The purposes of this study were (1) to examine the direct and indirect effect of school-level testing policies on reading achievement though changes in amount and types of reading instruction, (2) to investigate the reading trajectories moderated by school-level testing policies longitudinally, and (3) to examine the relationship between testing policies and the achievement gap by exploring whether certain student characteristics moderate the relationship between testing policy and reading achievement, using Early Childhood Longitudinal Study Kindergarten (ECLS-K) Cohort of 2010-2011 data. Findings from a multilevel full structural mediation model suggest that school-level frequency of state/local standardized tests had an indirect effect on student reading achievement through changes in both amount and the types of instruction at the school-level (cross-sectional fall kindergarten sample =12,241 children nested in 1,067 kindergarten classes). The findings from a three-level growth models indicated only children of Asian background and children from high socio-economic backgrounds who had frequent standardized tests in kindergarten accelerated in their monthly reading growth, whereas other children (e.g., low SES, non-Asian children) did not show any changes in the rate of the reading growth (longitudinal sample from fall of kindergarten to spring of first grade = 7,392 children nested in 744 kindergartens). The findings from the current study suggest that testing policy is not an effective means to reduce the achievement gap of children from disadvantaged family backgrounds, underperforming children or that children from low socieo-economic backgrounds. These children did not seem to benefit from frequent standardized tests longitudinally. Implications for supporting school assessment practices and instruction are discussed.