Comment by gnicholas | Hacker News Reader

Comment by gnicholas | Hacker News Reader

One of the most-cited stats regarding literacy is: a student who is not reading at grade level by the end of 3rd grade is 300% as likely to flunk out of school before finishing 12th grade.

As someone who works in literacy, I have seen this cited many times and referenced it several times myself. It was only this year, when I was reading Scout Mindset [1] that I realized this statistic doesn't say what so many people think it does.

People cite this stat to prove how important literacy is. After all, if a student can't read, then he is not going to be able to pass classes in a wide variety of subjects. This notion of causality is captured in the similarly-common saying "first you learn to read, then you read to learn".

But the stat doesn't prove causality at all. And when you step back and think about it, there is likely to be a very large overlap between people who have low reading ability and people who are more likely to drop out of school. This is perhaps obvious to people who do not work in literacy, where we tend to overestimate its importance.

For me, this was a perfect example of confirmation bias, where people — even people who are generally smart and well-meaning — credulously accept (or even tout) statistics without first analyzing them. Had we seen a stat that indicated that reading ability didn't affect educational outcomes, we surely would have dug in and tried to figure out what was wrong with the calculation. But this one sounded right, so now it's part of the accepted wisdom in the world of literacy. This statistic is cited wherever someone is asking for more money to be poured into early literacy programs, as if solving literacy will have this 300% multiplier effect.

I thought about posting about this on LinkedIn, and urging others to try to think about what beliefs are common in their industry, but have not been thoroughly examined. It would be great if we could call out some of these common-but-untrue beliefs, which may lead to misallocation of resources.

1: https://www.amazon.com/Scout-Mindset-Perils-Defensive-Thinki...

Correct. A lack of basic literacy by third grade is an indicator something else is wrong. Children will typically just learn to read with enough exposure to language by this age. Often this either indicates that a child has a learning disability or basic needs at home are not being met.

You can pour all the money you want into literacy, and that is fine and a good thing, but it does not address the needs of a child who is not progressing because no one talks to them at home or there are other basic needs not being met.

> Children will typically just learn to read with enough exposure to language by this age.

Is this universally true? Or is it only true for certain socioeconomic classes?

One of the biggest problems in education is the fact that during the summer vacation some students improve while others degrade. The differential is almost completely predictable by socioeconomic class.

I would assume that sufficient exposure is what varries with class. The way you're phrasing that seems to imply that some classes are just instrinsicly not as smart, which i don't think is true.

I didn't get that implication from bsder at all.

Seems to me their post implies all students learn during term time, showing there's nothing wrong with any of them. And anything outside of term time is extracurricular; everyone knows more extracurricular activities need more parental time and money.

I took the exact opposite reading: that socioeconomic conditions (rather than inherent abilities) were the first-order difference.

How could this not be true?

The only way for the poor to not be intrinsically less smart than the non-poor would be for there to be little-to-no impact of IQ on job choice.

But of course IQ affects job choice, which affects income, which determines class.

Consider the most common jobs in the upper income decile--physicians, managers, executives, and lawyers. Generally speaking, these jobs are inaccessible to people with average IQs, let alone low IQs. So of course these high-paying jobs are far more likely to be filled by folks with high IQs.

Now, consider the most common jobs in the lowest decile--nursing aides, cashiers, cooks, housekeepers. Jobs like this would be, generally speaking, boring for people with average IQs and absolute torture for folks with high IQs (who as a rule crave novelty and complexity to a greater degree than most). So of course these low-paying jobs are more likely to be filled by folks with low IQs.

This doesn't even get into the fact that ~15% of the population have IQs below 85. These folks will struggle to hold down any job, and are far more likely to sort into the ranks of the poor than into the middle or upper classes.

Ignoring objective truths is always dangerous, even if they are uncomfortable.

You are talking about something very different from OP. OP is consistent with "class is a predictor of availability of oportunities to increase the 'IQ' of a child, as 'IQ' can be more nurture than nature". You are focusing on the effects of low IQ without considering interventions to increase IQ. And our two posts completely disregard how bad of a measure IQ generally is in practice.

There are essentially no interventions that can improve IQ, only interventions that reduce it.

This seems like an incredibly strong claim presented without evidence (unless you believe in fairly debunked "skills are fixed at birth" nature-vs-nurture ideas).

The vast majority of pedagogical literature and science from the last couple of decades is talking about the importance of instilling a "growth mindset", depicting a whole zoo of interventions shown to improve cognitive skills in students, including as measured by SAT scores and IQ tests (even if they are imperfect measurement tools).

There is a relatively easy way to disprove your claim if you believe in IQ tests (an assumption I am making). You can improve your test score by practicing on similar tests (or even unrelated things like crosswords). Trivial exercises can improve your IQ test score or any other cognitive measure.

The data implies exactly the opposite.

A lot of "underperforming" students actually close some of the gap during the school year. The problem is that during the summer break, the gap reopens even wider for those from lower socioeconomic classes.

This is one of the primary arguments for "year round" school.

It could also be a disability that is not yet diagnosed or is otherwise untreatable by our current level of medical technology.

As a child who self-taught himself to read for the most part, I doubt this. I had to sit through many "see spot run" style lessons in kindergarten, 1st, and 2nd grade, while being perplexed with how many of my classmates struggled. And even my "self-taught" reading involved a fair bit of coaching from my parents. I remember alphabet, sound combination and grammar exercises around 4 years old before it clicked.

Plenty of kids don't even know the alphabet when they start school. It is interesting to compare those parents with my parents who felt that kids who couldn't full-on read and write by about 5 were being failed. Whether a kid will be a competent reader is probably determined by the home environment before school even starts.

Well it's statistically likely that the students who don't do well on reading comprehension might struggle with other things too, but it's not like reading is not a foundational skill for being able to do well later in school.

One outlying example is dyslexics - one can be arbitrarily bright, but have enormous difficulty reading - I've personally known extremely intelligent people who struggled finishing high school because of their poor reading comprehension - this is something that ought to be remedied.

I'd say it's far more likely these students struggle academically - but the problem is that the further they fall behind, the more of an uphill battle for them to keep up with the class, let's say X takes 50% more time to solve a text-based math problem than his peers, but due to his poor reading comprehension he takes again 50% more time just to understand the problem - which might make the difference between failing and passing.

And as lacking as public schooling is, I'd expect them to do their damnedest that people leave school with at least basic reading comprehension after 12 years of education.

This is unfortunately just one of many examples of generally widespread malaise of conflating correlation with causality.

It is especially common in certain "sciences" which typically involve a lot of data collection. Followed by attempts to crunch it up into some kind of popular and far reaching conclusions, intended to fool the examiners into awarding a PhD.

Speaking as a person with a PhD, I can say from personal experience that you don't need popular or far reaching conclusions to earn one. Most of the time the committee wants to see you've done the work and have, in some small way (sometimes surprisingly small) pushed the boundaries of knowledge out a bit.

The kind of Academic Inflation you're talking about is almost entirely the product of PIs who need to publish, publish, publish to get tenure and funding. The vast majority of people who are awarded PhDs understand they have almost no chance of even landing a tenure track position and consequently have little incentive to fluff up their work.

300X is 30000%. Which do you mean? (300X or 300%)

Sorry — meant to put 300% in the first time; I only got it right in my second mention. Fixed in original; thanks for flagging.

Distinguishing correlation and causation is the constant challenge and peril of all social science research, where mechanisms are complex and measurements indirect. There are a few methods which do offer some insight, the two most effective I'm aware of being identical-twin studies and qualification-test cut-off divisions.

In twin studies, the goal is to study twins (identical genetic inheritence) either from the same household (similar socioeconomic environment) or separate ones (often foster-parent settings or other forms of adoption), and comparing life trajectories, educational attainment, and employment history. I'm not specifically aware of education-related studies of this form, though I'd strongly recommend literature search (Google Scholar, etc.) on keywords such as "twin studies education".

On test-cutoff studies, there was one I ran across recently in which individuals who tested just at the cutoff of the qualification threshold for advanced academy entrance. That is, students scoring above the threshold attended the schools, those below did not. The findings were of comparable subsequent lifetime attainments (employment, income), suggesting that innate ability was a stronger predictor than educational history. I can't find the study presently, though I'm pretty certain it addressed the Specialised High Schools Admission Test (SHSAT):

https://en.wikipedia.org/wiki/Specialized_High_Schools_Admis...

Far more studies look at basic demographic or geographic data. To a large extent this constitutes an availability heuristic: the data are far more abundant and more readily available. How predictive they are on a causality basis is harder to determine. One example (turned up whilst searching for the entrance-exam study):

"Money, Race and Success: How Your School District Compares" (2016) https://www.nytimes.com/interactive/2016/04/29/upshot/money-...

Another school of thought (so to speak) is on what education's impact socially is, at least along economic-performance measures. (There are of course other legitimate outcomes of education, including its roles in culture and democratic institutions.) Such viewpoints often come from the right, though may also be found among more liberal economists. One such, Ha-Joon Chang addresses this in chapter 17 of 23 Things They Don't Tell You About Capitalism, "More education itself is not going to make a country richer" (2010). Dr. Ha-Joon notes that:

1. Many poor countries have high educational attainment.

2. Many rich countries have comparatively low attainment.

3. Numerous countries (particularly in east / southeast Asia) have made remarkable economic progress without especially high educational attainment.

4. Mechanisation has a far greater impact on per-worker productivity than education.

5. Businesses generally prefer de-skilling work in order to make labour more substitutable. This of course reduces the benefits of education.

Additional points (Ha-Joon does not address these directly) are that infrastructure counts for a great deal, and "brain drain" --- economic migration by the more educated --- can siphon off a country's most educated citizens, essentially subsidising the economies of other countries.

See: https://www.worldcat.org/title/twenty-three-things-they-dont...

Among Ha-Joon's more interesting citations:

L. Pritchett, "Where has all the education gone?", The World Bank Economic Review, 2001, vol. 13, no. 3. https://documents.worldbank.org/curated/en/85982146818027378... (PDF)

Harry Braverman, Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century (1974) https://nyupress.org/9780853459408/labor-and-monopoly-capita...

Stephen Marglin, "What do Bosses Do?", published in two parts in The Review of Radical Political Economy, 1974, 1975. https://scholar.harvard.edu/marglin/publications/what-do-bos... https://scholar.harvard.edu/marglin/publications/what-do-bos...

> The findings were of comparable subsequent lifetime attainments (employment, income), suggesting that innate ability was a stronger predictor than educational history. I can't find the study presently, though I'm pretty certain it addressed the Specialised High Schools Admission Test (SHSAT)

Isn't the SHSAT a test that measures a mix of innate ability and effort expended learning (both in general, and prepping for the SHSAT)?

That's not what the study is assessing.

Rather, it's looking at individuals with all-but-identical scores on the admissions exam, but on either side of the admissions line, one of whom attended a specialised high school, and one of whom did not. The difference in scores is well within the measurement error of the exam itself.

(Caveat: If I'm remembering the study correctly.)

That is: even if test prep is a factor in score, the scores considered were not differentiated, and those meeting or failing the cut-off could have equivalent access to test prep. I don't know whether or not such prep was specifically considered, and would have to track down the study to confirm or deny.

> But the stat doesn't prove causality at all

Does it prove that there is no causality?

Reply on news.ycombinator.com