I would add churches to that list. We love problem-solving approaches that sound like they should work. Because they were designed by pastors - an entirely different type of training, usually - who have really thought about this, and been a few places, and talked to people. So it only stands to reason that their ideas should work, right?*
Almost all programs fail when tested. That is, the vast majority of new "education, safety, and health" programs that appear to work based on non-experimental analyses cannot demonstrate statistically significant benefits in excess of costs in replicated, controlled experiments.There is additionally some reference to a favorite topic over at Bad Data, Bad: Failing to distinguish between correlation and causation. The full interview is here.
*All church growth numbers should be first examined in light of the population growth in the area, for example. Churches tell themselves that it's their music ministry, or their preaching, or their Sunday School, or their caring attitude that brings 'em in.
HT: Bird Dog at Maggie's Farm.