I can see a few possible problems with this.
1) Lag/delay in statistics. If the feature is abused as described in some of the posts above, an area considered safe can be unsafe for a while before the statistics catches up with reality. The opposite is also true; an area that has been "cleaned up" may be considered unsafe for a while.
2) Different types of violent crime. Not all violent crimes occur in the streets; domnestic violence is (at least where I live) considered a violent crime, and it is also a lot more common than unprovoked violence on the streets. At least that is what the police says - in the statistics they are bundled.
3) Seasonal / time of day differences. I live in a city that is flooded by tourists in the summer. Violent crimes increases significantly during those few months, and most of those crimes occur late evenings / nights when people at clubs/bars/pubs are drunk. Still, statistics for specific areas are compiled on a yearly basis.
4) In sparse areas, a single crime can have a huge impact in the statistics. Looking at statistics compiled "per capita", the area where I grew up had a 200% increase in violent crimes one year. It went from one case of domnestic violence to three - or 20/1000 per capita.