Risk-taking behaviors of homeless youth: moderation by parental monitoring and social support

ABSTRACT

Adolescence is a critical developmental period when youth are transitioning to be more independent while also being at heightened risk for negative developmental trajectories. This is especially true for adolescent youth experiencing homelessness. Using state population-representative 2017 Youth Risk Behavior Survey (YRBS) data, we examined the association between youth homelessness and engagement in risk-taking behaviors. The moderating roles of parental monitoring and social support were also examined. We found that 2.8% (8,922) of students were experiencing homelessness, and youth homelessness was associated with engagement in a greater number of risk-taking behavior types. The association between youth homelessness and risk-taking behaviors was not moderated by parental monitoring, which was high for stably housed youth as well as homeless youth. Social support was a significant moderator. For stably housed youth, the presence of someone to talk to when youth were feeling sad was associated with engagement in fewer risk-taking behaviors. Conversely, for homeless youth, having someone to talk to when the youth were feeling sad was associated with higher engagement in risk-taking behaviors. Findings extend our understanding of risk-taking behaviors among homeless youth, a particularly vulnerable group of adolescents. The study has implications for prevention and intervention efforts to support homeless youth.

KEYWORDS:

• Youth homelessness
• risk-taking behaviors
• parental monitoring
• social support
• Youth Risk Behavior Survey (YRBS)

Introduction

During the 2016–2017 school year, an estimated 1.35 million students in the United States were identified as homeless under the McKinney-Vento Act (National Center for Homeless Education 2019). Under this federal legislation, homeless youth are defined as students lacking a fixed, regular, and adequate nighttime residence. These students may be living in shelters, hotels or motels, doubled-up (i.e., living with others due to loss of housing or economic hardship), or in other temporary housing situations (e.g., cars, parks, and campgrounds). Homeless youth are a vulnerable population at a heightened risk for many negative outcomes, including mental health challenges (Armstrong, Owens, and Haskett 2018) and low academic achievement (Edidin et al. 2012). Of particular concern, and the focus of this study, is the elevated rates of risk-taking behaviors among youth experiencing homelessness compared to their stably housed peers (Busen and Engebretson 2008; Cauce et al. 2000). We sought to add to the understanding of these dangerous behaviors and to explore factors that might mitigate the link between homelessness and risk-taking.

Risk-taking behaviors of homeless youth

Overall, individuals tend to participate in more risky behaviors during adolescence than during any other developmental stage (Schulenberg et al. 2018). For our paper, we followed the “Theories of Adolescent Risk-Taking Behavior” by Igra and Irwin (1996), which defines risk-taking behaviors as a variety of potentially health-damaging behaviors, including risky sexual behaviors, substance use, and reckless vehicle use. Among adolescents, youth experiencing homelessness tend to engage in risky behaviors more often than do their stably housed peers (Busen and Engebretson 2008; Cauce et al. 2000). In regard to risky sexual behaviors, research indicates that many youth experiencing homelessness have multiple sex partners and engage in unprotected sex, survival sex, and prostitution at higher rates compared to stably housed youth (Edidin et al. 2012). Halcón and Lifson (2004) found that among homeless youth in their study, 55% of males and 74% of females reported engaging in sexual intercourse within the past 30 days. Almost 40% of the youth had two or more sexual partners in the past month and over 50% reported frequently not using a condom. Furthermore, Halcón and Lifson found that over half of their female participants had been pregnant at least once, and almost 30% had been pregnant two or more times. Although there was no comparison group of housed youth in this study, the authors did note that the rates of risky sexual activity were higher among homeless youth compared to their stably housed peers.

In addition to sexual risk-taking behaviors, unaccompanied youth (a subpopulation of homeless youth living without a designated caregiver) exhibit higher rates of substance abuse (including alcohol, marijuana, amphetamines, tobacco, and heroin) compared to stably housed youth (Baer, Ginzler, and Peterson 2003; Votta and Manion 2004). Across studies, prevalence rates of substance use among homeless youth range from 70% to 90%, and these rates are positively associated with age and experiences of stress or disequilibrium (Edidin et al. 2012; Ginzler et al. 2007; Halcón and Lifson 2004; Parriott and Auerswald 2009). Rhoades conducted a study in 2014 examining the rates of prescription drug misuse—another variable of interest for the current study—and found that almost 50% of homeless youth reported misusing prescription drugs, including opioids, sedatives, and stimulants. In a longitudinal study examining lifetime and 12-month prevalence rates of substance abuse disorders among homeless and runaway youth, Johnson, Whitebeck, and Hoyt (2005b) found that nearly 65% of participants met lifetime criteria for at least one substance abuse disorder, and almost half met 12-month criteria for at least one substance abuse disorder. Furthermore, nearly all youth who met criteria for a substance abuse disorder also met criteria for at least one other mental health disorder, including depression.

Although homeless youth, as a group, tend to engage in risk-taking behaviors at higher rates than housed youth, it is important to note that there are individual differences in risk-taking among youth who experience homelessness. To illustrate, Cutuli et al. (2013) found that about 45% of homeless and highly mobile youth maintained average or above average academic achievement. Other investigators found that some youth, despite being homeless, do not engage in risky behaviors or demonstrate adverse social-emotional outcomes (Kidd and Shahar 2008; Masten et al. 2008; Tevendale, Lightfoot, and Slocum 2008). Thus, there are youth experiencing homelessness who appear resilient in terms of risk-taking behaviors. Identifying factors that differentiate homeless youth who engage in risk-taking behaviors from those who do not can contribute to prevention and intervention planning for these vulnerable youth. Previous researchers have identified some important protective factors for homeless youth, including high self-esteem, ability to self-regulate, and high parenting quality (Kidd and Shahar 2008; Masten et al. 2008). We focused on two potential protective factors to explain variability in the relation between housing status and risky behaviors: social support and parental monitoring. A brief review of the literature follows, to explain why we selected these two particular variables for our study.

Potential protective factors

Previous research indicates that a greater level of social connectedness, including social connections with family, school, and prosocial peers, is associated with lower levels of psychological distress among homeless youth (Dang 2014). Additionally, researchers have found that the type of social support, identity of the social support provider, and depth of the relationship are important factors in the relation between homelessness and youth risk behaviors (Dang 2014; Haye et al. 2012; Rice, Milburn, and Rotheram-Borus 2007). Thus, having a strong social support network could mitigate engagement in risk-taking behaviors. In general, providers of support for homeless youth are frequently family members, sex partners, and service providers (Haye et al. 2012). When controlling for family abuse, Haye et al. (2012) found that family members were an important source of both tangible and emotional support for homeless youth. However, homeless youth who have experienced caregiver abuse are more likely to have smaller emotional networks due to fewer or rapidly declining home ties when compared to non-abused homeless youth (Falci et al. 2011). Although the literature highlights social networks as a potential protective factor, previous studies tended to focus on just one type of risk-taking behavior, with most examining sexual risk-taking and/or substance use behaviors. To illustrate, homeless youth who report having more friends who attend school and get along well with their families are at decreased risk of engaging in HIV risk behaviors and hard drug use (Rice, Milburn, and Rotheram-Borus 2007). Ennett, Bailey, and Federman (1999) found that homeless youth with larger social networks were less likely to engage in substance abuse, less likely to have multiple sex partners, and less likely to engage in survival sex than homeless youth with small or non-existent social networks.

In addition to social support, parental monitoring is associated with positive outcomes among both homeless and stably housed youth. Parental monitoring and supervision include the development of rules and expectations that parents demand from their children and the amount of time children are with their parents (Hayes, Hudson, and Matthews 2003). Adequate parental monitoring is associated with later onset of substance abuse and decreased odds of antisocial behavior (Chilcoat and Anthony 1996; Hayes, Hudson, and Matthews 2003). Among homeless youth, having a caregiver who is aware of where the youth are spending time and who they are with significantly reduces the likelihood of meeting substance abuse criteria (Johnson, Whitebeck, and Hoyt 2005b). In contrast, homeless youth with little or no parental supervision are at greater risk for substance use, decreased odds of using a condom, increased number of sexual partners, and lower levels of cognitive self-control (Kort-Butler, Tyler, and Melander 2011; Solorio et al. 2008).

Although the prior studies have made important contributions to our understanding of risk-taking behaviors and potential protective factors for homeless youth, these studies are characterized by several methodological limitations. The most serious limitation is the reliance on using convenience samples of youth from narrow geographic regions who were recruited from shelters, community centers, or agencies that served homeless youth. Thus, samples did not typically include youth who experienced the most common form of homelessness for families and youth—living doubled up, also referred to as “couch surfing.” It is estimated that more than 50% of homeless youth are doubled up with others due to being kicked out or running away from their homes (Hallett 2012). Studies that exclude these youth are limited in generalizability of findings. Indeed, there is evidence of a continuum of risk across forms of homelessness, with children and youth in shelters at higher risk on several dimensions (e.g., school mobility and attendance) compared to those who are doubled up (Deck 2016). Another limitation of prior studies is the focus on a single type of risk-taking behavior. Research indicates that there is high intersectionality among risk-taking behaviors (e.g., Van Ouytsel et al. 2017). Thus, a comprehensive measure that takes into account involvement in multiple types of risk behaviors might yield more meaningful findings.

The current study

We sought to close several gaps in the literature by using improved methods to advance the understanding of risk-taking behavior of youth experiencing homelessness. Our purpose was to measure the relationship between housing status and risk-taking behaviors of youth using a representative sample enrolled in North Carolina public high schools. Our sample included youth who self-reported experiencing homelessness based on the McKinney-Vento definition of homelessness and therefore included those with a diversity of experiences of homelessness. While previous research examining homeless youths’ risk-taking behaviors examined a single subset of risky behaviors (i.e., sexual behaviors or substance use), the current study is unique in its cumulative risk approach to understanding risk-taking behaviors. Our cumulative approach to measuring risk-taking behaviors created an index of risk-taking behaviors that combined nine variables (including sexual and substance use behaviors) into one all-encompassing variable. We hypothesized that youth identified as homeless under the McKinney-Vento Act would report engaging in a higher number of risk-taking behaviors compared to stably housed youth. We also hypothesized that the association between housing status and risk-taking behaviors would be moderated by parental monitoring and by social support. Specifically, we expected that adequate parental monitoring and social support would moderate the association between homelessness and risk-taking behaviors, such that fewer hours alone after school and having someone to talk to when feeling distressed would be associated with lower cumulative risk-taking behaviors scores.

Method

Participants and procedure

Participants were 3151 high school students who completed the 2017 North Carolina Youth Risk Behavior Survey (YRBS). Due to missing data on the housing question, 358 participants were excluded from the study, and 619 were removed due to lack of response to one or more of the items needed to generate the risk-taking variable. The sample consisted of 1226 (56.4%) females. In regard to race/ethnicity, 52.3% were White, 18.3% were Black, 22.4% were Hispanic, and 5.5% were Other (Asian, American Indian/Alaskan Native, Native Hawaiian/Pacific Islander). Of the 2174 participants included in the study (69% of the full sample), 54 (2.5%) met criteria for homelessness and 2120 (97.5%) were stably housed.

The YRBS was created by the Centers for Disease Control and Prevention (CDC) as a measure to monitor health-risk behaviors among middle and high school youth. In North Carolina, the survey is administered by the Department of Public Instruction. The YRBS utilizes a two-stage cluster sample design to generate a representative sample of high school students across the state. Schools were selected with probability proportional to school district enrollment size and representative classrooms were then randomly selected. The state met the CDC requirement for overall response rate of at least 60% for using population weighted data. The weighting adjusted for non-response rates and for students’ grade, gender, race, and ethnicity. More information about the sample procedure and weighting process, as well as the data that support the findings of this study, are available publicly on the CDC website (https://www.cdc.gov/healthyyouth/data/yrbs/data.htm). All analyses were conducted using IBM SPSS Version 25, and all procedures were reviewed and approved by the NC State University Institutional Review Board.

Using weighted data, youth reports of housing status indicated that 8922 (2.8%) students reported experiencing homelessness at some point during the past 30 days. Among the sample of homeless youth, 6290 (70.5%) reported living in the home of a friend, family member, or other person because they had to leave their home or their family could not afford housing, Another 1787 (20%) were living in a shelter or emergency housing, 292 (3.3%) in a hotel or motel, and 552 (6.2%) reported not having a usual place to sleep. Due to weighting, this prevalence rate differs from that based on data reported above in the description of participants. This is likely because homeless youth may have more difficulty attending school regularly and are therefore slightly overrepresented in the nonrespondents. Weighting accounts for this discrepancy.

Measure

The NC YRBS consists of a 117-item paper-and-pencil questionnaire. We examined race/ethnicity and gender for descriptive purposes and to assess for potential control variables.

Housing status

Participants were asked “During the past 30 days, where did you usually sleep?” The response “In my parent’s or guardian’s home” was categorized as stably housed. Using the McKinney-Vento Homeless Assistance Act to define homelessness, the following five responses were labeled as homeless: “In the home of a friend, family member, or other person because I had to leave my home or my parent or guardian cannot afford housing”; “In a shelter or emergency housing”; “In a motel or hotel”; “In a car, park, campground, or other public place”; “I do not have a usual place to sleep.” Participants who responded, “Somewhere else” were excluded from the analyses. It is important to note that this paper does not distinguish between homeless youth living with an adult caregiver and unaccompanied homeless youth (i.e., youth without a caregiver) due to the wording of the question we used to define homelessness.

Risk-taking variable

To compute the risk-taking composite score, responses to nine questions were dichotomized as Yes or No and the “Yes” responses were summed. Scores ranged from 0 to 9. For the following questions, students who answered “0 times” were labeled as not having engaged in the risky-behavior (No), and students who reported that they participated in the measured behavior one or more times were classified as having engaged in the risky behavior (Yes). The following questions were included: “During the past 30 days, how many times did you ride in a car or other vehicle driven by someone who had been drinking alcohol?”; “During the past 30 days, how many times did you drive car or other vehicle when you had been drinking alcohol?”; “During the past 30 days, how many times did you carry a weapon such as, a gun, knife, or club on school property”; “During the past 12 months, how many times were you in a physical fight?”; “During the past 30 days, on how many days did you have 4 or more drinks of alcohol in a row (if you are female) or 5 or more drinks of alcohol in a row (if you are male)”; “During your life, how many times have you used any form of cocaine, including powder, crack, or freebase?”; “During your life, how many times have you sniffed glue, breathed the contents of aerosol cans, or inhaled any paints or sprays to get high?”; “During your life, how many times have you taken prescription pain medicine without a doctor’s prescription or differently than how the doctor told you to use it?” Finally, students were asked, “The last time you had sexual intercourse, did you or your partner use a condom?” Students who answered, “I never had sexual intercourse” or “Yes” were classified as not having engaged in this risky behavior (No). Those who answered “No” were classified as having engaged in the risky behavior (Yes).

Parental monitoring

To assess parental monitoring, participants were asked, “On an average school day, how long after school are you alone without a parent or adult?” Students selected either, “I am not alone after school,” “Less than 1 hour per day,” “1 or 2 hours per day,” “3 hours per day,” “4 hours per day,” “5 hours per day,” or “6 or more hours per day.” Although researchers have used a variety of cut-off points to determine adequate supervision (Beyers et al. 2003; Chilcoat and Anthony 1996; Harris et al. 2007), there is strong support for defining inadequate supervision as four or more hours a day alone without an adult. According to the Office of Juvenile Justice and Delinquency Prevention (OJJDP 2016), approximately 62% of violent crimes committed by youth occur on school days. Because nearly 20% of these crimes occur in the 4 hours between 3 and 7 pm, the OJJDP suggested that efforts to reduce engagement in these behaviors focus on the 4 hours after school. Additionally, according to a report released by the U.S. Department of Education and Justice (2000), there is a gap of 20–26 hours a week between adolescents’ school schedules and parent work schedules; this equates to 4–5 hours of potential unsupervised time each school day. Based on these national reports, we defined inadequate parental monitoring as 4 or more hours each school day alone without an adult. Students who replied that they were alone “4 hours per day,” “5 hours per day,” or “6 or more hours per day” were classified as having inadequate monitoring.

Social support

To assess availability of social support, students were asked, “When you feel sad, empty, hopeless, angry, or anxious with whom would you most likely talk about it?” Students who answered the following were labeled as having social support: “Parent or other adult family member”; “Teacher or other adult in this school”; “Other Adult”; “Friend”; or “Sibling.” Students who answered “Not sure” were identified as not having social support. Students who answered, “I do not feel sad, empty, hopeless, angry, or anxious” were excluded from the analyses (n = 279).

Results

Multicollinearity diagnostics were within an acceptable range. See Table 1 for correlations among variables. First, we conducted two separate chi-squared analyses to test for group differences between homeless and stably housed youth on gender and race/ethnicity; no significant group differences were found. Analyses were then conducted to test hypotheses related to group differences between homeless and stably housed youth on risk-taking behaviors, parental monitoring, and having someone to talk to. Results of chi-squared analyses revealed non-significant differences between homeless and stably housed youth in social support and parental monitoring (Table 2). However, results of an independent samples t-test revealed significant group differences for engagement in risk taking behaviors, t = 11.93, p < .001. As expected, we found that stably housed youth (M = 0.79, SD = 0.04) engaged in fewer types of risk-taking behaviors than did homeless youth (M = 1.61, SD = 0.24).

Table 1. Bivariate correlations for unweighted variables.

Variable	1	2	3	4	5	6
1. Homelessness	–	−.01	.02	.11***	−.08***	.05*
2. Parental monitoring	–	–	−.06*	.11**	−.01	.02
3. Social support	–	–	–	−.05*	.09***	−.12***
4. Risk-taking	–	–	–	–	−.04	.01
5. Gender	–	–	–	–	–	.02
6. Race/ethnicity	–	–	–	–	–	–

Note: p < .05*, p < .01**, p < .001***.

Table 2. Group differences on demographic characteristics and study variables (weighted data).

	Homeless	Housed	Test statistic	p
Gender			1.44	.247
Female	3955 (44.3%)	164,684 (52.5%)
Male	4967 (55.7%)	148,835 (47.5%)
Race/ethnicity			1.18	.322
White	3788 (45.1%)	170,318 (55.0%)
African American	2632 (31.3%)	69,648 (22.5%)
Hispanic	1902 (22.6%)	58,783 (19.9%)
Other	79 (0.9%)	10,949 (3.5%)
Risk-taking behavior (mean; SE)	1.61 (.24)	0.79 (0.04)	11.93***	.001
Parental monitoring			0.05	.816
Yes	7512 (88.3%)	271,697 (86.8%)
No	997 (11.7%)	41,167 (13.2%)
Social support			2.99	.084
Yes	5810 (90.8%)	201,942 (82.3%)
No	591 (9.2%)	43,438 (17.7%)

Note: p < .05*, p < .01**, p < .001***.

Two separate regression analyses were performed to examine the possible moderating roles of parental monitoring and social support. We will begin by explaining the results of the parental monitoring moderation model (Table 3). In the first model, housing status along with gender and ethnicity (as controls) were entered. In the second model, parental monitoring was added to the previous model. There was a significant main effect for parental monitoring, b = 0.037, SE = 0.11, p = .002. In the final model, an interaction term was created (parental monitoring by housing status) and was entered into the model. Results indicated that parental monitoring was not a significant moderator (Table 2).

Table 3. Results of regression analyses predicting risk-taking behaviors with parental monitoring as a moderator.

	Step 1		Step 2		Step 3
	b	SE	b	SE	b	SE
Constant	1.35***	0.08	1.68***	0.32	1.85***	0.31
Homelessness	0.76**	0.23	0.82**	0.28	−0.12	0.62
Gender	−0.13*	0.06	−0.13*	0.06	−0.13*	0.06
Black	0.11	0.06	0.09	0.07	0.09	0.06
Hispanic	0.04	0.09	0.04	0.09	0.05	0.09
Other	−0.25*	0.11	−0.28*	0.11	−0.28*	0.11
Parental monitoring			0.34*	0.11	0.37*	0.11
Parental monitoring × homelessness					1.08	0.75
R^2	0.02		0.03		0.03

Note: p < .05*, p < .01**, p < .001***.

To explore whether social support was a significant moderator (Table 4), we began by entering housing status along with gender and ethnicity (as controls) into our model. In the second model, social support was added to the previous model. There was no main effect of social support, b = −0.17, SE = 0.08, p = .071. In the final model, an interaction term was created (social support by housing status) and was entered into the model. Social support was found to be a significant moderator (Table 4). To explore this interaction, simple slope analyses were conducted using the methodology recommended by Preacher, Curran, and Bauer (2006). Results revealed that the relation between housing and risk-taking behaviors was statistically significant for stably housed youth, b = −0.736, SE = 0.24, p < .001 and for homeless youth b = 0.767, SE = 0.23, p = .004, but the direction of “effect” differed for the two groups. That is, homeless youth who reported having social support engaged in more risk-taking behaviors than did homeless youth without support. Stably housed youth who reported having social support engaged in fewer reported risky behaviors than did stably housed youth without support.

Table 4. Results of regression analyses predicting risk-taking behaviors with social support as a moderator.

	Step 1		Step 2		Step 3
	b	SE	b	SE	b	SE
Constant	1.35***	0.08	1.32***	0.28	1.31***	0.27
Homelessness	0.76**	0.23	0.61**	0.27	−0.97***	0.09
Gender	−0.13*	0.06	−0.13*	0.06	−0.15*	0.06
Black	0.11	0.06	0.04	0.07	0.03	0.07
Hispanic	0.04	0.09	0.05	0.11	0.05	0.11
Other	−0.25*	0.11	−0.28*	0.11	−0.28*	0.11
Social support			−0.14	0.09	−0.17*	0.08
Social support × Homelessness					1.71***	0.31
R^2	0.02		0.03		0.03

Note: p < .05*, p < .01**, p < .001***.

Discussion

Our purpose was to explore the risk-taking behaviors of youth who experienced homelessness, with a special interest in potential protective factors that might buffer these youth from taking risks related to their health and well-being. Our study is unique in that our sample was drawn from statewide data collected in high schools (vs. shelters or service organizations), and we used a comprehensive definition of homelessness to include students in situations that are the most common for homeless youth—living doubled up. Thus, our findings are likely to be more generalizable to the broad population of homeless youth than are many prior investigations. Past research indicates that homeless youth engage in higher rates of risky behaviors compared to rates for matched youth who are stably housed (Busen and Engebretson 2008; Cauce et al. 2000). We used the YRBS—a unique data source—to contribute to knowledge of youth risk-taking behaviors of homeless youth and found results that were generally consistent with prior studies. Specifically, youth who reported being homeless (M = 1.61, SD = 0.24) engaged in significantly more categories of risky behaviors than stably housed youth (M = 0.79, SD = 0.04). It is noteworthy that the mean number of types of risky behaviors engaged in by both groups was very low.

Wide variability in risk-taking behaviors among the homeless youth was evident in the relatively large standard deviation for the cumulative score for homeless students relative to stably housed youth (although the small sample of homeless youth might also contribute to a large standard deviation). That variability is what we were particularly interested in understanding. Overall, and as expected based on prior research on adult monitoring of youth, we found that closer supervision by an adult during afterschool hours was associated with youth engaging in fewer types of risk-taking behaviors. This finding complements previous research that found an association between parental supervision or monitoring and engagement in sexual risk behaviors and substance use behaviors (Ennett, Bailey, and Federman 1999; Rice, Milburn, and Rotheram-Borus 2007). It is worthy of note that supervision accounted for very little variance in risk-taking.

In contrast to our expectation, supervision did not serve as a protective factor. Both groups of youth received very similar levels of afterschool monitoring, with the vast majority of youth in both groups reporting relatively close monitoring (i.e., they were alone for 3 or fewer hours). It is important to note that 70.5% of homeless youth reported living doubled-up and 20% reported living in shelters. Therefore, 90% of the sample had a high likelihood of monitoring by adults other than their own parents (i.e., shelter staff and other adults in shelters or in the shared housing). Thus, it would be interesting to explore which adults the students identify as monitoring them after school and to determine whether group differences become evident when the identity of monitors is considered.

We also expected that having someone to talk to when feeling sad or anxious would serve as a protective factor. Results indicated that this factor did indeed moderate the link between homelessness and involvement in risk-taking behaviors; however, when we probed the interaction, we found results were inconsistent with our hypothesis. Stably housed youth who reported having someone to talk to engaged in significantly fewer types of risky behaviors compared to stably housed students without someone to talk to. In contrast, homeless students with someone to talk to actually reported engaging in more types of risky behaviors than homeless students who did not have someone to talk to. We based our hypothesis on past research showing that social support could be a protective factor, but our measure was a single item that required youth to simply report whether or not they had someone to talk to when feeling distressed. The item did not measure size of the social network (Ennett, Bailey, and Federman 1999), quality of relationships such as getting along well with family members (Rice, Milburn, and Rotheram-Borus 2007), or overall social connectedness (Dang 2014); those dimensions of support were associated with low rates of risky behaviors of homeless youth in past studies. It is possible that the social support that some homeless youth sought out actually encouraged risk-taking behaviors rather than deterred them. Given that some risky behaviors occur in the context of social activities, such as substance use and riding in a car with someone who has been drinking, it may be that seeking social support was accompanied by engaging in risk-taking behaviors. Additionally, it is possible that homeless youth approached others for support when they were angry or sad after they had already engaged in the risky behavior. Thus, the conversations might not have been helpful in terms of reducing risky behaviors. It is also important to take into account that we have no information regarding whether the youth actually sought support from their social network. Perhaps the youth reported that they had someone to talk to but did not utilize that person in times of need. Future studies could examine when and under what conditions homeless youth seek support from their social network, especially in times of need. Our counterintuitive finding deserves further attention in investigations that examine the number and characteristics of people who are available for homeless youth to turn to as well as whether the youth considers the person with whom they are able to talk as a positive source of support.

For the current analyses, we examined whether or not youth had someone to talk to, but we did not explore who they sought support from. Given the unexpected finding related to social support, we conducted follow-up analyses to examine the type of person that youth in the sample were talking to when feeling sad or anxious (see Table 5). Post-hoc analyses indicated that stably housed youth and homeless youth generally sought support from similar networks, but homeless youth reported seeking support significantly less often from friends compared to stably housed youth who were more likely to seek support from friends. This is consistent with previous literature (Haye et al. 2012). In addition, previous research indicates that homeless youth often report that their social networks are comprised of transient and antisocial peer relationships (Johnson, Whitebeck, and Hoyt 2005a; Rice, Stein, and Milburn 2008). Thus, participants in this study who were experiencing homelessness may not have turned to peers for support when they were feeling sad because of the transient nature of their friendships. Future research should explore why homeless youth are less likely to seek support from peers, as lack of peer support in adolescence could be a risk factor for a host of negative outcomes (Laible 2007). It is important to note that having social support is just one protective factor to consider in the association between housing status and engagement in risk-taking behaviors.

Table 5. Group differences on sources of social support.

	Homeless	Housed	Odds ratio (95% CI)
Parent or other adult family member	2589 (40.4%)	56,019 (22.8%)	1.86 (0.85–4.1)
Teacher or other adult in school	367 (5.7%)	3561 (1.5%)	4.1 (0.89–18.58)
Other adult	736 (11.4%)	3616 (1.5%)	3.8 (0.61–23.6)
Friend	1897 (29.6%)	125,338 (51.1%)	0.36 (0.18–1.3)***
Sibling	221 (3.4%)	13,409 (5.5%)	0.56 (0.06–5.4)

Note: p < .05*, p < .01**, p < .001***.

Limitations and future directions

We examined two potential protective factors against risky behaviors of homeless youth, and both factors accounted for very little variance in risk-taking. Future studies should investigate the buffering effects of other potential protective factors such as youth’s executive functioning (Masten et al. 2014; Obradović 2010) and the degree to which they receive responsive, warm parenting (Herbers et al. 2011), factors that are known to be associated with relatively positive adjustment of young children experiencing homelessness. We also found a very low mean number of risky behaviors that both homeless and stably housed youth engaged in. The low means (i.e., a floor effect) could have limited the ability of the regression analysis to detect meaningful associations. Our risk-taking variable was limited in that it did not allow us to determine whether students engaged in risky behaviors prior to experiencing homelessness. Some of the questions used in the risk-taking variable asked students if they had ever participated in a particular behavior or if they had participated in certain behaviors during the past 12 months. Conversely, the homelessness question asked where students slept during the past 30 days. Therefore, it is possible that some risky behaviors could have taken place prior to experiences of homelessness and/or contributed to an individual becoming homeless. Our decision to define inadequate supervision as 4 or more hours alone on weekdays was well-justified by national studies. However, we tested the sensitivity of this decision by conducting a post hoc set of analyses using a lower threshold for inadequate monitoring (3 or more hours of unsupervised afterschool time); findings did not differ from the pattern we found using 4 or more hours as the threshold.

In our study, all variables were derived from youth responses on the YRBS, a single data source. It would have been beneficial to measure constructs using multiple methods and informants. For example, afterschool supervision based on parent report as well as youth report might have yielded a more valid indicator of monitoring. All youth were enrolled in school and were present on the day the survey was administered, so findings might not generalize to homeless youth who do not attend school regularly or those who have dropped out. These youth might be at even greater risk, so it is important to replicate the study with them. Additionally, data were collected at a single point in time. To understand the unfolding of risky behaviors and potential protective factors over time, longitudinal data will be needed in future studies. It is also important to note that our unweighted sample size was small for youth experiencing homelessness. Finally, this paper defined homelessness using the McKinney-Vento definition, thus both homeless youth living with a caregiver and unaccompanied youth were grouped together. Therefore, there is potential for between-group differences on our variables of interest. Future studies should explore if engagement in risky behaviors and our two moderators differ between homeless youth living with a caregiver and those living without one.

Conclusions

The current study examined the link between homelessness in adolescence and risk-taking behaviors using state population-representative 2017 YRBS data. We also examined the potential moderating role of parental monitoring and social support. Consistent with previous literature, homeless youth engaged in more risk-taking behaviors compared to their stably housed peers. Parental monitoring did not serve as a moderator. Interestingly, both homeless youth and stably housed youth reported similar rates of parental monitoring. Thus, even though stably housed youth and homeless youth both reported adequate supervision, homeless youth still engaged in more types of risk-taking behaviors. Therefore, homeless youth may benefit from more structured afterschool activities that work to strengthen important life, social, and educational skills (Passafaro, Gomez, and Weaver-Spencer 2016). The Youth Empowerment Strategies (YES!) program has emerged as an afterschool intervention that helps engage at-risk early adolescents (late elementary and middle school students) in community civic engagement. In this initiative, high school students are trained to serve as facilitators of the program, so YES! could be an empowering afterschool program for homeless youth (Wilson et al. 2006). Social support was found to be a significant moderator. For stably housed youth, students who reported having someone to talk to reported engaging in fewer types of risky behaviors. Conversely, homeless youth who reported having someone to talk to endorsed engaging in more risk-taking behaviors compared to those without an identified person. Our findings contribute to the literature on predicting individual differences in risk-taking behavior for youth experiencing homelessness, and we hope this study will encourage continued exploration of potential protective factors for these youth.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes on contributors

Caitlyn R. Owens is a doctoral student of School Psychology at North Carolina State University and a predoctoral fellow for the Carolina Consortium on Human Development. Her research has predominately focused on preventing child maltreatment and improving outcomes for underserved, vulnerable populations; her current projects involve further understanding the linkages between adverse experiences, specifically homelessness, mental health, and risk-taking behaviors in youth. She is also conducting research aimed at examining facilitators and barriers to implementing large-scale parenting interventions.

Evadine L. Codd, MS, is a doctoral student of School Psychology at North Carolina State University. Her research interests focus on barriers to academic achievement in underrepresented and underserved populations. Evadine is currently conducting research examining the influence of family functioning on emotional and behavioral outcomes of adolescents with histories of adverse childhood experiences.

Mary E. Haskett, PhD, is Professor of Psychology at North Carolina State University, where her research team focuses on factors that shape parenting, the influences of parenting on children’s social-emotional development, adjustment of children with a history of maltreatment and homelessness, and the understanding of approaches to support parents and children experiencing homelessness. She is currently conducting research on interventions to promote positive parenting and reduce risk of maltreatment among parents residing in emergency shelters and transitional housing. Dr. Haskett is a past president of the Section on Child Maltreatment in the APA Society for Child and Family Policy and Practice and a Fellow of the American Psychological Association; in 2016, she was honored to be the recipient of the National Association for the Education of Homeless Children and Youth Staci Perlman Research Award.