How 2023's Murder Decline Compares to Other Declines

Where would a 10 percent decline in murder this year rank with homicide declines in the 20th Century?

Murder is almost certainly going to fall nationally in 2023 and it is increasingly looking like murder will fall at a level not seen in decades (if at all). Our YTD Murder Dashboard shows murder is down nearly 13 percent in now 114 cities with available data. I previously wrote about how a 10 percent decline in murder this year would be the largest such drop ever recorded.

The operative words in the preceding sentence are “ever recorded.” The FBI has been estimating national murder rates since only 1960, and the largest decline in murder ever reported by the FBI is a 9.9 percent decline in 1996 at the height of the Great Crime Decline. Other sources of data are needed to compare a hypothetical 10 percent drop in 2023 to any year before 1960.

There is probably no way to estimate changes in the nation’s murder rate prior to 1900, at least in a way that let’s us compare to the present day with any confidence. But it is possible to calculate reported homicide rates for the first half of the 20th Century using data collected by the US National Center for Health Statistics (now part of the CDC).

The good news is that states have been reporting homicide figures since the late 1800s. The bad news is that reporting was incomplete until the early 1930s. This data source was summarized by the Census in 1975:

“Deaths were classified by cause according to the Revision of the International Lists of Disease and Causes of Deaths that were in use for the years shown. Data for the entire United States were not available until 1933. For the years prior to 1933 this series includes deaths only for the death registration States of the respective years. For 1900, 10 states and the District of Columbia are included, comprising 26 percent of the population of the United States. As States were added, the registration area gradually grew to include approximately 50 percent of the United states population in 1910, about 80 percent in 1920, and the entire United States in 1933.”

Death Registration States, 1900 to 1933. Source: Estimates of Early Twentieth-Century U.S. Homicide Rates: an Econometric Forecasting Approach, Douglas Lee Eckberg.

So we have homicide rate data for 1900 to 1959, it is just incomplete to various degrees prior to 1933. In 1900 there were 230 homicides in the 10 states (plus DC) from the reported death registration data, good for a homicide rate of 1.2 per 100,000. This obviously should not be taken as an approximation of the US murder rate for that year.

Fortunately, sociologist Douglas Lee Eckberg published a study in 1995 which attempted to create a more accurate estimate of homicide rates between 1900 and 1933. Eckberg’s work produced the below graph of homicide rate in the US beginning in 1900. The lower line shows the reported homicide rate while the upper line shows his estimated rate from 1900 to 1932 (with accompanying 95% confidence interval which is difficult to make out). This critical work normalizes homicide rates for years when reporting was badly incomplete.

Now we can build a graph showing homicide/murder rates since 1900. This combines Eckberg’s 1900 to 1932 estimates with reported homicide rates from 1933 to 1959 with FBI Uniform Crime Report murder rate estimates from 1960 to 2021. The graph is brought to the modern day by guesstimating a 3.5 percent decline in 2022 and a (still hypothetical) 10 percent decline in 2023.

Note that not all homicides are murders but homicide rates from CDC generally move in concert with murder rates. There’s a little wonkiness around 1960 when we switch from homicide to murder rates, but it’s not too bad.

US Estimated Homicide Rate (1900 - 1959) and Murder Rate (1960 - 2023)

So how would they hypothetical 10 percent decline in 2023 compare to other declines? Is it still the largest?

There are three years with a double-digit decline in Eckberg’s estimated homicide rates: 1902, 1909, and 1918. Eckberg’s estimates for 1902 and 1909 have fairly large confidence intervals because fewer than 50 percent of the US population reported death data in those years. Eckberg estimates a 27 percent decline in the US homicide rate in 1902 with the 95 percent confidence interval falling between +2.7 percent and - 49.5 percent. The 1909 decline has a similarly large — albeit slightly smaller — confidence interval that suggests that we don’t really know how much homicide actually declined in those years.

Homicide declined by 10.1 percent in 1918 per Eckberg’s estimate. The confidence interval is much smaller because nearly 80 percent of the US population was covered by reporting, so the decline was most likely between 8.2 percent and 17.9 percent. It would make sense that homicide declined in 1918 as millions of young men joined the US army and 2 million shipped were overseas to fight in Europe.

Eckberg estimated a large increase in murder in 1921, the first full year of Prohibition, which did not begin to abate until Prohibition ended in 1933. The entire US population was covered by that point, so the homicide rate estimates are pretty solid from 1933 until 2021 when the NIBRS switch adds some uncertainty.

The homicide rate estimates from the complete death registration data from 1933 to 1959 point to two years where homicide fell by 10 percent or more: a 12.6 percent decline in 1935 and a 10.5 percent decline in 1938.

Conclusion

This exercise suggests that there are probably three years where homicides fell by around 10 percent (or more) since 1900. A 10 percent decline in murder in 2023 would be amongst the largest over that span but possibly not the largest. We ultimately have no way to say for sure where exactly it would rank.

That said, a 10 percent decline in murder in 2023 would represent a decline of more than 2,000 fewer murder victims from one year to the next. Given how much smaller the US population was 100+ years ago, this would represent the largest one year decline in murders in US history.