It's very long but almost entirely focuses on what ratings or achievements the students are reaching in schools. Mentions compulsory attendance as a passing note but immediately dismisses the relevance of it later because some compulsory schooling systems have better results. It's all about results. Not the sitting and writing and not being able to speak or move without permission. Not the bullying encouraged by crowded corridors. Not the ratings teaching people to fight each other. Etc...Three, ranked by how structural they are:
**1. The batch-processing architecture itself.** The model is Prussian *Volksschule* — age-cohorts moving in lockstep at a fixed pace, exported to the US by Horace Mann after his 1843 tour of Prussia, and to most of Europe directly. It optimizes for administrative legibility, not learning. The consequence is mathematically unavoidable: teach to the median and you bore the top third and lose the bottom third. Bloom's 1984 "2 sigma" work put numbers on the gap between lockstep instruction and mastery-based tutoring, and while the two-sigma figure itself doesn't replicate cleanly, the direction is not seriously contested. Nothing else on this list matters as much, because every other reform gets poured into this container.
**2. Goodhart capture via high-stakes testing.** Once a metric becomes the target it stops measuring the thing. Campbell wrote this down in 1976; NCLB (2001) then ran the experiment at national scale in the US and produced exactly the predicted output — narrowed curriculum, test-prep displacing instruction, outright cheating scandals (Atlanta, 2009). The European versions are milder but the same mechanism runs through Abitur/maturita/A-level optimization. Downstream is credential inflation: Spence's 1973 signaling model, pushed hard by Caplan in *The Case Against Education* (2018), argues a large share of schooling's wage return is sorting rather than skill-building. Contested, but the weak version — that credentials inflate faster than content — is hard to argue against.
**3. No exit, and therefore no feedback.** Compulsory attendance plus funding tied to residence plus political rather than user-driven accountability means the system's error-correction loop runs through ministries and elections, not through the people being schooled. Bad practice persists for decades because nothing forces it out.
The honest complication on that third one, since you'd catch it anyway: the empirical case that *choice* fixes it is much weaker than the theoretical case. Louisiana's voucher program (2016 evaluation) produced significantly *negative* achievement effects; CREDO's charter studies show enormous variance — some sectors strong, many at or below district baselines. Meanwhile Finland is centralized, non-competitive, has no standardized testing regime to speak of, and outperforms — mainly by making teaching a selective, high-status profession. Which suggests teacher selection quality dominates governance structure. Decentralization is a defensible principle here, but "monopoly causes the bad outcomes" doesn't survive contact with the data as cleanly as the theory predicts.
So the value of big personal journalistic reports is still there.
