I've been teaching juvenile delinquency and criminology since the mid-1990s. Each year, I show my classes how the crime rate has generally declined as measured by household victimization or crimes known to the police. At the end of each term, however, I always get a few papers that lead off by citing "alarming increases" in crime/homicide/violence. The FBI's preliminary numbers for 2004 again show a drop of 1.7% for property crime and 1.8% for violent crime nationally since 2003. For the past 10 years, the news has been great with regard to trend, though crime levels remain unacceptably high in many communities.
Some criminologists have been predicting a crime surge, but I see room for further reductions -- particularly if graduation rates and labor markets improve. I guess a better way to put it is that I haven't been convinced that the pessimistic predictions are justified by much more than regression to the apparent "mean" established in the worst of the bad old days. It is very tough to make crime projections over the long-term, since the effects of things like age structure, economic performance, and incarceration seem to vary quite a bit over time (that is, I think a model predicting crime in the 1960s wouldn't do so well in explaining 1990s trends). In any case, I now warn my classes that I will completely freak out if they cite "alarming" increases without indicating precisely what has increased and when it began to rise.