Schools probably do do something

by Yair Halberstadt2 min read26th Sep 202127 comments


EducationWorld Optimization

It's a relatively commonly held view here on Less Wrong that school is mostly a waste of time.

I'm not going to comment on this directly. Instead I'm going to show that school probably has some impact on future life outcomes.

The evidence for this is the Relative Age Effect.

The term relative age effect (RAE), also known as birthdate effect or birth date effect, is used to describe a bias, evident in the upper echelons of youth sport and academia, where participation is higher amongst those born early in the relevant selection period (and correspondingly lower amongst those born late in the selection period) than would be expected from the normalised distribution of live births. The selection period is usually the calendar year, the academic year or the sporting season.

For example students in Oxford were well over 20% more likely to be born in September (the start of the academic year), than August (the end).

The classic explanation for this is that whilst there's not much of a difference in maturity between a 17 year old and an 18 year old, there's a large difference between a 4 year old and a 5 year old. So older kids shoot to the top of the class when they start school, and so are more likely to be engaged and interested in class, which keeps them at the top even once the difference in maturity stops being relevant.

This implies that either

A) School does something positive for those at the top of the class (pushes them up relative to kids who don't go to school at all).

B) School demotivates those at the bottom of the class (pushes them down relative to kids who don't go to school at all).

C) Given how difficult it is to get into Oxford, even 1 years worth of maturity can make a significant difference.

If we discount C, then school does have some impact on academic outcomes albeit we don't know if that outcome is positive or negative.

Further research could include:

  • is there a difference between students who took a gap year before applying and those who didn't? This would investigate option C above.
  • does this effect persist among home schooled children, where the classic explanation wouldn't apply.
  • does this effect persist into later life outcomes? Thai would measure whether the impact is short term or long term. According to Wikipedia this effect has been observed among CEOs of S&P 500 companies as well. However this is maybe just because they're more likely to get into a prestigious university, which is what really matters, not that school directly had an impact.


27 comments, sorted by Highlighting new comments since Today at 8:39 AM
New Comment

Breadlines do do something. Those in the front tend to become better fed than the rest.

I'm not sure what the point you're making is?

I think it's that yes, schools do something, but that something may have less to do with the things everyone usually thinks and hopes schools do, and more to do with "change how well they teach different students as a function of initial maturity in a way that locks them into better or worse future trajectories, based on an essentially random variable, namely their birthday."

Which is admittedly entirely consistent with your post as written, taken in isolation, but also very much at odds with the as-usually-understood spirit of what people around here who say school isn't that valuable are actually trying to say. Sorting people into buckets by birthday range and then using that to lock them into better or worse life paths, with little relationship to anything like the innate levels intelligence or conscientiousness that two different kids would have if each were evaluated at the same age rather than the same point in time, is technically schools "doing something." It is not schools doing what proponents of schooling, as it's currently done, want you to believe school does. It is not schools doing something socially or economically valuable.

I agree that that wouldn't be a valuable thing for schools to do. I would be interested though in a gears level explanation of how they lock in the older students, and whether the effect is a positive one for the education of those older students. If so then we have an obvious technique to improve schools - get that effect for all students not just the older ones.

I suppose that in the first few grades of elementary school, the difference of almost one years matters a lot. Going by the old model "IQ = 100 × mental age / physical age", being a 7 years old child in a group of 6 years old children is like getting +16 IQ points. You are also physically stronger, emotionally more mature, etc.

So it's kinda like getting a magical pill that makes you 16% better at everything when you start school attendance, and then the effect of the pill slowly expires... but you still get the secondary effects of prestige, self-confidence, special opportunities received when you won some competitions, higher motivation, etc.

To get that effect for everyone, you would have to somehow make everyone older than their classmates.

On the flipside framing, currently the younger kids have a magical curse that makes them 16% worse at everything, and ideally we would be able to get rid of that. 

Wouldn't this actually be evidence against school doing something?  I.e. we have discovered that relative age is more important than years of education?

Within this sample, years of education would be mostly fixed. Those with more/fewer would run into selection effects (e.g. getting held back a grade gets more years schooling, but presumably lower odds of getting into Oxford).

That's exactly my point. 

Suppose we were doing chemistry instead of schooling (so that I didn't already know the answer).

I have two beakers, one labeled 6.01 and one labeled 6.99 and I then pour 12 units of "school" into each beaker.  Afterwards I discover that the beaker labeled 6.99 has produced 2x as much "oxford" as the beaker labeled 6.01.  The conclusion I would come to would not be that "school" is pretty good at producing "oxford".

Even more fun, I challenge you to predict the amount of "oxford" produced by adding 12 units of "home school" instead of "school".

It' not about relative age (either as in age of one person divided by age of another or one age substracted from another), it's about their month of birth. So it's evidence for relevance of amount of received sunshine during pregnancy, relevance of age of being admitted in school and relevance of astrology.

Since it seems to somewhat align with different kinds of education starting in different times of year, my personal bet is on schools, though I wouldn't completely discount differences of pregnancies in different times of the year (sorry, astrology, but I need a lot more evidence to seriously consider you).

Did they compare regions with different cut-off dates? For example, in NY it's in December, and in MA it's in September. Actually looks like it's 12/1 in NYC and 12/31 on Long Island, too. Those would help distinguish age vs sunlight or other seasonal effects on development.

Yes this effect is consistent across different countries with different cut off dates.

That would imply there's a huge difference in abilities between 17 vs 18 year olds applying for Oxford, which sounds too extreme.

I think "applying for Oxford" is doing more work in that sentence then you give it credit for - it isn't a random sample, it's already skewed to only include those 17 year olds for whom the relative age effect wasn't enough to stop them from performing well in school. One of my freshman year roommates at Harvard, and probably the smartest of us all academically, was 16, having skipped grades - obviously these people exist, and obviously people like parents, teachers, and guidance counselers are already steering students towards applying to schools that match their ability levels at the start of senior year of high school. 

The comparison I care about (and that age cohort studies are trying to approximate) for this question is how much more those 17 year olds would have learned, over the course of all their years of schooling, if they'd been a little older at the very beginning. To look at that, I'd want to see, for example, whether the proportion of students applying to Oxford, or becoming valedictorian of their high school class, are more likely to be older or younger.

An excellent argument -- though obviously not conclusive.

The first alternative idea that comes to my mind is that you could just be teaching the oldest kids in those classes to see themselves as high status and that you could get the same effect through any other intervention that encourages particular kids to see themselves as better than other kids.

Certainly this seems to be an example of school doing something.

This is self-validating on the part of academia. To the extent it demonstrates value for school (which doesn't seem proven to me), it's valuable in getting you more and better school. Much of the world believes that better school translates into real world value, and that's self-fulfilling (via better job offers), but it doesn't demonstrate people learn more at school than they would through other uses of their time.

Note this also has been observed in other spheres - E.G. CEOs of S&P 500 companies:

I would strongly expect CEO-ship of a S&P 500 company to be causally downstream of a top-5 business school, a top-20 college, and the kind of high-status professional+social network you get from "more and better school".

I hope it's an exaggeration to say that "school is mostly a waste of time" is a relatively common view here. I would propose that there are a minority of people who don't benefit much from school — perhaps just a small minority — and that this minority is overrepresented here.

My feeling is that many of the people which did not benefit tend to "generalise from one example" and assume that's true for most kids. Actually, I (despite being generally pro-schooling) would say something stronger than you: there is a minority of people who are actually harmed by school compared to a reasonable counterfactual (e.g. home-schooling for some). Plus, many kids can see easily where the system is failing them, less easily where it's working.

If we take "relative" to mean "relative to the general population", then I think "relatively common" is a fair description - even if it is a small minority

Good point. My model of "most school interventions don't do much and large effect sizes will probably go way down upon replication" takes a hit if the relative age effect is true. I hadn't connected these two ideas before.

C) Given how difficult it is to get into Oxford, even 1 years worth of maturity can make a significant difference.

Why are we discounting this hypothesis? It seems the most plausible to me. IIRC IQ increases until you are 25 or so, and it's probably a sort of s-curve, meaning that there's probably at least a couple points of IQ gain over the course of year 17... why wouldn't that be enough to make a 20% difference?

I'm not discounting it, just saying if we do, it would imply X. I give a couple of suggestions as to how we could resolve whether that's the case or not at the end.

I wonder if the relative age effect works in the other direction sometimes. I was born at the very end of August and was usually the youngest kid in my classes. Growing up, I thought of this as an advantage -- it meant I was in more challenging classes. If I had been born just a few days later, maybe I would have been even more bored at school than I was.

Apparently at least one study found an opposite effect within the cohorts of students at a single university:

This study, though reminds me of the SSC post on Searching for One Sided Tradeoffs. If everyone admitted to a typical school is pre-selected for a certain overall level of ability, as visible in their applications, then the younger students in a class may have greater ability simply because their official, legible results underestimate them. Maybe?

I wonder, if we had year-round schooling instead of long summer breaks, and less once-a-year standardized testing, if schools would have started staggering class start dates? My school district had about a dozen classes within each grade, so what if instead of 12 new classes each September they had one a month, or 3 a season?

I know that some schools in the US have historically had start dates in both January and September.