Prior to the mass virtual learning experiments of 2020-21, evidence of the relative merits of various 48 states’ K-12 schooling types indicated as follows:
- Public schools spanned the quality spectrum from nearly worthless, to the best for many students.
- Charter schools also spanned the spectrum, but because the worst ones could be shut down, you didn’t have so many persistently failing schools (according to standardized testing).
- Private schools were harder to assess, but we think they didn’t have so many worthless schools? But hard to assess because they often aren’t required to take standardized testing.
- Remote learning (often in the form of credit recovery) usually had poor outcomes.
- Home schooling was effective, but since so few parents were able to choose this option, confounders like parental involvement and aptitude were very difficult or impossible to isolate, therefore making any conclusions about a scale up difficult or impossible to draw.
The situation of 2020-2021 led to the following outcomes:
- Virtual learning gave inferior and in some cases disastrous results for the majority of students, relative to their previous educational setting (usually public school). We think the most disastrous outcomes were for the children already facing difficult parental situations. Some children did about as well in virtual learning, and a significant, although we think small, minority clearly did better in a virtual setting.
- Some private schools stayed open for in-person learning while public ones did not. Given the previous finding about the relative impact of virtual learning, the open private school options were almost guaranteed to improve student achievement over the virtual public schools.
- Many parents were commandeered into assisting with their children’s virtual education, or went to the home schooling track. Assisting with virtual education is clearly not the same as the conventional home schooling approach; it was something that was done because it was needed, not because it was desired. I don’t think the evidence base at the end of 2021 is clear enough to draw any firm conclusions about the relative performance of the momentarily homeschooled children.
- Hybrid models where one teacher attempts to teach in person and virtually were clearly more work, and (we think) don’t seem to have been as effective as splitting into all-virtual or all-in-person.
To this understanding we need to re-iterate the fundamental principles driving the organization of education in the 48 states:
- The government forces all parents to ensure their children are educated.
- The government taxes everyone to pay for public and charter schools, whether or not they are any good, and regardless of the curriculum.
- The requirements of schools are driven by national, provincial, and local requirements, so individual parents in a given area without the financial means to home school or private school their children don’t get a choice about the method of education unless they move (which may not be financially feasible anyway). This also includes items such as required school vaccinations, dress and grooming codes, etc.
- In recent years, the per-pupil cost steadily has been increasing, particularly in heavily populated coastal cities, so that the average cost may be on the order of $15K per student. (the “Massachusetts model”) Increasing spending to this level seems to achieve significant marginal learning improvement.
Previously, the success of charter schools such as KIPP and Success Academy had demonstrated that public schools were not necessarily the most effective at teaching children. This indicated that at least choosing between schooling providers was the optimal policy. Ongoing reports of social promotion and suboptimal tracking further reinforced this conclusion.
With the additional evidence gained from the situation of 2020-2021, we also see that the in-person group educational setting is not optimal for a significant minority of students. Moreover, if the in-person and virtual settings don’t make a clearly detectable difference for another significant minority of students, there’s no reason to incur the expenses of transportation, physical plant, etc.
The 2020-2021 situation added to the previous disagreements about tracking pace, school schedules, which schools were tracked appropriately, teaching methods and curriculum; to these they added in-person/hybrid/virtual, with their attendant gross effects on student achievement and required parental involvement. Individual educators and interested scholars disagree on the relative influence of many of these factors, but the general consensus is that different students, especially in pre-K->3rd grade education, respond better to one set of these conditions than another.
In order to achieve net economic gains for the majority of the population (i.e. the population that does not have two upper middle class+ working parents) by sending their children to schools, per-pupil spending associated with tutoring, must be reduced by increasing class sizes. This limits the number of slices into which you can place students. If you have an elementary school with 500 children across 5 grades, and you need an average pupil-teacher ratio of 15:1 to make the economics work, you have ~7 different possible tracks. The possible combinations of (2) paces, (2) schedules, (2) methods already amount to 8, requiring the student balance between these options at this school to be almost perfectly distributed to even approximate the optimal. If you then add a virtual option, now you are at 16. And, with the exception of virtual vs. in-person, no one believes those are the only possible options.
To the above perfectly executed options, we must then add the possibility that the student body is unbalanced, so that you can only allocate 2 of your 7 tracks to anything other than the most intense or the slowest paced learning. In that situation, you have no chance to teach all the students optimally. On top of that, you put IEP learners in the classroom and tailor the lessons for them, don’t teach on the tracks you said you would, the school facility is garbage, the school is riddled with gangs and bullying…
Finally we must recall that the increased cost of per-pupil education means home schooling and small group education is always going to be a financial option. For these tranches of students that don’t learn optimally in the conventional schooling tracks available in the area, such instruction therefore is cost-competitive, and helps relieve the other schools of the inefficiency of teaching non-full classrooms.
Therefore the usual response of “make the public schools better via parental involvement/funding/public praise/sending your kids there” as a means to improve academic achievement, is not even possible in a well-executed environment. Moreover, the usual complaint about “you’re taking away money from the public schools by pulling your kids out” is incorrect from an execution perspective (vs. the funding formulas imposed by the government that people elect), it actually is neutral or cost-saving depending on the situation.
But there’s more to the story than just giving parents schooling options. We also need to force mobility in order to determine which of these settings is optimal for an actual child.
We also know from the 2020-2021 situation that different children responded differently to the virtual schooling. From prior experience, we had the anecdotal evidence that some home schooled kids wanted to attend high school when they got older, and some didn’t. Likewise, sometimes parents pulled their kids from the public school to give them the individualized attention they needed.
The optimal means of determining the schooling approach that maximizes individual student achievement is not clear. It will take decades of experimentation to have more confidence; however, here are the principles I suggest (with good reason) that such experimentation needs to incorporate:
- In order to compare the relative benefits of each schooling approach, we must have standardized tests. Practically speaking, they can’t be any less frequent than quarterly, because you would not want a child to go off the rails for more than 3 months. There’s a natural experiment we already run on learning loss called “summer” and the effects of that loss are well understood.
- If a student is achieving and progressing on the most rigorous (usually labeled as “gifted/talented”) track, there’s no need to change the school setting.
- If the student wants to stay on the most rigorous track and is getting by, let them stay where they want, even if they are struggling.
- If a student is on the lower tracks, they need to be rotated into alternative schooling settings, to see if they perform better in those. We would guess maybe once a year for a quarter? but that’s something we will have to figure out with experimentation.
Fortunately, changing the educational setting isn’t a life-altering decision (e.g. summer break), so we can try different things for each child and not be too concerned about making mistakes as we work towards identifying optimal approaches.