I think its only advisable to do the endline survey on the same people you did the baseline survey on. Unless you are dealing with a nomadic population and you are very certain that its the same people that moved to the other village, the real change in the indicators being measured should be observed among the same people, community or village where the first assessment was done.

Dear Mr./Ms./Dr. Ahmed:
I am sorry this response is so late; I was out of town on vacation for the past 2 weeks. Whether or not to try to include the same individuals or villages in an end-line survey as were included in the baseline survey depends on the objectives of your program. If the program aims to improve the community's level of an outcome, then you probably want to independently measure the status of the community at the two points in time, which means you need independent samples for each survey.

As you know, observer bias can produce spurious results; that is, the individuals and villages included in the baseline have received additional program input (the baseline survey itself) and may be more likely to respond to the intervention because of this prior sensitization, thus producing an overestimate of the effect of the intervention. This is especially true of KAP interventions.

Selecting separate, independent samples will measure the program effectiveness of your intervention, while selecting the same sample for both surveys will measure something more like efficacy. Program effectiveness is always lower than efficacy because it takes into account real-world constraints. I think it is the better evaluation of a community-targeted program.