There are two groups of people who could slow down the epidemic by increasing their sexual activity: first, those who have had unusually few partners in the past; and second, those who expect to have unusually few partners in the future. (In practice, these two groups are likely to overlap substantially.)
As I've argued in the body of this column, those in the first group are relatively likely to be uninfected, so by increasing their activity, they can improve the pool of safe potential partners for everybody else.
But those in the second group perform a different, and more macabre, social service. To visualize it, imagine for a moment that you are an AIDS virus, and think about the worst fates that could possibly befall you. High on your list would have to be getting passed on to somebody whose future plans don't include much sex. One way to make life more difficult for AIDS viruses is for more of those people to be catching the virus (instead of other people), and thus serving as "dead ends" for chains of infection.
The body of this column discusses only the benefits of increased activity by those in the first group. Professor Kremer's estimates account for the benefits of increased activity by both groups. To keep the calculations manageable, Professor Kremer assumes that the two groups coincide exactly, and therefore that each sexual conservative is able to confer either type of benefit.