That isn't to say that a policy of requiring helmets is good on net (because people may ride less), but in any accident that you hit your head, you would greatly benefit from wean a helmet.
This is junk science. The authors simply did the ANSI drop-test in a lab test. In the real-world, 99% of bike crashes with death/severe injury are the result of car-crashes -- which the ANSI drop-test does not model correctly at all.
They are claiming 90% risk reduction based on a laboratory model. That model obviously does not track with reality (show me any country where bike helmets reduced death/injury by that amount). If a model does not correlate with the real world, then it is by definition...junk.
That isn't to say that a policy of requiring helmets is good on net (because people may ride less), but in any accident that you hit your head, you would greatly benefit from wean a helmet.