Journalists: How to Responsibly Report on Local Crime Data

Read this first

The information collected and produced by local police departments and sheriff’s offices can change the world. These crime statistics are used by politicians, researchers and the police themselves to advocate for public policies, funding, resources and reforms to the justice system. Journalists are often tasked with contextualizing this data, but separating the truth from political bluster can be deceptively tricky.

While crime statistics released by your local police department may seem straightforward, the numbers can be incomplete or misleading. For example, information logged by police officers is often based on their own perceptions, which can be inaccurate.

When newsrooms don’t factor in the very real limitations of crime data, their coverage can contribute to fundamental misunderstandings about crime trends locally and nationally. In 29 of 32 Gallup surveys conducted over the past few decades, at least half of U.S. adults have said they believe there is more crime occurring nationally in the current year than in the previous year, despite the downward trend in crime rates during most of that period. Crime across all major categories has dropped substantially from peak rates observed in the 1980s and early 1990s.

Ultimately, since misinterpretations of crime statistics can perpetuate unwarranted fear, the stakes for getting it right are high.

This toolkit will explore the nuances of working with crime data, including common pitfalls to avoid. To learn more about the FBI’s crime reporting system, which offers data from more than 15,000 U.S. law enforcement agencies, check out our companion toolkit. If you’d like to talk to The Marshall Project about crime data for a news story or research project, you can request a consultation.

Join us on Tuesday, October 20, 2026 at 1 p.m. ET for our webinar about how to report on crime data this election season. Learn more and register here.

Understanding local police crime data

Reported crimes are not the full picture

What constitutes a crime is determined by local, state and federal laws, and is interpreted by the individual judgement of law enforcement officers. Police crime data includes the incidents that police are aware of, either because someone reported the events to them or they uncovered them during patrols or investigations. Crimes that don’t get captured in police records don’t appear in the data — and fewer than half of all crimes are reported to police.

Some offenses, including sexual assault and hate crimes, have especially low reporting rates because they come with societal stigmas or are concentrated among vulnerable people fearing consequences from contact with police or immigration officials. Murders have high reporting rates because of the severity of the offense and the physical evidence in the cases. Carjackings are also frequently reported, as insurance companies often require police reports to process claims.

Sometimes, the number of crimes in a community appears to rise or fall because of shifts in how frequently those crimes are reported. For example, programs that encourage hate crime victims to come forward may increase the numbers in police data, even though the actual crime rates have not changed.

Communities that trust police are more likely to report crimes. Victims are generally less likely to come forward if they are non-White, low-income, immigrants or have had troubling interactions with law enforcement in the past.

Police definitions and priorities can change

A change in how a crime is defined may make it appear that changes on the ground have taken place when they actually haven’t.

In 2013, the FBI adopted a broader definition of rape that includes sexual assault with an object and sodomy, leading to a huge increase in the number of rapes reported nationwide. Check your department’s data dictionaries, record layouts or “README” files, which may include details on what definitions have changed. If you can’t find a data dictionary, ask for one.

Police agencies also change priorities and protocols over time. They may decide to focus patrols in new neighborhoods or invest more resources on stopping shoplifting, which could result in a bump in the data without there necessarily being a shift in how frequently people in a given community are breaking a specific law.

Police data may be inaccurate

Information logged by police officers is frequently based on their own perceptions, which may not match reality. Officers in some places, like California, are instructed to visually assess and record people’s races during pedestrian and traffic stops, which can lead to inaccuracies.

Other times, police are limited in the information they can record, since some law enforcement databases do not allow officers to input certain details. For example, an officer may not be able to enter a person’s gender identity as nonbinary or transgender. As a result, a person’s listed gender in a police report may not match how they identify.

Two officers may differ in how they interpret and document events. Where one officer may classify a fight with a broken beer bottle as disorderly conduct, another may call it assault with a deadly weapon. Police agency policy manuals and codebooks could explain the instructions officers are given about entering details of an encounter, but individual judgment is almost always a factor.

Sometimes misrepresentations are deliberate. There are many examples of police issuing statements that fail to match body camera footage and deceptively altering crime data to appear more effective.

Clerical errors in data entry, like misspelling a name or writing down a wrong number in an address, are inevitable. Police agencies often do not retroactively correct their data if new facts emerge later on, such as if they determine that a person was arrested wrongfully. This may make it appear that police have solved more crimes than they actually have.

The data is often incomplete

Police agencies can express missing data in different ways. Sometimes they leave fields blank, while other times they enter a “0,” “-99” or “N/A.” Only by closely examining the values for outliers and unexpected entries will it become clear how missing data is represented.

Widespread, life-impacting events like natural disasters can affect how many people report crimes and how police record them. The COVID-19 pandemic led to significant policy changes across jurisdictions nationwide that impacted how departments responded to various crimes. Asking agencies if they followed different protocols during periods of notable turmoil can help prevent making erroneous assumptions about the data.

Whenever possible, it’s important to vet the data to see how much is missing and how that could impact the conclusions you’re trying to draw — or if the data is too full of holes to use at all.

Check what percentage of your data has incomplete values and consider why the information may not be present. Before moving forward with a story, it could help to estimate how much the results of your analysis could change if the missing details were filled in. You can do this by figuring out what the minimum and maximum values could be for the missing data to show the potential range of outcomes. You could also choose to report the information you have available and disclose how much of the data is missing, unclear or contains errors.

Crime data is cyclical

Certain crimes like assault and motor vehicle theft tend to happen in the evening more often than in the morning. Many crimes, including murders, occur more often on weekends and in the summer.

These patterns don’t hold in all situations or across all categories of crime, but keep in mind that they can impact the way the data looks when examining crime data over time. A rise in murders from January to July isn’t necessarily a sign of a new trend, but rather, something that likely happens every year.

Looking at the same data across several years can help identify crime rates that are actually changing over time. If you’re trying to determine whether the most recent year of crime data is following unexpected trends, a good rule of thumb is using at least five years of data. It’s also best to use consistent time periods for your comparisons, such as all crimes from January to June each year, rather than relying on partial data for the current year and a complete set of data for prior years.

Crime data naturally fluctuates

We wouldn’t expect an identical number of thefts to occur multiple years in a row — there will naturally be some fluctuation.

Scientists can measure whether a rise or drop in crime data is part of this normal fluctuation or something they refer to as “statistically significant,” as in, the numbers are unlikely to be the result of natural ups and downs in the data.

While these calculations can be complicated, keep this framing in mind: Some change from year to year is expected, and it doesn’t always indicate a new trend is emerging. If you’d like to learn more about statistical significance calculations, speak to a statistician or researcher. Often, these experts work in the criminology or statistics departments of local universities.

Crime and clearance rates

What is a crime rate?

Crime rates are usually calculated per capita — per 100,000 residents — which allows for fair comparisons across jurisdictions of different sizes. It’s typically best to calculate rates rather than use simple totals to compare the likelihood of crimes occurring in different places.

Number of crimes in an area / population of the area = crime rate

Incidents are often documented in the jurisdiction where the crime occurred, regardless of where the offender or victim actually lives. High-tourism cities and college towns can appear to have higher crime rates than other locations, because there are a lot of nonpermanent residents in those areas. This makes more crimes appear in the numerator of the calculation, but doesn’t add to the number of people represented in the denominator, making the crime rates higher.

To understand how much the data may be skewed by these factors, ask the police agency if they have fields in the data indicating whether victims and offenders are local residents. Read more about how crime rates can be distorted in Jeff Asher’s newsletter, Jeff-alytics.

What is a clearance rate?

A clearance rate determines how many crimes a law enforcement agency has “cleared” through arrests or other means. They are typically calculated using the number of cleared cases that occur in a given year. The calculation looks like this:

Number of cleared cases in a year / number of offenses that year = clearance rate

While this seems like a straightforward equation, there are important nuances. Someone can get arrested in January for a crime they committed in December, which can distort clearance rates. In small geographic areas, it’s even possible to have more clearances than crimes for a particular year.

An arrest isn’t the only way a crime can be cleared. Clearance by “exceptional means” can happen if an offender dies, the victim refuses to cooperate or the prosecution decides not to pursue the case. Police have also used exceptional clearances as a way to inflate clearance rates when crimes have not been solved. Check whether exceptional clearances are included in your local police’s clearance rates, and what fraction of cases are cleared by arrest versus exception.

Tips for analyzing crime data

Check that data is regularly reported through the whole time period. Some police agencies begin posting crime data for the most recent month or year before they have a complete dataset. Double-check the most recent incident date for crimes in your dataset and beware of an unexpected drop in crime counts over recent days, weeks or months, which could suggest incomplete data. It’s also possible a police agency only reports newer data and that older crime data is unavailable or incomplete. It’s often best to eliminate time periods with partial data from an analysis.

Use rolling averages rather than year-to-date calculations. Since crime is seasonal, year-to-date calculations can be misleading. When possible, calculating rolling averages — such as the assault rate for each of the most recent 12-month periods — often produces more accurate findings because it determines longer-term trends in the data. Learn more about rolling averages here. If year-to-date calculations are the best option in your case, be sure to use the same window of time in each year you examine (e.g., March to July data each year).

Dig deeper than basic crime rates. Rather than just looking at whether the number of crimes is rising or falling across a whole city or state, dig deeper into the trends in different communities or among people of different demographic backgrounds. If you’re looking for ideas, check out this list of great crime data stories from local news outlets across the country.

Compare crime data from different sources. Sometimes, the data police agencies report on their own website differs from what appears in other sources, such as the FBI’s national Crime Data Explorer or data collected by advocacy groups or researchers. Comparing multiple data sources can help you to catch errors, pinpoint differing definitions or identify situations where police have manipulated statistics.

Tips for writing about crime data

Offer caveats and explanations of missing data. When possible, explain in your story that you are working with police-logged crimes and that many crimes are not reported to police. Telling your audience what is unclear is valuable information in and of itself.

Be careful with small numbers. Reporting a percent change can be misleading if the underlying data contains small values. An increase from 8 assaults to 10 assaults is technically a 25% increase, but the actual number of crimes has only risen slightly. In these cases, consider whether percentages could create a false impression of the data, and if so, provide readers with the simple totals instead.

Give clear time frames. Be sure to include what time period is covered by your data. Saying “motor vehicle thefts are up 15%” or “assaults are at a record high” without a starting point leaves people without key information.

Steer clear of language like “skyrocketed,” “soared” or “plummeted.” It’s rare that crime trends actually increase or decrease exponentially. Instead, use specific descriptions that better reflect the scale of the changes, such as “doubled” or “fell for the third year in a row” or “dropped by 10%.”

Pitfalls to avoid

Don’t assume you know why crime rises or falls

Police or politicians may take credit for falling crime rates, but they are not always responsible. Experts have identified dozens of factors that influence crime rates outside of police interventions, like economic conditions, climate, population density and commuting patterns.

Consider whether similar crime trends are occurring in other areas. If crime is falling across the U.S., like the post-COVID crime drop that began in 2023 and has continued since, then police interventions in a single location don’t fully explain the situation.

Avoid ranking cities based on “safety” or “danger”

Countless web posts rank cities on factors related to public safety and danger. However, these articles are often misleading and based on flawed analysis. Crime data quality and collection efforts vary greatly from place to place, making it hard to compare one location to the next with the level of certainty that would allow for accurate rankings. If you’re trying to check trends in your local area against other benchmarks, look at national or statewide averages. It can also be useful to gut check what an audience would actually gain from comparing data across places. (Here's a silly example of why comparing data across places isn't always useful.)

Center discussions around crime rather than disorder

Many people confuse the presence of dirty city streets or empty shelves in retail stores with signs of criminal activity. Disorder like this can make people feel unsafe, but it isn’t actual evidence that crime is rising — or that more police activity is the appropriate solution. Redirecting interviews to focus on criminal offenses and policy solutions can help dispel the narrative that crime and disorder are interchangeable concepts. Learn more by reading the newsletter of criminal justice researcher John Roman.

The visible presence of unhoused people can be perceived as a sign of crime or danger, but a large unhoused population does not necessarily indicate there is an elevated crime risk. In fact, people who are unhoused are more likely to be victims of crimes, and it’s worth noting that many cities and states have instituted policies to criminalize being unhoused.

It can lead, even if it doesn’t bleed

It’s important to report both on increases and decreases in crime rates, even if the latter may not seem as attention-grabbing. Stories about falling crime rates can focus on successful public safety or community intervention programs and can highlight important progress being made on criminal justice issues that matter to your audience.

If you’re choosing to write about a crime that was unusually violent or perpetrated by a stranger, it can be helpful to add context explaining how often these crime characteristics actually occur. For example, homicides are often committed by family members and acquaintances, even though stranger-based crime can spark lots of media attention.

Even the word “crime” itself may not be specific enough when communicating with readers and viewers. Someone may be inclined to think of murder when they hear the word “crime,” while others may think of shoplifting or another kind of offense. Offering more nuance can prevent oversensationalizing.

Accessing crime data

Police agencies make crime data available in a variety of ways. Here is a tipsheet breaking down key strategies for accessing local crime data. Learn more about NIBRS and other national crime data sources in The Marshall Project’s companion toolkit here.

Expert sources

Criminologists and other experts can help you think through an analysis or offer valuable context about wider trends and criminal justice history. These experts have agreed to talk to journalists for articles about crime trends.

• Jeff Asher, AH Datalytics, co-founder. Asher runs AH Datalytics, a criminal justice data analytics firm, and he is a nationally recognized crime data analyst. He previously served as a crime analyst for the City of New Orleans and publishes a newsletter, Jeff-alytics, that explores the nuances of working with crime data. Email: jasher@ahdatalytics.com
• Shani Buggs, UC Davis Centers for Violence Prevention, associate professor of emergency medicine. Buggs works with a variety of stakeholders, including federal, state and local officials, and nonprofit organizations on community-driven violence intervention and prevention efforts. Her research has examined state-level firearm policies and local-level violence reduction initiatives. Email: sabuggs@health.ucdavis.edu
• Jeffrey A. Butts, John Jay Research and Evaluation Center, executive director. Butts is a research professor at John Jay College of Criminal Justice who studies the effectiveness of justice system policies and programs. He has consulted with journalists on the framing and methodology of crime data analysis. Email: jbutts@jjay.cuny.edu
• Jill Castellano, The Marshall Project, data reporter. Castellano reports on the uses and misuses of criminal justice data in American society. Her stories cover crime statistics, policing, the perception of crime and the broader criminal justice system. She is the author of this toolkit, along with our toolkit on how to report on FBI crime data. Prior to journalism, Castellano studied criminology and worked in a crime research lab. Email: jcastellano@themarshallproject.org
• Thaddeus Johnson, Council on Criminal Justice (CCJ), senior fellow. Johnson is also an associate professor of criminal justice and criminology at Georgia State University, a former law enforcement official in Tennessee, and an expert on policing, crime control, and governance equity. His research focuses on police lethality and coercion, street violence, recidivism, algorithmic and predictive bias, and correctional control. Email: tjohnson230@gsu.edu
• Ernesto Lopez, Council on Criminal Justice (CCJ), senior research specialist. Lopez has authored more than a dozen reports for CCJ on crime trends in large U.S. cities and served as research staff for CCJ’s crime trends working group. He presents to policy groups regularly, including the White House Domestic Policy Council and the National Governors Association. Email: elopez@counciloncj.org
• John Roman, National Opinion Research Center (NORC) at the University of Chicago, senior fellow. Roman is a nationally recognized expert on justice systems, firearms data and evaluating safety and justice programs. He also offers commentary on potential explanations behind rising and falling crime rates. Email: roman-john@norc.org