Ignoring the ethics factor, I disagree this would be a bad experiment. It would measure the ability to crack a password intended to secure xx dollars.
IMO it's still horrible experiment, because
1. Sometimes we don't know much BTC or USD stored.
2. Strength of the password mostly depends on the user, not amount of the BTC or USD stored.
3. It's also possible those hashes belong to fake wallet. See
Re: Fake walle.dat can trick bitcoin core client?