Thursday, August 16, 2012

How to read graphs: The public policy version

I follow The New Teacher Project on ye olde Facebooks, and lately they've been on a rant/rave/something about what they call irreplaceable teachers. This morning, TNTP posted this statement and the following graph:


"For decades, principals have assumed that more years in the classroom mean better teaching. But while teacher experience helps, it doesn't guarantee excellence."



Here's what I think this graph really says: everyone improves over time, and partly because of that the lowest-performing teachers likely won't catch up with the highest performing ones.

Let's parse this, shall we?

From a scientific point of view, this graph makes two statements. First, the average low-performing teacher is worse at teaching than the average brand new teacher. What does that actually mean? It means that if you take the worst of the experienced teachers and compare them with all new teachers, the new teachers are better — the problem is, how did TNTP define "worst"?

If you dig deep into TNTP's report, you'll find out that in each of the four — yes, four — school districts they analyzed, they used a different methodology to define low-performing teachers, with the result that they classified between 10 and 24 percent of teachers as low performing.

Now look at the graph again. Despite making up at most a quarter of the teachers supposedly at the bottom of the pile, "Experienced Low Performers" are nonetheless doing better than about 35 percent of all teachers on average? And why is that? Without spending a lot more time on this than I'd like to, I can't be sure, but I'll bet it's because there are some new teachers that are really, really bad — but you wouldn't know that looking at this graph, because it compares bad but experienced teachers with all new teachers.

Takeaway part one: some experienced teachers are worse than the average new teacher, but likely there are many new teachers that are even worse or at least no better.

The second thing to take away from the graph is something TNTP sort of doesn't want you to pay attention to: over a three-year time period, both new teachers and low-performing experienced teachers improve. Why do I say TNTP sort of doesn't want you to pay attention to that? "For decades, principals have assumed that more years in the classroom mean better teaching." And, guess what, they're right.

(By the way, did you notice how the first claim above doesn't actually contradict the second? That's an elementary public relations, um, method.)

So, takeaway part two: everybody gets better with experience.

For me, the conclusion is this: there's a lot of variation in teacher quality, but everybody gets better with time. And, probably because everybody gets better with time, the low performers are going to continue to be low performers compared with everyone else — though, if you read the graph carefully, they won't continue to be quite so low performing as time goes on — in percentile terms, they actually improve a decent amount.

Now, I don't want to be too hard on TNTP, who are doing some interesting and maybe even useful work. What they're trying to say is not inaccurate, really — it really is true that there there are teachers that remain near the bottom of the barrel even with experience.

What I'd encourage, however, is that you think critically about what a graph like this means, and think about the methods — and the motivations — that someone put into this in order to try to persuade you of something.

Now back to work...at another education nonprofit that shall remain nameless.


EDIT (18 Oct 2013): Oh, it was Teach For America. Whatever.



5 comments:

  1. I think this is interesting. There is a lot going on with the testing mechanism and how one decides that the teachers are inferior. Do the children in their care fail to meet milestones (which I think is a fair way to evaluate)? Do other teachers do that better? Also, are the kids evenly distributed (meaning, do the bad teachers get stuck with the "bad" kids)?

    Theresa

    ReplyDelete
  2. All interesting questions — especially that last one. How teachers get assigned to students — location, grade level, subject, whatever — is an important issue when interpreting data like this. I also don't know how they defined percentiles, i.e., how were teachers ranked?

    ReplyDelete
  3. I question the premise that everybody gets better over time. Surely there are teachers who are out there who have been teaching the exact same thing for decades and have no aspiration to be better. I can think of a few teachers in my personal education who meet that description.

    Christine

    ReplyDelete
  4. Do you know what they are percentiling against? If the median new teacher is already at the 50th percentile or so, then either the rank is amongst only new teachers, or, as you say, there are some fantastically bad new teachers.

    But there's a deeper problem with this: it is almost certainly always true. To wit: teachers "added value" will fall on a distribution. New teachers also have some distribution. Unless these distributions do not overlap (i.e., all new teachers are worse than experienced teachers), the graph you posted will always be true, independent of any statement on teacher quality. It is simply a mathematical statement that the distributions overlap, and that teachers at the bottom of the pack do not improve fast enough to "catch" the best of the new cohort of teachers. You can always pick your groups so that will be true, even if all of the teachers were of excellent quality on an absolute scale. The only way it isn't necessarily true, I think, is if teachers improve faster and faster over time, which seems absurd in the long term.

    And to Theresa: yes, there's feedback. In my indirect experience, bad kids do frequently get shuffled off to "lesser" teachers. The "good" teachers have more pull, and some less than ethical types get those kids transferred out of their class, which also tends to make them look like they're a more effective teacher than they actually are.

    ReplyDelete
  5. Christine, I question that premise, too, though on average teachers seem to get better over time — the interesting question is whether there are a lot of teachers who don't grow at all, don't you think?

    Marcus — the scary part is that the bad and inexperienced teachers are assigned to the students who are performing at the lowest levels, and often those are the kids in the least wealthy schools.

    ReplyDelete