Because standardized tests can be applied to such a large set, the biases of one test maker can be spread to many test takers. Contrast this with individual ass

Because standardized tests can be applied to such a large set, the biases of one test maker can be spread to many test takers. Contrast this with individual assessment where the biases of many test "makers" have an opportunity to cancel each other out. This is similar to the logic behind randomized assignment in randomized-control experiments. It's not that the mice are the same, it's that asymptotically their differences cancel each other out. I'd recommend this article: http://stat.wharton.upenn.edu/~rosenbap/heteroReprint.pdf http://stat.wharton.upenn.edu/~rosenbap/heteroReprint.pdf