THE BUREAU OF JUSTICE STATISTICS (BJS) and the Federal Bureau of Investigation (FBI), at the suggestion of the U.S. Office of Management and Budget, jointly requested that the National Academies of Sciences, Engineering, and Medicine create this Panel on Modernizing the Nation’s Crime Statistics. Our panel was given a detailed and broad charge to assess and make recommendations for the development of a modern set of crime measures in the United States and the best means for obtaining them. Specifically, the charge (the full text of which is in Appendix A) obliges our panel to pursue three major lines of inquiry:
- Substantive, developing a new classification of crime by weighing various perspectives on how crime should be defined and organized with the needs and demands of the full array of crime data users and stakeholders;
- Methodological, suggesting the best (conceptual) means of collecting data based on the suggested crime classification, including the integration of data on specific crime types from non-BJS or FBI sources and enabling the identification and measurement of emerging crime types going forward; and
- Implementation, recommending how crime data collection should actually proceed, practically and effectively, including suggesting how to leverage available information technology assets, accommodate the demands of crime data stakeholders, and moderate the burden placed on crime data providers.
This first report from the panel addresses the first plank of this charge, deferring detailed discussion of methodological and implementation plans to our second and final report. The challenges of implementing new measures of crime are sufficiently great that there is a natural temptation to jump straight to them. But the core task of this first report is hardly negligible; it is, in fact,
foundational and essential to development of an effective crime statistics system. What has been lacking for nearly a century is a comprehensive reassessment of what is meant by “crime” in the United States. A modern classification of crime, such as we propose in this report and as a product in its own right, provides a framework for conceptualizing crime and a blueprint for the construction of statistical measures.
In fulfillment of the panel’s charge, a centerpiece of our work thus far has involved inviting comment and open discussion with a full spectrum of crime data users and practitioners. This work, which was particularly concentrated in two workshop-style meetings held during summer 2014, is rare in that information from such a wide cross-section of stakeholders has been elicited and weighed. Moreover, we focus heavily on BJS, FBI, and the nation’s two principal, nationally compiled sources of crime data—the FBI-coordinated Uniform Crime Reporting (UCR) Program and the BJS-sponsored National Crime Victimization Survey (NCVS). Such focus is appropriate, yet it is also in keeping with our charge that we do not focus exclusively on those sources.
OVERVIEW AND CORE PREMISES
Our argument in this report builds in several, detailed steps:
- Addressing the fundamental questions of what is meant by “crime,” and suggesting why a classification is a valuable framework (Chapter 1);
- Summarizing both the current primary sources of nationally compiled crime statistics and an illustrative sample of data collections—not currently considered part of the nation’s crime statistics infrastructure—that may provide useful “crime” or contextual information (Chapter 2);
- Identifying primary uses of and demands for crime data by a wide array of users and stakeholders (Chapter 3); and
- Reviewing extant and historical classifications of crime, including those used in other nations, for appropriate exemplars for crafting our proposal (Chapter 4).
Though all of these pieces are integral to the broader argument, we compress or omit much of the detail in this Summary for brevity’s sake.
However, it is important that we establish two premises—that inform and are woven through the detailed chapters—clearly up front. First, the classification that we develop is a conceptual framework, relative to which data (crime statistics) will eventually be collected. That is, we anticipate that the classification will be used as a blueprint for designing data collection, and not as something used only as a post-processing, labeling step. The upshot is that the classification stands and has value as an organizing framework in its own right—quite separate from the formidable challenges that remain in implementing new data collection based upon it. Second, it should be made
clear that we are dealing with two different units of analysis. For purposes of building the classification, we think in terms of the criminal offense as the relevant unit; the classification attempts to partition the entire space of “crime” so that a particular offense behavior corresponds to one and only one category. Yet the practical unit of analysis on which we concentrate (and on which we anticipate data collection to be based) is the incident, comprising one or more offenses and “linking” one or more offenders with one or more victims. Crime is a complex phenomenon, and working with incidents (that may be amalgams of individual offenses) is essential for flexibility.
CURRENT PRIMARY SOURCES OF NATIONAL CRIME STATISTICS
Since 1930, the FBI has served as central coordinator of data on offenses known to the police, combining reports originating from some 18,000 local law enforcement agencies under the auspices of the UCR Program. Most public references to “the UCR” and its numbers are actually to the collection of known-offense counts by the Summary Reporting System (SRS), but the program has expanded over the years to cover other law enforcement measures—and efforts to build a more detailed National Incident-Based Reporting System (NIBRS) have progressed relatively slowly over the past three decades. And, since the early 1970s, the NCVS (as it is now named; originally, the National Crime Survey) sponsored by BJS has used the power of personal interviewing to view crime from the perspective of the victim and to estimate levels of total crime—including that which is not known to or reported to police. Despite their prominence in the field, though, neither the UCR nor the NCVS is designated by law or regulation as the nation’s “official” measure, at least in part because both possess unique strengths and limitations as measurement tools.
The assembly of police chiefs that defined the content of the UCR in 1929 did valuable—and highly pragmatic—work, focusing heaviest attention on a very small set of criminal offenses that they judged to be both most salient in the public eye and most tractable for uniform tabulation (in that variations in definitions across federal and state legal texts would be minimized). The crux of the problem that motivates this first report is that this useful standardization rigidified over time. The short list of UCR-defined offenses constrained the view of “crime” made possible from the resulting statistics, disproportionately weighting selected “street” or violent crimes over other offenses. Moreover, the UCR’s strictures helped set the content of subsequent data collections such as the NCVS. To be sure, changes in UCR protocols and coverage have occurred over time as the result of enacted law or the result of extensive deliberation and vetting by an advisory board structure.
Accordingly, we reach two related conclusions that inform the development of our proposed classification:
Conclusion 5.1: The definitions and concepts in the current U.S. crime statistics system were developed primarily from categorization of statutory language, which varies by jurisdiction. Reliance on statutory language is inflexible and not comprehensive, and it is unduly focused on limited input sources (reports from police/law enforcement or individual victims).
Conclusion 5.2: “Crime” continues to evolve and take different shapes. Accordingly, there is a need for an expansive framework for crime classification that is amenable to periodic revision.
Our user and stakeholder engagement suggested great desire and demand for reliable, consistent statistical series concerning crime that support the fundamental goal of comparability between areas and over time. This focus on comparability holds whether the comparison is implicit in the allocation of federal and state funding for criminal justice programs, explicit in the evaluation of the effectiveness of local law enforcement initiatives or state policy changes, or simply in the diagnosis of emerging problems by the media, researchers, or the public at large. Another common and critical use of crime data is their increasing use (and, certainly, capacity for misuse) as a tool for accountability—primarily by law enforcement agencies in administering their own daily operations, but also by advocacy groups and the public. For these and other high-level uses, existing crime data appear to be adequate (though users noted many ways that the extant data could be improved)—but for other uses and crime types, existing data are incomplete, inadequate, or unavailable. From these considerations, we draw the clear and basic conclusion that developing adequate crime statistics for the United States will require more than a single data collection and more than minor modifications to the current UCR Program or NCVS:
Conclusion 3.1: There is strong demand for comprehensive, yet detailed, information about crime by a broad range of users. No single data collection can completely fulfill the needs of every user and stakeholder, providing data with sufficient detail, timeliness, and quality to address every interest of importance. Any structure devised to measure “crime in the United States” should necessarily be conceptualized as a system of data collection efforts, and informative details about the collection and quality of the distinct components in this data system should be included to help ensure proper interpretation and use of the data.
DESIGN PRINCIPLES AND OBJECTIVES FOR A NEW CRIME CLASSIFICATION
An essential first step in developing such a comprehensive system is a modern classification of crime that is—importantly—a classification for statistical purposes. The current practice of labeling offenses through reference to a standard, but fairly unstructured, list is “classification,” but only in the weakest sense. By classification for statistical purposes, we mean a framework that is exhaustive of all crime (not just restricted to violent crime or selected property crimes) and that uses clear definitions to partition offenses into mutually exclusive categories.
After considering crime classifications developed in recent years by other countries, we further concluded that a hybrid approach between a pure attribute-based classification and a traditional definition-based classification would be ideal. In 2016, as in 1929–1930, there remains substantial variation in the wording and extent of content in federal and state criminal codes. Accordingly, developing a classification scheme that emphasizes the basic criminal behavior in category definitions (as opposed to trying to precisely match penal code definitions or what constitutes an offense) seems like a better approach. But we also think it important that the classification’s category listing be accompanied by collection of a set of fairly detailed attributes of the incident itself, and of the victims, the perpetrators, and their relationship. The value of the added attributes is that they permit end users to disaggregate (or even reclassify) criminal events as they see fit, focusing on relevant subclasses of particular events.
PROPOSED CLASSIFICATION OF CRIMINAL OFFENSES FOR STATISTICAL PURPOSES
In recent years, the United Nations Office of Drugs and Crime (UNODC) has been the organizing body for an expert work group in this area, drawing from the experience of countries such as Australia and Ireland that pioneered comprehensive overhauls of their crime classification systems. The UNODC group’s first version of an International Classification of Crime for Statistical Purposes (ICCS) is an attribute-based classification that is very consistent with the structures we wanted in a U.S. classification scheme.
Conclusion 5.3: The International Classification of Crime for Statistical Purposes (ICCS) framework, proposed and maintained by the United National Office on Drugs and Crime (UNODC), meets the desired criteria for a modern crime classification, and the use of shared, international frameworks enables studies of transjurisdictional and locationless crime.
Accordingly, our suggested classification draws heavily and extensively from the ICCS, and we suggest its use in developing a new set of crime indicators.
Recommendation 5.1: The attribute-based classification of offenses described in brief in Sections 5.2.1–5.2.2 and in detail in Appendix D should be used as an initial framework for developing modern statistical measures of crime in the United States.
Our suggested classification differs markedly from current U.S. practice by omitting “aggravated assault” as a category, and from the ICCS in delineating components of what was previously bundled together as “aggravated assault” into specific categories meant to capture information on shootings and firearm involvement in nonlethal assaults and threats. Aggravated assault is one of the original crimes identified for UCR coverage in 1929, but ambiguity has clouded its precise definition even from the outset. A clause in the UCR definition of aggravated assault that invokes the label when a firearm/lethal weapon is involved muddles things considerably, so that the label could be said to apply equally to a shooting that causes near-deadly or disabling injury as well as to a barroom confrontation where no physical harm is inflicted at all but a firearm is brandished. Misclassification—whether accidental or deliberate—between aggravated assault and simple assault is a longstanding source of frustration with current UCR police-report data, and is a primary issue when law enforcement departments are charged with improperly adjusting crime numbers so as to make a city’s violent crime totals appear better than they actually are. In the meantime, information on firearm-related incidents is obscured—on noninjurious threats (brandishing, or firing in victim’s general direction), assaults (nonlethal wounding), and homicide, alike. The nation as a whole lacks reliable measures of shootings, even though our stakeholder outreach suggested that the public expects and demands that their local law enforcement officials know, understand, and address shootings on a daily basis.
Our suggested classification delineates topic areas that are new in the U.S. crime statistics enterprise, such as several types of fraud and corruption, crimes against the environment or natural resources, and sexual exploitation of children. We broaden the concept of threatening or dangerous behavior to include separate categories for harassment and stalking that emphasize the “course of conduct” (pattern of repeated behavior over some period of time) nature of those crimes. We seek to remedy a point of ambiguity in current crime definitions by constructing a separate category of rape involving the inability of the victim to express consent or nonconsent. Previously, such cases might not have fit the precise definition of rape or penal code-based concepts of statutory rape. The suggested classification also includes a category for terroristic threats to buildings or critical infrastructure, where those threats convey clear intent to either directly harm or cause serious disruption.
If used as a blueprint for constructing a new set of indicators of crime in the United States, the classification we suggest in this report has the potential to be path-breaking in advancing the nation’s understanding of crime. The next phase of implementation and methodological issues is daunting and should begin naturally with a revisiting of our suggested crime classification and mapping it to current (or not-as-yet created) data sources that might supply the requisite information. We recommend this classification of crime as a preliminary step as it is not (yet) a list of offense codes that we think can or should be swapped into any current data collection immediately. It is a starting point for discussion and development and not the final word of a debate. Indeed, it is important in adopting a new crime classification that mechanisms for periodic revisiting and reevaluation of the classification be established at the outset.
FUTURE DIRECTIONS: TOWARD IMPLEMENTATION OF NEW CRIME STATISTICS
This particular moment in time is uniquely pivotal for U.S. crime statistics, coming as it does in the wake of public statements by the director of the FBI—noting frustration with the state of those statistics and indicating intent to sunset the UCR Program’s Summary Reporting System in favor of a full-fledged NIBRS. BJS, in partnership with the FBI, is now in the process of equipping a sample of law enforcement agencies to begin reporting data in NIBRS format (through the National Crime Statistics Exchange [NCS-X] Program). The NCS-X will showcase the analytic power of NIBRS because data will be available from a nationally representative sample of jurisdictions. The major organizational bodies of chiefs of police (including the International Association of Chiefs of Police, the Major Cities Chiefs Association, the National Sheriffs’ Association, and the Major County Sheriffs’ Association) recently issued a joint statement supporting the concept of the FBI’s transition from Summary Reporting to NIBRS. Against this backdrop, this first report can only speak directly to conceptual and classification issues and not the mechanics of implementation or revising practice.
Given the immediacy of the debate, this report’s discussion of classification supports a few relevant points that are important to state in this first report rather than wait for the final report. Among these: Even the most cursory comparison of our suggested classification with the current list of crimes covered by NIBRS, by the NCVS, and particularly by the SRS suggests that the gaps in current coverage and knowledge—trying to fill in our classification with current data—will be numerous and glaring. Accordingly, we conclude:
Conclusion 5.4: Full-scale adoption of incident-based crime reporting by all respondents or sources, that is sufficiently detailed to permit accurate classification and extensive disaggregation and
analysis, is essential to achieving the kind of flexibility in crime statistics afforded by a modern crime classification.
There are two basic corollaries to this argument that are supported by this report’s classification-based focus that we make now in order to contribute in a timely manner to current debates, even though more detailed discussion is reserved for our second report. First, the proposed transition away from the SRS format and content is sound and appropriate. The SRS was a major advance when created in 1929–1930 and proved instrumental for decades in shedding basic light on national crime trends, but it is simply inadequate to provide information of the quality or the level of detail demanded by modern crime data users. But the second corollary concerns the transition to NIBRS: It is important that the transition to NIBRS be cast strictly as an intermediate step. Even a full-participation NIBRS that holds to that system’s current design and content would have great difficulty satisfying all or most crime statistics user needs. Our concern is that NIBRS’ core development work and structuring took place in the late 1980s, and it is not clear that its design has kept pace with the times. Upgrading from a 1929-vintage crime data management system to a 1990-vintage system falls well short of the data infrastructure that crime statistics deserve, but it would indeed be a remarkable advance.