How much time should children be forced to spend in school?
More schooling but not much more learning
These results suggest that attending more classes causes students to learn more and thus get higher grades. But there are some problems. First, often class attendance itself is used to grade students, making the relationship trivial. Second, the relationship is strongly confounded since class non-attendance is related to other, unobserved student characteristics. What we want then, is something more causally informative about time spent in class and learning. Fortunately, we have quite a few such studies.
School duration extensions
The first type of study looks at expansions of school duration as a natural experiment. Cremieux tweeted a such study from Mexico recently:
Cabrera-Hernández, F., Padilla-Romo, M., & Peluffo, C. (2023). Full-time schools and educational trajectories: Evidence from high-stakes exams. Economics of Education Review, 96, 102443.
This paper estimates the effects of extending the school day during elementary school on students’ educational outcomes later in life. The analysis takes place in the context of a large-scale program introduced in 2007 that extended the school day from 4.5 to 8 h in Mexico City’s metropolitan area. The identification strategy leverages cohort-by-cohort variation in full-time enrollment in elementary schools. The results indicate that full-time elementary schools have positive and long-lasting effects on students’ performance, increasing high-stakes high school admission test scores by 4.8 percent of a standard deviation. The effects are larger for females than for males. The difference in the effects between males and females of 2.1 percent of a standard deviation represents 16% of the gender gap in the high school admission exam. Moreover, full-time schooling decreases the probability of delays in schooling completion.
So a whopping 3.5 hour per day increase lead to a tiny 0.048 d increase, equivalent to 0.72 IQ.
In Denmark, there was a major school reform in 2014 which, among other things, extended the amount of time pupils had to stay in school. The government commissioned a report on this, and it found that this lead to no changes in test scores in math and Danish:
The red line is public schools, the blue line are private schools. There is no change over time for either set.
It's actually difficult to figure out how much longer they made the school because the various sources about the reform don't say how long school days were prior to the reform, but asking GPT4, we get these values:
After reform (Wikipedia): up to 3rd grade 30 hours/week, 4-6th grades 33 hours/week, 7-9th grades, 35 hours/week.
Before reform (GPT4): "For the lower grades (0-3rd grade), it was typically around 20-24 hours per week, and for the oldest grades (7th-9th), it was about 25-28 hours per week."
So the extension was roughly 8 hours/week, or 1.6 hours/day. About half as drastic as the Mexican change, but it produced no detectable effects at all. Of course, one could cope and argue for sleeper effects that might show up later when teachers have gotten more used to the other parts of the reform, but don't hold your breath. For more background on this Danish school reform, you can read Helmuth Nyborg's and my article in Danish newspaper from 2020 (in Danish of course).
Sweden also recently tried such an approach with similar results:
The SACO association, Sweden's Engineers, has analyzed the state of knowledge in mathematics in primary school, upper secondary school and adult education. The report compares, among other things, the grades for the students who received an additional 330 hours of mathematics in primary school with previous cohorts. Despite 37 percent more math hours, the overall grade average among students in grade six is falling. At the same time, the association between higher grades and students with parents with higher education has strengthened during the time period examined. In addition, the percentage of students failing has risen.
I was able to find the Swedish rapport (Att räkna eller räknas bort), but it doesn't seem to add much of interest beyond the summary above. The decline in math grades is probably related to more foreigners in the younger cohorts, so the reform itself didn't do anything.
After writing these studies, I found that there a "Evidence for Learning" center, which has an online meta-analysis of these kinds of studies. I didn't look at the details of all 70+ studies they list, but almost all the largest ones show results like the above:
The Lee 2006 study is a study of Kindergarden pupils, so my guess is that it has teaching to the test issues, and of course regression towards the mean will apply.
Finally, the COVID-19 school closures lead to less time in school. Of course, there are other factors too, but we might consider this a large exogenous shock. There is a meta-analysis of the effect size of this:
To what extent has the learning progress of school-aged children slowed down during the COVID-19 pandemic? A growing number of studies address this question, but findings vary depending on context. Here we conduct a pre-registered systematic review, quality appraisal and meta-analysis of 42 studies across 15 countries to assess the magnitude of learning deficits during the pandemic. We find a substantial overall learning deficit (Cohen’s d = −0.14, 95% confidence interval −0.17 to −0.10), which arose early in the pandemic and persists over time. Learning deficits are particularly large among children from low socio-economic backgrounds. They are also larger in maths than in reading and in middle-income countries relative to high-income countries. There is a lack of evidence on learning progress during the pandemic in low-income countries. Future research should address this evidence gap and avoid the common risks of bias that we identify.
In other words, total disruption for weeks and sometimes months lead to a decline of only 0.14 d. This estimate is probably too high as COVID-19 also had all sorts of other effects that might affect learning, and because this meta-analysis includes various dubious studies from outside the developed world which show larger effect sizes (as they always do).
These kinds of findings about school duration in particular should be seen in the light of the large number of randomized controlled trials in education, that I received a meta-analysis of 2 years ago. The average effect size of these is very close to 0 too, a mere 0.06 d. Many of them involve more hours of teaching, so it is not so surprising that the large scale studies of school reforms and COVID-19 closures come to a similar conclusion: the number of school hours are not so important; what children bring with them from home is very important.
Of course, learning is just one thing of interest. Really, people are interested in increasing how much students learn because they know that learning
leads to statistically predicts other beneficial things later on life, such as income, longer lifespan and so on. So how about skipping the mediator and just looking at those outcomes? Greg Clark & Neil Cummins did this already using 3 large UK reforms that lengthened the compulsory schooling:
Clark, G., & Cummins, N. (2020). Does Education Matter? Tests from Extensions of Compulsory Schooling in England and Wales 1919-22, 1947, and 1972.
Schooling and social outcomes correlate strongly. But are these connections causal? Previous papers for England using compulsory schooling to identify causal effects have produced conflicting results. Some found significant effects of schooling on adult longevity and on earnings, others found no effects. Here we measure the consequence of extending compulsory schooling in England to ages 14, 15 and 16 in the years 1919-22, 1947 and 1972. From administrative data these increases in compulsory schooling added 0.43, 0.60 and 0.43 years of education to the affected cohorts. We estimate the effects of these increases in schooling for each cohort on measures of adult longevity, on dwelling values in 1999 (an index of lifetime incomes), and on the the social characteristics of the places where the affected cohorts died. Since we have access to all the vital registration records, and a nearly complete sample of the 1999 electoral register, we find with high precision that all the schooling extensions failed to increase adult longevity (as had been found previously for the 1947 and 1972 extensions), dwelling values, or the social status of the communities people die in. Compulsory schooling ages 14-16 had no effect, at the cohort level, on social outcomes in England.
The results for longevity look like this:
Children who were forced to stay one year longer in school didn't live any longer, and neither did they end up with more expensive houses:
These results are based on very, very large public UK datasets. So what good did these school extensions do? Their effects are literally undetectable when analyzed the most obvious way.
Going back to the initial tweet from Timothy Bates, we are now in a position to say that these results should certainly not be interpreted as showing that one can easily boost learning by merely adding more schooling. In fact, there is no easy way to boost learning. Everything has been tried. There are more resources on the internet to learn from than ever before in human history, yet general knowledge hasn't increased, and children and adults aren't any better at arithmetic. Clearly, opportunity to learn isn't an important variable in explaining differences between people. And dare I say, it never was?
From a cost-benefit perspective, more forced school for children means more teacher salaries to be paid. Since the effects of this on children's actual learning seem to be somewhere between nonexistent and minor, I suggest that this isn't worth the price. School is a kind of child slavery. We threaten parents with fines if their children don't show up to work camp where they are forced to listen to boring and often irrelevant things for years on end. When the children are put to work, they aren't making anything actually useful, but are doing mock work, which after a quick assessment will be thrown away. Sometimes they are obliged to do more of this at home too. They don't get paid for this labor either. In fact, the parents have to pay for it. Clearly, such a slavery policy can only be justified on utilitarian grounds if the later benefits of schooling are large. The research above, and all the other evidence reviewed by Bryan Caplan in his The case against education, clearly show that schooling isn't that useful by itself. Because of the high costs and low benefits, the cost-benefit analysis must be a very negative. The question is only how much of mandated schooling we should abolish.
Of course, some people will bite the bullet and say: the primary function of schooling is daycare. Great! That means we can easily lower the price. Looking after kids and making sure they don't kill themselves or each other is an easy job. It certainly doesn't require a 4 year college degree. After all, parents already provide daycare themselves when schools are closed without any such education.
Since people ask, I should say that I am not advocating literally no schooling, usually called unschooling. I'm a moderate, maybe we can start by cutting forced schooling by 50% and see how things go. This would merely take us back a few decades in time insofar as policy is concerned.