We summarized the history of crime statistics in the United States in Report 1, albeit with a focus appropriate to the substantive focus of that volume: the meaning of crime and the range of offenses covered in crime statistics. But history affords valuable lessons, both aspirational and cautionary, in reinventing U.S. crime data collection, through emphasis on the establishment and implementation of existing systems. As context for discussion of the structure of a modern crime statistics program and its implementation, this appendix traces a number of important historical themes, among them:
- A transition from initial enthusiastic use of crime statistics as a tool for understanding the dynamics of crime to an inertial production-oriented mode that unduly limited the public concept of crime;
- Promising steps in more recent decades toward broader concepts of “crime” in national crime statistics and to measuring or estimating crime (rather than giving the appearance of completely enumerating it), including the emergence of the National Crime Victimization Survey (NCVS) and advent of the National Incident-Based Reporting System (NIBRS);
- Tendencies from the earliest days of national crime statistics collection to collect some additional information about crime beyond crude, total
offense count—as well as a tendency to greatly misjudge, also from those earliest days, the information resources available to local police departments and agencies; and
- Lessons learned from detailed operational reviews of crime statistics based on offenses reported to law enforcement in 1958 (when they were the only available national crime data) and 1984.
This historical review provides background that demonstrates how ingrained some concepts and challenges have become, and so is very useful to understanding the importance of making improvements to crime data collection such as those we propose in this study.
Crime is a weighty, serious, and painful topic, and rightly so. But, for aspirational purposes and to appreciate the importance of crime statistics as a basic social indicator, it is instructive to step back and consider the earliest days of nationally compiled crime statistics. One of the most striking things about leafing through the earliest publications of crime statistics is that crime data were presented and met with a sense of distinct enthusiasm—bordering on exuberance—at the new and major insights that the data could yield and the policies they could inform. The very first issue of the Uniform Crime Reports bulletin—the publication that would evolve into the annual Crime in the United States—sounded naturally cautious notes on the source and nature of the new data (Bureau of Investigation, 1930a:1–2):1
All contributions of crime data [are] are purely voluntary, solicited solely for the purpose of making available statistical information concerning crime to those officially interested organizations and individuals who desire to receive and analyze the statistics. . . . Every effort will be exerted by the Bureau of Investigation to secure from its contributors statistics which are not only accurate, but which will become more comprehensive as time goes on and the methods of compilation are more widely and thoroughly understood.
The early issues of the Uniform Crime Reports are replete with detail on the number of new police-department data contributors: marveling in the first issue at larger-than-anticipated initial participation by local departments (“the size of the registration area for crime reporting has more than doubled in the short period of eight months,” with the number of cities returning one month or more of crime data growing “from 400 in January to a total of 895” in August;
1 The bulletins would bear the full name Uniform Crime Reports for the United States and Its Possessions from 1930 through the first (of 2) issues for 1953, dropping “and Its Possessions” from the title thereafter. The first several issues were produced monthly, but that gave way to a quarterly cycle, then semiannual, and finally (in 1959) annual.
Bureau of Investigation, 1930a:1), celebrating the milestone of “every State of the Union” (then 48) “as well as the District of Columbia” being “represented” in the reports submitted in October 1930, by at least one local department (Bureau of Investigation, 1930b:1), and continuing as the number grew over time.
By the third bulletin in the series, the Bureau of Investigation analysts began assessing the content of the new data they had accumulated. “One of the purposes of collecting and publishing crime statistics is to show, as adequately as can be determined from the data available, the rise and decline in crime, or in certain selected crimes, over a given period of time,” the bulletin noted. The analysis launched in the report was admittedly modest and crude: a look at 9 months of robbery and burglary data, disaggregated by population size of the reporting city. But it was insight that was newly possible—and so was taken with equal notes of caveat and enthusiasm. The bulletin cautioned that the two crime types were selected “because they are reported probably more widely and accurately than any other reportable offenses,” and care was taken to further qualify the results (Bureau of Investigation, 1930b:4):
The number of cities used [(695)] was determined by the number contributing returns on those offenses for the entire period [January–September 1930]. It does not purport therefore to include all cities in the country, nor even all that have at any time submitted a return. It is merely a miscellaneous group, drawn from cities of all sizes and in various sections of the country. It does include, however, approximately 80 percent of all cities of 100,000 population and over, 70 per cent of all cities of from 50,000 to 100,000 population. . . .
But, with those caveats, the results were striking for the day: putting into stark relief the sheer concentration of the volume of robbery and burglary offenses in the nation’s largest cities (100,000 population or greater) and providing a first look at month-to-month fluctuation that would inform later study of seasonal effects of crime.
With the initial excitement over the possibilities of the new data came tensions over how much detail should be released and how uniformly concepts were being measured. In its first issue for 1932, the Bureau commented that “an effort has been made to present the statistics in a more desirable form”—dispensing with a city-by-city table of reported data due to “numerous requests for summarized data” (Bureau of Investigation, 1932:1); this level of detail was partially restored (for cities of population 100,000 or greater) by first quarter 1934, based on further feedback from the International Association of Chiefs of Police (IACP) advisory committee on the UCR. In 1935, puzzlement over the accuracy of homicide reporting led to the deployment of the first data collection to bear the name Supplementary Homicide Reports, this time “provid[ing] for the listing of detailed data on individual cases” of homicide on
a short-term basis for selected cities (with population greater than 100,000). The supplemental collection revealed “that many cities were including, as actual offenses of murder, cases in which felons had been killed by a police officer in the line of duty,” and so prompted formal change to protocols and specification of a standard definition for “justifiable homicide” (Federal Bureau of Investigation, 1939:70–71).2 Early issues of the UCR bulletins are replete with concern about a potential point of ambiguity first expressed in the original UCR Manual (International Association of Chiefs of Police, 1929): the degree to which reported motor vehicle thefts were in fact true thefts of that specific type of property, rather than short-term instances of “joy-riding” and unauthorized use of a vehicle. Similarly, a 10th-anniversary history of the UCR program (Federal Bureau of Investigation, 1939:77) noted early and persistent concern over the consistency with which local police departments were heeding instructions that “all assault with a deadly weapon or with intent to kill should be listed as aggravated assaults;” that “simple assaults, assault and battery, fighting without weapons, and similar minor assaults” should not be so counted; and that ambiguous cases (such as “cases of assault with deadly weapons even though the assault was more or less the outgrowth of a fight”) be handled consistently from month to month and year to year.
More to the point, enthusiasm over discoveries from shifts in the volume of crime types over time led to the push—within constraints of volunteer contribution of data from local police departments—for some finer categorization and detail. “In order to comply with suggestions received from police administrators[,] a supplementary report of known offenses” was distributed to large-city police departments (over 100,000 population) in January 1935; the first quarterly report for the year tallied “reports received from 19 cities” with a combined population of 3.7 million, as that small set of cities returned supplementary information that was “apparently complete and correct” (Bureau of Investigation, 1935:13). What the 1935 supplementary collection did was to divide most of the UCR Part I crimes into at least two finer subcategories, subdividing:
- rape as forcible or statutory;
- robbery by location (highway, commercial house, oil station, chainstore, residence, bank);
2 Years later, a similar “special questionnaire” for select cities (over 25,000 population) was deployed in August 1960, to assess consistency of reporting of aggravated assault (Federal Bureau of Investigation, 1960:5). No detailed analysis is provided and no apparent change to definitions or concepts resulted, but the few summary statistics printed from the special study suggests initial, cursory study of the extent of later judicial follow-up on aggravated assault cases (e.g., whether charges were filed or convictions won).
- burglary by location (residence, all other) and time (during night, during day);
- larceny by value of article stolen (under $5, $5–50, over $50), by general type of property, or by mode (pocket-picking, purse-snatching, other); and
- auto theft as joyriding or all other.
To be clear, this was not collection of incident-level detail, but rather a separate accounting of offenses by the finer subcategories—and collection was initially limited to contributions by an unstructured sample of large police departments. Over the years, participation was requested from police departments in smaller cities, but still limited to only a relative handful of police departments, and some different categorizations added (e.g., basic weapon type used in homicide). A section tabulating these slightly-more-detailed data from “Supplementary Offense Reports” became a regular fixture in Uniform Crime Reports bulletins.
Eventually—and without substantial revision to the subcategorization since at least 1958—the “supplementary” information was folded into the regular UCR Summary data collection. In most cases, the subcategories first requested in 1935 became count subcategories on the basic “Return A” intended to be collected monthly from participating law enforcement agencies. Others developed into a permanent Supplement to Return A that remains part of the UCR Summary system, principally collecting summed estimates of the monetary value of property stolen (and property recovered by authorities) in the course of committing UCR Part I offenses.
The point of this appendix’s historical discussion is to suggest themes in the development of national crime statistics. One such theme that recurs frequently is that planners and analysts tend to assume—often correctly, but sometimes radically overestimating—that local law enforcement agencies have ready access to highly detailed data and records internal to their own departments. In the early days of national crime statistics, this assumption was voiced by the FBI and lauded as a key feature of the national program because it ensured that local departments would easily be able to contribute. As the 10th-anniversary history of the program recounts (Federal Bureau of Investigation, 1939:46):
It goes almost without saying that the practical basis of the uniform crime reporting system is one of its strongest features, inasmuch as all police executives should have constantly available for local administrative purposes up-to-date figures of the types to be included in the monthly and annual uniform crime reports. . . . There are widespread evidences of a continuously growing and expanding professional spirit among the entire
law enforcement personnel of the United States, particularly among police administrators, [in the compilation of] records and reports, and should contribute in future years toward the building of even higher standards of accuracy and uniformity.
Numerous UCR bulletins over the years anticipated the availability of richer local information in positing what it believed to be a likely use—and in some instances perhaps the most appropriate use—of the nationally compiled statistics: as a rough guide as to where to target analyses of those local police data to address the “local crime problem.” One bulletin for 1934 (Division of Investigation, 1934a:9–10) took pains to emphasize that “the tabulation of crime rates for individual cities should not, in the opinion of the Division of Investigation, be used to discredit or to glorify individual police departments. They should serve as a source of information regarding crime conditions in individual cities which would otherwise be more or less unavailable to interested individuals.” Instead, “it is believed that the following compilation may help to raise questions which will lead to further study of crime problems in individual cities which will indicate desirable remedial action.” A second bulletin for 1934 went further, adding that “variations in the data for the several cities” actually heightened the importance of more detailed local study: “It is believed, therefore, that the proper function of the following compilation of crime data for cities with more than 100,000 inhabitants is to furnish the information which will serve as a stimulus to further study of local crime problems on the part of police administrators and interested civic organizations” (Division of Investigation, 1934b:9–10). Imparting more specifics, the final UCR Bulletin for 1953 (Federal Bureau of Investigation, 1954:72) anticipated that:
A city police chief can compare his own figures for 1952 and 1953 and note the trend in his city, increases or decreases in the various crime classes. Then he can check his local trend data against the national figures given here in tables 25 and 26. This basic survey of trends gives direction to his additional studies of the local crime problem, such as: special types of robberies or burglaries; areas in his city with the greatest frequency of these crimes; time of day that they occur; possible rearrangement of patrol to combat the problem; and the like.
B.4 OSSIFICATION OF CONCEPTS, DIVERGENCE OF CRIME STATISTICS FROM THEIR USE, AND THE 1958 PROGRAM REVIEW
Just as UCR planners assumed that local law enforcement agencies might benefit from comparison of their local crime information with national numbers, they also recognized what the overriding interest in the nationally compiled data was likely to be. As Wellford (1982:81) put it in retrospect,
“each agency is interested in its [own] performance. From the beginning the UCR[s] have been oriented toward the development of census procedures so that all agencies could be described in the report and so that agencies could review the data for jurisdictions and agencies with which they felt they could be compared.” The operative phrase—and hint of the broader theme that took firm hold in the mid-20th century—is “census procedures.” Over the decades, the concept of estimating crime and measuring it amidst uncertainty slipped away in favor of the appearance of completely enumerating it.
The earliest Uniform Crime Reports bulletins are remarkable for a lot of their language, as we have already described, but here it is useful to very briefly trace the evolution of two threads in those narratives. The first is a very real sense of the inherent limitations of the data flowing into the national level, which come out in occasionally blunt and candid statements. This began from the start, with the first issue of the Uniform Crime Reports bulletin and its concise description of the work-in-progress nature of national crime statistics, quoted above in opening Section B.1. By the following year (Bureau of Investigation, 1931b:1–2), the bulletin began urging that “caution should be exercised [in] arriving at any conclusions or deductions from the figures presented,” for a variety of reasons. In addition to this being because “all cities of the country are not represented” in the tabulation of offenses known to the police, it was also because “wide divergencies and the total number of particular crimes for various cities of approximately the same population may in some cases not be indicative of a variance in the amount of crime in those cities.” In turn, those differences “may be charged to inadequate record systems, or a lack of understanding of the [UCR] classification.” The very next issue of the bulletin (Bureau of Investigation, 1931a:1) boiled things down further still, into a terse disclaimer—actually, more of a disavowal—that was repeated in several subsequent issues before being moderated somewhat:
In publishing the reports sent in by chiefs of police in different cities, the Department of Justice does not vouch for their accuracy. They are given out as current information which possibly may throw some light on problems of crime and criminal law enforcement.
At the same time, the Uniform Crime Reports Bulletins began an escalating, internal rhetorical battle of sorts against that primary projected use of the compiled crime data: comparison between jurisdictions and over time. When raised as a concern for the first time in 1935, the discourse was remarkably polite (Federal Bureau of Investigation, 1935a:13):
The Federal Bureau of Investigation desires to point out that comparisons between the figures of two or more cities should be made with caution. Just as peculiar local conditions may have a tremendous bearing upon the mortality rate, so similarly peculiar local conditions may cause the crime rates to be above or below average. In other words, in making comparisons
with the average figures or the figures of a single community, consideration should always be given to local conditions which may have a large amount of influence with regard to causing the amount of crime to be unduly large or unusually small.
UCR writers and editors clearly grappled with how they wanted to present the point, with the very next bulletin simplifying the advisory to the single sentence (Federal Bureau of Investigation, 1935b:12): “Comparisons between the figures of two or more individual cities should be made with caution because there may be present any number of peculiar local conditions which may cause the crime rates to be above or below average.” By 1940, though, the sentiment seems to have switched to the advisory not being specific enough about relevant “peculiar local conditions.” In that year, the disclaimer expanded (Federal Bureau of Investigation, 1940a:11):
A great deal of caution should be exercised in comparing crime data for individual cities, because differences in the figures may be due to a variety of factors. The amount of crime committed in the community is not solely chargeable to the police but is rather a charge against the entire community. The following is a list of some of the factors which might affect the amount of crime in a community:
- The composition of the population with reference particularly to age, sex, and race.
- The economic status and activities of the population.
- Educational, recreational, and religious facilities.
- The number of police employees per unit of population.
- The standards governing appointments to the police force.
- The policies of the prosecuting officials and the courts.
- The attitude of the public toward law-enforcement agencies.
- The degree of efficiency of the local law-enforcement agency.
Comparisons between the crime rates of individual cities should not be made without giving consideration to the above-mentioned factors. It is more important to determine whether the figures for a given community show increases or decreases in the amount of crime committed than to ascertain whether the figures are above or below those of some other community.
And, by 1950, the bulletin took pains to present, as a paragraph of underlined text, this caution (Federal Bureau of Investigation, 1950:13):
Caution should be exercised in comparing crime data for individual cities because the differences in the figures may be due to a variety of factors. Such comparisons are not desirable even though the figures for individual communities are converted into terms of the number of offenses per 100,000 inhabitants.
As a coda (albeit not underlined), the warning returned to the same “middleman” argument that had been advanced as early as 1931: “In publishing these figures the FBI acts as a service agency. The figures published are those submitted by the contributing agencies.”
Nearing the 30th anniversary of the UCR Program’s establishment, the FBI commissioned a three-member expert Consultant Committee on Uniform Crime Reporting in 1957 to review the conduct of the UCR Program;3 its results were published in a special issue of the UCR bulletin in 1958 (Federal Bureau of Investigation, 1958), and we summarize their recommendations in Box B.1. Though its discussions were limited principally to a set of specific issues raised by the FBI, the committee’s review was very thorough and touched on some of the themes raised heretofore in this appendix and report. In particular, the committee noted “a regrettable amount of misinterpretation by some” upon every release of UCR numbers, “[causing] unnecessary concern on the part of local governmental agencies, private organizations, and the public, frequently about imaginary problems” (Federal Bureau of Investigation, 1958:11). The committee expressed particular worry about the index/Part I crimes being misconstrued as reflecting “all crime” and suggested that part of the problem was the UCR Program’s delineation between Part I and Part II offenses. The argument was, essentially, that distinction between the “serious” Part I crimes and the set of Part II offenses implies that Part II is nonserious. Instead, should a distinction have to be made, the committee suggested that the line be drawn between crimes that manifest or appear first through report to the authorities and those that manifest or appear first through the making of an arrest. As to how to collect the data, the committee was not worried about the core “offenses known to the police” (Part I offenses) because, “assuming that the trend toward increased coverage continues, more or less complete reporting will be a reality in a very few years” (Federal Bureau of Investigation, 1958:39). But those offenses that are first seen through arrest (typically, the Part II offenses but potentially including others) suffered from irregularity in data collection. The committee suggested that either concerted effort be taken to bring participant reporting of Part II offenses up to the Part I norms or that estimating Part II offenses through a sampling procedure be considered instead. In publishing the committee’s report as a special edition of the Uniform Crime Reports bulletin, the FBI noted general agreement with the principles embodied in its recommendations—most notably the call to cut from semiannual reporting to a single annual volume.
But the actions taken beginning in 1959 belied much of the guidance of the expert committee and exacerbated some of the tensions and developments
3 The committee was chaired by sociologist Peter Lejins, University of Maryland, and its members were Charlton Chute, Institute of Public Administration, New York City, and Chief Stanley Schrotel of the Cincinnati Police Department.
recounted earlier. Concern having been expressed about Part I-heavy, UCR data being unduly portrayed as a full accounting of crime, two things in the opening pages of the very next issue of UCR data in 1959 are interesting to note: Crime in the United States emblazoned on the cover of the volume in bold capital letters4 and a new summary section conveying highlights of “crime volume” and “crime totals” with little to no condition being put on what was meant by “crime” in those assessments. Granted, the new summary began with a preamble that properly credited the voluntary nature of crime data contribution—but did not really describe the limitations of the data and, worse, characterized the actual UCR tabulations as almost superfluous detail:
You have here a wealth of information on crime in the US, due to the cooperation of police and sheriffs who voluntarily send crime figures to the FBI.
This summary is for the reader interested in the general crime picture. Technical data, of interest primarily to police, social scientists, and other students, are presented in the following sections.
Reports of the era continued the unusual practice of touting the top-line summary totals while, again, signaling a distinctly hands-off approach to the processes by which data were submitted. To wit, part of the boilerplate disclaimer that ran in Crime in the United States volumes beginning in 1961 reads (Federal Bureau of Investigation, 1961:27–28):
It is clear, of course, that regardless of the extent of the statistical verification processes used by the FBI, the accuracy of the data assembled under this program depends upon the degree of sincere effort exerted by each contributor to meet the necessary standards of reporting, and, for this reason, the FBI is not in a position to vouch for the validity of the reports received.
What followed for the next several decades was fairly predictable. Wellford (1982:78) lamented that “UCR data have been used by some as if the data were a valid social indicator of the total incidence of crime,” even though “the UCR was not created to be such a social indicator,” generally. But a major reason for this is not hard to intuit—branded as Crime in the United States, the UCR Summary data came to define “crime in the United States,” to reduce the national conversation on crime mainly to the reporting of short-term changes in a few, broad categories of crime rates, and to lend themselves to less-than-careful comparisons across jurisdictions because they were all that were available. Clear statements of the uncertainty associated with UCR figures have never really been attempted, and what few hints about individual agencies’ participation (in part or in full) over the reporting years were largely obscured over time. Even today, the UCR data publications themselves do
4 Internal pages still bore the usual Bulletin nomenclature, but the report itself was eventually renamed to fit the new cover name.
little to illuminate the error properties of the police-report data, with some clues peppered into footnotes (in the days of paper publication) or auxiliary documents (the separate “data declarations” accompanying the more recent electronic releases of Crime in the United States).
The Crime in the United States-as-“crime in the United States” effect is pervasive and still continues in some form, though a very powerful corrective to the UCR as the only available measure of crime began its evolution in the late 1960s. A prototype survey (Ennis, 1967) fielded in support of the work of the President’s Commission on Law Enforcement and Administration of Justice (1967) demonstrated the potential of crime measurement through survey interviews, and led to the implementation and development of the National Crime Surveys (NCS) in the 1970s.5 In its design—largely to ensure direct comparability with the UCR’s police-report data—the new NCS deliberately adhered to some of the UCR’s definitions and concepts. But it was revolutionary in providing a means to quantify the vast amount of crime not reported to the police and shed light on the so-called “dark figure of crime” that is, by definition, excluded in the Crime in the United States police-report tallies.
Sponsored by the Bureau of Justice Statistics (BJS) and using the U.S. Census Bureau as the data collection agent, the survey was redesignated as the National Crime Victimization Survey (NCVS) after a comprehensive redesign in the early 1990s. The NCVS has expanded its reach over the years through the implementation of topic supplements and revision of core content (for instance, querying respondents about experiences with identity theft or with their interactions with the police or other actors in the criminal justice system).
Among the reasons why we do not devote a great deal of page-space in this report to the NCVS and its conduct is that it has been the principal focus of several previous National Academies studies. The panel study that produced National Research Council (1976) served as a comprehensive early-phase review of the surveys; our predecessor panel to review the full portfolio of BJS programs dedicated a first report volume exclusively to the NCVS (National Research Council, 2008) and considered the NCVS’s fit into that portfolio in its final report (National Research Council, 2009); and a study of the measurement of rape and sexual assault in federal surveys (National Research Council, 2014) considered the implicit limitations of an “omnibus” survey vehicle like
5 It is “Surveys,” plural, because the survey program initially included a separate sample of businesses as well as essentially stand-alone surveys for some individual surveys; these did not long survive initial review, and the program consolidated behind the single National Crime “Survey” of households thereafter.
the NCVS in eliciting information about personally sensitive victimization experiences. More generally, going beyond the National Academies’ projects, the reason why we spend relatively little time talking about the NCVS in modernizing national crime statistics is that “modernization” is effectively the continual norm for the survey. Though it has endured seasons of budget-induced cuts in sample size, the NCVS is almost constantly in a state of evaluation and development related at improving the survey’s efficiency. This includes active research and development to find ways to overcome the chief liability of the NCVS, which is that most respondents reached by the survey at an given time will not have experienced crime in the past 6 months, much less experienced highly specific offense types. Hence, because specific instances of crime victimization are (statistically) a relatively rare event in the whole population, the NCVS is necessarily limited in the degree to which it can provide estimates of specific offense types or victimization measures among particular demographic or geographic subgroups. Nevertheless, BJS continues research work on generating some regular subnational estimates from the NCVS.
In the early 1980s, BJS sponsored a task force study to reimagine the collection of police-report data. The centerpiece of that task force’s work was a conference on the future of the UCR, held in early 1984, and culminated in a final report entitled the Blueprint for the Future of the Uniform Crime Reporting System (Poggio et al., 1985). The task force used the terminology “unit-record reporting” to describe its primary vision for next-generation UCR data collection, though the term that ultimately took hold would be “incident-based reporting.” The task force’s recommendations are shown in Box B.2 and suggest—like the 1958 committee before it—a thoroughgoing review of operations but one that stopped short of considering crime data collection more generally. It did not challenge the basic content of the UCR Summary system but it did add considerably to the number of covered offenses and the level of detail requested about each incident for several offense types.
In the wake of the Blueprint’s release, pilot work began with several law enforcement agencies in South Carolina in 1987, and new concepts for the proposed National Incident-Based Reporting System (NIBRS) were announced at a national UCR conference in March 1988. But the Blueprint task force shares another distinction with the 1958 UCR committee that preceded it: Its central recommendations were followed only in part, and not exactly in the intended direction. As outlined in Box B.2, the task force had incident-based reporting as an aspirational goal for all law enforcement agencies, but took care
to define two levels of participation. First-phase implementation of incident-based collection would be targeted at a sample of law enforcement agencies (the nation’s largest and hence, arguably, its best resourced police departments, plus a representative sample of smaller departments). However, in implementing the new system, the FBI opted to open the database and its new format to purely voluntary contribution from the full set of usual UCR participants.
The FBI’s first revision of its UCR Handbook after the establishment of NIBRS—the “NIBRS Edition” of 1992—did a commendable job of explaining the new program. Indeed, 25 years later, it is difficult to craft a better capsule summary of the potential benefits of NIBRS than that advanced in the handbook and quoted in Box B.3. Yet that statement also includes a strong instance of the recurring assumption noted earlier in Section B.3: greatly
overestimating either the depth of data in local agencies’ internal databases or their willingness/ability to share those data. On how NIBRS would actually come into existence, the manual observed that “the implementation of NIBRS will be at a pace commensurate with the resources, abilities, and limitations of the contributing law enforcement agencies” (Federal Bureau of Investigation, 1992:1). Like other UCR/NIBRS-related documents from the early 1990s, the manual sounded very optimistic notes on NIBRS take-up. One page after the comment on implementation pace, the UCR handbook offered reassurance that “Crime in the United States will continue to be published in its current format until conversion from the traditional summary UCR system to NIBRS is complete in most of the country”—implying that the wait would not be excessively long. Meanwhile, BJS’s first bulletin analyzing NIBRS data, Using NIBRS Data to Analyze Violent Crime (Reaves, 1993, using 1991-vintage data), included a foreword projecting that “an estimated 40% of the Nation will report to NIBRS by the end of 1994.” Both the FBI handbook and the BJS bulletin seemed to project quick take-up rates based on greater-than-expected submissions (and promised submissions) of test data from several states, as well as full-fledged early adopters such as South Carolina, Alabama, and North Dakota—as well as the critical assumption that NIBRS was simply going to be able to take advantage of crime data already residing in local law enforcement records management systems. To wit, the UCR handbook commented (Federal Bureau of Investigation, 1992:1):
NIBRS data are designed to be generated as a byproduct of local, state, and Federal automated records systems. Thus, an agency can build its own system to suit its individual needs, including all the information required for administrative and operational purposes. Only the data required by NIBRS are then reported to the national UCR Program.
In practice, though, take-up rates have been much slower. As of 2016, NIBRS had yet to achieve the projection of 40 percent participation, whether measured as a percent of population coverage or by proportion of UCR-reporting agencies submitting NIBRS data, that was originally forecast for 1994. In 1996, the FBI received NIBRS-format data representing 1,550 departments—under 10 percent of the more than 16,000 law enforcement agencies then contributing to UCR programs. For a system that the Blueprint authors suggested should focus first and heavily on the largest police departments, those 1,550 departments included only one with a service population of more than 500,000: Austin, Texas (Roberts, 1997). In subsequent years, BJS’s grant programs offered assistance to some additional state agencies to plan and implement NIBRS-format collection at the state and local levels. But a major consequence of the limited NIBRS take-up was that the nationally compiled NIBRS data could not be said to be “national” at all: that is, they could not be deemed statistically representative of the nation as a whole or, ultimately,
to major geographic or demographic subpopulations. Though BJS and some researchers6 performed limited analysis of NIBRS-compiled data—hinting at the analytic possibilities—NIBRS has lacked the ability to “sell itself” based on the utility of its data, and it took until 2013 for the FBI to issue a full report and release (however incomplete) of NIBRS data.
We will return in detail to the barriers to NIBRS adoption in Section D.1, but one contemporaneous summary warrants mention in this historical narrative. Commissioned by BJS, SEARCH (the National Consortium for Justice Information and Statistics) convened ad hoc surveys and focus groups in the late 1990s to try to document the obstacles to NIBRS adoption. The
cost of converting to NIBRS-format reporting was the most cited barrier; the focus groups also spotlighted concerns about artificial “spikes” in local crime that would be created simply by the switch from reporting multiple offenses rather than a single offense (under the UCR Hierarchy Rule). Law enforcement executives also expressed concern that completing incident-based reports might be too large a time drain for officers—a concern related to a lack of clarity concerning the direct benefits of national data compilation to reporting agencies. Also tellingly, the focus groups suggested more structural barriers: respondents noted concern about the arduousness of meeting FBI specifications for certification to submit NIBRS-type data, and several expressed a basic worry that federal officials were not entirely committed to developing NIBRS (Roberts, 1997; Maxfield, 1999).
On the UCR/police-report side of national crime statistics, a firm axiom developed over time, being restated several times by the FBI at several meetings of our panel: The only way to alter the content of the UCR Program is either through formal recommendation of the FBI’s Criminal Justice Information Services (CJIS) Advisory Policy Board (APB) or by enactment of federal law. Both of these channels for effecting change have strong or compelling arguments behind them, not least of which is that the multitiered APB process—vetting proposals and policies through an elaborate set of regional and topic-specific working groups (including a dedicated UCR subcommittee)—ensures that the perspectives of numerous law enforcement stakeholders, at least, are reflected in board recommendations.
What warrants mention as a historical theme, and lesson for moving forward, is that both of the principal channels for change in the nation’s current crime statistics have major and perhaps unintended consequences for the perceived integrity of those statistics. More specifically, the APB process is so cautious as to appear sluggish or unresponsive in answering questions of high national interest, and the federal legislation route forces “new” (and very different) offense types into “familiar” UCR (and, to a lesser extent, NCVS) templates at the expense of accuracy.
The signature example on the APB side is the decades-long effort to modernize the UCR’s definition of rape, which had been set in 1929 as “the carnal knowledge of a female, forcibly and against her will” (International Association of Chiefs of Police, 1929:195). Specific proposals to make the definition gender-neutral and to remove the “forcible” concept date back to, at least, the Blueprint review of the UCR Program in 1985 (see Box B.2). Yet even simple wording changes in definition can be costly to implement at the local level, and difficulty of implementation (including reconciliation with
state criminal code and recording practices) undoubtedly slowed the process. Ultimately, it would take until 2011 for the APB to recommend and for the FBI and Justice Department to approve a revised definition. But other examples are also telling: for instance, it took until 2015—very well into the computer age—for hacking/computer invasion to be added as a NIBRS offense category. This is perhaps not surprising given an observation repeated by several of the users/stakeholders the panel engaged in its workshops: The content of the Supplement to Return A (see Box C.2) has not been systematically reviewed since at least 1958. Accordingly, the limited offense and property loss estimates available through the UCR Summary have long been sorely out of date because they do not include at least one very major target of modern larceny and theft: consumer electronics that do not neatly fit the category heading of “televisions, radios, stereos, etc.”
The simplest way to appreciate the way in which legislative mandates have been shoehorned into existing UCR templates and structures is to review the list of forms and data collections in the UCR in Box 1.1 (Chapter 1). Four of the most recent and most important legally mandated UCR topics are almost literally tacked on, as separate forms, to the UCR Program. The addition of arson (1978; 92 Stat. 933), bias-motivated offenses/hate crimes (1990; 104 Stat. 140), cargo theft (2006; 120 Stat. 240), and human trafficking (2008; 122 Stat. 5083) each occasioned the development of a parallel data-reporting form as an add-on to existing UCR Summary collection. New codes for these offenses and for bias motivation could more readily be added to NIBRS though not—as in the rape definition revision—without difficulty. Typically, the new forms were also deployed to UCR respondents without much (if any) training in how to recognize relevant incidents and record them properly. The NCVS has been subject to similar change through legislation, as described in Section 2.2.1 of Report 1 and more completely by National Research Council (2008)—the most recent and arguably the most curious one being the provision etched into fiscal year 2015 appropriations law requiring the NCVS to produce estimates of honor violence.
The coda to this historical theme is that the current channels by which national crime statistics can be changed are predisposed to try to fit “new” and difficult concepts into “old” structures, rather than begin with the question of what is the most appropriate and accurate method by which statistical information on the topic could be derived.
An important but as-yet unspoken theme from the history of crime statistics is that the UCR’s police-report statistics are of great and natural
importance to BJS’s central mission, even though the UCR is administered by the FBI. So it was neither accident nor typographical error to say that BJS sponsored the Blueprint review of the future of UCR in the early 1980s, that it commissioned SEARCH’s investigation of the barriers to NIBRS adoption, and that it tried to encourage NIBRS adoption through grant monies in the 1990s and 2000s. This is due in large part to BJS’s administrative home in the Office of Justice Programs (OJP) in the U.S. Department of Justice, which in turn inherited the functions of the former Law Enforcement Assistance Administration (LEAA). Put simply, BJS and OJP have the capacity and mission to provide grants and technical assistance that the FBI as a law enforcement agency generally does not.
Accordingly, one of the most recent, positive developments in moving toward modern national crime statistics is the National Crime Statistics Exchange (NCS-X), announced by BJS in 2012 with funding from the FBI. NCS-X leverages BJS’s grantmaking authority to directly fund local and state agencies to implement NIBRS-compliant information management systems, with the first targeted efforts applied to a carefully designed sample of jurisdictions. When complete, and those new NIBRS participants’ data are combined with data from existing NIBRS reporters, the resulting dataset will finally be nationally representative. In this way, analysis of national crime dynamics will finally be able to benefit from standardized cross-jurisdictional data to inform solutions—and, in principle, convincing other local agencies of the virtues of NIBRS participation should be easier. This is, in some sense, a return to the original Blueprint vision of seeding detailed data collection in an initially small sample of jurisdictions, to permit the data system to function fully as other law enforcement agencies switch to incident-based reporting over time.
Early NCS-X activities included interviews with state UCR programs, a survey of all 400 sampled agencies, and readiness assessments and coordination with both state programs and sampled agencies. Importantly, the early work also included outreach to pivotal groups including the Association for State Uniform Crime Reporting Programs (ASUCRP), the IACP, the Major City Chiefs’ Association (MCCA), and the Major County Sheriffs’ Association (MCSA).
The final note in this retrospective is arguably the most dramatic and important single step forward, one taken during this panel’s period of work and around the time of the release of our Report 1. Acknowledging the inadequacy of UCR Summary-only count data to address important questions of interest, the FBI publicly announced its intent to shutter the UCR summary system by 2021 in favor of full NIBRS-format reporting. Formally, the mechanism for this “sunset” of the SRS was the approval by FBI Director James Comey on February 9, 2016, of a recommendation from the APB:
The FBI UCR Program will transition to a NIBRS-only data collection by January 1, 2021, and will evaluate the probability of achieving that goal on an annual basis. Federal, state, local, and tribal agencies unable to meet the five year transition and who have committed to transitioning to NIBRS will collaborate with the FBI CJIS to develop a transition plan and timeline for conversion.
The decision was endorsed in a joint statement of the IACP, MCCA, National Sheriffs’ Association, and MCSA. The imminent phaseout of the decades-old Summary Reporting System was announced in the most recent releases of UCR data, and the state UCR programs that coordinate inputs to the national level were formally apprised of the change in a June 10, 2016, letter from Director Comey.
The UCR Summary sunset is the most fundamental plank—but not the only one—in an ongoing UCR modernization program in process at the FBI. As our panel began its work, the UCR program was working to upgrade its computational infrastructure and break the constraints of having to work within the confines of processing and tabulation routines then still being performed on a 1960s-vintage computer mainframe housed in the FBI J. Edgar Hoover Building. (The UCR program only switched away from permitting paper returns in 2014.) In a foreword message to the 2014 release of Crime in the United States, then-Director Comey alluded to another major improvement being made in the program: 27 years after enactment of the Uniform Federal Crime Reporting Act mandated that federal law enforcement agencies report offenses to the UCR Program, there would finally be a concerted effort to get them—including the FBI itself and its field offices—to do so. As Director Comey closed his note:
Most worthwhile changes involve growing pains. Even though we are taking steps to minimize the anticipated bumps and bruises, this undertaking will require tremendous effort and resources by our law enforcement and government partners, as well as the understanding of the media and the public. But to continue in our current system without comprehensive data only stalls meaningful conversation and fuels empty debates, both within law enforcement and in the communities we serve. Developing the UCR Program into a collaborative effort that gives all of us useful information and clarity can tell us more about where we may have problems and how we can improve.