Wednesday, August 6, 2008

Teachers Have It Easy

While at the lake, I read Teachers Have It Easy: The Big Sacrifices and Small Salaries of America's Teachers. It's an excellent examination of one of the big problems facing education in this country, and it offers realistic concrete solutions which I always appreciate from "the current system isn't working" books. And it convinced me to change my mind about an issue I was dead set against: merit pay for teachers. I still think there are lots of merit pay issues to work out, but I believe that none of these are inherent in the idea of merit pay - they're sadly the ideas that are easiest for people to jump to. My biggest pet-peeve among these ides is following "logical" progression that causes many to suggest basing teacher pay on students' standardized test scores:
  1. Teachers should be paid based on how "good" they are.
  2. We should measure teacher effectiveness by how much their students learn.
  3. We should measure student learning through standardized tests.

Thankfully, most people would now not be surprised to hear an argument against 3 following logically from 2, so I don't need to make that argument. Less common is the understanding that 2 does not follow from 1 - a good teacher may still not cause students to learn. This faulty step comes from the idea of students as vessels into which the teacher pours their knowledge (the "banking" idea of education attacked by Paulo Freire), a notion which is now outdated. Instead, current educational theory recognizes that student learning requires work on the part of the student as well as the teacher. Student learning requires an atmosphere in which the student can concentrate and be stimulated. Student learning requires a previous knowledge base for the student to build on. And there are many other variables which affect student learning and take into account the reality that a student without enough to eat is less likely to learn than one who isn't distracted by hunger and other basic needs. None of these other dimensions of the problem have anything to do with the teacher (despite the teachers who go above and beyond to pay for food for their students out of their own meager salaries, knowing that their work is ineffective otherwise). Paying teachers based on much their students learn is assuming correlation without any evidence supporting it. (See Medley's 1982 article "Teacher Effectiveness" for more on this.)

Still, despite these issues, teachers can be evaluated in class by principals or other supervisors. It will require more and sustained work on the part of these administrators, so that teachers are not condemned by one bad day. But these steps are worth it. Teachers need to be paid more. The profession needs to be viewed as such to attract the best and create the best school systems. We're capitalists - it all comes back to pay.


Jack said...

I have no problem with (1) "good teaching" being evaluated by (2) "student learning" provided that "student learning" is tested 5 or more years later.

Dan said...

I don't disagree with the statement that "Teachers should be paid based on how 'good' they are.", but I also think that any method of assessing teachers raises a ton of hard questions: what is "good"---what are the goals of education?---and how does one assess how well teachers meet these goals? Who even has the authority to answer these questions (administrators, teachers, students, parents?)? I don't see how one could answer these questions in enough generality to create a national/statewide/schoolwide framework for merit pay---e.g., there are many different valid goals for education; teaching "well" for one student can be terrible for another (Do you get paid more for teaching to the top-third and losing the bottom or boring the top and helping the bottom?).

Supposing, hypothetically, that we had unusually talented and engaged administrators, teachers, students, and parents, I could picture something like this: at the beginning of a course, all four parties together work out goals specific to that course and the students in it. During and after a course, all of the parties involved provide feedback on how well they think those goals are being met. Then we just need some way of aggregating this feedback into a number of dollars of merit pay, balancing the different motives of the different parties (e.g., who knows better what the students learned, the students or the teacher?). This seems tricky but perhaps possible. On the other hand, this whole hypothetical requires an extremely unrealistic degree of earnestness, engagement, humility, and self-awareness to set the goals and get the feedback in the first place.

PS: yay, Freire! If you like race/gender/class/...-conscious philosophy of education, Lisa Delpit's book Other People's Children is a good read.

sorelle said...

dan - I agree, these are precisely the problems with any method based on this. But I do think that there are ways around defining "good" and its corresponding pay exactly. For example, the book describes school districts which seemed to have a binary satisfactory/ not satisfactory evaluation system, where satisfactory teachers received the entire bonus and non-satisfactory teachers received zero. The teachers' standing was determined by an interaction like the one you described (apparently, it's really happening!).

In addition, most of these systems also had bonuses for teaching "hard to staff" classes (SPED, ESL, math, etc), "hard" schools (i.e. not the magnet school), etc.

And yes, Delpit is awesome. (: I don't think I've ever read Other People's Children cover to cover... perhaps I'll add that to my list.

Dan said...

I probably should read this book before having a discussion about it. ;)

I don't have any problem with paying different amounts for different classes/different schools in the district. Assuming teachers have some choice about where/what they teach, that seems like a reasonable way to use a market to allocate people.

Re binary merit pay: I guess one could implement a system that gives the "correct" answer sufficiently often if the teacher is being graded pass/fail, as it were. It still makes me nervous, though. My nightmare scenario for this sort of thing is (a) the difficult hard-ass who makes people work, grades hard, but gets students thinking vs. (b) the empty shirt who grades easy and goes through the motions in a pleasant sort of way. In the latter case, the teacher is happy, the administration is happy (more students passed!), the (majority of) students are happy (less homework!), but no one learns much. In the former, the customer satisfaction scores may be lower, but it is still the better class. I'd worry that even a binary merit pay system would, over time, reward the latter over the former: everyone thinks the class went really well! Unless you have administrators who can spot this, or students who demand better---but how often do you have those?

sorelle said...

dan - my hope in that situation is that the administrators (who I think would be sitting in on the classes directly) would see the difference in the quality of the in-class instruction. I think a lot in this plan relies on having good administrators which is something I find myself naturally skeptical of, but which is definitely possible - so I've been suspending my disbelief on that point and assuming a responsible dedicated administrator with an understanding of teaching. Hopefully they also have oversight which ensures this.

Josh said...

Dan's last comment reminded me of an idea of Barry Nalebuff's (I don't have a reference for the idea, but I remember it came from him).

If you want to use grades as part of a way to evaluate teaching quality, the grades should come in context. For example, if a student got an A: does this student get A's in all her classes? Does she get A's in all classes she's taken in the same subject area (e.g. math)? On the flip side, how many students in the class got A's? How many students got A's when the class was taught by other teachers, or by the same teacher in previous terms?

Obviously all of these questions can be made into numerical statistics. These statistics *should* distinguish between the hard teacher that makes students think and the laid-back teacher that hands out A's like candy. On the flip side, if half a class get's Cs, but those students get Cs or lower in all their classes, then maybe it's not the teacher's fault (or at least no more that teacher's fault than any other teacher's).

These statistics might also pick up things like student improvement, particularly if these stats are kept for each assignment and examined over the period of a course.