James Surowiecki argues in the New Yorker that improvements over the last 30 years in professional sports and in such diverse arenas as chess and professional music are due largely to a great increase in the intensity of training regimes. Training is not only more intense, it’s now very focused on skills specific to the capabilities needed for success: “if you’re a baseball player, you work on rotational power; if you’re a sprinter, on straight-line explosive power.” Pro athletes not only get intense, early training, they keep on training throughout their careers.
Surowiecki’s asks why we don’t apply the same game plan to teaching.
we do little to help ordinary teachers become good and good teachers become great. What we need to embrace instead is the idea of teaching as a set of skills that can be taught and learned and constantly improved on. … More accountability and higher pay for teachers would help, too. But at the moment most American schools basically throw teachers in at the deep end of the pool and hope that they will be able not only to swim but also to keep all their students afloat, too. … To make gains, schools should take advantage of the training techniques that other countries have mastered: record classes so that teachers can study their own work and that of colleagues; let teachers observe each other; measure performance; and deploy a staff of full-time trainers.
Q-Comp, Minnesota’s large-scale experiment in teacher pay-for-performance, is evaluated in a new paper by Aaron Sojourner, Elton Mykerezi, and Kristine West. Bottom line seems to be that the program works.
The authors show that pay-for-performance led to gains in test scores. What’s more, by looking at both high-stakes and low-stakes tests they show that the gains are real–as opposed to coming from “teaching to the test.”
While the gains are quite real, they are also quite small. Student gains due to the program are about 0.03 standard deviations. While the gains are small, the costs of the program were also small. The authors estimate that the gains in terms of future productivity from better education are on the order of five times the costs.
The authors raise the background question of how much of gains are due to performance incentives and how much are due to changing who becomes a teacher and who stays a teacher. There is at least some evidence that these gains are mostly due to the former. The authors write,
At least with respect to changes to the pay schedule of this magnitude and this time span, P4P [pay-for-performance] did not induce important changes in the observable composition of the teacher workforce.
I suspect the authors are correct in concluding that the changes were not due to bringing in different teachers. The available bonuses under Q-Comp are minimal compared to the financial gains from choosing another career over teaching. Here’s a table using the authors’ numbers on the average bonuses available to teachers.
|school or district-wide goals (usually based on test scores)
|student achievement other than test scores
These prizes just aren’t big enough to have much effect on career choices.
If we want to prevent crime, should we invest in getting kids through high school or should we spend the money keeping the bad guys in jail longer? Giovanni Gallipoli and Giulio Fella offer a good argument that it’s better to invest in education.
The authors make a back of the envelope calculation which begins with estimates of the effect of just plain out subsidizing high school students to stick in school. (Turns out there have been some experiments which show subsidies work.) Then they use what we know about how much less likely high school graduates are to end up in prison. (One interesting number is that in 1997 70 percent of inmates had failed to graduate high school.)
Once they have a relation between dollars spent on education and the reduction in crime, the authors ask how much you would have to spend to get an equal effect by keeping an inmate in jail–and off the streets–for a longer period, noting that an inmate-prison-year costs something just under $30,000.
(I’m oversimplifying here…”back-of-the-envelope” in this kind of work is a complicated dynamic mathematical model that can only be solved by a computer.)
The answer is we’re much better off subsidizing high school graduation if that’s what it takes to get more kids to earn the diploma. Not only is education route more cost-effective, it also gives the graduates more skills so that they are more economically productive.
Maria Donovan Fitzpatrick calculates that teachers are willing to pay only 20 cents to buy an extra dollar’s worth of pension benefits. The obvious implication is that we ought to take some of the money now put into pension benefits and use it to raise salaries. I suspect this is a good idea, but some caveats are in order.
First, how did Fitzpatrick go about the research? In 1998, many Illinois teachers were offered a chance to pay to “upgrade” their future retirement benefits. How much the upgrade cost and how much it was worth varied from one teacher to the next. Fitzpatrick worked through all the intricacies and then statistically estimated that a 20 cent increase in price would have a just offsetting effect of a one dollar increase in benefits.
Just to be clear about this. It appears that teachers are willing to pay only 20 cents for a future benefit that is worth one dollar in present value. In other words, they’re leaving money on the table.
Fitzpatrick also gives us another relevant number. Employer pension contributions per hour worked are almost three times as much for public school teachers as for private sector professionals. That’s another hint that relatively too much goes into teacher pensions compared to salaries.
Now the caveats. Fitzpatrick reminds us that other studies where employees are given a choice between money now and pension later also reach the conclusion that they pick “money now” even though this “leaves money on the table.” One of the clearest findings of behavioral economics is that people don’t make rational decisions about providing for the future.
One more piece that worries me. While Fitzpatrick’s finding certainly point toward switching some of compensation from pension to salary, I can just see some politician switching money from pension and forgetting about the “to salary” part.
Today, I want to do some simple arithmetic which suggests we could be paying classroom teachers a heck of a lot more if we weren’t having to pay all sorts of extra teachers as “teacher coaches” of various sorts. I am not knocking the contribution of the extra teachers. The idea here is that if we paid all teachers top salaries we wouldn’t need to supplement the work of classroom teachers nearly so much. Lot’s of teachers with non-classroom assignments would return to the classroom…with a big raise.
The basic arithmetic is as follows. Right now there are about 18 students per teacher (adjusting for special ed). Elementary room classes average about 21 students and secondary schools average 27. Let’s call the overall average 23. If we assigned every teacher a regular sized classroom, then every teacher could get a 22 percent raise without the taxpayers spending another penny.
Is that going too far? Probably a little, as we probably should have a few teachers supplementing the work of regular classroom teachers.
But I can’t help that think that students, teachers, and everyone would be better off with high paid teachers running classes and fewer teachers on non-class assignments.
Here’s another picture of class sizes (for the most recent data) versus the average number of students per teacher. Note the red lines are well above the blue line.
There are about 16 students for each teacher, although class sizes are much larger. What are the “extra” teachers doing? The explanation that leaps to mind is that they’re engaged in special ed. If so, we might think that a pretty good use of “extra” personnel.
Turns out this isn’t the case. Special ed teachers are about 13 percent of the teacher corps. If you assume that special ed teachers pull out 7 students each (that’s the average pupil/teacher ratio in special ed schools), then the pupil-teacher ratio for non-special-ed students to non-special ed teachers is 17.3. If we assume that special-ed teachers pull out zero students, then the ratio is 18.3.
In other words, no matter how we account for special-ed, the pupil-teacher ratio is a lot lower than typical class size.
This week I present a short series on the suggestion that if we required all teachers to handle a classroom of students, we could pay teachers a heck of a lot more. The basic point is that we use huge numbers of teachers labelled “teacher coaches,” “subject specialists,” and similar titles. These teachers, who one suspects are often particularly good, “help out” regular classroom teachers. There’s nothing wrong with this in principle. In practice, it uses up a huge fraction of the budget.
To get started, here’s a little historical comparison of “class size” versus the “pupil-teacher ratio.” Class size measures the number of students in a typical class. Pupil-teacher ratio is simply the number of students divided by the number of teachers. You might think the two would come to about the same thing. They don’t, because many teachers are coaches, etc., who don’t have class rooms.
Secondary school class sizes have been steady to rising. Elementary class sizes have been falling. Pupil-to-teacher ratios have been falling steadily. (All three took something of an uptick following the Great Recession.)
The important question to think about is why is the green line so much lower than the blue and red lines? If there are about 16 students for each teacher, how come each classroom teacher has 21 (elementary) or 26 (secondary) students. What are all the “extra” teachers doing? Might it be better to have a lot fewer “extra” teachers and pay classroom teachers a lot more?
I’ve been reporting this week on The Equity Project’s (TEP’s) success at achieving remarkable student outcomes by massively raising teacher salaries while making other cuts to remain budget-neutral. Even great success does not mean that everything went perfectly.
Perhaps the most surprising issue TEP has had is incredible teacher turnover. In the typical New York city middle school about 27 percent of teachers do not return for a second year. At TEP–and note that TEP pretty rarely hires novice teachers–the number is 44 percent. I’ve made a picture, which isn’t pretty–the bars are just too, too high.
A second issue that TEP has had is that it seems to take several years in the program for students to really benefit. In math, there is an immediate, large gain (relative to comparable students). But the gain after four years is quadruple the gain after two years. And after a single year, there doesn’t seem to be much gain in language arts, although the gain is substantial by the four-year point. Here’s Mathematica’s chart:
Except in math, the early returns aren’t great–even though the four-year returns are excellent.
TEP also underwent growing pains. In the first two years of operation, TEP students actually lost ground compared to comparison groups. TEP made adjustments which led to big gains for these students in later years. This suggests that when other schools emulate TEP, they’d best be prepared for some stumbling during start-up.
Two more points, not about TEP but about scaling up the TEP model to widespread deployment.
- TEP raised salaries above the competition. To some extent this will have gotten TEP first pick at the best teachers. When you scale up, everyone can’t have first pick. This could be the bane of wide-scale emulation.
- When one school raises salaries, there’s not much effect on who becomes a teacher. If the high-salary TEP model becomes widespread, the pool of top-notch teachers will expand. This could be a great assist in wide-scale emulation. (But this teacher pool upgrade will take a few years to roll out.)
On net, TEP has scored a major victory for children’s education. Nonetheless, those who plan to emulate should understand that there are going to be bumps in following TEP’s path.
Monday I reported that The Equity Project’s plan to pay teachers $125,000 a year has produced remarkable student gains. Today I’ll combine information from Mathematica’s report with other data to give you an idea how TEP manages high salaries without busting the budget.
First thing–comparable teacher salaries in New York outside TEP are around $75,000. (Salaries are high in New York. Of course, New York is an expensive place to live.) So TEP teachers earn roughly two-thirds more than their counterparts.
Here are the cost-saving biggies:
- TEP classes are 15 percent larger than classes in typical middle schools in New York, about 31 students versus 27. This means TEP hires fewer teachers per student.
- Teachers work longer hours than elsewhere. TEP teachers take care of most of the administrative duties in the school. The assistant principal is a teacher wearing a second hat. Other teachers wear second hats as specialists in math or language or special ed coordinator. Attendance, lunch, and parent involvement are all handled by teachers. Handling extended-day activities is part of regular teaching duties.
- While TEP contributes to insurance coverage and a retirement plan, benefits are a notably lower fraction of compensation than would be seen in most schools. TEP benefits cost in the range of 8 to 13 of salary. In the typical school, benefits are 35 percent of salaries. (More like 42 percent in New York state.)
TEP uses a few other tweaks. Teachers aren’t paid extra for professional development activities (most of which are done in-house). Educational consultants aren’t hired. Teachers who mentor other teachers also teach full loads.
TEP does not fund raise to support operating expenses, although they do have a capital campaign in hopes of getting a building to get themselves out of portable classrooms. Contributions and private grants amount to $75 per student.
In sum, TEP has produced a financially scalable model.