It’s so far unclear whether extra school in middle adolescence benefits or harms those affected—some studies find a benefit to cognitive or non-cognitive skills, others don’t. Some find benefits to earnings. These are all affected by the usual problems: issues with identification, lack of controls, fade-out, and publication bias. But the evidence on earlier schooling is much less divided—and it almost universally finds that going to school too early stunts child development.
What’s more, “too early” is well within the range of when we currently send kids to school. In Britain kids go to school at four or five. But a Danish study (pdf) found that even at around age seven starting school later led to less crime and delinquency through life. This study—and most of the others I present—used a “quasi-random” study design.
For example, the authors might use arbitrary cutoffs. If someone is born on 31st August and another person on 1st September it’s likely that a jump in some variable between them that isn’t seen between 30th & 31st August birthdays, or between 1st & 2nd September birthdays, is down to the effects of the cutoff.