My student took the SAT yesterday, and I read on the College Board site that the March 14 SAT was canceled in many countries along with many U.S. test centers being closed. I’m wondering how a drastically smaller testing group, without many international students and a number of domestic students, will affect the scoring of this test. Does anyone know anything about this?
Good question! And wouldn’t the internationals be more likely to do well on the math, especially.
I read that 122 test sites were cancelled including our school. It makes sense though for our school since our school was closed from Friday to who knows when
My reading indicates that the curve isn’t determined by the current set of test takers. Instead, it apparently is predetermined based on an assessment of the difficulty of questions.
That’s not correct. While CB attempts to make a test of equal difficult to all historic tests, this is obviously impossible. Based on what is typically a very large number of test takers that have taken previous test(s), they statistically determine the difficulty of this test and adjust the scoring mechanism to make equal performance result in equal scores across tests. The process is called equating.
The situation above is a good example. If only a set of students that had scored 700 previously took the test, then theoretically they all would score the same - the test average. They do not all receive a 500, which CB theoretically defines as the “average” of the SAT. Assuming they did as well as they did before, they would all receive 700 (in reality, they would probably be distributed above and below about 710-720 to account for inter-test improvement and random performance variations).
The concern with scoring this test would be the number of repeat students with which they can equate the test to previous tests with statistical validity. In my estimation, not based on factual data, this number is likely much lower than the number they will have, unless a vast majority of tests were cancelled - I have no idea how high the number was. CB has been doing equating for decades, when test populations were much lower than they are now. Statistics can converge on a fairly tight confidence interval with fairly small numbers. Note that most political polls have populations of 500-2000 and margins of error in low single digits. I suspect there will still be tens of thousand of test results, but CB will need to determine if it meets their statistical thresholds.
(Fwiw, this is why complaints of a “harsh curve” are also incorrect)