Hacker News Reader

Comment by WalterBright | Hacker News Reader

There is a simple solution, but nobody will like it.

Give the teachers a base pay. Then, they get a percentage of a bonus amount. The percentage is calculated based on the percentage of the students in the teacher's class meeting grade level standard at the end of the year. The teachers will figure out how to teach them. Money, especially if the bonus could be 20-30000, is a big motivator, but those bonuses are still cheaper than the failed programs usually tried.

Q: what if a teacher gets assigned a poorly performing cadre of students?

A: assign the kids randomly to teachers. The ups and downs will average out.

Q: what if the teacher cheats on the year end tests?

A: have an independent authority do the testing

Q: what if the teacher teaches to the test rather than anything valuable?

A: change the test so that passing it reflects learning

View on HN

Check out "The Tyranny of Metrics" book for why this won't work.

Weird to say it's simple.

I’m sure I already know the answers from experience after becoming obsessed with KPIs as the solution to startup problems to seeing how companies just game vanity metrics. But could you provide some examples of why it doesn’t work in a school environment?

My best guess: It is possible to weed out non-productive KPIs, they are still extremely useful for breaking down reality distortion fields, but it takes an honest and thoughtful person with a good grasp of both the business and how the stats are collected - combined with management buy in. I have a feeling highly bureaucratic gov organizations won’t have either the talent, iterative approach, nor incentives to actually find the right metrics that matter.

It's too variable. A teacher has a variety of classes with a variety of children that changes completely every year. The outliers are enough to impact the metrics significantly but leaving them out paints an unclear picture. As well you have additional high variability of an individual students performance at any given time. This variability is greatest in the non-outliers which matter the most.

I answered that in the OP.

Metrics work when kids are prepared to learn in school, but many are not. Black boys in our worst neighborhoods have a >5% chance of being shot at least once by age 21. In the worst cities, 1/200 Black men and boys are shot every year. If you consider only those aged 13-55 in the worst areas it's more like 1/100 -- per year. This is not a simple environment to educate in. Kids don't have a safe place to play. Kids are in gangs. Kids are obese. Kids don't go to the dentist. Kids fight. Kids have high lead levels, asthma, and various trauma.

I don't follow your point. Are you saying we shouldn't commit to education metrics because a subset of the population aren't capable of fully engaging with education system? What about the kids who are engaged in education? Should they have to put their systems on hold or change it as a way to help(?) those who can't fully participate?

Those seem like two different problems with two different solutions. For ex: solving the environmental issues for the smaller subset and better correlating grades with actual outcomes in education for everyone. Especially if it risks masking problems and may ultimately harm more than it helps.

Oh I think we are referring to two different things. One poster suggested that performance-based pay for teachers is a simple solution and that it could be implemented with metrics. I countered because that is not simple. In addition, many teachers are doing support roles, library, phys ed, social work, reading specialist, etc.

OTOH metrics and objective standards are definitely worthwhile indicators. Agreed. In fact, in the modern age of data I think we ought to look at more things to see if there is a latent pattern positive or negative. Reading and math scores are not enough. I want to see physical fitness data. I want to know how many words per minute kids can type. How many dental cavities do they have. Can they pass a civics quiz on state or federal government. Do they know basic geography and have some literacy about the world.

Make it a sliding scale, performance incentives in the tougher inner city schools should be greater than a easy suburban school.

Yep, or target the incentives based on the composition of the situation on the ground. If your classroom is mixed with 1/3 IEPs for autism and other issues you have different achievable goals.

But this would affect any method. So its not really relevant when deciding a method right?

Not sure what you mean. You don't want disincentives for teachers -- it's already an issue where good teachers trying to improve rough schools get associated with poor performance through no fault of their own.

Incentives need to be based on obtainable goals and targeted improvements. Payouts (bribes!) to students for good grades and scores actually really work. Money helps kids internalize the societal value placed on their performance. Similarly, targeted bonuses for teachers work too, but only if formulated relative to the characteristics of enrolled students. See the literature by R. Fryer, for example

Math tests are hard to game.

The current metric is gamed pretty thoroughly, wouldn't you say?

Besides, businesses operate on metrics every day. It's the bottom line. It works. You just have to pick metrics that are the goal of the organization.

> Q: what if a teacher gets assigned a poorly performing cadre of students?

> A: assign the kids randomly to teachers. The ups and downs will average out.

This punishes students who can't get streamed to classes that challenge them at a degree appropriate for their aptitude.

The current system punishes them much worse.

Even simpler, but less likely to be popular. Pay teachers competitive (six figure) salaries, put a cap on student:teacher ratio (20:1, or even less?), and get rid of tenure.

Uhh,teachers in major cities make 100k with 10-15 years experience on 180 contract days per year, and also get pension. Student:teacher ratio is below 20 in all major cities.

Not sure how you’re defining major cities, but starting salaries in the top 5 CoL cities are definitely not near $100k

Boston $60k https://btu.org/wp-content/uploads/2019/12/Salary-Grids.pdf

Seattle $65k https://www.seattleschools.org/wp-content/uploads/2021/09/Ce...

LA $55k https://achieve.lausd.net/cms/lib/CA01000043/Centricity/doma...

SF $60k https://uesf.org/members/2017-2020-contract/

NYC $60k https://cdn-blob-prd.azureedge.net/prd-pws/docs/default-sour...

Looks like you can add $10-20k with a masters and a few grand per year of tenure.

Key word in my comment was “competitive”, which these aren’t for trained/skilled workers with other options.

In the most expensive locations I would agree that it is not very competitive, and the same goes for entry-level police and other needed services. (Though nimbys reap what they sow here.) The entire education system is inefficient and anachronistic. School should run all year, simply emphasizing more sports and nature in the summer. Then you don't need to fit the career and compensation structure of the personnel into seven contract hours a day working 180 days a year. You can pay a true salary, end regression and lower criminality in the summer, while also helping working parents be more productive.

Don't get me started though! College is even dumber. 36 weeks of class per year for 18 year-olds? Wtf. Kids graduate high school and ought to be in class or an apprenticeship the next Monday. Most young adults go to non-selective universities to get simple credentials like nursing or education. It's not about fancy internships at Google or other nonsense like gap years. That's relatively decadent stuff most young adults can't indulge in.

Both suggestions are very right-on. I would add: get rid of school boards. They are political dinosaurs who 'know nothing ... nothiiing'. Cut administration to essential (non-educational) needs A superintendant's income could be far more effectively invested in better teaching materials. A couple of good people tuned fulltime to student ecology and parent relations.

Education: Every school has dozens of well-educated and committed teachers; together they can focus on improving quality and streamlining away the BS.

Do higher teacher salaries correlate to student outcomes? Not to even mention causation.

Private school teachers earn less than public school teachers, but private school students get much better educational outcomes.

Bad or mediocre private schools—which are the vast majority—pay less than public schools.

Good ones may pay more.

Teaching in private schools also tends to be an easier job, mainly thanks to selection bias giving you an easier set of students to teach, and (oddly enough) school admin being both willing and able to tell parents to fuck off and go away if they or their children are being jackasses and won't shape up. Public schools can't do that. Turns out decent-or-better private schools would rather lose one paying student than let them make things worse for the rest, risking several paying students (and the good ones all have waiting lists anyway—they'll fill the slot, no problem).

If you can write a test (or other combination of measurements, metrics etc) that actually measures whether someone is well educated, and tells us how much of that is down to the teacher and not the overall school, family, home life, culture, etc, and cannot be gamed etc, then you will have solved one of the worlds longest standing social issues.

Then you will just need to convince people to vote to (more than) double what we pay teachers (and eat the resulting tax rises) to fund the resulting bonuses for at least some of them.

Clearly grades are worthless here. But that doesn't mean there is a better system. Especially given what people are willing to pay and the complex other inputs...

Complaining that tests don't measure anything is popular, but tests work.

I didn't say they don't measure anything. Just that they don't measure everything...

About 25 years ago, The Washington Post carried an item on Algebra I grades in the high schools of Montgomery County, Maryland. In the west of the county, one needed to score at or above 90% to get an A in Algebra. Mid-county the cutoff for an A was in the mid-70s. Assigning the kids randomly to teachers, across schools, is called "busing", which has never been popular among prosperous families--just the people likely to turn out at school board elections.

(Not that I don't see the appeal of your proposal.)

Modern public schools cannot stand elitism like gifted programs. The Seattle school district eliminated all gifted programs because they were inequitable.

My proposal caters to that desire for equity. Random assignments to teachers.

There are gifted programs on the east side. Also, what about Running Start?

> change the test so that passing it reflects learning

How?

If you randomize the types of questions enough every year, the best preparation for passing would be to just teach the general subject.

Yup. I always thought that people who spent all this effort cheating would expend less effort if they just bothered to learn the stuff. And besides, you get a bonus from having a bit of pride that you did it honestly.

    2 + 2 = _

Fill in the blank.

I’m not sure I understand what you mean. Is this an example of a question you feel teaches understanding?

I don't understand how one can do arithmetic with no understanding of it. It's an example of a question that tests understanding.

How would you determine if a student understands arithmetic?

You memorize the answers? Like, I had an entire week of third grade where they wanted us to do a table of 100 basic arithmetic problems, and it was really a test of whether you remembered that 8+9=17 rather than some sort of actual understanding of what “eight”, “nine” and “the combination of eight and nine” meant.

I think arithmetic in general is actually one of the most difficult things to test understanding of, because there’s not a whole lot of understanding involved. It’s kind of axiomatic for the rest of math to build on top of. I would attribute “understanding” more to being able to synthesize useful results from properties like addition and subtraction being inverses of each other, being able to form relationships like “two small numbers added together are always less than two big numbers added together”, etc. But I did actually start this thread by asking how you would test for this, because I believe doing so is quite difficult. I don’t think I’ve seen many a test that people haven’t tried to game without “understanding” the subject.

For the longest time we (and in many places probably still do) literally taught students multiplication tables. That is, hardcoded answers for a * b with a, b <= 10. As soon a double digit integers are involved - tough shit.

If you don't memorize the tables, you're going to have a tough time multiplying N-digit numbers.

Besides, work enough arithmetic problems and you'll inadvertently memorize them anyway.

So you've memorized what 23+69 is?

Nobody likes it because it has a ton of problems.

Jumping to "my solution's just too edgy, and that's why nobody will like it" before having a good grasp on the issue is perhaps premature.

Everything has problems. But it will be far better than the current system.

> edgy

It's not about edginess. It's about the public teachers union that blocks any and all attempts at merit pay.

Just let teachers do whatever they want with periodic evaluations, if the teachers suck move them around. Do the same with students. The shuffle will get the good teachers with good students.

This is a remarkably bad idea. No, the differences between students will definitely not “average out” at this scale. And how do you expect to handle poorly performing schools as a whole? Do you not expect the quality of teachers there to just plummet?

Paying people for results works very, very well in the private sector. Why wouldn't it work in the public sector?

> the differences between students will definitely not “average out” at this scale

Of course it will. Assigning kids at random will produce an average over time.

> how do you expect to handle poorly performing schools as a whole?

The teachers in that poorly performing school will have a powerful incentive to do better.

> Do you not expect the quality of teachers there to just plummet?

I expect if you dangle $20-30000 if they do better, they'll do better. Monetary incentives work.

And in businesses that pay commissions, the people who are good at getting those commissions (i.e. they're good at their job) want to be there. They're not satisfied with just a salary. High commissions attract the best.

The problem with saying that teachers will have an incentive to make a bad school better is that they can not hope to compensate for massive socioeconomic differences. If you pay solely on absolute results, all the good teachers are going to go to schools with favourable intake profiles, which are likely to be strongly correlated with house prices. This would be a further driver of inequality. There would have to be a realistic scaling factor, looking at improvement, perhaps, rather than absolute results.

The current system is an abject failure. Time for some old ideas that work, like paying for results achieved.

Yes, it’s clearly not working in some regards. Your basic suggestion, however, is also unrealistic - a school in isolation can not make up for all the problems of an area. Similarly, a brilliant teacher can only do so much. If they can move to a better area and bank their bonus for much less stress, they will do just that, leaving a failing area, and its children, with even worse provision. Frankly, there’s no straightforward solution. Better pay is probably part of it, but it’s a far more complex issue and one that has to be considered as part of a wider programme of improvement.

“Paying people for results works very, very well in the private sector.”

Can you provide evidence of this? CEO compensation is a terrible model that encourages short term tricks like stock buybacks, front-loading revenue, and other tricks to garner short term performance. As for rank and file employees, measuring results is difficult … just like in schools.

Every company that has salesmen paid on commission.

> measuring results is difficult

And the better job of that a company does, the more profitable it becomes. Companies have a very strong monetary incentive to do an effective job of that.

> CEO compensation is a terrible model

Is it? Executive bonuses etc. at Boeing crashed along with the MAX crashes.

Better idea: fully fund schools and stop blaming teachers.

Reply on news.ycombinator.com