Testing is a good idea, if it's done well. Measurement is always a good thing, unless it's expensive (tests aren't), or the measurements are used in a stupid way (which ... they are).
It's like measuring productivity by SLOC. Good programmers write more code. If you reward programmers for writing more code, though, they'll punch out boilerplate code which won't really do anything. They'll actually get less done, because they are writing meaningless lines of code, instead of actually solving problems. Steve Ballmer claims this a big reason why Microsoft beat IBM when OS/2 was coming out - as IBM rewarded their programmers for more SLOC, so they screwed everything up.
If you reward people for good measurements, they'll find ways to screw with the measurements.
Big tests mean, the students and teachers will focus on beating the test. They'll teach exam techniques, and stick to what they think will be on the test. It's fairly easy to teach a "shallow" way of solving problems, if you can guess the problems. Learning to solve unseen problems is hard - you actually have to understand the material. But most tests won't have a lot of difficult questions, so teachers will focus on getting their class to memorise the algorithm used to solve the basic questions.
Rote learning is good, if it's an enabler. Learning your times tables helps you do higher mathematics. But if you only worry about rote learning, then you can go through school without actually understanding what you are doing - you just follow the rules, and hope there aren't too many trick questions. Rote learning can be the foundation of higher learning, but creating nothing but foundations is a waste of effort.
It's also a huge waste of resources, as teachers will be spending all their time trying to figure out the test parameters, instead of worrying about whether their students actually understand the material.
Finally, there's the opportunity cost. Instead of saying "the good teachers are the ones which get good test results", they could do some deeper analysis. How do the "good" teachers control their class? What kind of homework do they set? What is the format of their lessons? What can the "bad" teachers do, to emulate their success? The "bad" teachers mostly just don't know how to teach well. There's probably a few idiots (who can't teach because they don't understand the topic), a few ones whose personality makes them ineffective, and some who are lazy or burnt out. But I bet most teachers would be much more effective if they knew what the "good" teachers were up to.
If the tests are high stakes, you'll just find that the "good" teachers are the ones who care most about gaming the test, not the ones who are actually good teachers. If the tests aren't high stakes, then the good teachers will naturally do well, and it's worth studying how they became so effective.
It would be justifiable, if the incentive of high stakes tests was strong enough to actually bring the kids up to a certain level, but I don't think it has. Instead, it just creates bad incentives, and is a distraction from serious quantitative research into teaching.
It's like measuring productivity by SLOC. Good programmers write more code. If you reward programmers for writing more code, though, they'll punch out boilerplate code which won't really do anything. They'll actually get less done, because they are writing meaningless lines of code, instead of actually solving problems. Steve Ballmer claims this a big reason why Microsoft beat IBM when OS/2 was coming out - as IBM rewarded their programmers for more SLOC, so they screwed everything up.
If you reward people for good measurements, they'll find ways to screw with the measurements.
Big tests mean, the students and teachers will focus on beating the test. They'll teach exam techniques, and stick to what they think will be on the test. It's fairly easy to teach a "shallow" way of solving problems, if you can guess the problems. Learning to solve unseen problems is hard - you actually have to understand the material. But most tests won't have a lot of difficult questions, so teachers will focus on getting their class to memorise the algorithm used to solve the basic questions.
Rote learning is good, if it's an enabler. Learning your times tables helps you do higher mathematics. But if you only worry about rote learning, then you can go through school without actually understanding what you are doing - you just follow the rules, and hope there aren't too many trick questions. Rote learning can be the foundation of higher learning, but creating nothing but foundations is a waste of effort.
It's also a huge waste of resources, as teachers will be spending all their time trying to figure out the test parameters, instead of worrying about whether their students actually understand the material.
Finally, there's the opportunity cost. Instead of saying "the good teachers are the ones which get good test results", they could do some deeper analysis. How do the "good" teachers control their class? What kind of homework do they set? What is the format of their lessons? What can the "bad" teachers do, to emulate their success? The "bad" teachers mostly just don't know how to teach well. There's probably a few idiots (who can't teach because they don't understand the topic), a few ones whose personality makes them ineffective, and some who are lazy or burnt out. But I bet most teachers would be much more effective if they knew what the "good" teachers were up to.
If the tests are high stakes, you'll just find that the "good" teachers are the ones who care most about gaming the test, not the ones who are actually good teachers. If the tests aren't high stakes, then the good teachers will naturally do well, and it's worth studying how they became so effective.
It would be justifiable, if the incentive of high stakes tests was strong enough to actually bring the kids up to a certain level, but I don't think it has. Instead, it just creates bad incentives, and is a distraction from serious quantitative research into teaching.