Homicides continued to rise during 2021

As many people have reported, there was a drastic increase in violent crime (particularly driven by homicide increases) from 2019 to 2020. Not surprisingly, people are interested to know whether this pattern continued into 2021, and if so, to what extent.

A recent report published by the Council on Criminal Justice (CCJ) attempts to shed light on this by collating information on crimes occurring in 2021 from a subset of 27 United States cities. Trends for individual crime types presented in the report do coincide with other estimates in terms of the overall pattern witnessed in 2021. However, the sample size of 27 cities is very small, meaning that their estimates (especially for homicide) are smaller in magnitude and likely do not reflect the full extent of the problem.

There were a total of 27 cities that had crime data posted online, which the authors used to gather counts of different crimes. However, not all cities reported on all types of crime, so many crime types are estimated based on even smaller subsets of cities. For example, only 11 cities had data posted regarding domestic violence cases (this was the smallest analytic sample), whereas 24 cities had data available on robberies (this was the largest analytic sample).

It is unclear exactly where the data are coming from. The authors mention it came from city’s police department websites and open data portals, but there are no links to specific sections of websites that were used. The reason why this matters is because there are different ways that police departments tabulate and display their data, so it is hard to know exactly how estimates were computed for each city. For example, weekly CompStat reports are often displayed on a department website that present a snapshot of the crime data at a specific point in time, but these numbers are subject to change as cases are sometimes updated or reclassified. In addition, these same departments will usually have some type of annual report with the “official” numbers, which sometimes differ from the initial CompStat numbers. Thus, it is important to note that preliminary police department data as reported on department websites is not always finalized and can be somewhat dynamic in this regard.

The authors collected data from individual police department and city websites because the official data for the full year of 2021 are not available. While this approach makes sense on its face, it also has drawbacks — importantly, only 27 cities had crime data posted online for the full year of 2021, so this inclusion criteria meant the sample was severely restricted.

Another way to look at this is to examine data that encompasses more cities but fewer months of data. This can be done by looking at year-to-date (YTD) comparisons, which compare a specific time period from one year to the same time period from another year. Thus, even though national-level data may not be available through the end of 2021, YTD comparisons can be used to effectively examine the national 2021 data that is available. For example, the FBI released quarterly data that showed YTD numbers for the first three quarters of 2021 (i.e., through September) alongside the YTD numbers for the first three quarters of 2020. Data are available for download and can be examined, which we have done briefly here.

Although the FBI data does not include the full year, it does include 159 cities — a much larger sample size than the 27 included in the CCJ report. With such a large difference in the number of cities, there are many more crimes reported in the FBI data and the coverage is more wide-ranging across the United States. This creates a larger sample size and also increases the representativeness of the sample. In other words, the FBI data is much more likely than the CCJ data to be representative of the nation as a whole. Thus, it is probably the more accurate estimate in terms of the “big-picture” look regarding 2021 crime numbers.

That being said, it’s important to think about how these numbers can vary across sources. Clearly, the CCJ data is not representative of the United States as a whole, yet their findings do slightly mirror those seen in the FBI trends. When comparing the two sources though, one of the most notable distinctions is that the CCJ numbers appear to be vastly undercounted despite the fact that they represent the full year of data. This is likely due to the overall small sample size, so it makes sense that numbers would be lower. Otherwise, the CCJ data are actually following generally the same pattern depicted by the FBI data, but they are probably not capturing the full magnitude of the problem.

For example, the CCJ report says that homicides increased by an average of 5.4% (+218 murders in the 24 cities that provided data). But the FBI estimate is more than double that, showing an average increase of 12% as of September 2021 (+461 murders among the 159 cities). However, both estimates do show generally similar findings: i.e., they both saw an upward trend from 2020 to 2021, though the increase was smaller than the 29% increase seen from 2019 to 2020.

In addition, both data sources showed that homicide rates are the highest they have been in a decade, yet not quite as high as rates witnessed in the 1990s. The CCJ report gives high credence to this finding, emphasizing that homicide numbers are still low compared with the 1990s. Yet, the CCJ numbers for homicide specifically appear to be undercounted by nearly half, so is it possible the problem is worse than they think it is? In addition, regardless of undercounting, the report seems to be minimizing the fact that murders increased by nearly 30% from 2019 to 2020, a spike that we have not yet recovered from. Murder rates are continuing to increase, evidenced in both sources of data (albeit to a lesser extent in the CCJ data), which is concerning especially after the large increase seen last year. This is not something to be minimized.

When it comes to other offenses, data sources again found similar trends, but the  CCJ numbers are lower than the FBI numbers. For example, both sources revealed a decrease in burglary and larceny offenses from 2020 to 2021. However, the estimates differ slightly: the data from CCJ’s study found that burglary and larceny offenses decreased by 6% and 1%, while the FBI data shows a decrease of 8.6% and 2% (respectively). Both reports also showed a fairly sizable increase in motor vehicle theft, which increased by 14% per the CCJ report and 16.5% per the FBI data. So essentially, the data sources show similar findings, but the FBI numbers are slightly inflated compared to those reported by CCJ. Interestingly, this disparity seems more prevalent to homicide specifically.

Though, it is also important to note that some large high-crime cities are not included in either source of data, which could also affect the accuracy of both estimates. For example, data from New York City are not reported in either dataset. Though, preliminary numbers from news reports show that New York saw 485 homicides in 2021, a 4% increase from the 468 tallied in 2020. Similarly, Oakland CA is also excluded from both datasets, despite the fact that the city saw 134 homicides in 2021, the highest count since 2006.

Taken together, one must wonder why it is so difficult to obtain accurate crime data. Perhaps part of the reason is due to the burden required by law enforcement to clean and submit the data, or perhaps it is purposefully not reported for reasons unknown. Either way, it seems that data from some of the largest cities reflect similar trends to those seen at the national-level, though they likely do not capture the true magnitude of the problem.