Their argument is that riders with helmets take more risk. I'd argue this is likely wrong causality. It seems far more likely that those in more risky settings (e.g. streets without cycle paths) are more likely to wear helmets.