It's not going to be perfect, but it gives a very solid snapshot of peoples experience with crime without the massive distortion we know we get from looking at similar sized samples asked what they think crime levels are.
So? Sample size only addresses sampling error, not nonsampling error, for nonsampling error its exactly as bad as the dinkiest little poll on the same topic (and for sampling error, it's not much better; polls are the sizes they typically are because it doesn't actually take a very large scale to be fairly reliable when you only consider sampling error, and, again, adding more size doesn't help at all against nonsampling error.)