Teacher Effects on Student Achievement and Height: A Cautionary Tale
nber.org
nber.org
Excellent work by those that wrote the paper.
The reason is, for one, the use of value-added measurement does not have a significant role as many would think. There is not a signle state in the U.S. use VAM as a single measure of teacher performance, nor a determinant factor in teacher firing or promotion. It is always mixed with other quantitative and / or qualitative measures (which usually are no better than VAM).
The second reason is, the research community has not find a better replacement yet. We always know we cannot directly use student achievement, now we know we cannot trust a more complicated measure such as VAM which takes into account of student history and other factors.
At system level however the data is very robust: in a developed education system the best you can do is to give autonomy to schools and teachers, assure some kind of accountability (not necessarily exams, but eg European-style 'supportive' inspection) and strengthening teachers through training and support is the most effective education intervention, except maybe for free school meals.
Value added measurements of student achievement will always fail.
In affluent areas, where student achievement is already high, schools can afford to hire the best teachers, but they will realistically only be able to boost achievement by a small fraction, given the child's prior attainment.
In deprived areas, where student achievement is lower, schools can only afford to hire less qualified teachers, and the boost to achievement will be a small fraction of previous attainment.
Either way, the measured effect that the teacher has will be marginal, while common standards and common testing are the only measured outcomes. This does not equate to the teachers not being invested in student outcomes, or trying their best for their students.
Value added measures were only introduced as a form of target setting, for teacher performance. I'd encourage everyone to watch Adam Curtis' The Trap[1] to see the effects of target setting.
Teachers in early years education set a target, then primary school teachers need to show value added so game the system to show improvement. This trickles up the system, and the net result is unrealistic targets for almost every child as they exit formal education.
More explicitly, there are many excellent teachers in poor areas. Teaching privileged children in a well funded school is much easier.
I would love to see a toolset like this come into being, but I have no ability to develop such tools.
Assessment can give me that kind of information; hence, my belief that making instruction-based assessment more globally trustable is the key component.
Again, I would love to see something like this come to be. I’d love to jump ship and work on it full time. It would not only provide significant academic advantages, but also make increased teacher compensation politically palatable. If we can measure teacher effectiveness, then we can reward it. If we can’t, we use tenure.
So perhaps the teacher measures were still valid because all the tall people were more successful because of other biases in the system?
I'm mostly just being cheeky here. To be clear, I still think it's nearly impossible to measure how good a teacher is based on student performance, since the teacher is only one part of the equation, the parents (and the student's home life) being a much bigger part of the equation. The teacher can only do so much if the child comes home to parents that don't feed them and tell them they are worthless.
IIRC, while it has been shown that taller people are more successful, and everyone says it is due to "hidden biases", those hidden biases were not in fact shown.
A competing theory that I am familiar with, which provides just as good (if not better) explanation is this:
Height is a (bad) proxy for healthy upbringing. Your maximum height is essentially genetically predetermined. But to actually attain that height, you have to have good everything growing up -- good food, good health. If you do not have access to nutritious food, or you are plagued with diseases that impact your growth (.. or treatment for them [0]), you will be shorter. So:
Tall = Tall Genes + Good Environment
Short = Short Genes + Good Environment ; or Tall Genes + Bad Environment
Thus, tall people are correlated with good environment growing up, and shorter people less so -- and the theory is that it is the environment you grow up in that explains your success, rather than some hidden biases.
It seems clear-cut, and there are studies that link certain genes to height on a statistical level. But I suspect we're going to find that height has little to do with genetics in the traditional sense, and outside a few notable outliers, that there are no "short genes", just environments that fit specific genetic profiles that are or are not available to the people with those genes. Europeans are tallest in a generalized global monoculture where Western diets and lifestyles are the norm, and where European-descended populations have enjoyed a generally higher quality of life for several generations. Color me surprised.
> outside a few notable outliers, that there are no "short genes", just environments
Predictions based on about 20000 different genes gave correlation 0.65, in the best paper I can find in 5 minutes. (Which was from 2017, a long time in this game.) This is for differences within a European-descended population.
They were further working to tie these effects to a group of epigenetic markers. (possibly successfully, I'm unclear.)
It may have been this study: https://en.wikipedia.org/wiki/%C3%96verkalix_study
There's a lot of academic interest in such things, precisely because they go against the grain of everything that's well understood. Finding one, and proving that it's really certainly not zero, could make your name.
You'd need to run studies on separated twins to cleanly split the two.
That's probably the meta-analysis that I was alluding to, which unfortunately gets extrapolated out to humans in general in most discourse. There are two caveats I see.
1) The population described was largely European-descended, and presumably in relatively wealthy, relatively egalitarian environments. Inequality is known to lower a population's average height - not with a general shift downward, but by lowering the floor and increasing the shorter population - so it could be that there is a generally smaller range of heights in that population than in humanity as a whole; and, with environmental variables less variable, the only thing left to determine height is genes. But saying that most of the difference between a 6'2" German and a 6'4" Dane is mostly genetic is very different from saying the same of the difference between a 5'11" African American and a 5'4" El Salvadorian.
2) Even in the population described, it's likely that class mobility is not super high. That means that the relative fortunes of families would not have changed over several generations, even as wealth accrued and the gaps between classes diminished. I'm not a statistician: how likely is it that genes associated with certain families and communities, entrenched in their class as they are, could be flagged as "tall" or "short" genes, respectively, but ultimately have little to do with the physiological processes of growth and development? I realize that those confounding variables are supposed to be controlled for by the analysis, but what if they're so tightly bonded in practice, across a large-enough population, that it's too difficult to separate them until you can actually say what each gene does and how that plays into growth?
1) This is true, and uncontroversial. Nobody doubts that bonsai are small for environmental reasons. More technically, the heritability of traits is inversely related to how variable the environment is. Once everyone is well-fed & well-cared-for, you expect higher correlations between parents' height and children's, compared to the earlier world where many were stunted. (And likewise you expect a stronger influence of people's innate interests on what jobs they do, once they are rich enough to turn down better-paying alternatives which they hate.)
2) This is a real concern, but it also depends a little bit on what you hope to achieve. My extreme example: If a freak accident on the way home from the 1964 olympics had stranded the Dutch men's & woman's gymnastics teams on the same island as (m&f) Nigerian basketball teams, and you visited today, then in this population genes for blue eyes would be a pretty strong predictor of being short. Because they would mark ancestry, not because of a more direct link.
This may not be a terrible thing. If what you want is a test to see who is more susceptible to some illness, for the purpose of giving them screening from age 40 instead of 50, then if what you are detecting is ancestry-markers not heart-muscle-genes, do you care?
But for other purposes you may really want to know the functional genes related. One thing you can do is to look only at differences in height (etc) between siblings, who by definition have the same ancestry. This is also done, but it's harder to collect this data, so studies for humans are smaller & noisier. But for plants you can do this very well, control everything. And I think that for dairy cows this is done very well (you know their ancestry perfectly, so the only point of DNA testing is to get information which that doesn't capture).
But children's height might just be a proxy for age, or stage of development. 8th-graders are taller than 6th-graders, and do much better on 7th-grade tests.
I am, as always, quite convinced that teachers aren't that valuable. At least not the kind in public school.
What I'd expect from a situation where there is wide variation in teacher ability is for them to form private schools where they extract massive value from the relationship because everyone will want the best tutor for their kids.
This can probably happen. But normal teachers aren't it. It's the specialized tutors who get this value. Which means that perhaps all really good teachers become private tutors so you'd expect everyone else to be essentially unexceptional.
What I'm saying is that if student success is already correlated to student height, then anything that measures student success will also measure student height.
The "Parachutes" paper is demonstrating that parachutes make no difference when used out of a stationary grounded plane.
I think the point is that both papers are demonstrating that the numbers alone don't tell the full story.
When beliefs regarding the effectiveness of an intervention exist in the community, randomized trials might selectively enroll individuals with a lower perceived likelihood of benefit, thus diminishing the applicability of the results to clinical practice.
So... was there some kind of selective enrollment in the teacher study? I don’t think that’s the issue here.
That skews the results of these studies. This is what they try to illustrate with the jumping out of plane study. The result won't be reliable if you don't have a control group that actually might have benefit from the treatment.
Cancer patients (jumpers), will or will not get a new type of chemotherapy (parachute). But since it is known that the sickest (hightest jumping hight) certainly will die without treatment (parachute), they decide to only test the treatment on patients with only early stage and benign cancer (close to the ground).
Both this article and its predecessor were published in the Christmas issue of the BMJ, where funny papers with a semi-serious point (in these cases, the ongoing debates in medicine about the superiority of RCTs over other forms of clinical evidence) are common.
My favourite of this genre is "Does Peppa Pig encourage inappropriate use of primary care resources?" (https://doi.org/10.1136/bmj.j5397) critiquing the clinical practice of Dr Brown Bear, a fictional cartoon GP.
So is it really that far-fetched to think teachers could have an influence on eating & sleeping habits that could definitely affect height?
The study supposedly checks for one aspect of this – seeing if there's a cross-correlation between achievement-boost and height-boost, and thus some single general "positive influence" – and finds no correlation. But I haven't looked closely, and it's quite possible they're over-controlling to find that no correlation – which seems quite common in educational studies on ideologically-fraught topics.
I could also imagine teachers independently affecting academic achievement & height. Imagine a 2x2 model: teachers can be either 'fun/popular' or 'severe/unpopular', but also 'instructionally-effective' or 'instructionally-inept'. Fun/popular teachers might keep students in attendance, cheery, eating well at school lunches, not stressed/overworked with homework – and thus more likely to grow in height. But, only some of these 'fun' teachers (who promote mood/physical vitality) are also moving academic measures higher. Meanwhile, some of the academically-effective teachers do so via a harshness & heavy-workload that impairs height.
Likely untrue, given a few moments search. For example, "On average, teachers spend $37 a month and principals spend $59 a month for food for their students." [1].
There's lots more data on teachers feeding kids in perhaps causal amounts.
[1] http://neatoday.org/2013/09/10/report-teachers-spend-37-ever...
Given the strong correlation between some breakfast vs none, and school performance, it would not take much to affect the outcomes by feeding hungry kids little snacks.
Keep in mind though that the correlation, while very highly statistically significant and replicated over many data sets, is not very strong: at r = 0.2, the r^2 is 0.04, so only 4% of variance in intelligence is explained by height. Thus, it is not clear whether the height-IQ correlation plays significant role as a confounder in the study above.
But I'm saying something much simpler. In a school class, there's another effect, that the oldest may be a year or more older than the youngest. The mix of ages in a classroom will produce a correlation all by itself. It seems to me this would be a much bigger effect: 7th-graders are quite a bit taller, and better on tests, than 6th-graders.
Given that, i doubt height measured at any one (age)point during childhood strongly correlates with iq. You'd need sufficient data to be able to reliably predict height as an adult for that, would be my guess.
More directly to your question, yes, as I said in my previous comment, the effect is very real, and is not driven by some outliers. There have been dozens studies on this, with Ns in tens of thousands. Wikipedia has a sample of them: https://en.wikipedia.org/wiki/Height_and_intelligence
Is the number (teacher score) useful? Maybe not. But you can't simply take away a number, resulting in no feedback. You have to either come up with a new number that is useful, or localize it enough that people can see the results. Otherwise people won't want to pay.
That's why these weird bureaucratic numbers usually come from state or federal requirements. People see the big pile of money going somewhere and they want "accountability" because they can't see the results for themselves.
Let's stop shuffling around so much money, laundering it through state and federal agencies. The money, if it comes back, always comes with strings attached. This goes for schools, highway funds, and many other things.
I read a book not too long ago entitled “The Nurture Assumption,” by Judith Harris. It influenced me more than anything I’ve read as a parent, because it is so obviously true. Her thesis is that everyone assumes that adults have a significant impact on a child’s future, yet there is basically zero evidence to support such an assumption.
Peer groups are the dominant influence on a child’s personality and development, full stop. This is probably hard for educators to accept because there’s really nothing that can be done about it from their perspective.
As an aside, the mistake in nutrition was simply that it wasn’t true. And it also wasn’t a “mistake,” it was a lie that the sugar industry paid to spread.
if you pick n numbers at random, it's not likely that one of them is bigger than the half of their sum.
but my intuition says that if you pick all variables from the same distribution, this should still hold.
At first, we don't know anything about the factors, so they are assigned a random importance. or am I missing something?
What's missing here are facts.
To dismiss something based on something anecdotally unrelated seems short-sighted.
A dominant factor can also only account for less than half of an explanation if there are enough factors. E.g. Peer group is 40%, parents are 30%, teachers are 20%, TV/videogames/groups-seen-but-not-interacting-directly-with 10%. Here peer groups would still be dominating but parents and teachers are significant factors as well. These percentages are close enough that it would be ridiculous to just focus on one (or even the first two alone).
I happen to believe peer groups are really important. But other things are important, too, including the adults.
Remove all the middle-aged and older men from a city, but still provide other necessities, and you'll see some terrible outcomes. (This is true in other species as well, like elephants where older males are hunted for tusks.)
We all have stories about a great teacher or a terrible parent. But these are anecdotes. You don't actually know what would have happened if you had different parents, or different friends, or went to a different school.
It's perfectly possible that if circumstances were different, you would have gotten to the same place, by a superficially different path.
The specific stories would vary, of course. But when we talk about society-level policy, it's not specific personal stories that we focus on. It's things like income, marital status, job field, health, criminality. A lot of evidence suggests that even if your specifics were very different you would have actually ended up largely in the same place on these metrics.
Yes, it’s all of them. And they all need to be addressed instead of trying to pick the one all-dominant factor.
> ... these are anecdotes.
Combine a lot of anecdotes and add in more scrutiny and we can call that data.
IMO this no longer holds true with the advent of the internet era. You're now spending equal amounts of time exposed to a countless number of people. Even if you aren't physically with them, and even if it's a one-way relationship (i.e. they are not aware of your observance and existence), they are still influencing who you are.
Having said that, saying "Peer groups are the dominant influence on a child’s personality and development, full stop." is over simple. Look at the children at school. They all share a peer group. Is there personality and development all aligned? Absolutely not. Diverse friendship groups develop with distinct interests and moral codes. How do you account for that?
As one example, you may have a kid who has a shitty family life who will act up to demand attention. If you have inexperienced teachers in a classroom of 35, it's going to be difficult to negate the impact of that kid on the classroom.
Everyone in the survey thought there was a strong effect except one group: Mothers with two or more kids.
For what it's worth, I'm a father of 3 and our oldest is in school 2 years now and could easily be described the same as yours. However, he is in a "good" school and we still see all kinds of new behaviours being brought home. Personally, I see it as being natural for his behaviour to change while experiencing all these new social dynamics. I see him having to go to school, put on a new personality that fits into the existing social dynamics, and then of course he comes home and has to unload the effort and stress that took. It can manifest as anger, exasperation, etc. - I try to remind myself what he's going through and just to be there to "receive" these behaviours, rather than becoming upset and trying to course-correct (I'm far from perfect in this). It's up and down, but in the macro trend, he's getting better at these transitions at a remarkable rate.
I view this stage as normal learning and it's something we all still do subconsciously as adults who go out into the "real world" and come home. How many of us are exactly the same in both situations? How many of us feel more relaxed when we get home? Learning to transition from the "outside world" to home and simply relax is a learned skill, IMHO, and hard at first for little ones and they cope with the transitions differently. Honestly, if he behaved the exact same at school and home and handled that transition with ease, I would be combination of worried and blown away.
Genetics are likely the dominant influence, with peer groups (shared environment) being the second largest influence.
At least, that's what the Three Laws of Behavior Genetics would say.
Our Catholic friends recently moved their young ones out of the public schools and into a Catholic school.
The songs from the ~6yo changed from ‘shake your booty’ to latin ‘Ave Maria’ at home :)
There's a lot that your boss never need know...
Then again, you might reject developmental psychology and rely on sociological studies of child rearing like Annette Lareau’s book, “Unequal Childhoods.” Lareau outlines two competing parenting styles based on parental income: natural growth and concerted cultivation.
Or, you can pass on both of those and read Critical X Theorists who complicate the narrative of both developmental psychology and sociology by providing studies based on narratives. For an accessible take check out Ijeoma Oluo’s work. For in-depth check out Gloria Ladson Billings.
The other point mentioned in this post is about “good” and “bad” schools intersecting with the geography of opportunity. What makes a school good? Well, since the late 90’s the main criterion for a good school has been high test scores (SAT’s or state proficiency scores, depending on the school). But test scores don’t tell you a whole lot about the school’s quality, even as a proxy measure (see Dan Koretz’s “Measuring Up.”) In that case, then, what makes a “good school” good and a “bad school“ bad? Relative income and race seem to be deciding factors (see Finnigan and Holme’s most recent book, or Rothstein’s book on redlining, or Nikole Hannah-Jones’ series on her own daughter’s education).
Long story short: it’s complicated and this article is about VAM, not parenting, school quality, or peer effects.
Edit: fixed spelling and punctuation
Twins reared apart studies would dispute this.
You should get your kid in a better school using any method you possibly can. I mean it. There's a reason Zuckerberg went to Phillips Exeter Academy and Gates went to Lakeside. It doesn't have to be private, just a better ranked school. Any better ranked school.
We too saw our child go through this. He was bullied because he didn't play the right video games, watch the same youtube channels so didn't know the right jokes...a kid can stay an outsider for so long or adapt to the environment both impacting behavior.
We wrote a letter, got him transferred, a majority of it has changed.
If you choose this route, come with hard facts. It's not that they don't care about your child, but you are one of several voices. Make it easy for them to help you.
This is the reason why a house near a good school in North America can easily cost $200-300K over a very similar house in a lousier neighborhood. I would seriously advise moving to a smaller property in a better school district, if you care for the future of your kids.
Bad kids -> your good kid
but also
Bad parents/trauma -> one child with a dominant personality -> outsized influence on peer group -> your kid.
All that said, if I had a kid I'd spend every penny to segregate him/her with other kids away from average dodos. I had a terrible experience as a kid surrounded by imbeciles and bullies, which changed after I went to a better school.
Above a certain threshold, it’s not a matter of resources or teacher quality. Baltimore schools spend as much per student as the wealthy suburban counties around it (and cost of living in Baltimore is cheaper).[1] Nobody from Montgomery County would send their kids to Baltimore public schools, but a ton of parents in Baltimore would love to send their kids to Montgomery County schools, but can’t afford $700k+ for a starter home.
[1] https://www.bizjournals.com/baltimore/news/2019/05/21/baltim...
And it's one I have the least hope to see changed, for exactly this reason:
>a ton of parents in Baltimore would love to send their kids to Montgomery County schools, but can’t afford $700k+ for a starter home.
Changing things would require middle/upper-middle class voters to destroy the very thing responsible for half the value of their biggest asset. It's a self-reinforcing problem.
Why would Baltimore parents want to send their kids to Montgomery County, if not for better teachers and more resources?
That has interesting implications for the idea of getting everyone to go to college. Maybe a lot of the benefits of a college are the peer group, and if there's no selection of the peer group, the benefits will evaporate.
Completely unethical and unlikely to be effective of course.
Whether parents are highly involved or not is an even higher factor. Think about the amount of time a parent COULD spend with a child over the course of their development years vs their peer groups.
I think this is way overstated, but aside from that, who is it that chooses where to raise kids and hence, what type of peer groups will be available to those kids?
We really don't know what is true unless we test it. I know it sounds ridiculous but it's not beyond the realms of possibility that teacher opinion of student capabilities influences social status which in turn influences release of growth hormones which influence height. Aren't there other species which grow when they rise to the top of the pack as growth hormones are released to assist in maintaining that status? Not saying it's likely but if left untested we cannot rule that out. There have been plenty of surprises in science when we went and actually tested what we previously presumed to be implausible.
https://xhxhxhx.tumblr.com/post/189336550187/teacher-effects...
He constantly configures his classroom to maximise learning continuity for the entire class. It is a massive challenge balancing out the problems students face physically or outside the classroom. Teaching has become the easy part of teaching.
The problem with student learning is that is affected by three things.
Emotional Health Physical Health Teaching Quality
to build an educated adult you need to optimise all three variables.
The problem with rewarding teachers is that their students aren't objectively assessed on all three dimensions.
In every district in the world there should be three person educational assessment teams that consist of an emotional health assessor, a physical health assessor and an academic progress assessor that work as a team to objectively assess the students and work with the teacher and support agencies to eliminate learning barriers.
Any students that need help should get referred to the correct service or program as quickly as possible. The teacher should focus purely on teaching and integrating support agencies.
Teachers deserve to be rewarded for their teaching ability in an objective way. As it happens there is a lot of unnecessary subjectivity and people blame teachers for not being social workers, counsellors and doctors. The education of every child in a class deteriorates as these three dimensions become less homogenous because the maximised learning continuity of the group depends on a limited resource. It is teachers who can creatively balance them while engagingly delivering the curriculum that are creating unrewarded alpha in the educational space.
Paying $250,000 a year in salary to a group of 3 specialist objective assessors with the data linked to the health and social services would get a better return on investment than any other expenditure a government could make in lifting the health, education and future earning potential of children.
If a child isn't learning there is always a reason. The faster you address that reason the more value you create. Sometimes it's the teacher, sometimes it isn't but for the child's sake we should definitely diagnose the reason.
There's also a correlation between IQ and academic achievement.
So, if we did decide for some (undoubtedly bad) reason to reward teachers for happening to have more than average tall students it would be comparable to rewarding them for having more than average high achieving students.
> The increased availability of data linking students to teachers has made it possible to estimate the contribution teachers make to student achievement.
There was some data available before (or the sentence would not have used the word "increased"). Why wasn't it possible to estimate with that?
> By nearly all accounts, this contribution is large.
It goes on to talk about what "large" means:
> Estimates of the impact of a one standard deviation (σ) increase in teacher “value-added” on math and reading achievement typically range from 0.10 to 0.30σ, which suggest that a student assigned to a more effective teacher will experience nearly a year's more learning than a student assigned to an less effective teacher (Hanushek & Rivkin 2010;...).
(Typo: "an less effective" should be "a less effective".)
A "range from 0.10 to 0.30σ" doesn't make sense. A Greek lowercase sigma (σ) is used to represent one standard deviation, but the sigma is used only on the upper end of the range. Should it have been from 0.10σ to 0.30σ?
And how are they measuring the impact on achievement of an increase in teacher "value-added", anyway? It says that estimates of the impact "typically range from 0.10 to 0.30σ", but it doesn't say what units those figures are in.
The sentence goes on to say that those unit-less estimates "suggest" that "a student assigned to a more effective teacher will experience nearly a year's more learning than a student assigned to an less effective teacher". Over what time period? That is, how long does a student have to study under a "more effective teacher" to get "a year's more learning"? 1 week? 12 years? It doesn't say.
And finally, how do those unit-less estimates "suggest" an impact measured in learning time? It doesn't say.
>A "range from 0.10 to 0.30σ" doesn't make sense.
Out loud, you might read it as "oh point one to oh point three sigma." And it's fine. Ever written a phrase like "it's very expensive, costing somewhere around $3-4M" with one "$" and one "M" but two numbers?
It's a terrible paper in general though, for example it references Bitler et al. 2015 but never provides a full citation.
But hey, let’s publish a pointless paper... because academia.
Like cost of hiring? Seriously, if teachers are not as critical, why not hire less or pay less salary... any half decent person in the classroom will do.
I think the study is about right that teachers are not as critical as they'd like to believe. I just don't know what to think of the implications...
When I was a teacher, I had zero power to remove a disruptive or inattentive student from my class. I had 9th graders who could not deal with negative numbers being expected to do well on state Algebra tests that are of arguable quality. Even if they made substantial progress, they would not be to grade level.
I can still remember one student who after nearly a year of not being bothered to pay any shred of attention in class, during the final review, paid enough attention during the last steps of solving an equation: Me: Alright, and combining like terms, what is 10x - 18x? Student, suddenly paying attention: But wait, you can't subtract a bigger number from a smaller number!
I can't stop the class on nearly the last week of 9th grade algebra to help this student understand negative numbers. And they are not interested in coming in to get help outside of normal class. Parents were not interested in making sure this kid got through school.
Not so fun fact: at this school, less than 4% would go onto any post-secondary education. Of them, about 2% would go on to finish a degree. Something like a 90% transitory rate (meaning that most students who started 9th grade at this school would transfer to a different school before graduation, if they made it that far). I could go on, but there was an underlying cultural issue that did not value education. That trend is hard to buck: I had third generation gang members, kids raising their siblings because parents (often a single parent) were working multiple jobs, kids dealing with daily violence, one kid was stabbed to death right off of campus. The majority of these kids and their families have no experience seeing what an education can do for their prospects. Add onto that a tough employment market (inner city San Bernardio!), and those who did have siblings who did get a degree, they often couldn't find a job.
All this to say, I don't think lowering salary of teachers would help. Maybe lowering the cost of administrators would help. At this same time, there were more administrative personnel in the district than teachers, something I could never understand.
You are 100% right, the world has yet to catch up. But mostly because politics, politicians, voting... etc. Not because we don't understand the problem/solutions.
Schools are funded incorrectly, schools end up Segregating on income, which sadly ends up segregating on race.
This ends up with rich schools and poor schools. Rich kids get good outcomes poor kids get bad outcomes. The cycle continues.
We know what we need to do, just no political will to do it.
From the introduction:
> In this paper, we provide a stark illustration of the limitations to using value-added models to identify high-and low-performing teachers. We do this by applying commonly estimated models to an outcome that teachers cannot plausibly affect: student height. Aside from the implausibility of teacher effects on height, student height is an attractive measure for this exercise since it is symmetrically distributed, interval measured, and arguably less prone to measurement error than achievement. We find that the estimated teacher “effects”on height are nearly as large as the variation in teacher effects on math and reading achievement.
It's also an interesting approach: take a model that is apparently predictive and see if it's also predictive of something that we know is unrelated. By showing that value-added models are displaying spurious correlation maybe policymakers will take note.
Could maybe take a similar approach to validating our ML models?
https://scholar.google.com.au/scholar?q=relationship+between...
If you'd actually bothered to understand the domain we are commenting on...
https://scholar.google.com.au/scholar?q=relationship+between...