In my world, we rely on practical tests of functionality *with the actual users* rather than theoretical measures of effectiveness. It seems the CDC agrees with me on this one.
As to one of your other points:
"The bottom line is that a single act of intercourse between a young couple has on average a one in 20 chance of pregnancy – this assumes the opportunity presented itself on a random day, as these things tend do when you are young."
So the answer is based on how often you fuck. The CDC numbers do not map precisely to data of this sort, however.
Last point: "measuring the effectiveness of encouraging people not to fuck" sounds like the craptastic questionnaire-based research I saw at (not to pick on them) the Lehigh psych department in the early 2000s. People lie, and they can't simulate the paths not taken effectively. There would literally be no way to know if someone decided to not have sexual intercourse based on an abstinence campaign.
Besides, my kind of abstinence campaign would be "Blowjobs for Everyone" or "Real lovers wank each other".