ECE Program Supports and Teacher ‐ Perceived Support from Families: Are They Connected?

: According to the Conservation of Resources theory of stress, early care and education (ECE) teachers who receive greater tangible and interpersonal supports from their workplaces will be more positive and effective in their roles. This may translate to them perceiving or eliciting greater support from families, which is a key component to family engagement, a growing area of study in the ECE landscape. This study explores whether four program ‐ level supports (benefits, professional development supports, teacher social supports, program ‐ level family involvement ac ‐ tivities) are associated with teacher ‐ perceived support from families. The hypothesis was that all four will be positively associated. This study uses survey data from 102 preschool teachers and 13 preschool program directors in urban areas of two US states. We use ordinary least squares regres ‐ sion with cluster ‐ robust standard errors and a stepwise build ‐ up modeling procedure to determine associations between independent and dependent variables. While teacher social supports had the expected positive association with teacher ‐ perceived support from families, family involvement ac ‐ tivities were negatively associated. Our findings suggest that programs looking to improve family engagement may consider interpersonal/cultural supports for teachers and the larger school com ‐ munity. All else equal, simply offering more family involvement activities may not improve en ‐ gagement culture.


Introduction
The family-teacher relationship is the core of family engagement in school (Nzinga-Johnson et al. 2009). It describes the mutual efforts and collaboration of families and teachers to support children's academic learning and overall development (Kim and Sheridan 2015). In US early care and education (ECE) settings-which comprise preschool, centerbased child care, family-based child care, Head Start, and other educational and care programs for children before kindergarten-the family-teacher relationship is associated with a range of outcomes, including attendance (Waanders et al. 2007) and literacy growth (Nix et al. 2018). Pomerantz et al. (2007) state in their review of family engagement literature that that when families are authentically engaged in their children's schooling, their children may develop more academic and behavioral skills and have greater motivation. Moreover, an ECE behavioral intervention studied in randomized control trials (RCTs), the "gold standard" of scientific studies in which participants are randomly assigned to conditions and their outcomes compared to test causal effects, suggests that the familyteacher relationship may be a mechanism of improving young children's social skills (Kuhn et al. 2017;Sheridan et al. 2010Sheridan et al. , 2017. However, despite these benefits and increased attention to family-educator relationships in recent years (Nitecki 2015), little is known about the building blocks of the relationship, such as mutual support, communication (Vickers and Minke 1995), and perceptions of shared understanding and similarity (Hinde 1997).
Understanding how to facilitate positive relationships between families and teachers is critical. As these relationships are bidirectional (Kim and Sheridan 2015), strong familyteacher relationships may enable information-sharing, mutual support, collaborative solutions to children's problem behavior, and a positive affect extending to the child-teacher relationship (Dearing et al. 2015;Shpancer 1997;Zulauf and Zinsser 2019), which is associated with children's academic and behavioral development (Lippard et al. 2018). Teachers' and families' perceptions of each other are important to the relationship, as qualitative research suggests, teachers and families may withdraw from each other if they perceive they are unsupported, devalued, or disrespected (Lareau 2011;Lasky 2000;Puccioni 2018).
Generally, positive mutual perceptions enable trust-building between teachers and families Christenson 2000, 1998), while negative perceptions are a barrier to trust and goodwill (Adair and Barraza 2014;Epstein and Becker 1982;Lareau 1989Lareau , 2011Lareau and Horvat 1999). Zulauf and Zinsser's (2019) mixed methods study suggests that supports from ECE programs are important in shaping teacher perceptions of parents, or providing a buffer to the stress that family relationships can entail (Corr et al. 2014;Cumming 2017;Faulkner et al. 2016). This study extends their work in a quantitative analysis of program-level supports and their association with an element of teachers' family-related perceptions (i.e., teacher-perceived support from families).

Family-School Dynamics and the Role of Teachers' Perceptions
Although family-school relationships are bidirectional (Kim and Sheridan 2015), US K-12 literature suggests that educators often take the lead in establishing and maintaining the relationship (e.g., Hoover-Dempsey and Sandler 1997), and in ECE settings, most home-school communication is initiated by educators (Rimm-Kaufman and Pianta 2005). In addition, teachers' invitations play a powerful part in families' construction of their involvement role and can influence parents by making them feel important and valued (Hoover-Dempsey and Sandler 1997). Thus, when teachers more actively request involvement from families, families' school-based involvement activities tend to increase (Green et al. 2007). In addition, parents consider teachers who proactively request involvement as having higher quality (Epstein 1984).
Qualitative research suggests that teachers may feel that the onus is on them to engage with families (Mahmood 2013). Thus, while schools overall bear responsibility for initiating family engagement (Clarke et al. 2009), teachers can be seen as the drivers of this relationship and the "glue" that holds it together (Patrikakou and Weissberg 1999, p. 36). Given that the perception of mutuality is important to relationships generally (Hinde 1997), and that perceived support is particularly important to family-teacher relationships (Elicker et al. 1997;Lang et al. 2016;Vickers and Minke 1995), a high level of teacher-perceived support from families may support a healthy family-school relationship. This may be positively associated with children's success, as family-school relationships are positively associated with children's academic and non-academic outcomes ( Van Voorhis et al. 2013). Healthy family-school relationships are important especially in ECE settings when these relationships are being established (Epstein 1992).
Family relationships are a significant part of ECE teachers' work also because ECE programs are often two-generational (Chase-Lansdale and Brooks-Gunn 2014; Schmit et al. 2014), requiring that educators be skilled in serving both children and their families. In addition, ECE may naturally entail more family contact, as young children often depend on adults to verbalize needs and advocate for them to a greater extent than older children. The two-generational demands mean that ECE workplace supports are all the more important so that teachers have "bandwidth" to cultivate healthy family-school relationships (Hobfoll 1989), which is associated with children's outcomes ( Van Voorhis et al. 2013).
At the same time, US teachers often lack preservice and professional development training in cultivating family relationships (Clarke et al. 2009;Patrikakou and Weissberg 2000) and can perceive the obligation to do so as overwhelming or an "add-on" (Brown et al. 2009;Mahmood 2013). They may feel ill-equipped to meet families' needs, particularly when serving a materially disadvantaged population (Faulkner et al. 2016). Kotaman's (2016) qualitative study of kindergarten teachers in Turkey suggests that teachers' perceived inability to meet parents' expectations may cause stress. In addition, educators may define family engagement as school-centric activities, such as classroom volunteering (Gross et al. 2019;Lareau 2011;Lasky 2000;Lawson 2003), while lacking familiarity with "invisible" forms of engagement, such as family talks about manners and behavior (Arias and Morillo-Campbell 2008;Jarrett and Coba-Rodriguez 2018;Koury and Votruba-Drzal 2014). In other words, a teacher may perceive more support from a father who regularly volunteers than one who focuses on behavior management at home. Such misunderstandings can stymie family-teacher goodwill and the development of a healthy relationship (Lareau 2011;Lasky 2000;Puccioni 2018).
Thus, the importance of family relationships in ECE settings-and the challenges in managing them-point to a need to understand program-level supports that may be associated with teacher-family relationships. Responding to calls for more research into the processes of family engagement (Walker et al. 2010), particularly relationship aspects in ECE settings (Hall-Kenyon et al. 2014;Moorman Kim et al. 2012;Sheridan et al. 2012), this study seeks to deepen understanding of the family-teacher relationship by examining associations between various program-level supports and teachers' perceived support from families.

Theoretical Supports
This study builds on the Conservation of Resources (COR) theory of stress (Hobfoll 1989(Hobfoll , 2001, which links individuals' real or perceived resources to stress. COR states that when employees perceive adequate resources, they are "freed up" to acquire greater resources. For instance, teachers with paid time off may experience greater well-being and energy, psychic resources they could use to learn new techniques and cultivate workplace relationships, which spur well-being and energy, continuing the cycle. In contrast, teachers with inadequate resources may struggle to preserve what they have, forgoing education opportunities and avoiding others, further losing psychic resources (Lane and Hobfoll 1992). For instance, teachers who lack time for socializing at work may experience loneliness and low mood and preserve their limited psychic resources through further isolation. The theory aligns with self-determination theory, which holds that when workplaces provide adequate supports, employees perform better and with greater energy and well-being (Deci and Ryan 2008;Ryan and Deci 2000). Applied to an ECE family engagement context, teachers with adequate workplace supports may be better equipped to cultivate actual support from families or have positive attitudes that allow them to perceive greater support from families.
These theories inform our quantitative exploration of associations between programlevel supports and teacher-perceived support from families. While several studies find associations between teachers' perceptions of climate and their work quality within the classroom (e.g., Dennis and O'Connor 2013;Guo et al. 2011;McGinty et al. 2008), and Zulauf and Zinsser (2019) suggest teachers' climate perceptions influence their perceptions of families, less is known about how actual program supports for teachers may associate with their perceived support from families. Such associations may provide new insight into which resources may best support teachers in their family engagement efforts and responsibilities. Moreover, a deeper understanding of these associations will provide insight into which resources may best support teachers in their family engagement efforts and responsibilities.

The Role of Program-Level Supports in Family-Teacher Relationships
According to COR theory (Hobfoll 1989), teachers in more supportive programs may elicit greater support from families by an increased mood, energy, and work quality. They also may have "freed up" resources (e.g., time, energy) to devote to cultivating family engagement (Hobfoll 1989), which in turn is associated with children's academic and behavioral outcomes ( Van Voorhis et al. 2013). In addition, teachers who perceive that their programs support them-regardless of the number and type of actual supports offeredmay be more likely to have an empathetic affect manifesting in positive perceptions of families (Zulauf and Zinsser 2019). This speaks to the importance of teachers' perception of a positive, supportive climate at their centers. In this study, we explore four types of supports that may be associated with teacher-perceived support from families via "freeing up" resources or contributing to a positive climate, and thereby teacher affect. They are compensation benefits, support for professional development, social supports for teachers, and program-level family involvement activities.

Non-Wage Compensation as a Potential Predictor
Teachers in US ECE settings have notoriously low compensation, with wages often considered to be "unlivable" (Whitebook et al. 2014). Child care providers can earn less hourly than animal caretakers (Whitebook et al. 2014), and teachers may rely on other sources of income (e.g., spouse wages) to make ends meet (Mcdonald et al. 2018). Compensation studies consistently point to the importance of wages (Arndt 2018;Bellm and Whitebook 2006;Phillips et al. 2016;Whitebook et al. 2014), suggesting the possibility of similar association with benefits. Hall-Kenyon et al. (2014) suggest that, given the low wages that US ECE teachers typically earn, benefits may take on even greater importance, constituting a financial guardrail that higher paid individuals create through saving. Indeed, increasing non-wage compensation (e.g., paid time off), or benefits, is recommended to improve ECE (Smith and Lawrence 2019; Whitebook et al. 2018). Moreover, according to COR theory (Hobfoll 1989(Hobfoll , 2001, benefits may constitute a resource that reduces teachers' stress, enabling psychic resources that they can invest for greater resources, including support from families. Thus, in this study, we extend literature on US ECE teachers' wages by examining compensation benefits as an independent variable predicting teacher-perceived support from families. ECE wage literature suggests that benefits may be associated with teacher-perceived support from families through teacher quality. For instance, in an Australian study, Leigh (2012) demonstrates that a 1% increase in starting teacher's salary is associated with a 0.6 percentile rank increase in the aptitude of students taking education classes. Moreover, higher teacher pay is associated with student academic achievement (Hendricks 2014) and behavior (King et al. 2016). Thus, the notoriously low compensation of ECE (Ackerman 2006;Mcdonald et al. 2018; National Survey of Early Care and Education Project Team 2013; Shdaimah et al. 2018;Bellm and Whitebook 2006) may make it difficult to retain preservice teacher students (Early and Winton 2001) and hire and retain high-quality teachers (Barnett 2003), especially as the demands and market for ECE grow (Moss 2006;Whitebook and Ryan 2016). Moreover, Lerkkanen et al. (2013) found that mother's trust in teachers is positively associated with the quality of the teacher's practices, underscoring Xu and Gulosino's (2006) statement that family-teacher relationship-building may be a component of teacher quality. Thus, compensation may be associated with teacher and family-teacher relationship quality, and in this study, we explore if these wage findings translate to benefits.
A second potential mechanism is through teacher stress, with which low pay is associated (Phillips et al. 2016). Qualitative research suggests that low wages take a psychological toll on teachers, with teachers finding low wages to be demoralizing and part of their overall sense of being perceived as unskilled or mere "baby-sitters" instead of professionals (Gerstenblatt et al. 2014;Moloney 2010). According to COR theory, high-stress individuals may alienate potential interpersonal supports (Hobfoll 1989;Lane and Hobfoll 1992). Thus, if low compensation benefits are associated with high stress, teachers with few benefits may be less likely to cultivate or perceive family support.

Program-Level Support for Professional Development as a Potential Predictor
Professional development (PD) refers to a variety of practices (e.g., teacher training, mentoring, workshops) that promote the development of educators' abilities to generate positive outcomes for children (Sheridan et al. 2009;Snyder et al. 2011). Sheridan et al. (2009) state that a long-term goal of PD is to improve teachers' ability to promote families' attitudes and abilities to create the best supported learning environment for the child. Policymakers find that successful family engagement programs incorporate ongoing training (Stark 2010).
PD activities may be associated with family-teacher relationships by improving teacher quality generally or by training teachers to collaborate with families. Teachers' credentials, training, and education are associated with teacher quality (Torquati et al. 2007) and sensitivity to children (Gerber et al. 2007), as well as children's outcomes (Saracho and Spodek 2007). Moreover, ECE teacher education programs often provide little training in cultivating and maintaining family relationships (Evans 2013;Patrikakou and Weissberg 2000), particularly practical experience (Early and Winton 2001). Thus, ongoing PD may provide such training, with evaluation studies suggesting PD can improve teachers' practice, including their perspectives in working with families from minoritized groups (Evans 2013;McMillan et al. 2012;Trumbull et al. 2003).
In addition, PD opportunities may be associated with family-teacher relationships by improving the overall climate of the school. Qualitative studies suggest that ECE teachers have a strong desire for PD, viewing it as a much-needed and valuable way to raise the reputation of the profession (Harwood et al. 2013;Nelson and Lewis 2016) and implement interventions and improvements to classroom practices (Kilgallon et al. 2008). In addition, in a study of program-level social-emotional supports including training, teachers at highly supportive programs were less likely to be depressed and reported higher job satisfaction (Zinsser et al. 2016). As a welcoming climate is an important factor in eliciting parent support and engagement (Baker et al. 2016;Hoover-Dempsey et al. 2005;Nitecki 2015), to the extent that PD improves the overall climate of the school, it may have a positive association with teacher-perceived support from families.

Program-Level Social Supports for Teachers as a Potential Predictor
Social support refers to the help, positive affect, and information that individuals receive from those with whom they have interpersonal relationships, and among teachers, it is negatively associated with burnout and positively associated with emotional wellbeing (Greenglass et al. 1997;Russell et al. 1987;Sarros and Sarros 1992). In a qualitative study with 20 Head Start teachers, Wells (2017) found many benefits of social supports among teachers. Teachers derived the most benefits from their co-teachers and program directors, but also, teachers across classrooms showed support by covering each other for bathroom breaks, celebrating birthdays, and eating lunch together (Wells 2017). Moreover, those with plentiful and high-quality social support systems at work had better attitudes and could counteract stress (Wells 2017). Similarly, Kupila and Karila (2019), in a project analyzing the development of teacher identity among ECE teachers in Finland, found that social supports among teachers enabled them to feel less isolated, share their emotional experiences, brainstorm or role-play to address challenges, and absorb norms and tips on the profession through peer mentoring. Likewise, Kilgallon et al. (2008) found that ECE teachers used social supports to implement pedagogical changes, share knowledge, and deepen their understanding of their work.
Consistent with these qualitative findings, Cramer and Cappella (2019) found that assistant teachers who seek out social support are more likely to be satisfied with their work. Jeon and Ardeleanu (2020) found that ECE teachers with a positive work situation-including interpersonal relationships-use more positive emotional regulation strategies and feel less stress. McGinty et al. (2008) also suggest that teacher-perceived collegiality in their program is associated with their attitudes towards teaching. Thus, teachers at programs with strong social supports may be more positive, perceiving greater support from families, or more effective, able to elicit greater support from families.

Program-Level Family Involvement Activities as a Potential Predictor
Program-level supports for family involvement are associated with greater levels of involvement (Ansari and Gershoff 2016), and qualitative work suggests that families appreciate repeat invitations for involvement (Bruckman and Blanton 2003). Moreover, program-level family involvement activities (e.g., classroom volunteering, attending school events, participating in a parent council) may associate with teacher-perceived support from families because teachers often define family engagement as participation in schoolbased involvement activities (Gross et al. 2019;Lareau 1994;Lareau and Horvat 1999). In other words, teachers may believe supportive families are those who assist them in their work (Christianakis 2011;Lareau 1994). Hoover-Dempsey and Sandler (1997) also pointed to the importance of school-level efforts, suggesting that schools create climates that can promote parents' role construction as involved parents.
Furthermore, frequent contact is an important factor in relationship development between families and teachers, and in particular, the development of trust Christenson 2000, 1998). The more families and teachers interact, the more opportunities they have to demonstrate consistency and trustworthiness and know each other as individuals (Adams and Christenson 1998). Thus, to the extent that program-level family involvement activities facilitate more contact, they may be associated with teacher-perceived support from families. For instance, a grandmother who volunteered in the classroom weekly would have regular opportunities to establish a trusting relationship with the teacher and would visibly demonstrate her support for the teacher. However, other studies find a non-significant association between program-level family involvement activities and family involvement (Green et al. 2007;Walker et al. 2011), leaving an opening for further exploration. Thus, in this study, we explore program-level family involvement activities as a potential predictor of teacher-perceived support from families.

The Present Study
We hypothesize that program-level supports for teachers will be positively associated with their perception of support from families. Self-determination and COR have identified possible mechanisms by which this may occur: (1) promoting teachers' work engagement and quality by meeting their psychological needs, and (2) by reducing stress and facilitating teachers' use of energy for the acquisition of greater resources. Moreover, qualitative (e.g., Nitecki 2015;Tebben et al. 2021;Zulauf and Zinsser 2019) and quantitative (e.g., Goddard et al. 2015;Jung and Sheldon 2020;Lubienski et al. 2008) studies suggest an interplay between program climate and culture and teachers' attitudes about parents. The present study extends such work by examining associations between programlevel supports, as reported by ECE program directors, and perceived support from families, as reported by teachers. Specifically, we examine: "To what extent are program-level supports (measured by compensation benefits, professional development supports, social supports [program-level social supports and teacher-perceived collegiality], and program-level family involvement activities) associated with teacher-perceived support from families?"

Materials and Methods
This study uses a dataset with 102 teachers (53 lead, 8 co-lead, and 41 assistant) from 13 ECE programs in two states. Six programs are Head Start, five are for-profit, and two are public preschools. Among teachers, 45% had a bachelor's or greater, and the median salary was USD 20,000-25,000, with range of less than USD 5000-75,001 or greater. The sample was mostly Black (49%) and White (43%), with the remaining 8% combined in a category called Other, comprising Asian/Native Hawaiian/Pacific Islander (n = 3), Multiracial (n = 2) and Other (n = 3). The mean years of ECE experience was 11.85 with a range of 0 (less than one year) to 34 years.

Procedure
These data were collected as part of a larger study on teachers' social-emotional capacity. They were collected in spring of 2016. To recruit ECE programs, research team members coordinated with local ECE agencies to obtain a list of programs. They randomly selected programs to participate. The eligibility criterion was that the program had to have at least four classrooms with full-time programming. We selected 13 programs. Team members called program directors and invited them to participate. After obtaining consent from directors who agreed, they sent directors a survey about directors' professional background program-level practices and policies, which took about 10 min to complete. To recruit teachers, team members visited the schools and described the study. All lead teachers were eligible, and all assistant teachers of all participating lead teachers were eligible. After obtaining consent from teachers who agreed to participate, team members distributed a paper teacher survey asking about teachers' professional background and perceptions regarding support and climate, which took about 20 min to complete. Team members collected completed surveys at a later date.
All respondents gave informed consent before participating. We conducted this study in accordance with the Declaration of Helsinki. The Institutional Review Boards of The Ohio State University (2015B0181) and Johns Hopkins University (HIRB00003643) approved this study.

Dependent Variable: Teacher-Perceived Support from Families
To measure teacher-perceived support from families, we used a subscale of the Scales Measuring Aspects of Child-Care Quality (Emlen et al. 2000), which has been used to measure teachers' perception of family support (e.g., Jeon and Ardeleanu 2020). Emlen et al. (2000) tested validity in the original scale, which is reported by parents, and found that the full scale had a 0.69 correlation with parents' perception of overall ECE quality. Our version was the child-care provider measure, adapted by the Fragile Families and Child Wellbeing Study (Reichman et al. 2001). The subscale has six items, and items have adequate internal reliability (α = 0.85 in the current sample). The items assess frequency (1 = never, 4 = always) of relationship perceptions. An example item is: "Family members are supportive of me as a caregiver". Note that these items reflect teachers' perception of support from families, and not perception of families in general. We summed as suggested by previous studies (Emlen et al. 2000;Jeon and Ardeleanu 2020), and as the completion rate was high and all items were positively phrased.

Independent Variables: Program Supports
We measured program support variables using director responses to survey sections about program characteristics. First, we measured compensation benefits by asking which of four different types of benefits their program provided to teachers: paid sick days, paid vacation days, tuition coverage for own children, and mental health supports (1 = yes, 0 = no). We calculated the sum of these four items. Second, we measured professional development supports by asking which of nine types of professional development supports their program provided. Supports (binary, 1 = yes) were finance-related (e.g., paid days off, reimbursement, compensatory time, college courses, college books), direct (e.g., on-site programs), or personnel-related (e.g., mentors, coaches). We used the sum of these nine items.
Third, directors reported which of four social supports (e.g., shared release hours for planning time, binary, 1 = yes) their program offered. We used the sum of these four items to represent program-level social supports. Furthermore, we measured teacher-perceived social supports by using the collegiality subscale from the School as a Caring Community Scale (Battistich et al. 1997, α = 0.91 in this sample). The scale has six items (e.g., "Teachers frequently consult with and help one another"), with responses on a 5-point Likert scale (1 = Not at all true; 5 = Very true).
Finally, to measure program-level family involvement activities, directors were asked which of 10 program-level family involvement activities (e.g., "We ask parents to help with assignments so that children's learning continues in the home", "We hold regular parent-teacher conferences", "We have parent meetings/council") their program used. We dropped three items that were selected by all directors ("Parents can visit and stay in the classroom", "We ask parents to volunteer for class activities", "We encourage teachers to communicate with parents regularly"). We used the sum of the remaining 7 items to represent family involvement activities.

Analytic Plan
This study examines associations between compensation benefits, professional development supports, teacher social supports, and program-level family involvement activities after controlling for demographics and experience. Because teachers' data are nested within programs' data, we explored the intraclass correlation (ICC) within a hierarchical linear modeling (HLM) design with teachers at Level 1 and programs at Level 2. The ICC was less than 0.01%, indicating that there was almost no variance in the dependent variable between programs. Thus, we used multiple linear regression analysis using ordinary least squares (OLS) regression in Stata 16.0.
To check for homoscedasticity, we performed statistical and visual tests. Specifically, we used White's (1980) and Breusch and Pagan's (1979) tests, which test the null hypothesis of homoscedasticity. Table 1 shows the results, with p values surpassing the 0.05 threshold, indicating a failure to reject the null. However, visual tests showed some narrowing of residuals at high values of perceived support, so we used cluster-robust standard errors (Colin Cameron and Miller 2015;Croux et al. 2003) using the Stata specification cluster. For this exploratory study, we used a stepwise build-up modeling procedure. Specifically, Model A is a null model with the dependent variable only, showing the intercept for comparison. In Model B, we added conditioning variables (teacher educational attainment, experience, lead status, salary, race, sense of community, and program Head Start status). We included teacher race first because employee supports may be useful or perceived differently according to an employee's race (Sloan et al. 2013). Second, if families internalize widespread biases against people of color (Wilkerson 2020), knowingly or not, they may be more likely to hold common perceptions that ECE is not "real" school (Moloney 2010) or that teachers are mere "baby-sitters" (Nelson and Lewis 2016) if their teacher is a person of color, which may influence their support.
Model C adds benefits, Model D adds professional development supports, Model E adds social supports (i.e., program-level social supports and teacher-perceived collegiality), and Model F, the full model, adds program-level family involvement activities. We performed sensitivity analyses, testing the models separately for Head Start and non-Head Start centers, for centers with high-and low-average student family incomes, and for the two states. Table 2 shows descriptive statistics of key variables and Table 3 shows bivariate correlations. Correlations between the dependent variable, teacher-perceived support from families, and key independent variables were small in magnitude, ranging from −0.13 (benefits) to 0.10 (social supports). This suggests a limited association between center-level supports and teacher-perceived support from families. Some independent variables were correlated (e.g., program-level social supports and involvement, 0.84, p < 0.001), so we examined multicollinearity using variance inflation factor (VIF) values. We retained all independent variables because all VIF values were less than 10 (Wooldridge 2016). The VIFs for involvement and social supports were 6.90 and 5.77, respectively, notably higher than other variables, which were less than 2.6. In addition, the F test of joint significance, which tests the significance of the model overall against a null hypothesis of a model without independent variables fitting the data as well, shows that the model is jointly significant, with F = 2.00 (4, 89) and p = 0.10.    Table 4 shows regression results. Model A shows a statistically significant intercept of 19.30 (p < 0.001). In Model B, we added conditioning variables (experience, lead teacher status, salary, race, and Head Start status), none of which were statistically significant, and which yielded an R 2 of 0.10. In other words, these covariates explained 10% of the variance in teacher-perceived support from families. In models C, D, and E, we added benefits, professional development, and social supports (i.e., program-level social supports and teacher-perceived collegiality), respectively. The coefficients were not statistically significant, and the variables did not substantially change the R 2 . In other words, benefits, professional development, and social supports were not associated with teacherperceived support from families.

Regression Analyses
In Model F (full model), we added involvement, which had a statistically significant negative coefficient of −0.99 (p < 0.05), meaning that teachers perceived less support from parents when their programs had more involvement opportunities. In addition, in Model F, program-level social supports became statistically significant (1.56, p < 0.01), suggesting that, when controlling for number of involvement practices, teachers perceived greater support from families at programs with greater social supports. The R 2 of Model F was 0.18. This indicates that an additional 8% of the variance was explained by the key independent variables compared to Model B (covariates only). No covariates were statistically significant in any models.

Discussion
In this study, we examined the associations between four types of program-level supports (program compensation benefits, professional development support, social support, and family involvement activities) and teacher-perceived collegiality, and ECE teacherperceived support from families. We hypothesized that all five key independent variables would be positively associated with teacher-perceived support from families, in accordance with the conservation of resources theory of stress (Hobfoll 2001). These theories posit that teachers in well-supported work environments may have more positive attitudes and work quality, which may facilitate greater actual or perceived support from families. Importantly, the dependent variable is teacher-perceived support from families, and as such, could reflect actual support, perceived support, or some combination thereof.

Benefits and Professional Development Supports: Null Findings
The null effects for benefits and professional development supports may speak to the limits of a program simply offering supports. First, an offered support may be infeasible. For instance, if directors provide reimbursement for PD expenses, but do not advertise that fact, or teachers cannot afford the up-front costs, the support is, practically speaking, nonexistent. Similarly, operational challenges could stymie the usability or take-up of ostensible supports. For instance, if paid PD days or reimbursements take months to process, teachers may be reluctant to participate in them. This is especially true given the typical US ECE worker salary, which is so low that its median qualifies those with a family of three for food stamps in every US state (U.S. Department of Education 2016).
Second, the offering of supports does not necessarily mean supports are high-quality, frequent, or intense, all of which may be necessary to realize benefits (or even associations). Hall-Kenyon et al. (2014) state that supports that are poorly tailored or not executed well can backfire and have unintended negative consequences. In this way, these results echo those of Harding et al. (2019), who propose that null associations between ECE professional development and teachers' well-being, attitudes, and teaching practices are due to variation in PD quality and content. Underscoring this point is that program directors reported program-level resources, while teachers reported their perception of support from families. It is possible that teachers were not aware of the supports that their pro-grams had, or that they perceived ostensible supports as burdens. For instance, PD supports, shared meeting times, or even mental health benefits all could feel like add-ons to full slates of work.
Together, these findings point to the importance of the quality, intensity, and feasibility of supports, and of climate overall. Structural supports may mean little if they are not usable or perceived as beneficial. Moreover, climate may "override" the presence or absence of program-level supports. In other words, a program that offers robust PD supports but has a climate where PD feels like an "add-on" may realize fewer benefits for teachers than a program with few PD supports but a warm and growth-oriented climate. Ultimately, the null findings suggested limitations of structural markers of teacher supports and the need for exploration of process markers of teacher supports.

Positive Significant Association with Social Supports
Among the hypothesized independent variables, only program-level social supports were significantly positively associated with teacher-perceived support from families (Model F). It may be that social supports were significant because this type of relationshipbuilding support is more effective than benefits, professional development, and family involvement at creating a friendly and encouraging environment. Zulauf and Zinsser (2019) conducted a mixed methods study comparing ECE teachers who expelled students to those who did not. They found that non-expelling teachers reported greater support from their program administrators, not in terms of explicit professional development or benefits, but from friendliness, helpfulness, and advocacy. Moreover, non-expelling teachers described positive and proactive parent communication norms that are "part of the programs' culture" (Zulauf and Zinsser 2019). In addition, these teachers were more likely to express empathy with families and have positive reactions to children's problem behaviors. Thus, social supports may promote healthier family perceptions (or even relationships) through improvements to overall climate. Sensitivity analyses suggest this positive association is driven by Head Start centers and centers serving children from relatively low-income families. As these children are more likely to present challenges to teachers (Reardon and Portilla 2016), which are associated with teacher stress (Friedman-Krauss et al. 2014), social support-given the warmth and knowledge-sharing with which it is associated (Kilgallon et al. 2008;Kupila and Karila 2019;Wells 2017)-may take on a greater importance in these settings.
Therefore, if social supports are successful at building connections among teachers, this variable may speak more to the overall climate of programs than relatively discrete supports such as benefits, professional development, and involvement activities. In other words, helping teachers connect socially may contribute more to their attitudes and workplace perceptions above and beyond "check the box" supports such as benefits. Interestingly, however, teacher-perceived collegiality, which is teachers' perceptions of the social support climate in their program, was not significantly associated with the outcome in any model. It is possible that when it comes to how teacher think about families, intentional supports by programs to create connections among teachers may be more important than what teachers just feel about their climate. In addition, it could be that different climate-related factors may be associated with teacher-perceived support from families and other family engagement-related constructs (e.g., welcoming environment, feeling that administrators "have your back"; Hoover-Dempsey et al. 2005; Zulauf and Zinsser 2019).
However, note that in Model E, social supports had a much smaller coefficient and was not statistically significant. Thus, social supports were significantly associated with teacher-perceived support from families only when conditioning on program-level family involvement activities. This could reflect the high correlation between collaboration and involvement (see Table 2), indicating that programs with high social supports tend to have many family involvement activities, as well. High correlation would make the effects of social supports indiscernible without parceling them out across values of involvement activities, which would explain the changes between models E and F. These coefficients were robust in a sensitivity analysis in which models E and F were switched, so that Model E included involvement but not social supports, and Model F had both. Further suggesting that involvement generally has a negative association with teacherperceived support from families independent of social supports is that when it is not in the model, the constant is higher.

Negative Significant Association with Program-Level Family Involvement Activities
Contrary to our hypothesis, family involvement activities were negatively associated with teacher-perceived support from families. In other words, at programs with more opportunities for families to be involved, teachers perceived less family support. This interesting finding contradicts other work that has examined family involvement and the family-teacher relationship. For instance, the quantity of involvement opportunities is positively associated with levels of family involvement (Ansari and Gershoff 2016), and family-teacher trust grows as involvement and interactions increase Christenson 2000, 1998). Underscoring these findings are Brown et al. (2009) andMendez (2010), who conducted qualitative evaluations of ECE school readiness interventions that operate through the family-teacher relationship. Both studies found that teachers perceive greater connectedness with families as families' involvement increased. While these studies' outcomes are not precisely teacher-perceived support from families, they indicate a generally positive association between families' involvement behaviors and family-teacher relationships.
We conducted a sensitivity analysis of this finding in which we compared the two states in which data were collected. The analysis showed that the negative association was driven by one state. This may signal that state regulations around family engagement may incentivize program directors to "check a box" regarding family engagement opportunities at the expense of the less certain work of creating a positive work climate and a welcoming environment for families. Moreover, the measure reflected reports from program directors on the number of involvement activities offered. Therefore, it does not reflect families' actual involvement. It also may not reflect the actual opportunities, as directors may overreport programs' offerings (Lower and Cassidy 2007). In addition, the activities offered may be difficult for families to participate in. Common barriers include scheduling conflicts and time constraints, rigid working hours, lack of transportation, language barriers, parents' negative school experiences, and different understandings of parents' role in their children's education (Baker et al. 2016;Loughran 2008).
Moreover, if greater involvement activities correspond to greater involvement expectations on the part of teachers, then the same rate of involvement may seem "worse" at a program with many involvement activities than at a program with few. In other words, families who participate in two activities may be perceived as "less supportive" at a program with seven activities than at a program with four activities. As negative perceptions and experiences are processed and remembered more thoroughly than positive ones (Baumeister et al. 2001), larger gaps between expectations and actual or perceived involvement may pave the way for lower teachers' perceptions of support from families. Furthermore, qualitative work in US K-12 settings finds that parents, especially those from minoritized groups (Kim 2009), may withdraw if they perceive that educators are condescending to them or think they have little to offer (Lareau 2011;Lasky 2000;Puccioni 2018). In addition, educators and families may have different perceptions of what constitutes "family engagement" (Gross et al. 2019), meaning families' family engagement actions may not be perceived as such by educators, while educators' expectations may not align with families' perceptions of family engagement.
Another potential explanation is that a poorly executed involvement activity may push families away, especially families who are low-income and therefore already more likely to be facing barriers to involvement (Grant and Ray 2018). For instance, if programs regularly send home newsletters in a language that some families cannot read, or with outdated information, then families speaking the non-dominant language may feel left out or even turned away. Given that the family-school relationship is bidirectional (Kim and Sheridan 2015), it is im-portant that schools both make overtures to families and are places where families feel comfortable "engaging themselves". If a greater number of activities corresponds to a greater number of unintended exclusionary messages, then a negative association would occur, and furthermore be more likely in schools serving lower-income students. Hoover-Dempsey et al. (2005) found that a welcoming environment is foundational to establishing successful familyschool relationships, which Nitecki (2015) underscores in a case study of a successful ECE family-school partnership. Thus, a family perception of poor quality or accessibility may undermine a programs' intended positive effect of involvement activities. It is worth to explore this contradictory finding in future studies because K-12 and some ECE research have emphasized family involvement activities as a way of improving family engagement and children's outcomes (e.g., Ansari and Gershoff 2016;Hayakawa et al. 2013;Henderson and Mapp 2002). If it does not support teachers' feelings of relationships with families, however, it is important to re-visit the current family engagement practices.

Limitations and Future Work
First, as this study's analytic method is OLS regression and its data are cross-sectional, it cannot support causal claims. The analysis shows associations only, meaning findings do not show that one variable affects or influences another. In addition, although the independent variables in this study met VIF thresholds (Wooldridge 2016), some independent variables were highly correlated, which raises the possibility of multicollinearity. It is possible, for instance, that generally, the same kinds of programs tend to have both high collaboration supports and family involvement activities, or that a third unobserved variable drove both, generating their high correlation.
In addition, while the sample featured randomly selected programs, and a variety of program types, the number of groups (J = 13) is low. This limits generalizability, particularly given the wide-ranging landscape of early care and education, which operates under widely varying regulations, curricula, and funding models (Chaudry et al. 2017), and is not compulsory, as is K-12 education. Moreover, the sample was mostly Black and White, and had relatively few participants from other racial or ethnic backgrounds. Future studies should incorporate larger samples from more diverse backgrounds and represent a larger number of programs.
A further limitation is the lack of variables that would provide context for the supports. For instance, data about the content of professional development, teachers' awareness of social supports, perception of benefits, and family take-up of involvement opportunities would enable more nuanced and deeper understanding of the associations between program-level supports and teachers' perceived support from families. Although the covariates explained 10% of the variance, none were statistically significant, suggesting that other variables account for the variation in teacher-perceived support from families. Likewise, data about the quality of the supports would help contextualize the findings. Similarly, the dataset did not contain measures of the quantity and quality of training teachers had received about how to engage families, which could have explained variation in their perceived support. Moreover, the dependent measure has only six items and thus is limited in its capture of teacher-perceived support. For instance, it does not allow separate analyses of material and emotional support. Together, these limitations mean that the results should be interpreted with caution.
Thus, future work should expand this study through both larger samples and with contextual variables. Analyzing director-reported supports with teachers' awareness or uptake thereof would address the possibility that ostensibly offered supports are infeasible, impractical, or unknown by those they seek to benefit. In addition, qualitative research may help us understand unexpected findings on family involvement activities. Such work will necessitate more complex studies with additional research questions but will facilitate a deeper and more nuanced understanding of the interplay between contextual supports within programs and teachers' perceptions of family support.

Implications
As both the absolute and proportionate number of US children using ECE programming grows (National Center for Education Statistics 2019), stakeholders are exploring ways to expand, improve, and equalize programming (e.g., Chaudry et al. 2017). At the same time, ECE practitioners remain chronically underpaid (Whitebook et al. 2014), and often misconstrued as baby-sitters (Nelson and Lewis 2016), even as their role grows in importance and demands (Cassidy et al. 2019). Thus, stakeholders have called for improved supports for the ECE workforce (e.g., Schilder 2016;Ullrich et al. 2017).
This study contributes to family-school relationship literature first by finding that program-level social supports for teachers is positively associated with teacher-perceived support from families, while program-level involvement activities are negatively associated. Put another way, the only positive association with teacher-perceived support from families in this study was with social supports-a measure that included collaborative and relationship-building items such as shared planning time, staff meetings, and social events. This is an understudied area in the family engagement literature. As such, this study points to the possibility that interpersonal relationships among teachers have a role not only in creating positive environments for those within the program (i.e., teachers and children), but also for the wider community (i.e., families).
Moreover, it may have a larger role than other types of supports, such as benefits and professional development, commonly listed in the ECE literature (e.g., Phillips et al. 2016;Whitebook et al. 2014). Specifically, this study points to the possibility that climate factors, above and beyond supports "on paper", may contribute to teachers' perceptions of support. Thus, policymakers and program leaders may consider the overall environment, and not merely "check boxes", when considering ways to support teachers. Program leaders may carve out dedicated time for teachers to socialize or be mindful when creating schedules. Together, the findings in this study suggest that the existence of supports for teachers, such as paid days off or reimbursement for workshops, may not be enough to actually support teachers.
Thus, this study is a caution against assumptions about impacts or effects of program-level teacher supports, especially because it raises the possibility that offering many ways for families to be involved in school may predict negative teacher perceptions. This negative association suggests that the how of interventions bears on their success, as much or more than the what of interventions. Rather than simply adding opportunities for families to be engaged, such as a Facebook page or newsletter, school leaders should attend to how educators talk about family engagement, perceive families, and cultivate expectations within the school, and be mindful of the families' needs and perspectives when designing engagement opportunities. In this way, this ECE study aligns with recommendations from Epstein (2016) regarding school, family, and community partnerships in K-12 settings.
Thus, interventions seeking to improve family-school relationships for Head Start and other ECE programs may start with climate surveys and qualitative work exploring how teachers-and families-perceive the climate, structure, and ostensible benefits of their workplace. Such work could inform interventions that are responsive and relevant. In addition, it may create more efficient supports, as reallocating resources from low-uptake or irrelevant supports could be helpful to ECE programs, which often operate on lean budgets.
In centralized programs such as Head Start, teacher supports may be prescribed and regulated. However, given the vast variation in ECE program settings, sizes, scopes, aims, geographies, populations, and degree of formality, there is hardly a standard of support across the ECE landscape (Pianta 2007). Thus, policymakers at federal, state, and local levels would do well to consider the quality, relevance, and accessibility of supports as more children attend formal preschool (Friedman-Krauss et al. 2020) and states seek to refine their Quality Rating and Improvement Systems (Boller and Maxwell 2015).
Above all, this study points to the need for nuanced analysis into the environments in which ECE educators labor, and the many relationships they manage to do so. As such, it contributes to an expanding body of work that highlights the large (Bainbridge et al. 2005) and growing (Harper 2018) importance of these educators, and the need for attention into if-and, if so, how and how well-they are supported. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data are not publicly available due to IRB restrictions for confidentiality.