Contextualizing the Impacts of Homelessness on Academic Growth

by Alexandra E. Pavlakis, Peter Goff & Peter M. Miller - 2017

Background/Context: Students experiencing homelessness are also often living in poverty and may share many of the same characteristics and experiences with children in low-income housing. Scholars aim to understand the impacts of homelessness above and beyond the effects of poverty, but studies are mixed. Contextual factors—such as the localized implementation of the McKinney-Vento Homeless Assistance Act (McKinney-Vento), which aims to reduce barriers to school success for students experiencing homelessness—are often overlooked by scholars but may play an important role in explaining inconsistencies between single-site studies.

Purpose/Objective/Research Question/Focus of Study: Our purpose is to examine the impacts of homelessness above and beyond poverty. We ask two questions: “To what extent does homelessness impact students’ academic growth?” and “To what extent does chronic homelessness impact students’ academic growth?” In making sense of our findings, we consider the unique context of our study site.

Setting: We draw data from Midtown, a pseudonym for a Midwestern city that has experienced rising homelessness. Midtown has a nationally recognized program aimed at overseeing McKinney-Vento.

Research Design: We conducted a secondary analysis of a longitudinal administrative district dataset. For our first question, we used ordinary least squares (OLS) regression, with and without student fixed effects, with standard errors clustered at the school level. To help isolate the impact of homelessness apart from poverty, we (a) limited our sample to include only those students who have experienced sustained poverty (history of free lunch status) and stable housing prior to fourth grade, (b) included relevant covariates to adjust for other between-group differences, (c) used student fixed effects to control for any remaining time-invariant, unobserved between-group differences, and (d) examined achievement growth (rather than absolute achievement). To examine our second research question (chronic homelessness), we used OLS regression, with standard errors clustered at the school level.

Conclusions/Recommendations: Notwithstanding the fact that Midtown devotes considerable resources toward McKinney-Vento, the impact of homelessness remains evident in our fixed effects model. Turning to chronic homelessness, our results showed no discernible impact on achievement growth. We speculate that as time goes on, the Midtown district may play an important role in buffering families. Our study suggests that in order to reduce the impact of homelessness on academic achievement, it is important to be aware of when students become homeless—so that their needs can be met right at this critical juncture. Specific recommendations, such as the ongoing use of residency questionnaires and surveys, are discussed.

BACKGROUND AND PURPOSE

There are over 1 million students experiencing homelessness in the U.S. (National Center for Homeless Education [NCHE], 2015). Research suggests they tend to score below average on tests, suffer from low attendance (Masten et al., 1997, 2012; Obradović et al., 2009; Rafferty, Shinn, & Weitzman, 2004), and move schools frequently (Murphy & Tobin, 2011). While students experiencing homelessness tend to exhibit high rates of developmental delays, depression, anxiety, learning difficulties, and behavioral problems (Bassuk & Rubin, 1987; Zima, Wells, & Freeman, 1994)all of which may interfere with school successthey may be underserved by medical and special education services (Buckner, Bassuk, & Weinreb, 2001; Zima et al., 1994; Zima, Bussing, Forness & Benjamin, 1997). Yet students experiencing homelessness are also often living in poverty and may share many of the same characteristics and experiences with children residing in low-income housing (Buckner, 2008). In light of these commonalities, scholars aim to understand the impacts of homelessness above and beyond the effects of poverty (Buckner, 2008; Miller, 2011). Gaining clarity on the distinct impacts of homelessness can help inform policy and programmatic responses aimed at best meeting the needs of these students and their families.

Studies, however, are mixed on whether students experiencing homelessness fare worse than low-income housed children on a range of outcomes (Buckner, 2008). Methodological differences may account for some inconsistency in findings, as scholars often use different sample sizes, comparison groups, measures, and definitions (Buckner, 2007, 2008). Yet as Buckner points out, even rigorous studies present inconsistent findings. Historical, contextual, and policy-related factors, such as increased response to homelessness over time, differences in local shelter conditions, or the uneven implementation of the McKinney-Vento Homeless Assistance Act (which aims to reduce educational barriers for homeless children) may play an important role in explaining inconsistent findingsparticularly between single-site studies (Buckner, 2007, 2008; Miller, 2011). Not only is there a lack of agreement around the independent impacts of homelessness, suggesting that more research is needed on this topic, but scholars also tend to overlook the local context when interpreting findings (Buckner, 2012). Using school longitudinal student-level data from one medium-sized Midwestern city, our purpose is to examine the impacts of homelessness above and beyond poverty from a contextualized perspective. In other words, we address this research gap by considering how the unique contours of our study site may shape our findings.

INVESTIGATING THE UNIQUE IMPACTS OF HOMELESSNESS

While most scholars have suggested that students experiencing homelessness clearly exhibit more risk than the general population, some argue that they appear similar on psychological, developmental, and cognition indicators when compared to low-income students (Schteingart, Molnar, Klein, Lowe, & Hartmann, 1995; Shinn et al., 2008). Examining cognitive functioning and emotional-behavioral adjustment, Rescorla, Parker, and Stolley (1991) found that school-aged children in shelters did not demonstrate significant differences from similar children who were housed and attending a medical clinic, although preschool children living in shelters exhibited delays in receptive vocabulary and visual motor development and emotional-behavioral problems. Similarly, Masten, Miliotis, Graham-Bermann, Ramirez, and Neemann (1993) found that although homeless children had more recent exposure to stress than housed low-income children, behavior problems were related to risk and stress more than to housing status or income. Focusing on academic achievement, Buckner, Bassuk, and Weinreb (2001) found that housing status was not related to achievement but that mobility, defined as the number of schools attended in the past year, played an important role.

Other scholars, however, have suggested that homelessness is related to higher rates of internalizing behaviors such as depression and anxiety (Buckner, Bassuk, Weinreb, & Brooks, 1999; Menke & Wagner, 1997) and lower scores on reading, spelling, and arithmetic measures (Rubin et al., 1996). Rafferty and colleagues (2004) controlled for earlier achievement and conducted longitudinal analyses on homeless and housed but low-income children. Housing status changes were related to declines on standardized achievement tests, but the association faded over time. According to mothers reports, students experiencing homelessness changed schools more often, experienced more grade retention, and described lower-quality schooling experiences. Likewise, self-reports suggested less ambitious plans for postsecondary education. Fantuzzo, LeBoeuf, Chen, Rouse, and Culhane (2012) showed that homelessness had a unique relationship with poor classroom engagement, while school mobility was uniquely associated with not only engagement but also academic achievement. Experiencing homelessness and mobility was the most damaging to both engagement and achievement.

Obradović et al. (2009) examined mathematics and reading achievement gaps and growth trajectories in four elementary school cohorts in a large urban district. Trajectories of homeless and highly mobile students were compared to those of low-income stable students (who were on free or reduced price lunch status) and all other tested students in the district. Obradović et al. found that homeless and highly mobile students had the lowest initial achievement and in some cohorts, homeless and highly mobile and low-income students also had slower growth than their peers. Using administrative data from a large urban school district, , Herbers and colleagues (2012) were able to examine the role of early academic achievement on later achievement among 18,011 students based on four exclusive categories of risk: homeless and highly mobile, students on free meal status, students on reduced meal status, and students who were neither homeless and highly mobile nor low income. Not only did homeless and highly mobile students demonstrate the lowest levels of first grade achievement (controlling for attendance rate, special education status, and English language proficiency), but risk status also predicted achievement and growth in third through eighth grade reading and math (beyond its relations with initial achievement). For homeless and highly mobile students, housing status and low initial achievement had unique negative consequences for future reading and math scores. While the extant literature examines a broad array of outcome variables relevant to the well-being of children, Table 1 highlights the select findings of studies that focus on academic achievement.

Table 1. Select Findings of Noteworthy Research on Homelessness and Academic Achievement

Citation	Academic Achievement Measures	Sample	Key Achievement Findings
Rubin et al. (1996)	Reading, spelling and arithmetic (Wide Range Achievement Test-Revised [WRAT-R]).	102 sheltered homeless children and their mothers; 178 housed children and their mothers from the homeless childs classroom.	Homeless children scored significantly lower on reading, spelling, and arithmetic than housed peers. Summary: Difference.
Buckner et al. (2001)	Basic reading, spelling, mathematical reasoning, and composite achievement score (Wechsler Individual Achievement Test Screener [WIAT-S]).	174 youth in sheltered homeless and low-income housed families.	Housing status not associated with achievement. Summary: No difference.
Rafferty et al. (2004)	Mathematics achievement (MAT; California Achievement Test [CAT]).	46 formerly homeless adolescents; 87 permanently housed adolescents whose families received public assistance.	Homelessness was related to declines on standardized achievement tests but the association faded over time. Summary: Limited difference.
	Reading achievement (Degrees of Reading power Reading Test [DRP]; Metropolitan Achievement Test-Revised [MAT-R]).
Obradović et al. (2009)	Mathematics and reading achievement gaps and growth trajectories (Northwest Achievement Level Tests [NALT]; some administered through Computer Adaptive Level Testing [CALT]).	14,754 students: homeless and highly mobile (HHM) compared to not mobile, low-income students, and all other tested students in district.	HHM had the lowest initial achievements and in some cohorts, slower growth than their peers. Achievement variation was also noted within HHM trajectories. Summary: Difference.
Fantuzzo et al. (2012)	Standardized math and reading achievement (Complete Battery Plus version of the TerraNova).	Third grade cohort in Philadelphia (n=10,841).	Unlike school mobility, homelessness was not uniquely associated with academic achievement. Experiencing homelessness and mobility was the most damaging to achievement. Summary: No difference.
Herbers et al. (2012)	Oral reading assessment (ORA) for first grade; Reading and math achievement and growth (Scaled scores from the reading and math sections of the Computer Adaptive Levels Test [CALT] and Measures of Academic Progress [MAP]).	18,011 students in four risk categories: homeless and highly mobile (HHM); on free meals; on reduced meals; neither HHM nor low income.	HHM had the lowest ORA scorestheir first-grade achievement as well as HHM status had independent negative consequences on reading and math trajectories. Summary: Difference.

CONTEXTUALIZING THE MCKINNEY-VENTO HOMELESS ASSISTANCE ACT

The McKinney-Vento Homeless Assistance Act (McKinney-Vento) provides a range of guidance to schools and supports to students experiencing homelessness to help them overcome barriers to academic success. Under the 2001 reauthorization, McKinney-Vento outlined an inclusive definition of homelessness that covered not only students in shelter and temporary residences but also students living in hotels/motels, staying in cars or other places not meant for human habitation, awaiting foster care placement, or doubled up with friends or family out of economic necessity.¹ This definition is more expansive than common conceptions of homelessness (i.e., living on the streets or in congregate shelter) and broader than the definition historically used by the Department of Housing and Urban Development (HUD) (McKinney-Vento, 2001; Miller, 2011).

In the 201314 school year, over 1.3 million students were identified as homeless under McKinney-Vento. These students get ready for school in diverse residential contexts15% sleep in shelters and await foster care, 6% live in hotels/motels, 3% are unsheltered, and 76% are doubled up (NCHE, 2015). The community context matterssome areas (particularly suburban and rural areas) can be void of shelters and may rely more heavily on hotels/motels. Likewise, the capacity at family shelters can also vary by community as well as by seasonal fluctuations. Research suggests that even within communities, individual school stakeholders can vary greatly in their awareness of these heterogeneous living contexts

and how they may shape practice (Miller, Pavlakis, Samartino, & Bourgeois, 2015; Pavlakis, 2015).

Fundamentally, McKinney-Vento aims to ensure that students experiencing homelessness are enrolled immediately in school (irrespective of administrative requirements and records), have equal access to school events, and are not segregated based on their housing status. To the extent feasible, students identified as homeless have a right to remain at their school of origin, even if residential instability relocates them outside of the school catchment area. In fact, if parents request it, schools must provide transportation to and from the school of origin (Canfield, 2015; Murphy & Tobin, 2011). This provision is particularly important because school mobility is linked to lower academic achievement (Kerbow, Azcoitia, & Buell, 2003; Mehana & Reynolds, 2004) and disengagement (Gruman, Harachi, Abbott, Catalano, & Fleming, 2008; Rumberger & Larson, 1998; South, Haynie, & Bose, 2007). Yet transportation can look dramatically different in communities with extensive train or bus lines compared to those that may even have limited taxicab options. Transportation is also expensive; in the 201314 school year, around 75% of local educational agencies (LEAs) did not receive McKinney-Vento subgrants to help support costs (NCHE, 2015; Wong et al., 2009).

McKinney-Vento also mandates collaboration, requiring LEAs to identify liaisons who are responsible for connecting students experiencing homelessness to outside agencies, identifying students, ensuring families understand their rights, and improving awareness of student homelessness (Canfield, 2015; Murphy & Tobin, 2011). Through relationships built with families, outside providers, and various school employees and stakeholders (including teachers, secretaries, bus drivers, nurses, and cafeteria staff), students experiencing homelessness may be better positioned to succeed in school. However, for many liaisons, McKinney-Vento related duties come on top of a host of other responsibilitiesdiluting their time and attention. Research suggests that some liaisons do not have the capacity to meet their McKinney-Vento duties and many school districts fall short of fully implementing McKinney-Vento (Hallett, Low, & Skrla, 2015; Miller, 2009, 2011; Miller et al., 2015). 

RESEARCH QUESTIONS

With a few notable exceptions (e.g., Herbers et al., 2012; Menke & Wagner, 1997; Obradović et al., 2009) most of the extant literature on the impacts of homelessness restricts its analysis to children and families living in shelters. Instead, our study takes an education perspective and defines homelessness according to McKinney-Vento. With our longitudinal administrative dataset and educational perspective, we ask two questions: To what extent does homelessness impact students academic growth? and To what extent does chronic homelessness impact students academic growth?

Some research has suggested that reading achievement is more sensitive to homelessness and school mobility than mathematics achievement (Fantuzzo et al., 2012; Obradović et al., 2009). This may be reflective of the strong role that families and the home context play in the advancement of language and literacy skills (Fantuzzo et al., 2012; Hart & Risley, 1995). However, because of the strong national policy focus on quantitative skills (Ketterlin-Geller, Chard, & Fien, 2008), and our desire to better understand the roles of schools, we chose to examine mathematics achievement over other subject areas. While we use students mathematics trajectories as a proxy for academic achievement, our focus is not on math content, curriculum, and pedagogy, but on better understanding the impacts of homelessnessabove and beyond povertyon academic achievement growth.

For both questions, we account for the fact that students are clustered in schools. To address the first question, we also run a fixed effects model by student. To make sense of our findings, we draw from our rich contextual understanding of the district and community. By contextualizing our analysis, we also hope to aid policymakers and school leaders in deciding how applicable our findings and implications may be to other localities.

CONCEPTUAL FRAMEWORK

Our study draws from the continuum of risk framework (see Figure 1). This framework suggests that the struggles students experiencing homelessness face cannot only be attributed to homelessness, but may more appropriately reflect their position at the far end of a continuum of risk, where homeless children face more risk than housed low-income children, who tend to be more at risk than middle-class children (Buckner et al., 2001; Masten et al., 1997; Ziesemer, Marcoux, & Marwell, 1994). From this perspective, students experiencing homelessness face three types of risk: the risks all students may face (e.g., biological factors, certain family events such as loss), poverty-related risks (e.g., neighborhood crime), and homelessness-specific risks, such as shelter conditions or stigma (Buckner, 2007, 2008). We wanted to consider the applicability of this framework to a unique educational context. Therefore, we use the continuum of risk framework in Midtown, a pseudonym for a Midwestern city. Midtown has a public school district that serves over 27,000 students and boasts a proactive and purposeful response to student homelessness.

Figure 1. The continuum of risk framework

THE MIDTOWN CONTEXT

The city of Midtown houses a handful of family shelters and runs a number of rehousing programs aimed at ending homelessness. Facilitating interagency collaboration, the medium-sized city is also home to county-level social workers that are well connected to the school district. About half of Midtown students are on free/reduced lunch status and 55% are students of color. The district has also experienced rising homelessnessfrom the 200304 to the 201213 school year, the number of students experiencing homelessness has increased over 167%.² Compared to other districts of its size, Midtown has been proactive in creating supports to address these challenges.

Importantly, Midtown has a nationally recognized, award-winning program aimed at overseeing McKinney-Vento. For instance, the district has four full time social workers who serve as sounding boards for building-level school staff. These district-level social workers also have the flexibility to hold information sessions in out of school spacessuch as in congregate shelters. To facilitate referrals to outside agencies, the district distributes binders of community contacts (such as numbers for food pantries, mental health services, and tutoring) to individual school social workers, and most schools also have an onsite closet of donated clothing, hygiene items, and school supplies for students in need. The district also provides a range of guidance materials to relevant school and community stakeholders and hosts professional development and trainings aimed at improving identification, reducing stigma for students and families, and overcoming the common barriers students experiencing homelessness face in achieving academic success.

Building-level practices, such as buddy programs, are in place to smooth transfers between schools. County-level meetings enable school social workers to collaborate across schools to ease the impacts of school mobility. Furthermore, the district is working on aligning curriculum across buildings so that school transfers are less disruptive to the educational trajectory of students who move. While our qualitative work suggests that there was still much that could be done to strengthen supports to families experiencing homelessness in Midtown, in a number of ways the district is unique and innovative (see Miller et al., 2015; Pavlakis, 2014, 2015). While informed by our larger, mixed methods study, this paper draws specifically from longitudinal Midtown school district data.

METHODS

Identifying the impact of homelessness on the student experience presents several methodological and logistical challenges. In keeping with the continuum of risk framework, the primary identification issue pertains to isolating the impact of homelessness apart from the impact of poverty, which often accompanies homelessness. Our analytical strategy addresses this concern by (a) limiting our sample to include only those students who have experienced sustained poverty (history of free lunch status) and stable housing prior to fourth grade, (b) including relevant covariates to adjust for other between-group differences, (c) using student fixed effects to control for any remaining time-invariant, unobserved between-group differences, and (d) examining achievement growth (rather than absolute achievement).

To noteand as is typical in districts across the countrythe majority of students who were identified as homeless in Midtown were captured through self-reports at registration at the start of the school year. McKinney-Vento implementation can be inconsistent, however, not only between school districts, but, as our qualitative work suggests, even within single districts and individual schools. As a result of this irregular implementation, the supports available to students may vary widely across contexts (Miller, 2011; Miller et al., 2015; Pavlakis, 2014, 2015). Likewise, the McKinney-Vento measure (as recorded by Midtown and many other districts) does not provide information regarding where students are staying (i.e., shelter, aunts house, or bridge overpass) or on the duration of condition (days versus months).

QUESTION 1

To answer our first question, since 88% of Midtowns students experiencing homelessness were on free lunch status the year before they became homeless, we created a cohort of students (across years) who were on free lunch status (below 130% of the poverty line) in first, second, and third grades but not homeless in any of these grades. We followed this low-income cohort in future grades in order to examine their math achievement in years in which some students did become homeless. We used their third-grade standardized math exam score as a proxy for prior achievement. Our sample began in 200910 and included fourth through seventh graders. All told, we had a sample of 3,980 yearly enrollment and achievement records (162 homeless records) representing 1,788 unique students (57 homeless) across 43 schools.³

Students math achievement for each grade (future math) is our dependent variable. Our independent variables include: third grade math score, homeless status, grade, year, ethnicity, attendance rate, special education status, and mobility (see Table 2). Because the math exam is taken in the fall, we use students attendance rate (days attended/days enrolled), homeless status, special education status, and mobility from the previous school year. In choosing our independent variables, we examined the scholarship on homelessness, poverty, and achievement from a continuum of risk perspective. Low attendance, for instance, is a barrier to academic success and students experiencing homelessness tend to have worse attendance than the general population. There is, however, less clarity around their rates in relation to stably housed, low-income children (Miller, 2011; Obradović et al, 2009; Rubin et al., 1996). Likewise, in order to define our measure of mobility, we drew from educational research on school transfers and residential change. Research suggests that within-district, midyear mobility may be more problematic for students than interdistrict transfers that occur over the summer (Beatty, 2010; Fantuzzo et al. 2012; Hanushek, Kain, & Rivkin, 2004). Yet low-income students and students of color are more likely to change schools midyear and stay within the school district (Burkam, Lee, & Dwyer, 2009; Hanushek et al., 2004; Schafft, 2006). Aligned with the continuum of risk framework, the very nature of homelessness creates conditions ripe for school mobility and classroom disruptions (Rafferty et al., 2004; Tierney, Gupton, & Hallett, 2008). Informed by the mobility literature, we define a student as mobile if he/she changed schools within Midtown between September and June of the previous school year.

Figure 2, which does not clearly support the continuum of risk framework, provides the average future math score by housing status. Yet to properly address our research question, we used OLS regression, with (model 2) and without (model 1) student fixed effects, with standard errors clustered at the school level. In a fixed effects model, we are comparing students to themselves, so time-invariant variables (which do not vary within any given student over time) drop out. We chose to use a fixed effects model because with OLS, we know there are some variables that are related to homelessness and achievement that we could not measure and therefore we could not include in the model. By using a fixed effects model, we can implicitly control for those omitted variables that are constant over time. In other words, the fixed effects model (model 2) contains not only all the time-invariant control variables of model 1 but also other omitted, time-invariant variables. For this research question, the reference groups for the binary/categorical variables were: not homeless, Grade 4, Year 200910, not mobile, Native American, and not in special education.

Table 2. Descriptive Statistics

Variable	Mean (Standard Deviation)
Future Math	458.16 (51.58)
3rd Grade Math	.50 (44.47)
Homeless Status	.01 (.12)
Attendance	.00 (.05)
Special Education	.19 (.39)
Mobility	.04 (.19)

Figure 2. Average math scores by homeless status

Both models include an interaction term between homelessness and attendance rate. The hypothesis we are testing with this interaction is whether the effect of attending school is different for students who are homeless. Ideally, a differential (positive) role of attendance rate for students experiencing homelessness would indicate that schools are providing additional services that are assisting students in these challenging times.

The results of the above models are presented in Table 3. Here we note the expected negative impact of homelessness on students achievement growth (-5.26 and -9.71); the variation in the effect of homelessness on achievement growth is such that the estimated impact is only significantly different from zero in the fixed effects modelan important difference between the two models. When we compare the impact of the model that is potentially biased by omitted variables (OLS, model 1) with the model that includes those omitted variables (fixed effects, model 2), we can assume that the omitted variables are causing the difference between the models. This implies that even with the nonparametric matching technique and the use of control variables, notable unobserved differences between homeless and nonhomeless (but poor) students remain. These unobserved factors create an upward (and positive) bias on the effect of homelessnessa bias that makes the impact of homelessness seem less harmful. In other words, when these variables are not considered, homelessness appears to be less detrimental than it is in reality.

Table 3. The Impact of Homelessness on Achievement Growth

	Model 1 (without student fixed effects)		Model 2 (with student fixed effects)
Variable	Coefficient (S.E.)		Coefficient (S.E.)
Constant	470.76 (13.37)	***		416.36 (1.59)		***
Third Math (centered)	0.71 (0.02)	***		0 (omitted)
Homeless	-5.26 (5.14)			-9.71 (4.75)		*
Grade 5	23.73 (2.45)	***		-1.26 (1.02)
Grade 6	40.10 (2.02)	***		-4.27 (1.43)		**
Grade 7	67.58 (2.11)	***		0 (omitted)
201011	-2.52 (1.82)			20.45 (1.73)		***
201112	-2.35 (2.11)			42.99 (1.86)		***
201213	-5.11 (2.00)	*		61.61 (1.94)		***
Mobility	-3.84 (3.09)			0.98 (2.85)
Attendance (centered)	39.56 (14.30)	**		28.34 (19.69)
Asian	-21.42 (12.90)			0 (omitted)
African American/Black	-30.08 (13.15)	*		0 (omitted)
Hispanic	-26.13 (13.80)			0 (omitted)
Two or More	-23.25 (13.01)			0 (omitted)
White	-24.46 (12.37)			0 (omitted)
Special Education	-13.76 (2.23)	***		2.23 (3.48)
Homeless: Attendance (centered)	-40.52 (63.92)			-54.18 (58.21)
R-squared	0.6012				0.8803
(adjusted)	0.5995				0.7817
No. Observations	3,980				3,980
Sign codes:* p<0.05, ** p<0.01, *** p<0.001

The interaction between homeless status and attendance rate is not significant in either model. This suggests that attending school for students experiencing homelessness does not grant any additional benefits to achievement, as compared to their nonhomeless peers. Further, we also note that mobility did not reach statistical significance in either model. We speculate on the reasons for these findings in our discussion.

QUESTION 2

Because homelessness, as we have operationalized it above, may be a brief, transitory experience for some students, we also examined outcomes for students who were identified as homeless for 2 consecutive years. We believe that these students experienced chronic homelessness and therefore the impact of homelessness may be more pronounced. To pursue this line of inquiry, we created a cohort of students who fit one of two categories: a) for 3 continuous years, they were on free lunch status but stably housed (sustained poverty); or, b) for the 1^st year they were on free lunch status and stably housed but for the following 2 years they experienced homelessness (chronic homelessness). While homelessness takes on a different definition here, our other variables were defined the same way as for question 1 (based on prior year). Taken together, our sample included 2,626 students (119 chronically homeless) across 51 schools (see Table 4). Figure 3 presents the average future math score by chronic homelessness status; at face value, it appears to support the continuum of risk framework.

Table 4. Descriptive Statistics

Variable	Mean (Standard Deviation)
Future Math	492.77 (53.45)
Prior Achievement	454.28 (58.13)
Homeless Status	0.05 (0.21)
Attendance	0.93 (0.08)
Special Education	0.22 (0.42)
Mobility	0.05 (0.21)

Figure 3. Average math scores by chronic homeless status

To examine our second research question, we used OLS regression, with standard errors clustered at the school level. The reference groups for the binary/categorical variables were: sustained poverty (i.e., not chronic homelessness), Grade 5, Year 200809, not mobile, Native American, and not in special education.

Although we expected the chronic homelessness variable to pick up compounding effects of long-term homelessness, the results presented in Table 5 show no discernible impact on math achievement growth.

Table 5. The Impact of Chronic Homelessness on Achievement Growth

Variable	Coefficient (S.E.)
Constant	160.14 (10.17)	***
Prior Achievement (centered)	0.68 (0.02)	***
Chronic Homelessness	-27.49 (27.81)
Grade 6	-4.02 (2.26)
Grade 7	4.94 (2.78)
Grade 8	-3.59 (3.08)
Grade 9	-9.40 (3.70)	*
Grade 10	-10.90 (3.32)	**
200910	8.85 (2.24)	***
201011	3.49 (2.69)
201112	2.21 (2.73)
201213	0.90 (2.08)
Mobility	-6.07 (4.20)
Attendance (centered)	41.40 (10.38)	***
Asian	-4.24 (4.17)
African American/Black	-10.30 (3.98)	*
Hispanic	-10.57 (3.93)	**
Null	-2.60 (7.81)
Two or More	-5.02 (4.60)
White	-.142 (3.28)
Special Education	-16.97 (1.64)	***
Chronic Homelessness: Attendance (centered)	28.16 (32.34)
R-squared	0.6126
No. Observations	2,626
Sign codes:* p<0.05, ** p<0.01, *** p<0.001

DISCUSSION

Aligned to the continuum of risk framework and in keeping with recent empirical work on homelessness, our fixed effects model (Table 3, model 2) shows a negative impact of homelessness on student achievement (Herbers et al., 2012; Obradović et al., 2009). Notwithstanding the fact that Midtown devotes considerable resources toward McKinney-Vento implementation, the impact of homelessness remains evident. The magnitude of the impact (9.71) represents almost a 10-point drop in math achievement and a 2.12% change. The average year-over-year math achievement gain for all students in the cohort is 19.66 points; therefore, our estimate represents a loss of about half a year of progress. While a magnitude of 9.71 is within a standard deviation (51.58), we only identified the lower bound; in other words, homelessness is at least this detrimental. Likewise, Obradović et al. (2009) found that homelessness and high mobility were risk factors for early school failure, yet nearly two-thirds of homeless and highly mobile students (HHM) had estimated math trajectories at or above 1 standard deviation from the national mean.

While it is challenging to isolate the impacts of homelessness from poverty, our fixed effects model (Table 3, model 2) makes important inroads in this direction. This model helps contribute to the mixed findings in the literature and supports the applicability of the continuum of risk framework. However, it is important to note that fixed effects cannot account for the nonrandom distribution of students. While McKinney-Vento prohibits the segregation of students experiencing homelessness, the fixed effects would be biased if administrators compensate for lower student ability by altering teacher assignments.

What types of the omitted variables might drive the difference between our OLS model (Table 3, model 1) and our fixed effects model (Table 3, model 2)? These variables (or set of variables) would need to have either a positive correlation with homelessness and a positive correlation with achievement growth or a negative correlation with homelessness and a negative correlation with achievement growth. For instance, parents social capital could have a positive correlation with the reporting (and therefore identification) of homelessness as well as a positive correlation with academic achievement.

It was interesting that the interaction between homeless status and attendance rate was not significant in either model. We speculate that this may be attributable to the both the narrow band of variation in attendance patterns and the substantial resources and services already directed toward high-needs students in this district. For instance, 80% of all observations in our sample have attendance rates between 89% and 99%, while students experiencing homeless have an average attendance rate of 90% and nonhomeless students average attendance rate is 95%.

Further, mobility did not reach significance in either model. Because school changes can occur for many different reasonssome of which signal social mobility and result in better schoolsthe school mobility literature (which does not always account for homelessness) is multilayered and nuanced (Beatty, 2010). However, we chose a measure of mobility (intradistrict midyear school transfers) that is often associated with negative outcomes (Beatty, 2010; Fantuzzo et al., 2012; Hanushek et al., 2004). Our mobility findings may partially reflect the fact that we used the McKinney-Vento definition of homelessness, which is broad. In other words, intradistrict midyear transfers did not predict achievement above and beyond our inclusive definition of homelessnessa definition that captures students who lack a fixed, regular, and adequate nighttime residence (McKinney-Vento, 2001). It is also possible that mobility would be more detrimental immediately after a student transfer; however, we measure achievement in the fall following the transfer (when state exams are administered). In retrospect, given the districts emphasis on school stability and smoothing the transition for students, it is possible that many of the short-term impacts of mobility may fade out by the following academic year. Future research should investigate the nature and impacts of delivered services around school mobility.

Math achievement is also only one outcome; even in an award-winning district such as Midtown, homelessness may enact more profound effects on other achievement tests, or even on noncognitive measures such as self-esteem and behavior. For instance, in the study by Fantuzzo et al. (2012), homelessness alone was not linked to significant differences in reading and math achievement, but students who experienced homelessness had more trouble with social and task engagement than their peers. Likewise, Obradović et al. (2009) found that HHM performed better in mathematics than in reading.

Measurement error also contributes to an attenuation of the homelessness effect; there are students within the not homeless sample who are homeless but who remain unidentified (Miller, 2011). To the extent that these students have notably lower outcomes as a result of their (undocumented) homelessness, the aggregate achievement of the not homeless students has been artificially deflated, making these two groups appear more similar than they are in reality.

Turning to longer-term, chronic homelessness, our results showed no discernible impact of chronic homelessness on math achievement growth (see Table 5). There may be psychological reasons that help account for the fact that chronic homelessness was not significant. It may be possible that parents experiencing chronic homelessness strongly stress the importance of studying and excelling in school as a pathway out of chronic homelessness. Or perhaps students experiencing sustained poverty (but not homelessness) do not receive the same level of encouragement from parents or teachers.

In particular, we speculate that as time goes on, some of the families experiencing chronic homelessness may develop coping mechanisms and routines that may be protective. Echoing Buckner (2008, 2012), we hypothesize that the Midtown district may play an important role in supporting and buffering families. In other words, chronically homeless families may have more time to build relationships with school and community actorsrelationships that may serve as conduits to protective resources.

Our study suggests that in order to reduce the impact of homelessness on academic achievement, it is important to be aware of when students become homelessso that their needs can be met right at this critical juncture. This point is an important contribution to the unresolved debate around the unique impacts of homelessness. Our findings imply that schools and districts should prioritize identifying students in a timely and continuous manner. While most students are identified through residency questionnaires at enrollment or at the beginning of the school year, we suggest that schools also consider administering these forms to current students at steady intervals throughout the year. This may help to better capture recent changes in housing status. States and districts should also consider including questions about housing status on routine school climate and engagement questionnaires in order to gain a deeper sense of scope and check for inconsistencies with official McKinney-Vento counts. For instance, starting in 2017, states and districts have the option to include questions on homelessness on the Youth Risk Behavior Survey (YRBS)a questionnaire on high school students health risk behaviors that is administered by the Centers for Disease Control and Prevention (CDC) (National Association for the Education of Homeless Children and Youth [NAEHCY], 2016). Disaggregating the YRBS by housing status allows localities to better understand both the risky health behaviors students experiencing homelessness may participate in as well as the protective factors available to them (Massachusetts Department of Elementary and Secondary Education, n.d.). If districts are concerned about minimizing the potential impacts of chronic homelessness, data such as the YRBS can provide insights into relationship building and service needs. States and districts should look for similar opportunities to better understand homelessness amongst elementary and middle school students.

To improve identification, families also need to understand the various ways in which McKinney-Vento can support their children. When families remain unidentified or do not understand their rights, the full potential of McKinney-Vento is thwarted. Fortunately, recent policy changes may facilitate timely identification and allow more families to forge the relationships and connections to resources that may be protective. Under the Every Student Succeeds Act (ESSA), which was signed into law in December 2015, a number of amendments were made to McKinney-Vento; many of these changes stress the importance of identification. For instance, ESSA stipulates that liaisons are required to post families McKinney-Vento rights in a wider range of community locations and must ensure that the rights are delivered in an understandable form (Duffield & Julianelle, 2016). These changes have the potential to improve self-identification and overcome stigma so that families connect to McKinney-Vento fasterand are also better able to reap its benefits. Future research should explore the impacts of these changes in supporting students academic growth.

Notes

1. After this study was conducted, the McKinney-Vento Act was reauthorized under the Every Student Succeeds Act (ESSA). Changes, which took effect in October 2016, included a phasing out of awaiting foster care placement from the definition. For more information, see the National Association for the Education of Homeless Children and Youth at www.naehcy.org. 

2. Although we believe that this increase represents a growing population of students experiencing homelessness, strategies to better identify homelessness have also increased over time.

3. As per our data agreement, the state (and standardized exam) must remain confidential.

References

Bassuk, E., & Rubin, L. (1987). Homeless children: A neglected population. American Journal of Orthopsychiatry, 57(2), 279286.

Beatty, A. (2010). Student mobility: Exploring the impact of frequent moves on achievement: Summary of a workshop. Committee on the Impact of Mobility and Change on the Lives of Young Children, Schools and Neighborhoods, National Research Council and the Institute of Medicine, Washington DC: The National Academies Press.

Buckner, J. C. (2007). Impact of homelessness on children: An analytic review of the literature. In Rog, D. J., Holupka, C. S., & Patton, L. C. Characteristics and dynamics of homeless families with children (pp. A1A36). Washington, DC: U.S. Department of Health and Human Services, Office of Human Services Policy, Office of the Assistant Secretary for Planning and Evaluation.

Buckner, J. C. (2008). Understanding the impact of homelessness on children: Challenges and future research directions. American Behavioral Scientist, 51(6), 721736. doi:10.1177/0002764207311984

Buckner, J. C. (2012). Education research on homeless and housed children living in poverty: Comments on Masten, Fantuzzo, Herbers, and Voight. Educational Researcher, 41(9), 403407.

Buckner, J. C., Bassuk, E. L., & Weinreb, L. F. (2001). Predictors of academic achievement among homeless and low-income housed children. Journal of School Psychology, 39(1), 4569.

Buckner, J. C., Bassuk, E. L., Weinreb, L. F., & Brooks, M. G. (1999). Homelessness and its relation to the mental health and behavior of low-income school-age children. Developmental Psychology, 35(1), 246257.

Burkam, D. T., Lee, V. E., & Dwyer, J. (2009, June). School mobility in the early elementary grades: Frequency and impact from nationally-representative data. Paper commissioned by the Committee on Impact on Mobility and Change on the Lives of Young Children, Schools, and Neighborhoods.

Canfield, J. (2015). School-based practice with children and youth experiencing homelessness. New York, NY: Oxford University Press.

Duffield, B., & Julianelle, P. (2016, January). The McKinney-Vento Act, as amended by the Every Student Succeeds Act of 2015 [PowerPoint presentation]. Retrieved from http://www.naehcy.org/educational-resources/essa

Fantuzzo, J. W., LeBoeuf, W. A., Chen, C. C., Rouse, H. L., & Culhane, D. P. (2012). The unique and combined effects of homelessness and school mobility on the educational outcomes of young children. Educational Researcher, 41(9), 393402. doi:10.3102/0013189X12468210

Gruman, D. H., Harachi, T. W., Abbott, R. D., Catalano, R. F., & Fleming, C. B. (2008). Longitudinal effects of student mobility on three dimensions of elementary school engagement. Child Development, 79(6), 18331852.

Hallett, R. E., Low, J. A., & Skrla, L. (2015). Beyond backpacks and bus tokens: next steps for a district homeless student initiative. International Journal of Qualitative Studies in Education, 28(6), 693713.

Hanushek, E. A., Kain, J. F., & Rivkin, S. G. (2004). Disruption versus Tiebout improvement: The costs and benefits of switching schools. Journal of Public Economics, 88(910), 17211746. doi:10.1016/S0047-2727(03)00063-X

Hart, B., & Risley, T. R. (1995). Meaningful differences in the everyday experience of young American children. Baltimore, MD: Paul H Brookes Publishing.

Herbers, J. E., Cutuli, J. J., Supkoff, L. M., Heistad, D., Chan, C.-K., Hinz, E., & Masten, A. S. (2012). Early reading skills and academic achievement trajectories of students facing poverty, homelessness, and high residential mobility. Educational Researcher, 41(9), 366374. doi:10.3102/0013189X12445320

Kerbow, D., Azcoitia, C., & Buell, B. (2003). Student mobility and local school improvement in Chicago. Journal of Negro Education, 158164.

Ketterlin-Geller, L. R., Chard, D. J., & Fien, H. (2008). Making connections in mathematics conceptual mathematics intervention for low-performing students. Remedial and Special Education, 29(1), 3345.

Massachusetts Department of Elementary and Secondary Education (n.d.). Homelessness in Massachusetts public schools. Retrieved from http://www.doe.mass.edu/mv/Survey2013.pdf

Masten, A. S., Herbers, J. E., Desjardins, C. D., Cutuli, J. J., McCormick, C. M., Sapienza, J. K., & Zelazo, P. D. (2012). Executive function skills and school success in young children experiencing homelessness. Educational Researcher, 41(9), 375384. doi:10.3102/0013189X12459883

Masten, A. S., Miliotis, D., Graham-Bermann, S. A., Ramirez, M., & Neemann, J. (1993). Children in homeless families: Risks to mental health and development. Journal of Consulting and Clinical Psychology, 61(2), 335.

Masten, A. S., Sesma Jr, A., Si-Asar, R., Lawrence, C., Miliotis, D., & Dionne, J. A. (1997). Educational risks for children experiencing homelessness. Journal of School Psychology, 35(1), 2746.

McKinney-Vento Homeless Assistance Act, 42 U.S.C. "" 1130111489 (2015).

Mehana, M., & Reynolds, A. J. (2004). School mobility and achievement: A meta-analysis. Children and Youth Services Review, 26(1), 93119.

Menke, E. M., & Wagner, J. D. (1997). A comparative study of homeless, previously homeless, and never homeless school-aged childrens health. Issues in Comprehensive Pediatric Nursing, 20(3), 153173.

Miller, P. (2009). Leadership and collaboration for homeless students: Examining community perceptions. The Urban Review, 41, 222250.

Miller, P. (2011). A critical analysis of the research on student homelessness. Review of Educational Research, 81(3), 308337.

Miller, P., Pavlakis, A., Samartino, L., & Bourgeois, A. (2015). Brokering opportunity for homeless students. International Journal of Qualitative Studies in Education, 28(6), 730749.

Murphy, J., & Tobin, K. (2011). Homelessness comes to school. Thousand Oaks, CA: Corwin Press.

National Association for the Education of Homeless Children and Youth. (2016). CDC to add new questions on youth homelessness to school survey: Advocates urge states and school districts to use them [Press Release]. Retrieved from http://www.naehcy.org/educational-resources/yrbs-press-release

National Center for Homeless Education. (2015). Federal data summary school years 201112 to 201314: Education for homeless children and youth. Retrieved from http://center.serve.org/nche/downloads/data-comp-1112-1314.pdf

Obradović, J., Long, J. D., Cutuli, J. J., Chan, C.-K., Hinz, E., Heistad, D., & Masten, A. S. (2009). Academic achievement of homeless and highly mobile children in an urban school district: Longitudinal evidence on risk, growth, and resilience. Development and Psychopathology, 21(2), 493518. doi:10.1017/S0954579409000273

Pavlakis, A.E. (2014). Living and learning at the intersection: Student homelessness and complex policy environments. The Urban Review, 46(3), 445-475.

Pavlakis, A.E. (2015). Reaching all families: Family, school, and community partnerships amid homelessness and high mobility in an urban district. Urban Education. doi:10.1177/0042085915613547

Rafferty, Y., Shinn, M., & Weitzman, B. C. (2004). Academic achievement among formerly homeless adolescents and their continuously housed peers. Journal of School Psychology, 42(3), 179199. doi:10.1016/j.jsp.2004.02.002

Rescorla, L., Parker, R., & Stolley, P. (1991). Ability, achievement, and adjustment in homeless children. American Journal of Orthopsychiatry, 61(2), 210220.

Rubin, D. H., Erickson, C. J., Agustin, M. S., Cleary, S. D., Allen, J. K, & Cohen, P. (1996). Cognitive and academic functioning of homeless children compared with housed children. Pediatrics, 97(3), 289294.

Rumberger, R. W., & Larson, K. A. (1998). Student mobility and the increased risk of high school dropout. American Journal of Education, 107(1), 135.

Schafft, K. A. (2006). Poverty, residential mobility, and student transiency within a rural New York school district. Rural Sociology, 71(2), 212231.

Schteingart, J. S., Molnar, J., Klein, T. P., Lowe, C. B., & Hartmann, A. E. (1995). Homelessness and child functioning in the context of risk and protective factors moderating child outcomes. Journal of Clinical Child Psychology, 24, 320331.

Shinn, M., Schteingart, J. S., Williams, N. C., Carlin-Mathis, J., Bialo-Karagis, N., Becker-Klein, R., & Weitzman, B. C. (2008). Long-term associations of homelessness with childrens well-being. American Behavioral Scientist, 51(6), 789809. doi:10.1177/0002764207311988

South, S. J., Haynie, D. L., & Bose, S. (2007). Student mobility and school dropout. Social Science Research, 36(1), 6894.

Tierney, W. G., Gupton, J. T., & Hallett, R. E. (2008). Transitions to adulthood for homeless adolescents: Education and public policy. Los Angeles, CA: University of Southern California, Center for Higher Education Policy Analysis.

Wong, J. H., Elliot, L. T., Reed, S., Ross, W., McGuirk, P., Tallarita, L., & Chouinard, K. (2009). McKinney-Vento Homeless Assistance Act Subtitle BEducation for Homeless Children and Youth Program: Turning good law into effective education. 2008 update. Georgetown Journal on Poverty Law and Policy, 16(1), 5398.

Ziesemer, C., Marcoux, L., & Marwell, B. E. (1994). Homeless children: Are they different from other low-income children? Social Work, 39(6), 658668.

Zima, B. T., Bussing, R., Forness, S. R., & Benjamin, B. (1997). Sheltered homeless children: Their eligibility and unmet need for special education evaluations. American Journal of Public Health, 87(2), 236240.

Zima, B. T., Wells, K. B., & Freeman, H. E. (1994). Emotional and behavioral problems and severe academic delays among sheltered homeless children in Los Angeles County. American Journal of Public Health, 84(2), 260264.

Cite This Article as: Teachers College Record Volume 119 Number 10, 2017, p. 1-23 https://www.tcrecord.org ID Number: 21840, Date Accessed: 3/18/2022 3:01:55 PM Purchase Reprint Rights for this article or review