Is under-representation of Black, Latinx, English learners (EL), and students from economically challenging communities in gifted programs due to inequality in early academic achievement or bias in the gifted identification process? Using three-level multilevel logistic models, we examine the degree to which the disparities in gifted identification are due to disparities in early achievement. Our datasets include 3rd, 4th, and 5th grade achievement scores and demographic variables from all students in a 3rd grade 2011-2012 cohort across three states. Students who were free/reduced lunch, EL, Black, or Latinx were between 2 to 8 times less likely to be identified as gifted compared to non-free/reduced lunch, non-EL, and White or Asian students. However, between 50% and 100% of gifted identification disparities could be explained by student-level differences in early academic achievement, which is consistent with an opportunity gap explanation of underrepresentation. [This paper was published in "AERA Open" v9 n1 p1-25 Jan-Dec 2023.]