Reconceptualizing the factor structure of school readiness: A bifactor model

• In Event: Early Self-Regulation and Long-Term Outcomes: Why Does Childhood Self-Regulation Predict Adult Success?

Abstract

School readiness is often conceptualized as a multidimensional construct comprising distinct developmental domains. Two common models of pre-kindergarten skills are multidimensional first-order models (e.g., executive functioning and academic skills predict later outcomes) and hierarchical models (e.g., first-order factors feed into a second-order construct of school readiness). However, these models may contain bias or cause theoretical confusion if the common variance among various school readiness skills largely account for any observed correlation between a single skill and a given outcome.

This study examines a bifactor model of school readiness. In our bifactor model, a general factor of school readiness is fitted alongside specific pre-k skills factors, allowing us to examine whether specific factors are unique predictors of later outcomes or if later outcomes are better explained by a general factor of school readiness. This model can help inform whether longitudinal models that distinctly consider each skill may be confounded by factors that cause common variance across all school readiness skills.

CSRP was a cluster-RCT implemented in Head-Start centers serving high-poverty neighborhoods. This study restricts the sample to students with data on at least one of the pre-k skills measured (n=552). Approximately 64% identified as Black and 29% as Hispanic. The average income-to-needs ratio was 0.69 (SD=0.56).

During the spring of preschool, 11 measures were used to assess pre-academic skills (math task, Peabody Picture Vocabulary Task, letter-naming task), self-regulation (Preschool Self-Regulation Assessment (PSRA): snack delay, tongue task, toy wrap, pencil tap, balance beam), and behavior (Behavior Problems Index, PSRA assessor report of attention and impulsivity control). Composite SAT scores were generated using administrative records (n=385) from Chicago Public Schools and College-Board. College enrollment data were provided by the National-Student-Clearinghouse for eligible students (n=411).

All measures were standardized, and behavior problems were reverse coded to ensure all measures were positively scaled. We first tested a first-order factor structure with self-regulation, pre-academic skills, and behavioral functioning as specific pre-k factors (Model 1). Next, we fitted a bifactor model with a general school readiness factor drawing on all measures (Model 2). Finally, we regressed early adult outcomes (i.e., SAT scores and college enrollment) on the resulting factors from the bifactor model.

Results:
The Confirmatory Factor Analysis for Model 1 revealed poor model fit, with all three factors significantly correlated. Model 2 showed substantially improved fit, although three measures (toy wrap, balance beam, attention and impulsivity control) did not load. While the factor structure remained unchanged for Model 3 and 4, model fit dropped from great to acceptable.

As Table 1 shows, the general factor consistently predicted all three outcomes. With basic demographic controls included, the general school readiness factor significantly predicted SAT scores and college enrollment (though 4-year postsecondary enrollment was marginally significant). Surprisingly, none of the three domain-specific pre-k factors were independently and consistently predictive of adolescent outcomes. These findings suggest that (i) a bifactor model of school readiness may better represent the inter-relationships among early skills, and (ii) general preschool readiness appears to be more predictive of later achievement than specific developmental factors.

Authors

• Xinyu Pan, Columbia University
  Presenting Author

• Caroline Botvin, Columbia University

• Tyler Watts, Teachers College, Columbia University