If my “bad quarter” hypothesis is correct, that one-fourth of a population will commit violent crime over a lifetime, then either the violent crime rate would vary randomly around an average value (adjusted for life expectancy), or the number of crimes per criminal would need to change to compensate for the observed change in crime rate.
The crime rate has not varied randomly over time. From 1980 to 2003, the U.S. violent crime rate appeared to drop linearly over time except for a “bump” in the 1990s. The lifetime rate (percent of the population who will be a victim of violent crime over a lifetime) dropped from 44 (in 1980) to 42 (in 1985), then climbed to a peak of 57 in 1991 and gradually dropped to 40 by 1999 before resuming its earlier trend downward, reaching 37 in 2003. The rate for murder and non-negligent manslaughter paralleled that for violent crime; as a percentage of violent crime it actually dropped (from 1.7 percent in 1980) from and appeared to start leveling off (at 1.2 percent by 2003).
The ratio of violent crimes to arrests for violent crime dropped precipitously from 1993 to 2002, from 4.3 to 2.3. The bad quarter hypothesis projects those ratios to be 2.3 and 1.4, implying that roughly half of perpetrators have not been caught.