Some economists are imperialists. They like to explain everything in the whole wide universe by using simple economic models, usually models which are static (with no time passing at all) and certain (with no "not-knowing" about anything). When you then add the ceteris paribus assumption (holding all other things constant) you indeed get some very shocking predictions. Usually they are shocking only because the basic model is too crude and simplistic to allow for what goes in reality or they are shocking for only some parameter values in the model and there is no real attempt to see if those are the values that actually prevail in reality.
One example of this: It's pretty easy to show by using the basic models of competitive markets and monopoly that there would be less crime under a monopoly, because a monopoly always produces less than a competitive market in those models. It also charges a higher price for its products.
I've seen this argument used to conclude that Mafia is a good thing to have, because it is like a monopoly (single provider of crime services) and thus it would cause less crime than a competitive market of many independent criminals. And there may be some partial truth in that.
On the other hand, the basic models I refered to assume that the product these types of crime firms sell is all the same, a homogeneous product like flour or sugar. But in reality Mafia is likely to branch out into crimes in totally new fields, and some of that is caused by the very fact of its power in the market. So we can't usually just compare the two situations by using the models that apply to homogeneous products.
Why all this boring econo-babble? Because of Steven E. Landsburg's new book, entitled More Sex Is Safer Sex. The Unconventional Wisdom of Economics. A chapter from that book is available for your perusal, the very chapter that talks about why having more partners for casual sex might reduce the rates of HIV in a population. Landsburg uses this argument to state that it might be a good idea to encourage currently uninfected people to have more casual sex. That way anyone in that market is less likely to "hook up" with an infected partner, and the disease won't spread as rapidly.
Now, this is unconventional wisdom. But the article (by Michael Kremer) from which it comes (or at least the working paper for that article) isn't quite as sweeping in its conclusions:
Under asymmetric information about sexual history, sexual activity creates externalities. Abstinence by those with few partners perversely increases the average probability of HIV infection in the pool of available partners. Since this increases prevalence among the high activity people who disproportionately influence the disease's future spread, it may increase long-run prevalence. Preliminary calculations using standard epidemiological models and survey data on sexual activity suggest that most people have few enough partners that further reductions would increase steady-state prevalence. To the extent the results prove robust, they suggest that public health messages will be more likely to reduce steady-state prevalence and create positive externalities if they stress condom use rather than abstinence.
Note all those "mays" and "suggests" and "few enough"? That's what drops out when economists start to sell their science as unconventional. Kremer's conclusions depend on the parameter values in the model.
I have another example by Landsburg, from his microeconomics textbook. In it he argues that polygyny benefits women, not men, and he uses an example like this:
For example, imagine a one-husband one-wife family where an argument has begun over whose turn it is to do the dishes. If polygamy were legal, the wife could threaten to leave and go marry the couple next door unless the husband conceded that it is his turn. With polygamy outlawed, she does not have this option and might end up with dishpan hands.
So. Of course if polygamy was legal the husband could just threaten back that he will get another wife unless the current one does the dishes. But the most serious flaw in the argument is that it works only if half-a-husband (from the couple next door) is not much different from a whole husband. This is an odd view of marriage, but not uncommon among the "polygamy benefits women" crowd. Men need multiple wives, women only slices of a husband.