavior outcomes, compared to the mostly limited pattern group.

These analyses represent an important contribution to the field as the first examination of the association between longitudinal patterns of material hardship and children’s outcomes. Still, these analyses faced the same concerns regarding endogeneity described above. Controlling for potentially confounding covariates should have improved the validity of the estimated associations between material hardship pattern and child outcomes but will not have fully addressed the likely possibility that unmeasured and thus uncontrolled characteristics of children and families may explain both material hardship pattern and children’s wellbeing outcomes. Nonetheless, this analysis represents an important first step, providing new evidence contributing to our understanding of the impacts of material hardship.

Results

Results of the linear regression models predicting child health status, internalizing behavior, and externalizing behavior from cross-sectional material hardship class are presented in Table 5. The results indicated that moderate and severe material hardship classes, compared to limited, were consistently and statistically significantly associated with worse health status (2% and 4% decreases compared to mean levels in the sample, respectively), greater internalizing behaviors (0.04 standard deviations (SD) and 0.17 SD, respectively), and greater externalizing behaviors (0.09 SD and 0.19 SD, respectively), accounting for baseline family characteristics and time-varying public assistance receipt.

Table 5.

Associations Between Material Hardship Classes and Child Wellbeing

	Very Good/Excellent Health (indicator)			Internalizing Behavior (standardized scale)			Externalizing Behavior (standardized scale)
	n=12,738 child-year observations			n=11,951 child-year observations			n=11,614 child-year observations
Predictor	Coefficient	SE	Sig.	Coefficient	SE	Sig.	Coefficient	SE	Sig.
Material Hardship Class (vs. Limited)
Moderate	−0.020	0.01	*	0.043	0.02	**	0.093	0.02	***
Severe	−0.040	0.01	**	0.174	0.03	***	0.191	0.03	***
Child’s sex (proportion female)	0.023	0.01	**	0.017	0.01		−0.069	0.02	***
Child’s age, 1-year wave (months)	−0.001	0.00		−0.004	0.00		0.000	0.00
Child born at low birth weight	−0.056	0.02	***	0.029	0.03		0.056	0.03
Mother’s age (years)	−0.004	0.00	***	0.003	0.00		−0.002	0.00
Mother’s race-ethnicity (vs. Black, non-Latinx)
white non-Latinx	0.021	0.01	*	0.060	0.02	**	0.031	0.02
Latinx	−0.005	0.01		0.004	0.02		0.008	0.02
Other	0.034	0.02		0.026	0.04		0.114	0.04	*
Mother’s education level (vs. college degree)
Less than high school	−0.032	0.02	*	0.076	0.03	*	0.039	0.04
High school or equivalent	−0.002	0.01		0.032	0.03		−0.001	0.03
Some college or technical school	−0.007	0.01		−0.047	0.03		−0.041	0.03
Mother’s housing tenure (vs. own)
rent, no gov. assistance	−0.001	0.01		−0.030	0.02		−0.040	0.02	*
rent, with gov. assistance	0.020	0.02		−0.015	0.03		−0.017	0.04
public housing	−0.004	0.02		0.010	0.03		0.056	0.03
other	−0.002	0.04		0.125	0.10		0.029	0.10
Mother’s relationship status to bio father (vs. married)
Cohabitating	−0.025	0.01	*	0.031	0.02		0.023	0.02
Other	−0.023	0.01		0.020	0.02		0.031	0.03
Mother US-born	0.066	0.02	***	−0.048	0.03		0.046	0.03
Mother total number of children	−0.003	0.00		−0.008	0.01		0.003	0.01
Mother good/fair/poor health (vs. very good/excellent)	0.073	0.01	***	−0.068	0.02	***	−0.044	0.02	*
Mother meets criteria for depression, 1-year wave	−0.019	0.01		0.104	0.02	***	0.118	0.03	***
Mother worked in year prior to child’s birth	−0.028	0.01	*	0.007	0.02		−0.014	0.03
Mother lived with both own parents at age 15	−0.012	0.01		0.019	0.02		−0.021	0.02
Income to poverty ratio (vs. 300%+ FPL)
0–49% FPL	−0.040	0.02	*	0.043	0.03		0.038	0.04
50–99% FPL	−0.046	0.02	**	0.063	0.03	*	0.049	0.03
100–199% FPL	−0.033	0.01	**	0.054	0.02	*	0.067	0.03	**
200–299% FPL	−0.018	0.01		0.007	0.02		0.017	0.03
TANF participation, in wave	0.005	0.01		0.063	0.02	**	0.036	0.02
SNAP participation, in wave	−0.021	0.01	*	0.018	0.02		0.043	0.02	*
Medicaid participation, in wave	−0.016	0.01	*	0.039	0.02	*	0.021	0.02

^***<.001;

^**<.01;

^*<.05;

^{^}<.10

Table 6 presents results of child fixed effects linear regression models, which examined within-child differences in the associations between material hardship class and child health status, internalizing behavior, and externalizing behavior. The results of these models indicated that accounting for all time-invariant factors and time-varying TANF, SNAP, and Medicaid participation reduced the strength of the association between moderate material hardship (versus limited material hardship) and child wellbeing. Specifically, moderate material hardship was no longer associated with increased internalizing behavior and was only marginally (p<.10) associated with worse health status and increased externalizing behavior in these models. By contrast, the association between severe material hardship and worse child wellbeing remained statistically significant, even accounting for child fixed effects. Severe material hardship class was associated with worse health status (4.4% decrease), with greater internalizing behaviors (0.12 SD), and with greater externalizing behaviors (0.19 SD) in these models.

Table 6.

Material Hardship Classes and Child Wellbeing – Fixed Effects Models

	Very Good/Excellent Health (indicator)			Internalizing Behavior (standardized scale)			Externalizing Behavior (standardized scale)
	n=13,754 child-year observations			n=12,870 child-year observations			n=12,503 child-year observations
Predictor	Coefficient	SE	Sig.	Coefficient	SE	Sig.	Coefficient	SE	Sig.
Material Hardship Class (vs. Limited)
Moderate	−0.019	0.01		−0.012	0.02		0.037	0.02
Severe	−0.044	0.02	*	0.121	0.04	**	0.192	0.04	***
Received TANF	0.020	0.01		0.050	0.02	*	−0.015	0.02
Received SNAP	−0.021	0.01	*	0.000	0.02		0.007	0.02
Received Medicaid	0.002	0.01		−0.007	0.02		−0.016	0.02

^***<.001;

^**<.01;

^*<.05;

^{^}<.10

Results of the linear regression models examining the relationship between longitudinal material hardship patterns and child health status, internalizing behavior, and externalizing behavior are presented in Table 7. The models compared each of five material hardship patterns – mostly moderate, mostly severe, improving, worsening, and inconsistent – to the mostly limited material hardship pattern, controlling for baseline family background characteristics and TANF, SNAP, and Medicaid participation at any time. These results indicated that four of these material hardship patterns had a statistically significant impact on internalizing and/or externalizing behaviors. In predicting internalizing behaviors, the mostly moderate pattern was associated with a 0.13 SD increase, the mostly severe pattern with a 0.35 SD increase, the improving pattern with a 0.09 SD increase, and the worsening pattern with a 0.19 SD increase. Examining associations with externalizing behaviors, the mostly moderate pattern was associated with a 0.17 SD increase, the mostly severe pattern with a 0.23 SD increase, and the worsening pattern with a 0.21 SD increase. In contrast, only the worsening (5.8% decrease) material hardship pattern had a significant and negative association with child health status.

Table 7.

Associations Between Material Hardship Patterns and Child Wellbeing (n=2,551)

	Very Good/Excellent Health (indicator)			Internalizing Behavior (standardized scale)			Externalizing Behavior (standardized scale)
	Coefficient	SE	Sig.	Coefficient	SE	Sig.	Coefficient	SE	Sig.
Predictor
Material Hardship Pattern (vs. Mostly Limited)
Mostly Moderate	−0.012	0.02		0.134	0.04	***	0.166	0.04	***
Mostly Severe	−0.059	0.03		0.348	0.06	***	0.225	0.06	***
Improving	0.017	0.02		0.094	0.04	*	0.083	0.04
Worsening	−0.058	0.02	*	0.185	0.05	***	0.206	0.05	***
Inconsistent	−0.012	0.03		0.095	0.05		0.076	0.05
Child’s sex (proportion female)	0.007	0.01		0.114	0.03	***	−0.052	0.03	*
Child’s age, 1-year wave (months)	−0.001	0.00		−0.007	0.00		0.000	0.00
Child born at low birth weight	−0.011	0.02		0.031	0.04		0.070	0.05
Mother’s age (years)	−0.007	0.00	***	0.004	0.00		−0.007	0.00	*
Mother’s race-ethnicity (vs. Black, non-Latinx)
white non-Latinx	0.026	0.02		0.239	0.04	***	0.068	0.04
Latinx	0.006	0.02		0.052	0.04		−0.003	0.04
Other	0.032	0.04		0.125	0.07		0.189	0.08	*
Mother’s education level (vs. college degree)
Less than high school	−0.058	0.03		0.088	0.06		−0.060	0.06
High school or equivalent	−0.008	0.03		0.063	0.06		−0.109	0.06
Some college or technical school	−0.026	0.03		−0.025	0.05		−0.114	0.05	*
Mother’s housing tenure (vs. own)
rent, no gov. assistance	0.006	0.02		−0.074	0.03	*	−0.030	0.03
rent, with gov. assistance	−0.021	0.03		0.022	0.06		0.084	0.06
public housing	−0.009	0.03		0.029	0.05		0.113	0.05	*
other	−0.119	0.11		0.170	0.19		0.088	0.20
Mother’s relationship status to bio father (vs. married)
Cohabitating	−0.052	0.02	*	0.087	0.04	*	−0.002	0.04
Other	−0.023	0.02		0.059	0.04		0.028	0.04
Mother US-born	0.051	0.03		−0.076	0.05		0.025	0.05
Mother total number of children	0.013	0.01		−0.011	0.01		0.007	0.01
Mother good/fair/poor health (vs. very good/excellent)	0.093	0.02	***	−0.073	0.03	**	−0.077	0.03	**
Mother meets criteria for depression, 1-year wave	0.014	0.02		0.157	0.04	***	0.122	0.04	***
Mother worked in year prior to child’s birth	−0.036	0.02		−0.025	0.04		−0.042	0.04
Mother lived with both own parents at age 15	−0.009	0.02		−0.005	0.03		−0.027	0.03
Income to poverty ratio (vs. 300%+ FPL)
0–49% FPL	0.000	0.03		−0.120	0.05	*	0.017	0.06
50–99% FPL	−0.048	0.03		−0.056	0.05		0.015	0.05
100–199% FPL	−0.033	0.02		−0.020	0.04		0.107	0.04	*
200–299% FPL	−0.032	0.02		0.008	0.04		0.014	0.05
TANF participation, in wave	−0.028	0.02		0.071	0.03	*	0.050	0.03
SNAP participation, in wave	−0.069	0.02	**	0.060	0.04		0.097	0.04	*
Medicaid participation, in wave	0.024	0.03		0.000	0.05		−0.033	0.05	0.476

^***<.001;

^**<.01;

^*<.05;

^{^}<.10

Discussion

Examining material hardship as an independent variable is still unusual in the literature (Heflin & Iceland, 2009; Mccarthy et al., 2016; Yoo et al., 2009; Zilanawala & Pilkauskas, 2012), so this study builds out the body of research which conceptualizes material hardship as a way to understand families’ deprivation rather than solely as a potential consequence which families may face. By investigating material hardship as a measure of economic need, one that can be considered in the same contexts as income poverty often is, this study expands the conceptualization of poverty and deprivation. Further, this study extends the examination of deprivation to a notably different and larger group of children whose families struggle to meet their essential needs but whose experiences are not captured in studies looking only at families experiencing income poverty (Rodems & Shaefer, 2020; Thomas, 2022).

As researchers work to understand if and how material hardship negatively effects child and family wellbeing, it is essential to explore various definitions and conceptualizations of material hardship in order to ensure this new evidence provides meaningful, robust, and nuanced information from which we can draw policy and practice conclusions. This study contributes to that effort by directly examining the associations of both cross-sectional material hardship classes and longitudinal material hardship patterns with child wellbeing. Moreover, this analysis also posits a direct association between material hardship and child wellbeing, conceptualizing material hardship as an independent experience of deprivation (Mccarthy et al., 2016; Neckerman et al., 2016; Zilanawala & Pilkauskas, 2012), different than other measures of economic privation.

In the results of this study, the associations between material hardship class and child health status, internalizing behavior, and externalizing behavior at a given timepoint were consistent: more intense material hardship class was predictive of worse health status and more internalizing and externalizing behavior challenges, accounting for both family social positioning characteristics and public assistance program participation. These results, which are consistent with limited prior work, suggest that understanding material hardship class provides valuable evidence about the deleterious impacts of deprivation on children’s health and behaviors (Yoo et al., 2009; Zilanawala & Pilkauskas, 2012). In child fixed-effects models, which addressed time-invariant confounding even by factors unmeasured in the study, associations between moderate material hardship (versus limited) and child wellbeing outcomes were attenuated, but severe material hardship (versus limited) was consistently associated with worse child health and more behavior problems. The persistent associations between severe material hardship class and worse child health status, internalizing behavior, and externalizing behavior, accounting for a number of family characteristics and employing robust modeling techniques, provide compelling evidence that children’s exposure to intense material hardship is likely to have broad-reaching and negative consequences for their wellbeing.

The study’s final set of results provides novel findings on the associations between longitudinal material hardship and child wellbeing. Other forms of deprivation, including income poverty, have different impacts on child wellbeing depending on timing and duration, and this study provides the first evidence about how differing timing, duration, and intensity of material hardship exposure impacts children’s wellbeing (Brooks-Gunn & Duncan, 1997; Magnuson & Votruba-Drzal, 2009; Ratcliffe & Mckernan, 2010, 2012).

First, it is important to note that child health status, at least as measured in this study, was at most weakly associated with material hardship pattern, and indeed this association may have been entirely spurious. Specifically, the lack of association between nearly all material hardship patterns and child health status may reflect the limited variability in the health status measure. Few children in the study were reported by their primary caregiver to be in poor health, and thus the analysis may have been unable to detect meaningful differences in children’s health status by material hardship experience. Future work might draw on the biometric information available in the restricted FFCWS data to examine other health outcomes in order to better understand how longitudinal material hardship may impact physical health. Additionally, because some physical health problems can take time to develop, data from the forthcoming, 22-year wave of the FFCWS may provide valuable information on the health status of focal children as they enter adulthood (Ratcliffe & Mckernan, 2012) and provide the opportunity to examine impacts of material hardship experienced across the life course on the gradual development of health problems over time.

While health status was not related to longitudinal material hardship pattern, greater internalizing and externalizing behavior problems were consistently predicted by more intense material hardship patterns, with the strongest associations related to the most intense (mostly severe) or most recently intense (worsening) patterns. These findings suggest that persistent material hardship may be particularly detrimental to child wellbeing. Additionally, the relative strength of worsening compared to improving material hardship patterns in predicting poor child outcomes suggests that the recency of severe material hardship may be an important factor in the consequences of hardship for children, as the worsening pattern is characterized by increasing levels of material hardship over time, with the most intense taking place closest to the measurement of children’s health and behavior. That not every pattern of material hardship was strongly associated with compromised wellbeing is promising, indicating that intervention to reduce or mitigate hardship may have benefits for children’s wellbeing, even if material hardship cannot be fully eliminated.

Limitations

The FFCWS dataset is limited in two ways important to this study. First, the study cohort is intentionally limited in its generalizability, designed to capture the experience of families facing precarious circumstances but therefore not representative of the US. Most importantly, the present study cannot speak to the experiences of material hardship among families in non-urban settings, although it seems plausible that the child wellbeing consequences of hardship may translate across place. A second limitation to the design of the FFCWS data is the inconsistent timing of survey waves, which take place anywhere from one to six years apart. Most important for this study is the six-year period between the 9-year and 15-year waves as this represents both an extended time period in which many families could have a range of exposures to material hardship and a substantial portion of middle childhood and adolescence about which the present study cannot draw conclusions. Nonetheless, the FFCWS dataset remains the sole US source of repeated, consistent measures of material hardship in combination with child wellbeing measures and therefore offered the best available data for this study.

Implications

This study strengthens the evidence on the harmful associations between material hardship and children’s wellbeing, and the specific results have implications for policy that might prevent or respond to material hardship. Namely, the relative strength of both persistently and recently intense material hardship in shaping child wellbeing suggests the value of intervening to reduce hardship whenever possible (rather than identifying a critical developmental window). New or expanded US social welfare policies designed to mitigate experiences of hardship directly, such as food and housing assistance programs, might well reduce exposure to hardship and improve child wellbeing. While distinct from income poverty, material hardship as measured in this study is still rooted in families’ lack of financial resources to meet their essential needs. Thus, a substantial and consistent expansion of economic resources for families with children could increase their capacities to meet their needs, reducing hardship and improving children’s physical and behavioral health. Whether through expansion of the current US child tax credit or adoption of a child allowance akin to those provided in many peer nations, a sizeable and consistent expansion of economic resources targeted to families with children could meaningfully reduce material hardship with expected benefits to child wellbeing.

Data availability

The data are publicly available to registered users through the Princeton University Office of Population Research data archive, accessible here: https://opr.princeton.edu/archive/restricted/Default.aspx