Standardized testing has nothing to do with science. There are studies in teaching and learning that do attempt to gauge the effectiveness of different teaching approaches, etc, and they typically use randomized trials and carefully controlled situations to be able to measure the effect.
But more crucially, the subjects (both teachers and students) in those studies are not rewarded based on the outcome of the study. Even if there is a correlation between the test and what we want schools to achieve (whatever that is), making teachers jobs contingent on that outcome will select for teachers who are good at achieving those results, which increasingly will diverge from your desired outcome. It's an example of dangers of "optimization by proxy" (http://lesswrong.com/lw/28r/is_google_paperclipping_the_web_...).
But more crucially, the subjects (both teachers and students) in those studies are not rewarded based on the outcome of the study. Even if there is a correlation between the test and what we want schools to achieve (whatever that is), making teachers jobs contingent on that outcome will select for teachers who are good at achieving those results, which increasingly will diverge from your desired outcome. It's an example of dangers of "optimization by proxy" (http://lesswrong.com/lw/28r/is_google_paperclipping_the_web_...).