ON STATE EDUCATION STATISTICS

Open Access
Publisher Website
Google Scholar
Abstract
SUMMARY: State Education Statisticsis a table of information, aggregated to the state level. Its goal appears to be that of a thermometer to measure the health of the American educational system. As such, its usefulness is limited by three problems: Some outcome variables–ACT and SAT test scores–are produced by a highly self‐selected sample of individuals, so that the differences observed in outcome can be caused by differences in performance, differences in selection ratio and/or differences in selection rules. We were unable to disentangle these through statistical adjustments. Some input variables were calculated on the population as a whole (i.e., per capita income, average teachers' salary) rather than on the group characterized by the outcome variable (i.e., those individuals taking the SAT). These are often profoundly different and can cause substantial biases in the estimation of statistical measures of the interrelationships among the variables. Some input variables are inappropriately measured or scaled for the purposes intended. For example, teachers' salaries are given in current rather than constant dollars. Thus, one might conclude that over the ten years considered teachers' salaries went up 81%, rather than the more correct conclusion that, in constant dollars, they declined 20%. Among the recommendations proposed to solve these problems is an expansion of the National Assessment of Educational Progress (NAEP) to allow accurate state estimates of students performance, as well as expanding the range of areas covered to more fully measure the section of the students' population intending postsecondary education.