This story by Chase Woodruff appeared on Colorado Newsline on April 5, 2023.
As voters in Colorado’s two largest cities returned their ballots on Tuesday in pivotal municipal elections, a news story containing a major factual error about the state’s crime rate was shared widely by Republican leaders, police officials and other political figures.
The story from Denver’s local CBS affiliate initially falsely claimed that a new Department of Justice analysis showed that Colorado “has the highest rate of violent crime victimizations in the country.”
In fact, while the report suggests that Colorado’s violent crime rate may be higher than reported, 28 states and Washington, D.C., were entirely excluded from its analysis, which came with a range of other caveats and limitations.
The story was quickly shared far and wide by Colorado Republican politicians and law enforcement officials. The Twitter account of the Colorado House Republican caucus shared a link to the story, which it said showed “it’s time to lock up criminals and uplift our law enforcement so that our communities our safe.” Former state GOP chair Kristi Burton Brown said it was “what happens under too many years of (Democratic) control.”
As voters headed to the polls for city elections in Denver and Colorado Springs on Tuesday — in a campaign season dominated by strong rhetoric on crime and homelessness — one Denver City Council candidate said the story was a “reminder” of the need for tough-on-crime policies, while a conservative group, Save Denver Now, linked to the story and urged voters to “vote wisely.”
After a Newsline reporter asked a station representative for comment, CBS4 partially edited the story and its headline on Wednesday to clarify some of the data’s limitations.
The DOJ report, titled “Criminal Victimization in the 22 Largest U.S. States, 2017–2019,” was published by the Bureau of Justice Statistics on March 6. It’s based on data from the National Crime Victimization Survey, a U.S. Census Bureau program that collects information from a sample of households regarding violent crimes and property crimes on an annual basis.
The BJS calls this self-reported victimization data a “complementary” picture of U.S. crime rates. Unlike crime estimates based on the FBI’s National Incident-Based Reporting System, the National Crime Victimization Survey data offers researchers a way to measure crimes that may be committed but not reported to police. Additionally, the two metrics “measure a set of offenses that overlap but are not identical, which leads to differences in estimates between the two data sources,” according to the BJS.
In recent years, Colorado’s reported violent crime rates have ranked near the middle of the pack compared to other states, according to NIBRS data. Though the data from the survey sample suggests that unreported crimes would push that ranking higher, they don’t show that Colorado is “Number One,” as the CBS story continued to claim as of late Wednesday afternoon.
For decades after its creation in 1972, the National Crime Victimization Survey collected data from a sample of households that was only large enough to accurately measure trends at the national level, rather than on a state-by-state basis. Beginning in 2016, the sample was enlarged and adjusted in order to enable state-level analyses, but only for the 22 most populous states.
“The decision to include 22 states, instead of more or fewer, was based on the NCVS sample allocation and on the cost of boosting the sample enough to produce precise, representative estimates of personal and property victimization for individual states,” the BJS report says.
The victimization estimates for Colorado are based on responses from roughly 16,000 eligible households between 2017 and 2019. Because the estimates are derived from weighted statistical samples, caveats apply.
The report found that Colorado’s violent crime victimization rate was higher than the U.S. average to a statistically significant degree. But its estimate of 45 incidences per 1,000 people reflected a range of between 36.2 and 52.8 at a 95% confidence level, making direct comparisons with other high-incidence states like Arizona and Washington less certain.
“When estimates are derived from a sample, as with the NCVS, caution must be used when comparing one estimate to another or when comparing estimates over time,” the report states. “When the sampling error around an estimate is taken into account, apparent differences between estimates may not be statistically significant.”