During trials, the doctor and the patient don't know that the patient is getting a placebo (it's called "double blind"). If people know that what they are getting is a placebo, I don't think it works as well. So the only way to get an effective placebo is to lie to the patient and tell them it's a good drug.
But there is a long history of abuses in medicine (Nuremburg, Tuskegee syphilis experiment, etc.), so "informed consent" is very very very important in medical ethics. This means you can't lie to patients. To use placebos as you suggest, you'd have to throw that out. This is a dangerous path.
from:
Placebo Effects: Biological, Clinical and Ethical Advances (Damien G Finniss, Ted J Kaptchuk, Franklin Miller, and Fabrizio Benedetti
PMCID: PMC2832199 / NIHMSID: NIHMS169379
http://www.guardian.co.uk/science/2010/dec/22/placebo-effect...