That's probably not unreasonable. An hour doing homework is an hour not doing basketball, for example.
They even point this out as a specific point in the abstract: You can't compare kids doing enrichment activities to kids not doing enrichment activities while controlling for every other variable to be the same, because it is manifestly true that time spent on enrichment activities can not be spent on other activities, such as sleep or socializing, which also have benefits.
Please double-check my reading comprehension of your comment. Are you saying that it is not possible to analyze the benefits of enrichment activities as opposed to what normal kids do, because there is no way to have a group of kids who do exactly what normal kids do + enrichment activities, because taking the time to do enrichment activities necessarily knocks out at least one activity normal kids do, thereby making control impossible?
When I was in high school I volunteered as a counselor at a summer program affiliated with my dads job in an inner city environment. The stuff we did help keep those kids, 2-4 grades younger than me, away from (or perhaps delayed) getting recruited into the drug trade.
Asking bankers and economists to measure educational and social outcomes sounds pretty dumb. Enrichment isn’t designed to produce ROI, it’s to enrich your experience as a human being. We don’t have the income mobility we once did, so folks are likely to land in whatever income strata they came from. Using income as a metric to measure whether enrichment activities are worthwhile or not is the same as evaluating the quality of apples based on the price of oranges.
At the same time kids at that age don’t know what their best interests are and could do too much of an activity with diminishing returns while not developing other skills.
Read slide 17. The key assumptions are that: the confounder present at bunching point is representative of the confounder present everywhere, and the confounder affects outcome linearly. The second assumption is what strikes me as too strong, and you can read the paper Appendix C for how they address it. The reasoning is quite weak and they even phrase it as such: "Our main empirical findings therefore do not seem to be an artifact of this linearity assumption". Overall I would take this paper with a grain of salt, from a cognitive science point of view it just makes no sense that doing more reading would have no cognitive benefits (in fact, we know the opposite to be true.)
Also, the paper hasn't even been reviewed yet. We should revisit it after it has been peer reviewed.