(WITHDRAWN) The Heterogenous Relations between School Support Staff and Student Reading Comprehension

Background: Nearly one-third of fourth grade students in the United States performed below basic achievement levels on the National Assessment of Education Progress in 2017 (McFarland et al., 2019). It is well known that child-level cognitive, behavioral, and social factors are related to reading achievement (Shonk & Cichetti, 2001; Pears et al., 2011). What has not yet received full attention is the extent to which the presence of support staff may uniquely contribute to our understanding of individual differences in reading outcomes. Authors adopted an ecological framework with a novel analytic approach that specifically models heterogeneous relations to examine how the presence of school support staff affected students’ reading achievement.

Methods: A sample of 5^th grade aggregated district-level reading comprehension (RC) scores (n = 37) in a southeastern state were used. Quantile regression allows for the estimation of coefficients for the relation between predictor(s) and the dependent variable along designated points along the conditional distribution of the dependent variable (Koenker & Hallock, 2001; Petscher, 2016). The strength of the relations of independent variables (reflecting the district-level presence of school nurses, school counselors and school social workers) to RC was estimated across 10 quantiles of RC. Quantile regression results were compared to OLS to understand the differences in estimates, conclusions, and implications of results given the choice of using a conditional means model (i.e., OLS) or a conditional tau model (i.e., quantile regression).

Results: Bivariate correlations showed weak relation between district-level factors and aggregated student RC (i.e., .16 < r < .26) with no statistically significant associations. OLS regressions of the associations of the presence of support staff with RC scores indicated no relation for school nurses, counselors, or school social workers. Quantile regression results indicated that availability of support staff were positively predictive of RC across all school districts. At the .90 and .95 quantiles of RC, the presence of school nurses was a significant predictor (β_.90 = 0.21; p < .01 and β_.95 = 0.25; p < .01), indicating that for districts with RC scores in the .90 and .95 quantiles, the presence of school nurses is a promotive factor for RC scores. OLS regression demonstrated no relation of school social workers RC scores. The quantile analysis indicated that for the .05 quantile of RC, school social workers within the district was a significant predictor (β = .37, p< 0.00), indicating that for districts with RC scores in the .05 quantile, the presence of school social workers was a promotive factor for RC scores.

Conclusions and Implications: Findings indicated that quantile regression analyses demonstrated nuances of district-level support staff factors associated with RC scores that OLS regression did not support. Availability of support staff, such as school social workers, was promotive of RC across quantiles, indicating that the presence of support staff matters for student learning. More importantly, the absence of school social workers is a risk factor for RC at the lower end of the distribution. Districts should consider policies that promote funding for additional support staff.