Or oversimplifying.
You dont want teachers to just give their own tests. I know teachers who teach a lot and teachers who teach little. A "C" in the first teacher's class means that a student learned more than an A student in the second teacher's class. This is inequitable.
Team-written tests can be manipulated. One reason for team tests is to make sure all classes teach the same content. "We didn't cover that in my class" means it's not tested. Everybody gets an A and the test is emasculated. The guilty are let off the hook, and it's hard to maintain standards if having low standards is rewarded.
District written tests can be divorced from the classroom. But keep districts on the same page. Problem is, a low-SES high-school held to the same standard as a high-SES classroom will always be failing. You can't catch up 3 grade levels in reading in 6 months without a lot of intensive, individually-tailored one-on-one work that a classroom teacher can't manage. There has to be some acknowledgement of this in how the tests are administered, otherwise all that happens is that teachers are pressured to teach the test. (Then you get a conversation I had recently: "Last year's _____ test at this time focused on economic ties and influence. Our kids didn't do especially well, so that's really what we focused on. This year--not a word about it. Military expansion and conquest was 80% of the test, and we barely mentioned it." That's a crap shoot.)
Same for state tests.
But at least state and district tests aren't manipulated to make teachers look good or make students look smart. And that's why they were instituted--to get past self-serving tests.
Ravitch is wrong in some ways. Our district and state tests are "leveled and TEKed", fancy edu-Tex-speak for "ranked by cognitive difficulty and explicitly tied to one content skill and often to one content and one process skill." So I can look at district and state test results and say, "Ah, 38% of my kids don't know the kinetic theory of gases, and the number jumps to 59% when called upon to evaluate numerical data in a table."
That the tests are divorced from the classroom makes some of the data meaningless. District tests are stilled known quantities, at least after the fact. Some data we dump. "Didn't teach that. And that question is gibberish." Some we keep. "How could my students both show that 80% know this standard and 75% have no clue?" I look at the questions and find that 80% know it when it's a simple question, but 75% get the calculation using scientific notation wrong.
Where she's stellar on this is not using the valid data in useful ways. Yeah, I could teach scientific method; it comes up over and over in my class. But if they missed a bit of content there's no way to scrape together enough time to reteach it. Same for state tests--we get the results during finals week. And by next fall, when they retest if they failed this spring, they'll have forgotten a lot of the content so the "diagnostic" value of the test isn't so great.
= new reply since forum marked as read