AS A NATION, we know much less about crime than we should:
- Word of new thefts of data from businesses or government agencies seems to arrive near daily—thefts that put thousands, if not millions, of Americans at risk of identity theft and serious financial harm. Cyberspace is not only an active and ongoing international battlefront, but also a perpetual domestic crime scene. The nation’s crime statistics have not been well equipped to measure cybercrimes such as the deployment of malware or coordinated attacks on computer networks.
- Drug production, sale, and possession are important crime categories, as is highlighted by today’s high rates of opioid overdose. But levels of these activities are not well characterized in current crime statistics; counts of seizures and arrests for drug law violations may say as much about law enforcement’s reaction to illegal drug activity than the levels and changes in the drug activity itself. Since 1991 the Office of National Drug Control Policy has periodically published estimates of the quantity (in metric tons) and dollar value of the major drug markets, along with estimates of the numbers of people using these drugs with various intensities. Those are arguably more informative metrics for characterizing the extent of drug-law violating than are arrests for those violations, yet they have not historically been part of conventional reporting on crime in the United States.
- Every day, inventory and resources “disappear” from the stocks of the nation’s retailers and businesses, some for explainable reasons but much through criminal acts such as theft/pilferage by employees or fraudulent resale. Collectively, the phenomenon is known as shrinkage and is poorly understood by the public, in part because businesses do not commonly
report all of their losses to law enforcement officials even though the billions of dollars in losses almost inevitably affect all consumers.
- Current national crime statistics lack the detail and the timeliness to address important concerns, even about offenses that are commonly thought to be well measured. For instance, in 2006 and 2015 in particular, news accounts and general perception fueled concerns about major increases in homicide despite the fact that the current homicide rate is less than half the rate of the 1980s and 1990s. Some local law enforcement agencies were equipped to use their own data resources to characterize the local problems, and databases created by media outlets and advocacy or other groups provided some additional insight. But the nation as a whole was hindered by the typical 10-month gap between the end of the calendar year and the release of current national crime statistics, unable to understand whether local patterns were part of broader regional or national patterns or whether increased homicide activities were limited to specific forms of homicide such as drug-related murder.
These are some examples of knowledge gaps in the nation’s current crime statistics. This panel’s Report 1 on Modernizing Crime Statistics (National Academies of Sciences, Engineering, and Medicine, 2016) focused on the definition and concept of crime; this final report considers how the nation can remedy these important knowledge gaps.
CURRENT NATIONAL CRIME STATISTICS, AND THIS PANEL’S WORK
For decades following its establishment in 1930, the Summary Reporting System of the Uniform Crime Reporting (UCR) Program (SRS, or UCR Summary) was the principal source of U.S. crime statistics, compiled and published by the Federal Bureau of Investigation (FBI)—since 1959, in a report bearing the prominent title Crime in the United States. The UCR data are aggregated tallies of offenses reported to or otherwise known to law enforcement agencies; the source data are voluntarily contributed to the national collection by way of a network of state-level coordinating bodies. Beginning in 1972, the nation’s crime statistics were joined by an essential, complementary data collection—what is now known as the National Crime Victimization Survey (NCVS) sponsored by the Bureau of Justice Statistics (BJS). The NCVS uses household interviews to produce estimates of criminal victimizations that are not necessarily reported to law enforcement, the so-called “dark figure” of crime. More recently, a mid-1980s evaluation led the FBI to institute the National Incident-Based Reporting System (NIBRS)—characterized as the future of UCR data collection but marred by decades of slower-than-anticipated adoption by state and local law enforcement agencies.
In 2016, the FBI formalized a decision to retire the UCR Summary system by 2021 in favor of a full-fledged NIBRS.
At the suggestion of the U.S. Office of Management and Budget (OMB), this Panel on Modernizing the Nation’s Crime Statistics was engaged to review the whole U.S. crime statistics enterprise. Specifically, the panel was tasked with a three-part charge by BJS and FBI, the full text of which is in Appendix A. Report 1 covered the first, substantive plank in that charge: performing a comprehensive reassessment of what is meant by crime in U.S. crime statistics, and recommending a new classification of crime to organize measurement efforts. This report covers the two remaining planks of the charge:
- Methodological issues, assessing the adequacy and appropriateness of current crime statistics collections and suggesting the best conceptual means of collecting data based on the recommended 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 issues, describing how national crime data collection should 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 while being mindful of the voluntary nature of crime data reporting.
By the nature of panel studies like this one, and given the sheer extent of the panel’s charge, this report is necessarily a conceptual blueprint for modernizing crime statistics rather than a precise and detailed implementation plan. But it serves to highlight the importance of the work to better understanding key national concerns.
REDEFINING AND MEASURING CRIME
Pursuant to our charge, we take a broad perspective on crime by defining it generally as socially unacceptable behavior that causes harm or threatens harm against another. Our behavioral definitions are not tethered to the specific legal language of federal or state criminal codes, and so treat as “crime” some things that might commonly be construed as tortious or civil because they are usually adjudicated in civil courts or regulatory proceedings. By this, and by construction of our proposed classification of crime offenses for statistical purposes, we do not suggest that any single offense or group of offenses is more important than another, just that all are essential for a complete, composite picture of crime.
This broader definition of crime is different from decades of crime statistics practice, and a brief trace of historical themes puts some fundamental issues
in stark relief. Following a short-lived initial period of genuine enthusiasm at what could be learned from national crime statistics, the content of those statistics and the routines for producing them ossified—became overly rigid and stagnant—during the decades in which the UCR Summary system was the sole source, with major consequences. Initial candor about the completeness of the UCR data and clarity about their limitation to offenses reported to law enforcement became overwhelmed by an undue sense of comprehensiveness. Once the annual report of UCR data acquired its still-current name in 1959, Crime in the United States came to define “crime in the United States” in many eyes, restricting attention to a narrow set of specific offenses and further still on those offenses known to law enforcement. More fundamentally, the mindset of estimating crime—measuring its levels, trends, and characteristics—gave way to one of appearing to completely enumerate it, reducing the study of crime to the reporting of short-term changes in a few, broad categories of crime rates. The emergence of the NCVS as a counterweight to police-report-only data, and the introduction of NIBRS two decades later, made it possible to further widen the view, but the historical restrictions remain strong.
To provide detail on the coverage gaps in current crime statistics, this report matches the crime classification described in Report 1 to the coverage of current nationally compiled crime statistics sources (NIBRS and NCVS), as detailed in Appendix C and summarized in stylized form in Box S.1. Categories 1–5—which loosely correspond to the “traditional” violent interpersonal crimes and property offenses that dominate current crime statistics—are reasonably well covered in current crime data, albeit with some major conceptual differences. But offenses in our categories 6–11 are almost completely new terrain for U.S. crime statistics. Some (such as cybercrime) are “new” and emerging types of crime; others (such as environmental crimes and many types of fraud) would be new only in statistical tallies, having received little to no standardized national collection to date even though they have long been serious problems. A central observation is that any report that wishes to claim to present a comprehensive view of crime in the United States must pay serious attention to categories 6–11, as is done for categories 1–5.
One of our conclusions is meant as a strong corrective to decades of crime statistics being characterized or interpreted as an enumeration of all crime: A system of modern crime statistics will be more useful if it can provide more information and more flexibility for detailed analysis than simple summary counts.
Conclusion 2.1: The aim of modern crime statistics is the effective measurement and estimation of crime. Accurate counting of offenses and incidents is important, but the nation’s crime statistics will remain inadequate unless they expand to include more than just simple tallies with no associated measure of
uncertainty or capacity for disaggregation. Through the collection of associated attribute data, the crime statistics we suggest should—at minimum—enable the analysis of data in proper geographic, demographic, sociological, and economic context, and provide the raw material for important measures related to an offense (such as the harm it causes) in addition to its count.
In this study, the brand of crime data collection that we propose is incident-based, recording attributes describing incident, offender, and victim characteristics that are essential for disaggregation and analysis. For any particular offense type, the motivating questions should be what does the nation need to know about this kind of crime, and what is the most accurate or most feasible way to collect that information, with the desired degree of analytical detail. Uniformity of definitions and underlying concepts is essential, but
absolute adherence to any one mode of data collection (e.g., police-report data versus survey-based data) is not; nor is observing any one portion of processing in the criminal justice system an imperative (i.e., some offenses may best or only be measured at the point of arrest or of charges being filed in court rather than earlier in the investigative phase).
Conclusion 2.1 is phrased as it is to convey that different measurement types may be needed for different specific offenses—and there are offense types for which counts alone may be inappropriate or misleading. This can be said of offenses falling on either side of our traditional/new divide, but is particularly acute for some of the new crime types:
- Deployment of ransomware involves malicious software that denies users access to their own data and computer systems until a ransom is paid. But there are material differences between a ransomware deployment that locks up an individual home-computer-user’s machine and one that hobbles use of the information systems of entire hospital/medical groups or law enforcement departments; an offense count of 1 does little to convey what actually happened and improve understanding.
- Likewise, counts of environmental violations may be deceiving. Improperly disposing of a laptop computer battery and spilling a large amount of toxic materials into the water or air might each count as one event. Conversely, two events of the same size might have quite different consequences: Spilling the same number of barrels of oil into a water supply may not be equivalent to spilling the same number of barrels of oil in a nature reserve or an industrial site.
A related challenge in crime measurement is the timeliness of estimates, and the tradeoff between timeliness and precision. For some offense types including homicide, it may be essential to generate more frequent, interim estimates at the declared expense of increased uncertainty in addition to more refined final estimates.
STRUCTURE OF NEW NATIONAL CRIME STATISTICS COLLECTION
A related conclusion recognizes that no single data source can span the full range of crime; rather, a new crime statistics system will necessarily be a system of multiple data collection efforts:
Conclusion 2.3: Improvement in the nation’s crime statistics will require enhancements to and expansions of the current data collections, as well as new data collection systems for the historically neglected crime types highlighted by the proposed crime classification.
It follows from this conclusion that there is an important role going forward for the core methodologies currently at work in national crime statistics, albeit with refinements to those methodologies, as well as new approaches.
Data on offenses reported to or otherwise known to law enforcement—such as those embodied in NIBRS—will continue to be essential for U.S. crime statistics. Both of the above conclusions show strong and enthusiastic support for both the retirement of UCR Summary reporting and for full implementation of the National Crime Statistics Exchange (NCS-X) initiative (managed by BJS including funding from the FBI). The NCS-X initiative is working to target NIBRS development in a sample of jurisdictions so that the combined NIBRS-format data will provide a nationally representative sample, something that has not happened given the system’s slow adoption rate for the past 30 years. But we emphasize a point made in closing Report 1: A full-local-compliance NIBRS is necessary but not sufficient. This is because a full NIBRS would still fall short of measuring all the offense types for which police-report data might be the most appropriate and accurate avenue, and it covers some offenses (notably assaults and threats, as well as the controlled substance offenses in our category 6) in a relatively uninformative way.
Another important point on the future of voluntary-contribution police-report crime data is that they will succeed and endure to the extent that they become a largely invisible by-product of daily operations at local law enforcement agencies. Individual law enforcement agencies differ in the priority they assign to the preparation of incident reports by officers and the tools available to produce them (ranging from hard-copy paper forms to mobile computer terminals in patrol cars). But, to the extent that more agencies see reporting (and the resulting crime statistics) as important means of both doing their work and of bolstering their accountability to the communities they serve, the more national crime statistics and their users will be able to benefit from a system where reporting to national collections is not a chore or something done out of a sense of obligation.
Self-report survey information on criminal victimization—including but not limited to the NCVS—will also remain a pillar of U.S. national crime statistics. Survey data on crime are necessarily limited in the degree to which the resulting data can be disaggregated along useful analytical lines; certain crime types are still, relatively speaking, rare events in the population at large, so the survey’s sample size would have to be unworkably large to support fine-grained estimates. But survey data should remain part of the crime statistics infrastructure because of their unique, conceptual strength—their ability to estimate the level of victimization that is not reported to law enforcement authorities. In this report and this project, we say less about the NCVS and survey measures than about police-report methodologies; this is due to the fact that the NCVS has been in a state of near-continuous refinement and improvement, rather than a reflection on the importance of the data.
The frontiers for methodological and implementation work on the survey information side of national crime statistics are fairly well known and already active. These new frontiers include finding ways to directly or indirectly generate some subnational estimates from the data and developing techniques for measuring crime types (notably rape and sexual assault) that are particularly sensitive and difficult to measure, as well as making more effective use of interview time (including with households with no victimization experiences to report) to generate the kind of community-context variables that might be used in drawing meaningful comparisons between crime levels in multiple jurisdictions. It must also be noted that there are some offense types for which survey-based measures of perceived victimization, even at the expense of geographic precision, may actually be the most accurate available, as well as the most salient to data users and stakeholders. Identity theft is one crime type where NCVS has already proven an effective measure, and others might include other classes of fraud as well as stalking and harassment offenses. Again, though, the quality of victimization survey data is contingent on eliciting sufficient responses on possible crime incidents in order to draw inferences, and so it is critical that the NCVS be adequately resourced to fulfill its purposes.
To these two core methodologies, we add a third: the aggregation of information through a crime measurement clearinghouse function, perhaps primarily administrative record-type data drawn from external resources including federal and state agencies. These new types of information are necessary for the new types of crime in our classification, and developing them will require some substantial initial attention and investment. This is merely because they have not been an active part of crime statistics infrastructure to date and their use in measuring crime is unfamiliar terrain, and not because they are more important in any sense than the more traditional methodologies. It is also true that these new data resources may prove to be effective measures of some older, more traditional offense types as well.
Generally, new data resources must be harnessed because there are many crime offense types for which neither police-report data nor survey data are apt or workable as a source of offense counts and characteristics. One obvious example are crimes against governments and businesses that are not specifically spatial in a way that is linked to a local police jurisdiction. If a particular convenience store is robbed, the local police will likely be contacted; if a corporation’s intellectual property is stolen by hackers attacking from overseas, the local police will not be involved. Indeed, the corporation may not notify any policing agency to avoid reputational damage done by public acknowledgement of its victimization. There are many other reasons that police report and victimization survey data are limited in what they can provide. For example, local law enforcement—traditional police—may either not have true jurisdiction relative to other regulatory agencies (e.g., environmental offenses) or simply may not be the first natural point of contact (e.g., fraudulent credit
card transactions). In others, the crime manifests later in the sequence of events in the criminal justice system, such as the filing of charges or resolution through settlement or plea arrangement (e.g., corruption, bribery).
The panel’s next conclusion returns to the base from which we work in suggesting the contours of national crime statistics going forward—the classification of crime developed in our Report 1:
Conclusion 2.4: The classification of crime for statistical purposes recommended in Report 1 should be used as a blueprint for the nation’s crime statistics. It should be used to guide revised measures of crimes currently included in existing data as well as pilot/initial data collection of as-yet-unmeasured crime types—even though it is impossible today for data collection to include every crime in the classification. While it is imperative that the classification undergo regular update and revision, initial data collection should begin with this as a base, rather than continue to refine the classification in the absence of data.
COORDINATION AND GOVERNANCE
The task of creating or modernizing any one of the subsystems we envision as part of modern U.S. crime statistics is difficult; implementing them all is harder still. And a guiding precept in constructing the classification of crime is that it should evolve, being revised over time to keep up with stakeholder needs and the changing nature of crime itself. In the panel’s view, the essential point is to get started, with some measures and refinements, and not let the perfect be the enemy of the good. Real change cannot happen productively until first steps are taken and prototype measures are produced and scrutinized.
We gave extensive consideration to the factors that led to much slower adoption of NIBRS-format reporting than was anticipated—factors that affect the successful implementation of changes to the collection and content of crime statistics in general. In our assessment, the single greatest barrier to change in the nation’s crime statistics system is there is no strong sense of ownership, or ultimate responsibility for, those national crime statistics. Both the FBI and BJS have strong, clear stakes in crime statistics, but the culture that evolved after 1930 was a relatively passive, hands-off mode of leadership at the national level. In important ways, this style is entirely predictable and understandable—stemming as it does from the reservation of most policing and criminal justice functions to the states under the U.S. federated system (and not the federal government), and also flowing out of abundant respect for the voluntary nature of contribution of data to national statistics, whether that is local law enforcement agencies filing UCR/NIBRS data or respondents completing an NCVS interview. But legal inability to impose on crime data
providers and deference to voluntary contribution has led to a reluctance to motivate or incentivize crime data provision in any way, whether through “carrots” (e.g., direct technical-support funding) or “sticks” (e.g., linking data-reporting completeness to grant fund eligibility). Arguments can be made on all sides, but the end result is a crime statistics system that lacks essential leadership.
Specifically, we note that there is neither a strong coordinating body nor a defined governance structure for the whole U.S. crime statistics enterprise, dedicated to the generation of data on all types of crime. There is no reason why these two functions, coordination and governance, need to be conducted either by the same agency or by two different ones; we distinguish the two roles because there are subtle differences in what we see as the ideal roles in both areas. A strong national-level coordinating agency for crime statistics would be more actively and directly involved in the routine daily tasks of data collection and processing; it would have to cultivate and maintain an honest-broker role in interacting with data providers as well as providers of computer hardware/software solutions. The chief entity in a governance structure would take the broader view; its duties would include setting policy for the system as a whole, establishing the process for updating and maintaining the underlying classification of crime, providing a voice for and attention to feedback from the range of crime data stakeholders, and facilitating ongoing methodological research and development.
Two more conclusions by the panel note the absence of both these functions and make the direct appeal for them to be created:
Conclusion 3.1: A stronger federal coordination role is needed in the production of the nation’s crime statistics: providing resources for information systems development, working with software providers to implement standards, and shifting some burden of data standardization from respondents to the state and federal levels. The goal of this stronger role is to make crime data collection a product of routine operations.
Conclusion 3.2: Having an effective governance structure for the complete U.S. crime statistics enterprise is critical. There is currently no entity responsible for reporting on the full range of crimes in the proposed classification (most notably for top-level categories 6–11).
The important issue of who should play these coordination and governance roles remains to be answered. Rather than make any recommendations about where the coordination and governance roles should be housed, the panel chose to identify requirements that any organization should fulfill in order to successfully carry out these two roles. What is critical is that
national crime statistics be produced, coordinated, and governed pursuant to the expected sensibilities of a statistical agency, one for which data collection and dissemination—and attention to issues of data quality—are essential to the agency’s mission. Specifically, this means adherence to the expected Principles and Practices of a Federal Statistical Agency that are regularly updated by the Committee on National Statistics of the National Academies of Sciences, Engineering, and Medicine (2017c).
Neither of the major agencies that currently collect crime data, BJS and the FBI, have all the necessary requirements, although both have some relevant attributes. BJS’s authorizing language currently contains a very detailed list of specific duties—several of which are directly pertinent or responsive to the ideal coordination and governance roles. However, BJS has never been given the authority or been resourced in a way in which it could begin to fulfill all of them. Conversely, the legal authorization for the FBI’s UCR Program is considerably more vague but has long endured and been supplemented over time. Addressing the new crime categories that have long been neglected requires paying attention to domains that fall outside the traditional highest priorities of either agency. Whatever agency (or agencies) serves as central coordinator for U.S. national crime statistics can only be effective if it can either work effectively with—or compel information resources from—myriad sources. This includes executive agencies spanning multiple problem domains and cabinet departments, independent agencies such as the Federal Trade Commission and the Environmental Protection Agency, quasi-independent entities such as the Postal Inspection Service, and federal and state law enforcement entities (including Offices of Inspector General). It is not clear if either BJS or the FBI is currently positioned to be successful in expanding the collection of crime statistics.
We are not aware of relevant research literature or other scientific footing that can serve as a basis for concluding that any particular organizational-chart approach is superior to another. Instead, we defer to the agency that may be best suited to manage the complex set of federal agency actors (and state and private authorities) who would ultimately be called on to contribute to crime data:
Recommendation 3.1: The U.S. Office of Management and Budget (OMB) should explore the range of coordination and governance processes for the complete U.S. crime statistics enterprise—including the “new” crime categories—and then establish such a structure. The structure must ensure that all of the component functions of generating crime statistics are conducted in concordance with the sensibilities, principles, and practices of a statistical agency. It should provide for user and stakeholder involvement in the process of refining
Underscoring the statement made early in this summary, that crime statistics should not be limited only to broad counts of offenses, this report also argues that modern crime statistics require consideration of completely different ways to measure crime (in addition to improving the ways that it is counted or its level is assessed). We conclude that alternative metrics such as the harm imposed by or costs associated with crime merit research and, eventually, measurement and reporting:
Conclusion 2.2: There is a problematic lack of solid information about the relative magnitude of the costs, harms, or importance across crime offense types or specific incidents. Further research could be sponsored on estimating the magnitude of societal harm associated with the major categories in the proposed classification—ideally, in a way that facilitates comparison over time and offense types.
Appendix D of this report includes longer-form discussions of some additional important topics, mentioned in brief in the main text but fuller consideration of which might otherwise disrupt the narrative flow. These include:
- Examining the commonly cited barriers that impeded adoption of NIBRS for 3 decades, including discussion of what has changed now to make success more likely;
- Arguing that improved data quality should be the major goal of voluntary contribution of data to national crime statistics;
- Observing that—rather than warn against overly simplistic comparisons of crime rates across jurisdictions and time—the dissemination platform for modern crime statistics should also provide easy access to the kind of contextual and community data that enables effective comparison; and
- Describing the challenges of getting a better reading on crime involving businesses and organizations as actors.
The path ahead for U.S. national crime statistics is neither easy, inexpensive, nor short in terms of time to completion. To modernize crime statistics will require a large investment in collecting data on things that are currently all but ignored in current crime measurement efforts. An adequately resourced
NCVS and a full-participation (and fully representative) NIBRS are important steps. Real benefit will arise from utilizing these tools and finally starting to shed light on long-neglected offenses, consistent with a rigorous classification and drawing from third-party administrative-type data resources that have not typically been thought of as part of the nation’s crime data infrastructure. That data infrastructure should enable longer-run research and development on both new auxiliary measures of crime (such as the harms done by offenses) and on innovative methodologies for measuring offenses that elude detection in the administrative-type data resources. Although modernizing crime statistics will be challenging, it is critical to encourage these efforts. Crime and societal responses to all forms of crime result in substantial costs to the nation. Improved measures of crime are imperative for establishing policy priorities, evaluating policy effectiveness, and providing accountability data for agencies charged with responding to crime. There is great benefit to be realized from better statistics on the nation’s crime problems; it is time to develop a more inclusive picture of crime in the United States.