The national main NAEP assessment samples were selected based on a two-stage sample design:
For fourth and eighth grades, the base public school samples were the state assessment's public school samples for each jurisdiction; likewise, the state student samples and assessment became the national student samples and assessment. This was the first year in which the design integrated the state NAEP sample into the national sample directly rather than drawing a separate sample of schools and students for the two surveys.
The state samples only cover public schools, and only the fourth and eighth grades. Thus, there was a need to supplement the state samples with a nationally representative private school sample for fourth, eighth, and twelfth grades, and a nationally representative public school sample for twelfth grade (for the 50 states and District of Columbia). In addition, there were a number of jurisdictions which decided not to participate in the state program:
For the national samples for fourth and eighth grades, subsamples were drawn from the state school samples in these jurisdictions, which were used to represent them in the national sample. These schools were recruited as part of the national sample. The school sample sizes for these jurisdictions were drawn so that the targets for student samples for each of these jurisdictions for the national sample were proportional to the number of eligible students in the nation who resided within the jurisdiction.
For the twelfth-grade public school sample, schools in a high-minority stratum (schools with 15 percent or more Black and Hispanic students) had a relative measure of size twice as large as those in low-minority schools (the complement set), to implement oversampling of minorities in the twelfth-grade public school sample, as has been done in previous national main NAEP assessments. The measure of size was also conditionally altered to minimize overlap with a sample of 400 grade 10 schools already selected for the 2002 Educational Longitudinal Study (ELS) to minimize burden engendered from being part of both surveys.
From the stratified frame of twelfth-grade public schools, a systematic random sample of grade-eligible schools was drawn with probability proportional to a measure of size based on the estimated grade-specific enrollment of the school (with doubled measure of size for high-minority schools, and altered to minimize overlap with the ELS).
Each selected school in the twelfth-grade public school sample provided a list of eligible enrolled students, from which a systematic sample of students was drawn. A total of 136 students were sampled, if possible, from each school. Minority students were oversampled in schools in the low-minority stratum.
The private school samples for each grade were explicitly stratified into five strata and implicitly stratified within the explicit strata (by ordering the schools within the explicit stratum by selected school characteristics). There was no oversampling or minimization control. For private schools, 60 students, if possible, were selected with no oversampling of minorities.