The authors used publicly available data from local governments and police agencies to examine crime by category from January 1, 2023, to December 31, 2024. The initial list of cities was created using the most recent year of U.S. Census data available. After compiling a list of cities with publicly available crime data in each of the eight categories described above, the cities were categorized by how frequently their data were updated, the type of data reported (i.e., raw incident data, aggregated monthly counts, or quarterly reports), and how and which crimes were reported (NIBRS offense codes or other categories). Project criteria required that cities report incident-level data and update the data frequently, with no regular severe lags. 

The data are reported here in a minimally adjusted format. All data from the 49 U.S. cities available were normalized such that the eight crime categories included the same data across all cities. Within the 49 cities included in the analysis, missing data for short periods was imputed in two ways. For gaps of less than five days or less than 15 percent of one month, missing data were assumed to be missing completely at random and imputed as the value of the previous valid day. There were 278,702 valid observations, comprised of up to eight observations per day per city, depending on how many categories each city’s open data portal shares. Other notes:

• Data from all 54 cities included in the Live Crime Tracker as of January 10, 2025, were pulled on January 10. Three cities—New Orleans, LA, Baltimore, MD, and Madison, WI—were removed for this analysis because their 2024 data was not recent enough to warrant a year-long estimate of crime rates (i.e., their most recent data available was August 2024 or earlier). Two cities—Honolulu, HI, and Louisville, KY—were further removed from this analysis because their 2023 data were not available through the open data portal at the time of analysis. This resulted in 49 cities with data in any of the eight crime categories for our comparison. 
• Each city’s data were filtered so that the dates of 2023 and 2024 data are directly comparable. If a city’s data are only available through December 5, 2024, 2023 data were filtered for that city to only go through December 5, 2023. 
• Data were pulled on January 10 to minimize the prevalence of missing data due to reporting delays in the final days of 2024. 
• Not all cities report all crime types, leading to differing numbers of cities used to calculate estimated rates. Most notably, fewer cities report sex offenses than other crime types, resulting in a smaller sample for those estimates.  
• All observations over three standard deviations from that city's year mean were given an indicator that the data were potential outliers. That resulted in about 2,700 outliers in a dataset of more than 270,000. Staff reviewed these records to distinguish outliers from errors. We note that unusual data points are expected in very large datasets. However, since NORC is pulling records from the only source of daily data (official police records), there is no secondary data source for comparison other than annual totals and news reports. For each crime type within a city where outliers were indicated, staff determined whether the outliers only appeared in one of the years, which would bias the sample. Among these outliers, homicide in Kansas City was notable. Homicides on two dates (January 1, 2024, and February 14, 2024) together summed up more homicides than Kansas City police recorded officially for the full year. Thus, the total number of homicides officially reported by Kansas City Police Department (from the same data source our data was originally derived) was used in the final analysis. Specifically, the Kansas City homicide total was 144, not 497, as indicated by summing daily data. 

An important caveat is that uncertainty is different from error.

The authors note that there is skepticism about the validity and reliability of national crime statistics. This is partly the result of the United States reporting national crime statistics from different sources, such as the National Crime Victimization Survey (NCVS) and the national reported crime statistics from the FBI’s UCR program. Because these reports describe different time periods, at times when crime data are volatile—such as in 2022 and 2023, when violence stopped increasing and began to decline—the reports can conflict. 

Additionally, the FBI does not traditionally report whether changes from one year to the next are statistically significant. Historically, almost all law enforcement agencies report data to the FBI and statistical testing is required. Beginning in January 2021, the FBI changed how law enforcement reports data, requiring detailed information about every crime rather than the summary data required before 2021. However, crime reporting to the FBI is voluntary, and the new, more intensive data collection led to a rapid decline in law enforcement reporting. As of May 2024, the FBI reports that about 82 percent of the United States population is covered by a National Incident-Based Reporting System (NIBRS) reporting agency. In several states, less than seventy percent of the population is covered by a NIBRS reporting agency (e.g., Alaska, Arizona, California, Florida, New Jersey, New York, and Pennsylvania). As a result of the lower reporting rate, the FBI data currently is more of a sample than a census of law enforcement agencies. Reporting the statistical significance of observed changes can improve understanding that some level of uncertainty exists in aggregated estimates.

The Live Crime Tracker fills this gap. The Tracker compiles crime data directly from law enforcement agencies to bridge the gap between the most recent crimes reported and national data and does it in a timely way.