A Severity-Weighted Index of Violent Crime

In debates over criminal justice policy, people are constantly referring to crime indexes for the question of whether crime is up or down and by how much. But there are problems with the official indexes. One of them is that indexes tend to be dominated by the least serious crime chosen for inclusion in the particular index. Crimes are simply counted, and because the frequency of crimes tends to be inversely related to their severity, the less serious crimes dominate.

For example, the FBI’s index of violent crime includes murder (and voluntary manslaughter), rape, robbery, and aggravated assault. Murder is the most serious, followed by rape, but the other two are much more common. As a result, the violent crime index is largely a measure of robbery and aggravated assault, and it relatively insensitive to changes in the rates of murder and rape.

One alternative is an index with crimes weighted according to their severity. I have seen such indexes in other countries and for some jurisdictions within the United States, but none for the United States as a whole. Here is a first cut at a severity-weighted index of violent crime in the United States.

First, data from the Summary Reporting System (SRS) for 1979-2020 were downloaded from the FBI’s Crime Data Explorer. Unfortunately, the FBI discontinued the SRS in 2021, so no data are available for last year. It has replaced the SRS with the National Incident-Based Reporting System (NIBRS), but the paperwork burden of that system has produced low participation rates. See Elizabeth’s post for the full story. We may be able to work over the transition later, but for now the index stops at 2020.

Second, the broadening of the definition of rape in 2013 had to be accounted for. We had data for rape using the “legacy” definition only from 1979 to 2012, both definitions for 2013 to 2016, and the revised definition only from 2017 to 2020. For the four-year overlap period the ratio of revised to legacy averaged 1.38, so the numbers from 1979 to 2012 were all multiplied by that figure to give an estimate of what the number of rapes would have been under the revised definition for those years. The resulting number is added to the murder, robbery, and assault numbers to get a total number of crimes, which will not be the same as the official UCR violent crime number for that year in the years that needed to be adjusted. This number is divided by the population for the year in 100,000s to get the unweighted violent crime index.

Third, weights needed to be assigned to the various crimes in the index. RAND Corporation looked at estimates of the cost of crime in a 2010 report which considered the costs versus benefits of increasing police forces. It is Hidden in Plain Sight: What Cost-of-Crime Research Can Tell Us About Investing in Police by Paul Heaton. It seems awkward to even talk about a cost in dollars of crimes that involve pain, suffering, psychological trauma, and even death. Even so, these estimates have practical uses, and the assignment of weights to different crimes is one of them. The RAND paper computed an average of three studies (Table 1 on page 5), and these numbers were used to assign weights.

To compute the weighted index, the number of crimes of each type in a year was multiplied by the cost figure described above. Dividing by population, again scaled in 100,000s, gives the cost of these crimes in dollars per 100,000 population. To get the weighted and unweighted indexes on the same scale, the result was divided by a number calculated to make the two indexes the same for the first year, 1979. That number works out to about 239,000. Here is a graph of the results:

Weighted and Unweighted Indexes of Violent Crime

In 1979, the two lines begin at the same point by definition. The crime rise of the late 1980s and early 1990s is less steep with the weighted index, but the post-1993 decline lines are nearly parallel. The gap begins to narrow in the late 2000s. Beginning in 2015, the rise in the weighted index is significantly greater than in the unweighted.

With the jump in homicides in the pandemic year of 2020, the weighted index is back where it was in 2003, while the unweighted index is 18% lower than it was in that year. The weighted index is up 21% since 2011, while the unweighted index is unchanged from that year.

When polls indicate that people believe that crime is up, it is common to hear pundits scoff that the official statistics show it is not and that the people who believe it is are just ignorant. But there are many ways to crunch the numbers, and the official index is not inherently better than the others.

As I mentioned above, this is a “first cut.” We will likely be revising and refining this weighted index in the coming weeks. But I thought this initial “back of the envelope” calculation would be of interest to our readers.