[WS:] Studies show that people tend to under-report undesirable experiences and over-report desirable ones in surveys. There is also a recall problem especially when the reference period is rather long (6 months in case of NCVS vis a vis 1 to 4 weeks used in labor force surveys). The reference period itself can also result in the undercounting of infrequent/irregular experiences - such as vicitimization.
Being a crime victim is by all means an undesirable experience and it is often associated with shame, especially if the perp is a family member or someone the victim knows. Frankly, I doubt that many people in this country would want to make to look their lives miserable to their own friends or family members, let alone an anonymous interviewer.
So with that in mind, while the NCVS is an improvement over the police records, it is very likely to undercount victimizations, especially "shameful" ones such as assaults by family members or rapes.
-- Wojtek
"Modern conservatism is just a neoliberal gloss on medieval domination."