NAEP Technical DocumentationStratification by Achievement Data and Median Income for the 2017 State Assessment

			Jurisdictions Using Achievement Data or Median Household Income in Stratification

The achievement data obtained from each jurisdiction are derived from the results of state assessment programs. The contents of the achievement data files varied by jurisdiction and included achievement measures for a variety of subjects, grades, and multiple assessment programs. One achievement measure was selected for each responding jurisdiction to be used in the stratification process. Where available, the achievement data were used for implicit stratification by grade. Since the achievement data are more current than the median household income data, as well as more likely to be well-correlated to NAEP assessment scores, they were judged to be a more effective stratification variable. The achievement measures were selected according to the following criteria:

• At both grades 4 and 8, achievement measures from state assessments conducted in mathematics and reading were under consideration. If both were available, the mathematics measure was preferred. As a rule, the most current measures available were used. For California, the measures were from the 2014-2015 state assessment. For all remaining jurisdictions, the source was from the 2012-2013 state assessments. The sole exception was Shelby County School District which recently had its boundaries redrawn. Changes to the NCES school IDs hampered the process of matching current score data to schools. As a fallback, data from the 2010-2011 state assessments were used.

• Achievement measures should match to at least 70 percent of the schools on the sampling frames.

• Achievement measures should differentiate schools from one another. For example, district-level measures or those with high missing rates (30 percent or more), were judged not to be useful for differentiating schools. In addition, achievement measures that did not have large enough dispersion, based on inspection, were not used for stratification either.

• All other things being equal, the possibilities for score types were average scale score, median scale score, percentile rank, median percentile rank, normal curve equivalent, raw score, index score, and percentage above a particular cut point or quartile. In general, the availability varied for any given state/grade/subject/year.

Achievement data used for implicit stratification were obtained for all 50 states and the District of Columbia for both fourth- and eighth-grade assessments. In Puerto Rico where achievement data were not available, median household income was used based on the zip code area in which the school was located. The source of median household income was the 2010-2014 American Community Survey (ACS). The estimated grade enrollment was used for the stratification for DoDEA and BIE schools, since neither achievement data nor median income were available. Estimated grade enrollment was obtained from the Common Core of Data (CCD) file developed by NCES. 

Printer-friendly Version