This story was produced in collaboration with Reveal from The Center for Investigative Reporting and Newsy.

As part of the FBI’s standardized Uniform Crime Reporting (UCR) program, thousands of law enforcement agencies across the country regularly tell the FBI how many serious crimes were reported in their jurisdictions and how many of them they “cleared.” Police agencies often point to the annual clearance rate — the number of cases cleared divided by the number of total crimes (excluding unfounded cases) — as a measure of their effectiveness. Clearance rates are often presented to the public as “solve” rates.

While clearance rates have been described as “cleared by arrest” rates in FBI documents, FBI rules actually allow the police to clear a case without making an arrest. According to the FBI UCR handbook, a case can be cleared by “exceptional means” if the answers to the following questions are all yes:

  1. Has the investigation definitely established the identity of the offender?
  2. Is there enough information to support an arrest, charge and turning over to the court for prosecution?
  3. Is the exact location of the offender known so that the subject could be taken into custody now?
  4. Is there some reason outside law enforcement control that precludes arresting, charging and prosecuting the offender (for example, suicide, deathbed confession, double murder, etc.)?

This means that the police departments have the ability to mark a case as “cleared” in a broad range of situations — if after talking to police the victim doesn’t feel comfortable proceeding, for example, or if the local district attorney tells the investigators he or she doesn’t want to proceed with prosecution. The outcome is that police clearance rates go up, making the agency look better to the public, while the suspect remains on the street. FBI rules state that cases cleared in this manner require enough evidence for arrest, but it is not evident that police always adhere to that standard.

Unfortunately, the system does not require most agencies to declare how many of their cases were cleared by exceptional means. More than 60 percent of law enforcement agencies reporting to the UCR program still use the older summary system, which does not distinguish between the two types of clearance. The FBI’s newer National Incident-Based Reporting System (NIBRS), however, does require law enforcement disclose arrests and exceptional clearances separately.

A review of the NIBRS data indicated that rape offenses are frequently cleared by exceptional means, yet neither the FBI nor the public has any way to know how many rape crimes are being cleared without arrest in the majority of cities across the United States. We hoped to uncover the arrest and exceptional clearance rates previously hidden from the public by requesting data from police internal case management systems in those places.

We used the UCR population data to determine every law enforcement agency in the United States that serves a population of 300,000 people or more. Twenty-six of those agencies report to NIBRS and the rest, 77 agencies, were summary UCR reporters. We sent requests under state open records laws to all 103 agencies asking for incident level data for all Part I crimes, the most serious crimes under UCR methodology, including the following for all cases from 2014 to 2016:

  1. The incident number of the offense
  2. Incident/offense date
  3. Type of offense
  4. Unfounded (yes/no), i.e., whether or not determined to be false or baseless complaint
  5. Unfounded date, if applicable
  6. Clearance date, if applicable
  7. Arrest date, if applicable
  8. Type of clearance (arrest, cleared by exceptional means, etc.)
  9. Exceptional clearance type, if applicable (offender died, victim uncooperative, offender deported, prosecutor declined case, etc.).

Not all jurisdictions were able and willing to provide each of these pieces of information, but we communicated extensively with each agency, even filing a lawsuit against one agency for what appeared to be disregard for local records laws. At the time of publication — more than a year after the bulk of our requests were sent — we had received nearly 70 datasets. Some were missing critical pieces of information. It was common for unfounded data to be missing from internal databases, as well as dates. We removed datasets that were missing clear disposition data.

When agencies were willing to communicate, we worked with the records departments to understand any caveats that may apply to the data and to navigate each agency’s encoding systems. The full Part I datasets were filtered to contain only rape crime data — these incident-level files have been made available by request, so local journalists can analyze the data and identify specific cases and their clearance statuses.

Most of the files contain incident or case numbers, which can be used to request the police reports from law enforcement agencies for review. In some cases, we combined or converted the files from other formats, such as PDFs. If included, we removed the names and addresses of victims. The final columns in our released files are nearly identical to those provided, with the exception of columns that were added during processing. Those columns are described in the documentation.

We identified the fields that contained case dispositions, including those that aligned with arrests and exceptional clearances, and calculated clearance rates by dividing the total number of cases cleared by arrest and exceptional means in the data by the total number of what the FBI calls “actual” rapes — reported rapes minus any unfounded cases. We also identified unfounded cases and, when available, the reason for exceptional clearances. We reached out to each jurisdiction prior to publication to give them a chance to comment on our interpretation of the data they provided.

To provide a point of comparison and to check the integrity of the data provided by each jurisdiction, we requested the summary UCR and NIBRS master files from the FBI. We parsed the summary UCR file as described in the provided documentation and the NIBRS file by adapting the NIBRS-to-UCR conversion methodology provided by the FBI. Most agencies had total rape counts and rape clearance rates close to the UCR data. When the difference was significant, we asked agencies to explain the discrepancies.

Understanding UCR and NIBRS Clearance Rate Calculations

Under the summary system, a rape offense is tallied as an “offense reported or known to police” in a reporting year if the incident was reported to the police between January and December of that year. Offenses are counted only toward the highest-ranked offense according to the UCR’s hierarchy rule. Then any cases that were unfounded in that reporting year are removed from the count to arrive at the number of “actual offenses,” even if the case that was unfounded was reported in a previous reporting period. The number of offenses cleared by arrest or exceptional means in the reporting year are then tallied, even if the cases being cleared were reported in a previous reporting period.

The NIBRS system counts individual incidents, which can comprise one or more offenses. NIBRS master files are converted back to the summary UCR system, including clearance rates, by counting a single offense for each incident according to the hierarchy rule, which generally is not used in the NIBRS system. Offenses are counted toward a reporting year based on the date the incident occurred, unless that date is missing, in which case the date reported may be used instead. Unfounded cases appear to be automatically excluded from NIBRS data, so they do not need to be removed in later stages. The number of offenses cleared by arrest is tallied by the number of incidents in the master file that have an arrest date in the reporting year. The number of offenses cleared by exceptional means is tallied by the number of incidents in the master file that have an exceptional clearance date in the reporting year.

The FBI documentation does not mention how unfounded cases, which are not tracked as part of the newer NIBRS system, are converted into the summary UCR dataset that is released in widely-used FBI products. . In the summary UCR master files, we found that the data shows NIBRS reporters as having zero unfounded cases, instead of blank or missing data. The result is deeply misleading, as we found that many of those places continue to unfound a high number of cases.

The UCR unfounding rate is calculated as the number of offenses unfounded in the reporting period (even if the offense was reported in an earlier year) divided by the total number of offenses reported or known to police in that period (including unfoundeds). This is different from the “actual offenses” denominator used for clearance rates, which excludes unfounded cases.

Understanding Agency Clearance Rates

While, in most cases, the total number of rapes and clearance rates are very similar in the requested data, summary UCR data and NIBRS data, it is important to note that there are subtle differences in how each rate is calculated.

Dates: Clearance rates are broken down by year, but the summary UCR and NIBRS systems differ on whether to count by the date the incident occurred or the date it was reported. This is how we prioritized date data:

  1. Some jurisdictions explicitly broke the the three years of data provided into three separate tabs or files. When this was available, this explicit distinction was used to count the offenses/incidents that fell into a year’s clearance rate.

  2. Most jurisdictions also provided a reported date field. This was used to group cases by year if the agency did not explicitly provide a year.

  3. In some cases, law enforcement agency data did not include either of these date elements, so the date in the incident number or the date of occurrence was used to tally crimes by year.

This methodology generally prioritizes the year a case was reported over the year it occurred, which is more in line with how summary UCR handles dates. NIBRS data, on the other hand, generally prioritizes occurrence date first and reporting date second.

Counting the clearance of old cases: Our clearance rates are a snapshot of the data for 2016 that was in an agency’s system at the time it was released. Case dispositions change over time, so a case that is open today could be closed by arrest tomorrow. Crimes can also be downgraded or unfounded to change the total count. The summary UCR methodology allows law enforcement agencies to count cases that were reported in previous reporting years but cleared in the current year toward the current clearance rate. The NIBRS conversion methodology also counts these cases to some degree. While it was requested in our records requests, many law enforcement agencies said they do not track clearance dates and arrests dates in their databases or that including that data would be prohibitively difficult. The current records data therefore looks at the “true” clearance rate for the specific cases that were reported to police in a certain year, not the rate of cases that are cleared during that period regardless of when they were reported. The dispositions in the data are the most up to date as of the date the jurisdiction provided the data to Scripps, which was generally in early to mid-2018. You can think of the summary UCR clearance as a backward-looking clearance rate and the rate we calculated from agency data as a forward-looking rate, but generally over time those two rates should be very similar.

Incidents versus offenses: Our records requests referred specifically to incidents, rather than offenses, but many provided counts of the number of offenses in an incident or included police reports with more than one rape offense twice in the data. As a matter of convention, we counted each incident row provided as a separate rape offense and used victim counts or offense counts to increase the number of total rapes when they were provided. NIBRS defines an “incident” as a “one or more offenses committed by the same offender, or group of offenders acting in concert, at the same time and place,” and a single incident can contain multiple offenses. We opted to count the crimes as they were provided, with a preference for counting individual offenses if that information was also provided. In data with duplicate incident numbers and in which the law enforcement agency did not provide clarification, duplicates were assumed to be purposeful indicators of multiple incidents or offenses in a single police report. The only exceptions to this convention were instances in which a jurisdiction clearly made an error in joining database tables and accidentally produced a large number of duplicates. In these cases, we consulted with the law enforcement agency and deleted the accidental duplicates.

Users of the agency incident data are encouraged to review our documentation carefully and to contact local agencies with any questions about how best to interpret specific crime and clearance codes. Questions about our interpretation of the data can also be directed to [email protected].

Do you have access to information about police handling of rape cases that should be public? Email [email protected], [email protected] or [email protected]. Here’s how to send tips and documents to ProPublica securely.