If the student evaluations correlate strongly with the standardized tests' measures of student progress, then:
a) What's the purpose of having both?
b) If the evals are preferred over the tests, will "good" teachers continue to teach to a predictable, standardized curriculum?
c) Is the correlation additional evidence in favor of "differential compensation", that is, a compensation program based at least in part on exam scores?
d) Even if the information supplied is similar, doesn't this extra test/survey administration detract from instructional time? Is the information gleaned sufficient to compensate for the loss of instructional time?
e) Atlanta (Georgia, USA) is still reeling from a years-long cheating scandal. If such evaluations become "high stakes" (and there will likely be a push to do so, despite likely union opposition), won't these results be exploitable as well? (And perhaps even more so, through campaigning, social engineering, etc?)
What's the purpose of looking at multiple polls when you're trying to predict the outcome of the upcoming election? More evidence gives you higher confidence and lower margin of error. And as the article says, these student surveys provide clean, stable data that doesn't fluctuate very much from year to year and doesn't require much correction for race and family income.
These surveys take on the order of 10-15 minutes. That's nothing compared to a battery of standardized tests. They wouldn't have to be very informative at all in order to be worth the small sacrifice of instructional time, and if they're the second-best predictor of class achievement, then they're certainly worth the time (if the results are actually used).
I don't think there's going to be much movement to stop paying attention to standardized tests and curriculums, since these surveys don't measure the same thing - roughly speaking, the standardized tests seek to measure how much was learned, and these surveys add a dimension of why.
The point being, if they're strongly correlated, the "why" can already be inferred (assuming adoption of this particular theoretical causal chain).
I agree the movement toward testing is likely to continue unabated. The point was the futility of additional questions; when a strong correlation is already known, the responses can be predicted with a high degree of accuracy, so the additional information is not really "additional". Yes, social scientist prefer multi-item measures and avoiding a single source bias (particularly when publishing theory), but in the end, if the correlation is maintained, not much new is really learned.
(And one of the reasons for multiple election polls is that the results change over time, and leading up to an election, that's relevant. Additionally, the entire electorate isn't polled each time, so the "cost" to the system is lower, relatively speaking.)
Let's make some simplifying assumptions: the survey identifies two causes of poor performance, a lack of academic rigor, and a poor classroom environment (such as a teacher that's mean and unresponsive to requests for help or clarification). Those two categories are weighted equally on the survey - a teacher who gets 100% on the survey is good in both categories, a teacher who gets 0% is bad in both categories, and a teacher has more than one way to score 50%.
So, if both of those factors affect student performance on standardized tests, then there will be a strong correlation between the overall survey scores and the test scores. But analyzing the details of the survey results can offer actionable guidance that the test scores can't - the survey does provide useful information for how mediocre teachers can improve, even when it matches the test scores in predictive power for future test scores.
a) What's the purpose of having both?
b) If the evals are preferred over the tests, will "good" teachers continue to teach to a predictable, standardized curriculum?
c) Is the correlation additional evidence in favor of "differential compensation", that is, a compensation program based at least in part on exam scores?
d) Even if the information supplied is similar, doesn't this extra test/survey administration detract from instructional time? Is the information gleaned sufficient to compensate for the loss of instructional time?
e) Atlanta (Georgia, USA) is still reeling from a years-long cheating scandal. If such evaluations become "high stakes" (and there will likely be a push to do so, despite likely union opposition), won't these results be exploitable as well? (And perhaps even more so, through campaigning, social engineering, etc?)