Every year, the FBI publicly releases reported crime figures across the nation to provide an objective measure for whether crime is rising or falling.
Fluctuations in reported crimes -- no matter how small or insignificant -- may be easily overblown without proper context. The most recent cycle of this played out after homicides across the nation rose in 2015 and 2016 after decreases going back to 2006 (nationally reported crime statistics lag a year behind).
I cannot say any better than you what caused the recent increase in homicides. What we can do a better job of though is objectively identifying the magnitude of those rises and falls relative to previous crime trends. In a recent paper in the academic journal Homicide Studies, my colleague Tom Kovandzic and I provided two statistical graphics to help assess homicide rates. The first is called a funnel chart, and this helps compare homicide rates in one city to homicide rates in another city of a comparable size. The lines indicate the rates one would expect a city would fall within 99 percent of the time if the city had an average homicide rate the same as the nationwide homicide rate, about five homicides per 100,000.
The idea behind the line getting wider is that it is easier to see large swings in the homicide rate for smaller cities. Take, for example, Plano, which had a recent domestic incident that resulted in eight homicides. With a city population of around 300,000, that one incident will increase Plano's homicide rate per 100,000 population by nearly 3. Smaller cities see more volatility in their homicide rates from year to year than larger cities.
The second type of graph I suggest are fan charts that show a city's homicide rate over time, along with a predicted band that the points should fall within 80 percent of the time. Below shows New Orleans, which had an outlying homicide rate the year Katrina occurred, but otherwise was always within the band.
Doing this exercise I came to two conclusions. First, the majority of cities' homicide rates in 2015 were contained within their predicted intervals. There was no unexpected crime wave given historical data. The second is that many of the cities identified as outliers in the funnel chart, places like St. Louis and Baltimore, had very high homicide rates relative to other cities of comparable size going back over 30 years. Any explanation of rising homicide rates in those cities needs to be consistent with their very high homicide rates going back decades. In this light, recent explanations for the homicide increase, such as the opioid epidemic or the "Ferguson effect," the idea that police became more reluctant to stop and question people because of increased scrutiny, fall short.
Andrew P. Wheeler is an assistant professor in criminology at The University of Texas at Dallas, Andrew.Wheeler@utdallas.edu. His personal blog is at