I would actually argue it does improve security since you're spending more time on the people who have a higher probability of being a terrorist versus those that don't.

What doesn't improve security is just random screening where you're frisking 4-year olds and senior citizens and letting the adult males walk on by.

In that case, the odds of a Yian being a terrorist would increase, thereby increasing the odds of a Yian getting searched. If we assume that the odds of being a terrorist varies based on demographics, then this would catch more terrorists than random searches.

It reminds me of the N-armed bandit problem: use random searches some small percentage of the time, and use a distribution based on prior samples to identify subsequent searches. (It's not exactly the same, but the same technique applies.)