Early childhood teachers’ lives in context: Implications for intervention effectiveness within early childhood education

• In Event: Lessons from teachers' professional development programmes

Proposal

Background: Teachers are the most consistent, strongest school-level predictor of student learning (Hanushek, 2011; OECD, 2005). They are also frequently tasked with implementing educational interventions. Yet, there is a dearth of research examining teachers’ professional and personal experiences and how these influence their ability to implement educational practices. This is of particular importance in pre-primary and under-resourced settings, where teachers are generally under-paid and poorly respected and are likely to be working in contexts with few resources and little pedagogical support (Neuman et al., 2015; Raikes et al., 2015; Sun et al. 2015; UNESCO, 2014).

Purpose: This study uses data from Quality Preschool for Ghana (QP4G) to explore how teachers’ experiences influence the impact of educational interventions. We examine whether teachers’ lives in context (i.e. teacher skills, knowledge, and attitudes; poverty and health; and contextual supports and barriers to teaching: Authors, in press) moderates the impact of this educational intervention on classroom processes and children’s learning. In doing so, we identify key aspects of teachers’ lives (e.g., poverty risks, low professional status) that promote or disrupt the intervention’s effectiveness. We further test whether this moderation takes place via increased adherence to training materials.

Intervention/Design: QP4G is a pre-primary teacher training program created by the National Nursery Teacher Training Center in Ghana. It was implemented in six of the most disadvantaged urban/peri-urban districts of the Greater Accra region. The goal was to develop and test a nationally scalable model for effectively, and at low cost, improving preschool classroom quality and children’s school-readiness skills through building capacity and support for the existing, child-friendly curriculum.

Schools (j=240) were sampled from participating districts, stratified by public/private school status, and randomized into three treatment arms: 1) teacher training only; 2) teacher training plus parent awareness; and 3) business as usual. The teacher training consisted of 5-days in September 2015, 2-days in January 2016, 1-day in May 2016, and in-classroom coaching. Topics included: 1) how children learn/developing a child-friendly environment; 2) classroom management; 3) language/literacy; 4) math; and, 5) assessment and planning. Parent awareness consisted of three meetings with parents, administered by the Parent-Teacher Association, on the importance of play-based learning, parents’ roles, and the benefit of parent-teacher and parent-school communication. The final sample included 443 teachers and 3,435 students.

Data Collection: Teachers’ lives in context measures are drawn from teacher self-report/assessment, head teacher surveys, and classroom observations conducted prior to treatment (Fall 2015). Outcome measures come from classrooms observations and child assessements conducted in Spring 2016. Our main outcomes are classroom processes – assessed using the Teacher Instructional Practices and Processes System (TIPPS: Seidman et al., 2013, 2018) – and children’s learning – assessed using the International Development and Early Learning Assessment (IDELA: Pisani et al., 2015). We also examine two sub-domains of TIPPS and four sub-domains of IDELA.

Teachers’ lives in context (Authors, in press) is a theoretically-rich but empirically-sparse concept. As such, there are numerous possibilities for how to operationalize and measure teachers’ lives as well as which moderation model to test. We employ a methodological approach to reducing post-hoc measurement and modeling decisions that mirrors pre-registration, yet still allows opportunity to explore multiple, related hypotheses. We first create an a priori list of possible models based on theory and prior literature (Authors, in press). These include four ways of measuring teachers’ lives from more specific (e.g., moderate/severe depression) to more encompassing (e.g., a count measure of all poverty and health risks, including depression) plus all single moderation models that build upon these four base models for a total of 35 models for each outcome.

Data Analysis: We employ 2-level models (teachers within schools) to predict classroom processes and 3-level models (students within classrooms/teachers within schools) to predict children’s learning. After imputing missing data, we run each model separately over each of 100 imputed datasets, saving the model AIC after each run. We then examine only the AICs. Using the median and mean AICs, we determine the best fitting model for each outcome. We then re-run and interpret only these best fitting models using STATA’s mi estimate routine (which uses all 100 datasets). This approach reduces researcher bias/researcher degrees of freedom.
All models control for teacher baseline adherence, teacher age, district, private school status, and baseline score of the outcome. Models predicting children’s learning further control for child gender, child age, and kindergarten level.

Results: We find that having a secondary degree moderates the effect of teacher training only on classroom processes such that there is a program impact (Cohen’s d = 0.54, b = 0.19, p = 0.027) for teachers with higher degrees assigned to that treatment arm and not for other teachers or teachers assigned to teacher training plus parent awareness. There is no main effect of having a secondary degree. We also find that the teacher training only has a positive impact (d = 0.08, b = 0.02, p = 0.003, per additional barrier) on early literacy, a sub-domain of the IDELA, for teachers with more contextual barriers than average (µ = 3.65; SD = 1.70) and no impact for those with average or fewer. This effect is partly offset by a marginally significant negative coefficient on contextual barriers, suggesting that barriers may have a negative influence on children’s learning absent the intervention, but that the intervention helps students in classes with teachers facing a large number of barriers. There is no significant moderation of any other outcome. We do not find evidence that these moderation effects operate via increased adherence to training materials.

Conclusions: Teachers’ lives in context is an under-explored and vital topic within education. It is also likely to be of particular importance for pre-primary teachers working in under-resourced settings (Neuman et al., 2015). This study suggests that overlooking the impact of teachers’ lives may be limiting our understanding of why educational interventions succeed or fail. It also suggests that, even when the intervention aims to increase teacher skills and improve teaching quality, teacher conditions and experiences at baseline may undermine its effectiveness. We discuss implications for research and practice.

Authors

• Kate Schwartz, New York University Steinhardt

• Marc Scott, New York University

• Elise Cappella, New York University Steinhardt

• J. Lawrence Aber, New York University Steinhardt

• Sharon Wolf, University of Pennsylvania

• Jere R. Behrman, University of Pennsylvania