For ad serving, tagging, tracking reasons, duplicating is a bad idea.
Also, two many things can go wrong when copying a new Campaign, and you create new elements (with no history) that create test bias.
Over the years, I've seen many bugs in the online gui, and desktop tool, that affect copying - care should be used here.
Once they're running side-by-side, normal variations in each, can cause additional bias, especially if using automated bidding tactics (like CPA, ROAS, etc).
The budget pacing of each, not being identical, is another example of why duplicate is the less statistically sound method of testing.
If these are high CPC campaigns, or low daily click volumes, oh joy, you're no longer testing the variable under test, but many other parameters that cascade under semi-automation, even ad rotation.
Most good PPC Managers are OCD about being organized.
When you delete a campaign, not everything gets deleted, depending on how you delete it.
This can also cause issues with analysis.
Plus experiments has random induction, while side-by-side is two campaigns competing, and very small natural variations can unbalance your testing.
Take their hand, turn it palm down, and sharply smack the back of their hand, at the same time say "no".
Try to do this as soon after they duplicate a campaign, so they make a good cause and effect connection.
After 30 or more minutes have passed, take them for a good walk, throw their ball, pet them, re-assure them that if they listen, they'll still be loved.
