Many studies highlight the importance of high quality classrooms for children enrolled in early education programs (e.g. Zaslow et al., 2010), yet measuring quality and defining high quality can be challenging.  The Head Start Impact Study (HSIS), a large randomized evaluation of the Head Start program, measured preschool classroom quality using the Early Childhood Environment Rating Scale-Revised (ECERS-R; Harms, Clifford, & Cryer, 1998) and the Arnett Caregiver Interaction Scale (CIS; Arnett, 1989).  However, as the early childhood education field transitions to newer measures of classroom quality, the question remains: How can we best utilize valuable data collected using these early tools to precisely measure meaningful dimensions of classroom quality? To do so, the current study seeks to identify dimensions of preschool classroom quality that can be measured by combining items from the ECERS-R and CIS tools; we validate these dimensions by comparing the impacts of Head Start on these new and original measures of classroom quality. 

To address our study questions, we utilize a number of approaches to data reduction: First, we conduct exploratory and confirmatory factor analyses on the combined ECERS-R and CIS items using data from 761 classrooms in the HSIS.  Based on those results, we conduct a bifactor analysis to pull out the shared variance of a general classroom quality factor and isolate residual quality domain-specific factors. 

Results from the exploratory factor analysis, using Geomin rotation and maximum likelihood estimation (to account for missing data), indicate three domain-specific quality factors: Materials and Space for Learning, Positive Teacher Interactions, and Negative Teacher Interactions.  The confirmatory factor analysis supports this three factor solution.  Sensitivity analyses revealed that the confirmatory three-factor model and model fit were robust across model specifications.  The bifactor analyses yielded similar domain-specific factors and a general classroom quality factor common to all items. Model fit statistics support this three domain-specific factor solution.

We then estimated the impacts of random assignment to Head Start on the original and domain-specific measures of preschool classroom quality (n=4,440 children).  Because random assignment to Head Start significantly impacts the probability that a child is missing observed quality data, we used Tobit models to estimate the “true” impact of random assignment to Head Start on classroom quality as distinct from the impact on having observed quality data.   Preliminary results yielded significant impacts of random assignment to Head Start on the quality of children’s preschool classroom as measured by the ECERS-R, CIS, and all three domain-specific factors. 

Impacts were largest on Positive Teacher Interactions, suggesting that combining ECERS-R and CIS affords a more sensitive measure of quality.  Positive interactions with teachers are one of the most important early experiences for children in preschool classrooms, and are strongly associated with numerous positive child outcomes (Burchinal et al., 2010). Thus, isolating impacts on Positive Teacher Interactions represents a significant step forward in utilizing rich data collected with older tools to answer pressing current and future policy questions regarding providing high quality effective early education experiences for young children.