> Surveyors call up people. They ask, "Were you the victim
> of a crime. If yes, what kind?" The technique isn't perfect,
> but how do you know it yields an undercount?
Yes. Let me try again. The CVS has different goals than this thread does. This thread is based on you asking the leading question "How many violent folk do we have, anyway?" -- leading because you're biassed toward a smaller number. A number explicitly smaller than 2M.
They have a need to be:
- Defensible - Repeatable - Statistically coherent, reporting period to the next
They want, as a primary goal, to -- for instance -- establish baselines and trends. The trend is way more important than the actual count. Agree? Now: we have three possibilities:
- They capture the actual violent crime count/rate exactly - They overcount - They undercount
Let's rank them by probability; I'll start by saying that it seems least likely (by far!) that they get it exactly right. I'll further postulate that they work incredibly hard to make sure they don't overcount: good sampling methods, good statistics, etc.
So what does that leave?
> People could also overreport to make their lives seem more
> dramatic than they are.
I'm sure that's even less likly than them getting the right answer.
/jordan