Microsoft did that research. Well, they didn't propose to kill anybody's mother but the test participants used their real bank credentials and Microsoft tested different behaviours in IE to see what would deter users from giving these credentials to a bogus site having accepted a task to log in and perform some basic operation.

Nothing.

Nothing deterred the users. Warning dialogs were clicked past, obvious problems or mismatched information was ignored. The only way to stop users from giving their credentials to bad guys was what I call Brick Wall UX. The browser has to stubbornly refuse to let you do it. Unable to complete their task the user at last gives up.

This is a teachable moment. Your users are probably not going to be smarter, better informed or more cautious at least on average than in this test.