Three quarters of a century ago Edwin Sutherland flatly declared, "statistics of crime are known as the most unreliable and the most difficult of all statistics”  (Sutherland 1924). The sentiment was not new then, and observations on problems with crime statistics remain staples of criminology and sociology textbooks today. This was not just an academic observation either. Crime statistics are important to policymakers and ordinary citizens as well as to academics. The amount of crime is a basic fact that allows us to compare our society with others, helps us know how our society has changed over time, and provides evidence to help determine what social conditions and what social policies are useful for minimizing disorder.

The quantity of crime - the total count, the categories, and the rates - remains shrouded to a degree that would seem intolerable to people in most other fields. We face a "dark figure”  of crime, where the number of unknown or unknowable criminal events "haunts all attempts to estimate crime rates even in the present; for the past it is multiplied enormously”  (Lane 1992, p. 39). This "dark figure”  is analogous to the one facing astrophysicists. They know that there is a vast mass of "dark matter”  in space, but not how much. How much dark matter determines whether the universe ends in fire or ice. For crime, the "dark figure”  mocks our ability to know important things about our society and its development.

Many scholars have worked to quantify the level and trend of crime in the United States. This essay sketches some of that research and discusses the strengths and weaknesses of the results. It barely touches the immense problems of counting crime before the advent of regularly gathered and published criminal statistics. There is a lively tradition of scholarship devoted to that topic, but it is beyond the scope of this essay.¹ This discussion does not discuss white collar crime either, which is even more poorly measured than violent and property crime. Except for a few historical series, this essay does not deal with city and state differences in crime rates and trends, though evidence suggests that these regional differences are large. It focuses instead on the sources of unrecorded crime, on data development efforts aimed at bringing crime to light, and on the U.S. system of national crime statistics.

Back To Top

The social world is, for the most part, invisible to its inhabitants. Any one of us can witness only a small fraction of the events that occur within it. Some important aspects of societies, such as average life spans, total homicides, cannot be witnessed by a single individual because they are summations of large numbers of events occurring in diverse places and times. Nothing that happens outside our immediate range of vision and hearing, whether because of distance, closed doors, the presence of gatekeepers, or privacy rights, can be known directly, and the more pronounced the distance or other barriers, the more difficult it is to see.

Because they can carry severe consequences, criminal and deviant actions are even more difficult to observe than are other social actions. Offenders are not the only participants who wish to conceal crime. Bystanders may want to avoid involvement; friends may want to avoid injuring their intimates; victims may fear retaliation or want to avoid the humiliation, pity, derision, or gawking curiosity that can accompany victimization. Victims may feel it is not worth the effort to report a crime.

People also misrepresent crime by reporting events that are not criminal or that did not occur. One reason is to hide a different crime, as when a theft is claimed in order to hide insurance fraud. People also make false accusations to exact revenge and to avoid embarrassment. Perhaps the best known historical example of false accusation is the case of the "Scottsboro Boys,”  where a group of nine young black men were convicted in 1931of raping two white women solely on the basis of the women's testimony. At subsequent retrials one of the women recanted her earlier testimony, but the men were convicted again (Goodman 1994; Carter 1979).

When there are great social strains, it is possible for even fairly bizarre accounts to be given credence by the public and by law enforcement officials. Witch crazes in eighteenth-century New England fall into this category. In late-twentieth-century society, the form of story called the "urban legend”  commonly involves an alleged recent, local crime that can be found to be an old story, existing in diverse communities, with no clear basis in fact (for example, the legend of razors in Halloween apples) (Best and Horiuchi 1985). Urban legends and beliefs similar to them seldom affect criminal statistics, but at certain times they can lead to arrests, prosecutions, and imprisonment. In the past two decades, there has been a surprisingly widespread belief in "Satanic”  cults that practice sexual molestation of small children, the memories of which the children repress. People have been convicted and imprisoned on such charges, even when the evidence was sparse and contradictory. In the social science community, the majority opinion is that most - if not all - of the alleged events never occurred (Richardson, Best, and Bromley 1991; Loftus 1979; Loftus and Ketcham 1994). These are extreme examples of the way in which shared social beliefs shape and distort crime statistics.

The police also distort crime statistics. They misrepresent the criminal behavior of others and engage in criminal behavior themselves. There are a number of reasons for this. Many police appear to believe that following official standards interferes with the performance of their jobs. In addition, the danger in police work and the need for mutual trust and reliance breed powerful internal solidarity. The "blue wall of silence,”  which draws much commentary, can hide such criminal behaviors of police as use of excessive force and manipulation of charges and evidence in ways that affect specific crime rates and racial and ethnic differences in crime rates (see, for example, Barker and Carter 1990). Moreover, police work offers opportunities for graft. Overall, then, the reason crime generates "the most unreliable and the most difficult”  statistics is that so many different actors have such strong reasons to mislead.

It is easy to see why many are motivated to misrepresent deviant or criminal behavior. What about those who want to give accurate accounts of what they have witnessed? One of the surprising findings of social science research is that they mislead as well. It has been known for over half a century that when people gossip or pass rumors, the accounts they pass become shorter and simpler - a handful of details are played up to the exclusion of other, potentially important points. Accounts change to fit the social expectations of the tellers. A biased and oversimplified picture replaces a more accurate account (Allport and Postman 1975). One cause of this is limitations in human perception. Eyewitnesses cannot agree on even basic information about offenders, victims, or circumstances. Witnesses are profoundly suggestible (Wells and Loftus 1984). The longer the chain of information-passers is, the greater the distortion and simplification of the original message will be. More new and incorrect elements enter. The final message comes to mirror social expectations.

Back To Top

Which behaviors are considered criminal depends upon culture institutions. Changes in culture and the law mean changes in measured crime. For example, in New York at the opening of the twentieth century, there was a large increase in convictions for rape (Sutherland 1924, p. 32). This was caused not by an increase in sexual assaults on women, but by a legal change that raised the age of sexual consent for girls from 10 to 18 years and by the institutional fact that statutory rape was included with forcible rape. Statutory rapes are not included in modern criminal statistics such as the FBI's Uniform Crime Report rape statistics reported in series Ec4. The modern statistics refer to forcible rapes only.

Cultural change can affect crime statistics even in the absence of legal change. Roger Lane argues that much of the increase in arrests in late-nineteenth-century Massachusetts was caused not by an increase in crime but by an increasing intolerance toward disorderliness (Lane 1968). Eric Monkkonen and Jeffrey Adler document similar developments at other times and places (Monkkonen 1981; Adler 2001).

Homicide is arguably the most consistently defined of all crimes. Yet the definition of homicide has changed substantially over the past century. Much of the surge in homicides in early-twentieth-century cities can be attributed to a growth in social sensitivity to death by violence and a change in the way many deaths were categorized. Auto accident deaths, which were on the increase, often were ruled homicides. Police changed their method of reporting deaths. In the nineteenth century, they had worked diligently to avoid calling an action a murder; in the early twentieth century, they began calling more deaths wrongful murder (Lane 1989; Weiner and Zahn 1989; Adler 2001).

Early vital statistics counts of deaths by homicide did not regularly distinguish between felonious and justifiable homicides. Most justifiable homicides are killings by the police. Today the FBI tabulates police killings separately, and vital statistics tables allow "legal intervention”  deaths to be broken out from homicide totals. In the 1920s, perhaps a quarter to a third of all homicides was "justifiable”  according to the definition then in use. Thus, the high level of the homicide rate in the early part of the twentieth century, as shown in series Ec192, does not necessarily imply a high murder rate. It could mean a large number of "justifiable”  killings by police, which are excluded from contemporary murder counts. But recognition of this problem requires recognition of another problem: standards of police conduct have changed greatly over time. Today, although the overall homicide rate is about the same as it was in the 1920s, justifiable homicides by police are only about 2 percent of the total (Federal Bureau of Investigation 1998, p. 24). Judged by today's standards, police of earlier eras would be judged grossly reckless and negligent in their use of deadly force. Many early "justifiable”  homicides would now be ruled felonious.

Changes in social organization affect crime statistics. This is because both the formal and the informal (therefore often hidden) criteria by which an event is assigned to some recognizable category vary by time and place. Cultural differences play a role here, but social rules of classification are also important. For one who views every abortion as murder, the U.S. murder rate is very much higher than official statistics suggest.

Public sensibilities are volatile and laws can change. The United States whipsawed from rates of drinking and drunkenness far above those of today to a change in the Constitution outlawing the sale of alcohol, then to Prohibition's repeal. Drunk-driving laws have grown much more stringent in recent years. Until recently, gambling was illegal in most of the nation; today many states promote it, at least in the form of state lotteries. Civil rights for black Americans were very different during slavery, Reconstruction, the Jim Crow era, and the era of the Civil Rights movement. It is again legal in several states to carry concealed firearms. The United States did, then did not, and now does execute criminals - in some states at least.

Laws and mores concerning sex and reproduction are perhaps the most volatile aspects of public and private life. Laws regulating acts, ages, circumstances, and appropriate partners seem always contested. This is the social basis of ongoing controversies about contraception, abortion, reproductive technology, divorce, homosexuality, and sexuality in the arts, where competing social movement organizations line up against one another in what have been called "culture wars”  (Hunter 1991; Bolton 1992). Until the 1960s, racial intermarriage was illegal in most Southern states; the ban was not removed from the South Carolina constitution until 1998. The age of sexual consent in most states is now typically 16 or older, and people are serving prison terms for statutory rape, but in some of the same states a child of 13 can (with parental permission) marry. In this social scene, it has to be hard to understand the meaning of even well-documented changes in rates of arrests, convictions, or incarcerations, in terms of the private activities of masses of people.

Back To Top

The U.S. criminal justice system is characterized by substantial decentralization. Each state runs its own system, with its own criminal codes, modes of organization, and funding. Within states, most counties, cities, and other governmental bodies have their own law enforcement agencies (police departments, sheriff's office, and so forth). As a result, determining lines of authority can at times be contentious. This is partly because authority is divided along geographic lines but also because social considerations must be taken into account (for example, state police typically will pursue different forms of crime than will local police, though this itself varies from state to state). Court systems and prisons likewise vary from state to state because of the evolution of their development. Thus, Michael Hindus found that by the late 1800s Massachusetts had developed a fairly unified statewide prison system, but that in South Carolina authority was jealously held at the county level (Hindus 1980). Overriding all of this is the separate federal justice system, which itself has a large number of agencies, with often crosscutting jurisdiction.

Gathering basic information on crimes is made difficult by this hodge-podge of agencies. State agencies cannot be compelled to provide criminal statistics to the Federal Bureau of Investigation (FBI); however, they do so in a voluntary arrangement. Some states require that all agencies submit criminal statistics to a central statewide agency, and those states submit crime data from that agency. From other states, though, individual local agencies provide data, and some, particularly very small police departments that encounter few crimes, find the process not worth the effort. Crimes that are reported must be coded to FBI standards, which may be difficult when the statutes of the state differ greatly from the norm.

Consequently, the highly decentralized character of the American criminal justice system is the source of much of the inconsistency in the U.S. crime statistics, as will become apparent later in this essay. The system requires that data be collected by local agencies and then forwarded to the FBI, sometimes going first to a state agency. Because local agencies differ in their official and unofficial approaches to crime, these numbers are not fully comparable with one another.

Back To Top

In the United States, crime statistics were first collected in a handful of large cities and in one or two states, notably Massachusetts, well over a century ago. The development of comprehensive crime statistics gained momentum in the second quarter of the twentieth century. The statistics have improved in detail and accuracy, and the speed of improvement has increased. Nonetheless, the obstacles to complete tabulation remain formidable.

The first comprehensive national measures of criminal activity refer to prison inmates. This is the phenomenon most distantly removed from the original criminal action; therefore, if one is interested in tracing crime trends themselves, this is the statistic most likely to misrepresent crime. On the other hand, incarceration is a phenomenon that is interesting in its own right.

The decennial censuses began publishing statistics on prisoners in 1850. (See Table Ec-A for a chronology of important events in crime and criminal statistics.) Since then, there have been several different data collection programs. The most important and comprehensive, the National Prisoner Statistics series, was begun by the Census Bureau in 1926. This is a voluntary program in which penal institutions supply statistics on prisoner admissions, releases, "movements,”  and executions (for descriptions of these data, see Cahalan 1986; Langan 1991).

It would seem almost too easy a matter to collect data on prisoners. They are - literally - a captive population, inherently easy to count. That is far from the case, though, mainly because of problems resulting from federalism. Over the years it has proven difficult to get the many prisons throughout the country - each operated by a local jurisdiction - to provide the required information. The situation worsened in the early 1970s, when often over a third of states provided no admissions-series figures (see Langan 1991).

There are, moreover, gaps in which one or another regularly reported datum simply ceases to be reported for a time. One finds these gaps in federal as well as state data. For example, there are no published admissions series data for 1951–1959, 1961, 1963, 1965–1969, and 1971–1973. From 1990 through 1992, Correctional Populations in the United States, the annual volume of National Prisoner Statistics, issued no release statistics or statistics on "returned”  prisoners (mainly people who had previously been paroled or on probation) for federal prisons. As "new”  data are found, updated figures replace older ones for some items but not others. This means that official data on prisoners is in a constant state of revision. The partial revisions of the prisoners data is the reason why many subtotals of crime statistics in the tables shown here do not always add to the total.

The available data, displayed in Figure Ec-B, bring to light two important facts about incarceration. First, there has been a profound increase in the rate of state and federal admissions of prisoners, beginning in the mid-1970s and continuing to the early 1990s. Second, the incarceration rate of nonwhites is much higher than that of whites, with the gap increasing since the mid-1970s.

Do these increased incarceration rates mean that crime, too, has increased? The answer is probably no. Scholars who have studied the matter believe that the steep rise in imprisonment reflects changes in legal policy. This has led to a greater number of arrests for behaviors that, in the past, would have been ignored or dealt with administratively. Prison sentences, further, are now prescribed for activities that in the past would have been dealt with via parole, probation, or fines.

The lack of good early data is similar for purely juvenile facilities. The census has provided counts of juveniles in various types of youth facilities since 1950, but the first trial survey of public institutions for delinquents was undertaken by the Children's Bureau of the old Department of Health, Education, and Welfare in 1953. The first comprehensive survey had to wait until 1956 (Children's Bureau 1958). It received completed questionnaires from only 78 percent of the institutions surveyed, though in later years, as it became an almost annual survey, the response rate rose to 90 to 95 percent. Following reorganization in the late 1960s and early 1970s, the surveys - now considered censuses of juvenile institutions - were taken over by the Justice Department, and data were collected by the Census Bureau. Coverage of what came to be called the "children in custody”  series was then expanded to include short-term holding facilities and private juvenile institutions, and became a biennial survey. Coverage of public institutions is almost universal today, but there remains some difficulty in collecting data from private facilities. For example, in 1991 the response rate for private juvenile facilities was only 84 percent, and there were about 8 percent outright refusals to cooperate (Table Ec402–425).

In the absence of nationwide data on crimes, scholars have looked for clues about trends in criminal activity in the vital statistics that include homicide and suicide as causes of death. Mortality questions were first asked in the Census of 1850, but the responses were judged at the time to be unreliable. The "death registration area”  was established in 1880, as part of a movement promoting continuous official reporting of deaths (Census Bureau 1906, p. lv).

Table Ec190–198 displays official statistics on homicide and suicide counts and rates beginning with 1900. For the early years, these official numbers are known to be misleading. The compilers admitted that data for the subcategories of violent deaths (deaths by "external causes”) - homicide, suicide, and specific types of accident - were poor, calling them "incorrect and absolutely misleading”  (Census Bureau 1906, p. lv). The problem was that it was common to list a physical cause of death (for example, poison) without specifying whether the death was accidental or intentional. Most death certificates provided no place to indicate crime, so the compilers at the Census Bureau had to make the determination. They categorized 97,805 "suspicious”  deaths as "accidents”  in the five years from 1900 through 1904 (Wilbur 1916). In some cases, they sent death reports back to the originating officials for more information. They also developed a new model death certificate, and by the time they published the death tables for 1907, they appear to have resolved this problem (Eckberg 1995).

A second problem is that the death registration area did not encompass the entire nation until 1933. The earliest states in the area were primarily Northeastern and upper Midwestern states that had relatively low rates of homicide. Southern and Western states, which had higher rates, were late to enter. The admission over time of these high-homicide states created a false impression of rising homicide rates.

Figure Ec-C shows the official national homicide rate and a separate series designed to correct for the two problems described previously. If the corrected figures are accurate, then the rate of killing early in the century was on the order of six times the rate originally published and not far out of line with rates today.

Figure Ec-C reflects a substantial increase in the homicide rate beginning in the 1960s, with several peak years of rates above 10 killings per 100,000 population. This high level would support imprisonment statistics in that it points toward a great increase in lawlessness in the latter part of the century. Note that homicide is a comparatively rare crime, normally claiming fewer than 20,000 victims per year and accounting for perhaps half that many incarcerations. By itself, it could not cause the explosion in imprisonment rates shown in Figure Ec-B.

Gathering data on crime from police departments themselves began in the early 1930s, at about the same time that the death registration area became complete. Today these statistics are presented in the FBI's Uniform Crime Reports or UCR, the most important element of which is the annual publication, Crime in the United States. The UCR is the single most closely watched measure of the crime level and trends in the United States.

The development of the UCR was led by the FBI in a cooperative arrangement with local and state law enforcement agencies. A major innovation was to record, along with records of arrests, "crimes known to police.”  To qualify for inclusion in the report, police do not need to arrest a suspect; they merely indicate that crimes have been reported. The FBI also instituted a focus on seven important and reasonably well-defined "index crimes.”  These index crimes are murder or non-negligent manslaughter, rape, aggravated assault, robbery, burglary, motor vehicle theft, and larceny or theft. In recent years the UCR has included arson as an index crime, but a great many jurisdictions do not yet report it. The index crimes let the FBI develop a standard measure of the amount of serious personal crime and monitor trends. The number of these index crimes reported each year since 1960 and the crime rate for the nation as a whole are presented in Table Ec1–10 and Table Ec11–20. Index crimes tabulated by urban police departments for the years 1937 through 1957 are reported in Table Ec21–29. The FBI keeps records on over thirty types of crimes. The statistics are published in annual issues of Crime in America.

Even though they are heavily used, the accuracy of UCR data has long been questioned. First, the UCR data are known to seriously underreport the total quantity of crime because only the most serious charge in a given incident is recorded (Maltz 1999). For example, an armed robbery might result in charges of robbery and carrying a concealed weapon, but the UCR would record only the robbery. This is called the "hierarchy rule.”  It was implemented to simplify record-keeping, but one of the consequences is that minor crimes are underreported.

Moreover, the UCR faces predictable problems securing regular and timely cooperation from local and state agencies in a decentralized governmental structure. For a number of years, the main tables in Crime in the United States drew on data from only a subset of cities because there was fairly slow progress toward cooperation with agencies in small towns and rural areas. There were also problems of categorization and organization. In an effort to solve these problems, the entire system was reorganized in the late 1950s. The FBI no longer considers data collected before 1960 to be reliable. To address remaining weaknesses in the data, the FBI has been developing a new system, the National Incident-Based Reporting System (NIBRS), to replace the present UCR. It will keep a complete record of each individual incident so that minor offenses will be better represented in national crime statistics (Federal Bureau of Investigation 1998, pp. 1–4).

Problems in gaining prompt cooperation from local agencies still remain. Crime in the United States never gets complete statistics from all jurisdictions. In 1997, for example, 7.1 percent of the population of Metropolitan Statistical Areas (MSAs), 22.6 percent of the population of cities outside of MSAs, and 22.9 percent of the population from rural counties were not covered by agencies that returned completed crime reports. For various reasons, complete data were not available for Illinois, Kansas, Kentucky, Montana, New Hampshire, and Vermont. Indiana, Mississippi, Missouri, New Mexico, Ohio, South Dakota, and Tennessee are notable for the large proportions of their populations not covered by reporting agencies. Kansas has not provided complete state totals since 1994. States converting from the older reporting system to NIBRS have run into problems gathering and presenting crime data that fit UCR protocols. There also remain problems with definitions used in categorizing crime. Since 1984, Illinois has not collected data on forcible rape that fit UCR guidelines. Many agencies in Illinois follow different rules than the UCR on the categorization of other offenses, so that a correction factor has to be employed before figures can be compared with those of other states (Federal Bureau of Investigation 1998).

A sizable fraction of UCR crime data for any given year, then, are imputed, which is to say that estimates are used in place of counts. These are calculated primarily by applying proportional trend figures from reporting agencies to nonreporting agencies that are similar in size and type of jurisdiction (Federal Bureau of Investigation 1998, pp. 401–2). There are ongoing discussions about the best imputing methods to use (Maltz 1999).

Figure Ec-D shows the trend of UCR crime rates for the violent index crimes (criminal homicide, rape, robbery, and aggravated assault) and the property index crimes (burglary, larceny-theft, and motor vehicle theft). There were significant increases in both property and violent crimes known to police in the period during which homicides and imprisonments began their accelerated growth in the mid-1970s. Since the 1990s, however, both crime indexes turn downward, while imprisonment counts continue to rise.

The trend for arrest statistics is similar to that of crimes known to police (Table Ec30–53, Table Ec54–97 and Table Ec98–141), namely, rising from the 1960s through some part of the 1980s or early 1990s, then leveling off or even dropping significantly. In all, it is clear that the increase in prison populations through the late 1990s is not attributable to an increase in the number of crimes reported to police or in amounts of victimization (see the discussion in the next paragraph). The issue has been subject to substantial discussion, and it appears to grow from a variety of changes in U.S. public policy toward arrests and imprisonment.

In the early 1970s, a second federal crime measure appeared, the National Crime Victimization Survey (NCVS) of the Bureau of Justice Statistics (BJS). The NCVS is a survey (in recent years of over 80,000 individuals in some 43,000 households) that has been conducted annually since 1973. Responses are shown in Tables Ec142-181. These surveys are administered following principles of scientific survey research in which probability samples of people are selected and then questioned about their experiences with criminal victimization and with delinquency. Interviewers tried to avoid the biases built into the rest of the criminal justice reporting system in which only some people report certain crimes. In addition, by going directly to the public, the data collectors ended reliance on law enforcement agencies entirely. Because it is an alternative or supplement to UCR's crimes "known to police,”  it employs categories similar to those in the crime index: rape, robbery, simple and aggravated assault, theft or larceny, burglary, and motor vehicle theft. It is similar to the National Household Survey on Drug Abuse (NHSDA) in that it covers only people age 12 and older. Additionally, "households”  do not include merchant vessels, military barracks, or institutions such as prisons, and the surveys do not cover commercial crime or homicide (Rand 1998).

Nonetheless, even these steps do not completely eliminate the possibility of bias. First, it is virtually impossible to select a perfectly representative sample from a complex society like the United States. Then, after the sample is drawn, those who refuse to cooperate further complicate the interpretation of the findings. Even though one may "weight”  data to adjust for over- or underrepresentation of people with various characteristics, there is always the possibility that those who do not cooperate are simply different from those who do.

Another issue is the reliability of self-reported activities. Research has shown that people misreport many of their activities, including their voting behavior, contributions to charity, amounts of photocopying, and church attendance (Bradburn, Rips, and Shevell 1987; Presser and Traugott 1992; Goldstone and Chin 1993; Bishop and Fisher 1995; Marler and Hadaway 1999). Gary Kleck argues that people underreport gun ownership to researchers (Kleck 1991, pp. 455–60). Thus, it seems likely that people would misrepresent their illegal activities and shameful experiences to strangers in surveys.

The concern with self-reported victimization and delinquency is not just that it understates the total amount of the activity, for if it could be shown that it did so by a certain proportion, then one could easily estimate the true amount. The concern is that the amount of concealment might be different for people of different races, ethnic groups, regions, classes, sexes, and so forth, either masking or exaggerating average group differences.

With many types of victimless crime, though, self-reports may be almost the only way to determine the prevalence in the general population. Also, since the first pioneering self-report survey in the 1940s (Porterfield 1943), survey respondents have seemed quite willing to reveal sensitive information about themselves, once rapport has been established, especially if methods are used to assure the respondents of the anonymity of their answers. Moreover, close inspection of the results does not reveal many systematic problems (Hindelang, Hirschi, and Weis 1981), though black youths are more likely than others to underreport officially recorded delinquency. Most questioning in self-report studies is fairly general and focuses on minor deviance that people might not feel uncomfortable reporting, though there are some exceptions (for example, Wright and Rossi 1986; Elliot, Ageton, et al. 1983). Thus, it is common to use the results of such surveys to trace general trends.

There are two fairly long-term self-reported delinquency series that focus on recreational drug use. The first, the Monitoring the Future (MFT) program, has been surveying the drug use of high school seniors every spring since 1975 (and youths of other ages for a considerably shorter time). The results of this program are displayed in Tables Ec1015-1126. The investigators are aware of potential biases, but they believe that false and misleading answers are minimized, and they give several reasons to think this; however, they note differences in the propensity to answer questions (for example, black males are more likely than others not to answer questions on drug use). About a third of schools in the samples refuse to participate and must be replaced. If these schools had particularly difficult drug problems, the effect would be to underreport drug use. Noncooperation by students is another problem. About 17 percent of sampled students did not complete questionnaires in the most recent year, mostly because of absenteeism and conflicts with school trips.

The MTF program surveys the same grades year after year, so results are comparable across cohorts. Because it surveys only high school seniors, however, it can tell us little about drug use among, say, middle-aged adults. Also, because the survey is done during class, it does not catch students who have dropped out, have been expelled or suspended, or skip school on the survey day. As a group, we would expect such people to be more involved with drugs than the group of teens attending school.²

A more broadly based drug survey is the NHSDA, which dates from 1971, with consistent statistics beginning in 1979. The results of these surveys are displayed in Table Ec951–962, Table Ec963–974 and Table Ec975–1014. The NHSDA surveys a representative sample of noninstitutionalized, civilian households, to provide data representative of the U.S. population age 12 and older (excluding military personnel, people in prisons, hospitals, and other institutions, and homeless people who do not live in shelters).

The two surveys attempt to provide figures on the amount of drug use, by specific types of drugs and for various subgroups in the population. Because their sampling is so different, their findings can be used as checks on each other. Nonetheless, both miss those who want to be missed or cannot be found, as for example those in criminal subcultures. Neither of them independently assesses the actions of their respondents. It is generally believed that the trends they uncover follow trends in the general population fairly accurately, though both probably understate the total amount of drug use.

Illicit drug use is one common form of delinquent behavior, and it can be used as an indicator of the general lawlessness of the population. If we think of illicit drug use in this way, we find that neither the MTF surveys of seniors nor the NHSDA support the idea that lawlessness was increasing during the period of rising "crimes known to police”  or imprisonments. Figure Ec-E plots the trend. It shows that both surveys indicate a remarkable decrease in the rate of illicit drug use in the general population since the 1970s, a drop of over 50 percent. MTF data do show a rebound in the early 1990s, though not to the level of use during the late 1970s.

In addition to the drug-use questions, the MTF program includes a series of questions about the types and number of other delinquent acts students have committed in the course of the year, including vandalism, trespassing, theft, and injury of another person. No details on incidents are gathered, and no items refer to sexual offenses or homicides, but students are asked about actions that could be charged as such serious offenses as assault, larceny, auto theft, and robbery. Responses are shown in Table Ec601–625 and Table Ec626–650.

It is difficult to know the extent to which the total amount of delinquency is understated by these measures. One calculation suggests that the understatement may be relatively small. If self-reported delinquency is compared with self-reported victimization in similar categories (for example, theft of possessions worth more or less than $50), one finds that the proportion of students admitting to having performed such acts is roughly equal to the proportion of students who report being victimized in school. The number of delinquent acts that are admitted, though, is far fewer than the number of victimization experiences reported.

For drug use there are some new, alternative sources that offer contrasts with the findings of general surveys, by surveying different populations using different methods. The most ambitious such project is the Arrestee Drug Abuse Monitoring program (ADAM, previously called Drug Use Forecasting) that surveys prisoners in several dozen cities. Persons just arrested answer a series of questions and submit to urinalysis. The program is voluntary, assurances about confidentiality are given, and uniformed officers are not used to gather the information. Although one might well ask how "voluntary”  this will seem to a person just arrested, about 20 percent of prisoners do refuse to participate.

For the general population, NHSDA data on drug use within "the past thirty days”  indicates that about 6 percent admit to using any drug, 5 percent admit to using marijuana, and under 1 percent admit to using cocaine. In contrast, 1998 ADAM data show a median city rate of cocaine use of 37 percent for men and 40 percent for women. Because cocaine is detectable for only about three days after ingestion, the share of arrestees who use cocaine regularly could well be higher. About a third of inmates tested positive for marijuana, and about two thirds tested positive for at least one drug (Arrestee Drug Abuse Monitoring Program 1999). These rates of drug use among arrestees are six to fifty times the level in the general population. One possible explanation for the extraordinarily high rates of drug use recorded for arrestees is that prisoners have fewer opportunities than those in the general population to disguise their drug use, but the most recent ADAM annual report finds that juvenile prisoners who are in school have substantially lower rates of drug use - 40 to 70 percent lower - than juvenile prisoners who are not in school (Arrestee Drug Abuse Monitoring Program 1999, p. 4). Most professionals interpret the findings to mean that serious, chronic offenders are also likely to be heavy users of many different illicit drugs (Weis 1986).

In principle, researchers should be able to "triangulate”  different, fairly independent methods to get some kind of summary figure on phenomena like drug use. In fact, it would seem that this would be essential for the measurement of hidden deviance. However, the differences in methods, in the subpopulations studied, in statistical assumptions, and so forth, have thus far yielded disappointing results. This was symbolized by a 1997 White House drug strategy report that presented the number of "habitual”  cocaine users in the United States as both 582,000 and 2.2 million, based on NHSDA and ADAM data, respectively (Leen 1998).

Victimization surveys face similar problems with sampling and questioning. Take, for example, the National Crime Victimization Survey, the most highly regarded and long-term survey of crime in the United States. The weaknesses in the NCVS system are well known. It misses that part of the population most likely to be victimized - those who do not have any permanent residence, or who cannot or do not want to be found. For aggravated assaults and rapes, the NCVS figures appear to exclude many attacks by non-strangers that do not result in injuries. Research has shown the opposite as well - that some of the crimes reported, particularly of assaults and rapes by strangers, do not fit the legal criteria for those crimes (Gove, Hughes, and Geerken 1985).

In an effort to correct these problems and following recommendations by the National Academy of Sciences, the NCVS was redesigned in 1992 to stimulate the recollection of respondents and to allow more precise, less subjective responses (Kindermann, Lynch, and Cantor 1997). Both the earlier and later forms were used that year, for sake of comparison. The effect of the redesign was dramatic. Only three types of crime did not show significant rate increases under the new research procedures: robbery, theft, and motor vehicle theft. The reported amount of rape, particularly, was 157 percent higher using the new procedures than using the old ones, reflecting mostly an increased reporting of non-stranger, attempted (but not completed) incidents that were not reported to the police. For violent crime overall, the increases were greater for whites and other racial groups than for blacks, for more affluent people than for poorer people, for middle-aged people than for younger or older people, and for suburbanites than for urban or rural dwellers. Overall, people who were thought least likely to fall prey to criminal victimization were most likely to underreport in the old system (Kindermann, Lynch, and Cantor 1997, pp. 2–4). A sizeable fraction of the victimization gap between groups was shown to be an artifact of differences in their response patterns. The impact of the redesign on reported race differentials for rape was particularly dramatic. The redesigned NCVS now finds relatively small black-white differences.

The NCVS redesign has also helped to resolve a long-standing puzzle regarding racial differences in the UCR crimes known to police and self-reported delinquency surveys. UCR data show blacks to be dramatically overrepresented in many crimes; self-reported delinquency studies find only small racial differences. Was the difference caused by a biased criminal justice system or by some other artifact? It could be a result of the fact that most self-report studies usually concentrate on more minor offenses or that black respondents excessively misrepresent their delinquent activities, or it might be a problem of sampling because many self-report studies use high school students and blacks have higher drop-out rates than whites. In the late 1970s, Michael Hindelang showed that victimization statistics mostly supported arrest statistics, though a portion of the racial gap in arrests for rape, simple assault, and aggravated assault could be attributed to the racial bias of the police and the criminal justice system (Hindelang 1978). Because the NVCS redesign led to a reduction in the racial gap in victimization, it probably will lead to a reexamination of the extent of racial bias among police and in the criminal justice system more generally.

It is interesting to compare the picture of crime painted by the NCVS and UCR data. We would expect the two measures to display different pictures of the level and perhaps trend in crime because they were constructed on different bases. The UCR measure, "crimes known to police,”  understates less serious crime through its hierarchy rule. Because the NCVS records all incidents, the number of minor crimes it shows is undoubtedly more accurate than is the number reported by individuals to the police. Because the NCVS includes criminal actions that were not reported to the police, this would lead to a difference from the UCR even without the hierarchy rule.

Series Ec11 along with series Ec160 and series Ec165 present the level and trend in the crime rate as measured by these two statistics. As we would expect, the NCVS reports a much higher level of crime in all years than does the UCR. The magnitude of the difference may come as a surprise, though. In some cases it is as large as a factor of ten. The large difference between the two measures does not necessarily mean that the police are doing a poor job. It may be the case, as Wesley Skogan put the matter, that most victimizations are not "notable events.”  The majority are property crimes in which the perpetrators are never detected, the financial stakes are small, and the costs of calling the police greatly outweigh the benefits (Skogan 1981). Another possible explanation for the large difference between the two measures is that many victimizations are considered by the victim to be private matters. We know that many rapes and assaults perpetuated by intimates are viewed in this way.

Except for the most recent years, trends in the two sorts of data also differ. Crimes reported to police rise while reported victimizations fall. The fall in the victimization rate is the most pronounced trend. (It should be noted that the 1992 questionnaire redesign does not lead to a "bump”  in victimization rates because a correction factor is now applied to all previous years' rates to ensure comparability.) These differences are apparent across the full range of property and violent crimes. According to the UCR figures, property crime rates increased until the early 1980s and then slowly subsided, while the NCVS finds declining rates of property crime across virtually the entire period. The NCVS shows a slight downward trend in overall violent crime, whereas the UCR finds an explosion in violent crime up to the early 1990s. The general findings are that the NCVS reports much more crime than does the UCR, but the quarter-century trends are the opposite until the decade of the 1990s.

The rape statistics deserve special discussion. In the past, the reporting rate for rape was extremely low. Police statistics may seriously misrepresent both rates and trends. This is apparently what happened in the 1970s and 1980s. From 1973 through 1992, Crime in the United States showed a 76.1 percent increase in the rate of reported rapes, which was taken by many as indicating a profound increase in sexual violence in the nation (series Ec4). The NCVS reports across the same period, however, showed a 55.6 percent decrease in rape rates (series Ec161). Because it is based on a sample, NCVS figures tend to fluctuate, but even using a three-year (1990–1992) average for rape shows almost a 39 percent decrease from the 1973 figure. Much of the rise in rapes known to police can be attributed to a dramatic increase in the proportion of rapes reported to police and treated by the police as assaults. Figure Ec-F shows the dramatic divergence between the two measures of rape up through the early 1990s, with UCR figures rising and NCVS figures declining. Beginning in the early 1990s, both measures begin to decline together. (It should be repeated that NCVS data do not record victimizations among children younger than 12, so its rate may be artificially elevated; this should not affect general trends, however.) Also, to return to the general question of trends in lawlessness, the rape statistics do not indicate the increase in lawlessness since the mid-1970s that appears indicated in the sentencing statistics. Like the drug-use data, they point toward decreased lawlessness.

Back To Top

Criminal statistics remain difficult and unreliable, but not completely so. The frustrations that people who got a glimpse of the hidden world of crime experienced led to technical innovations that introduced light. By the 1930s, homicide victimization was well documented. By the 1990s, we had good quantitative evidence on especially difficult-to-measure crimes such as rape and drug use.

Some cautions are warranted though. There are problems that are very difficult to address in principle, and even if all major problems with contemporary criminal data are overcome, we will still face problems tracing most crime trends very far into the past. The UCR statistics are not hopeless; they and the redesigned victimization statistics tend ever more to point in the same direction with respect to trends in personal crime - that it is downward in the United States. If one gives up the dream of counting every single offense and is willing to settle for figures that seem in principle reasonably close to the hidden reality, then one can say that a lot is known about the amounts of different sorts of crime, about the people who commit them, and about the people who are victimized.

Differences in regional laws and policies remain problems. Cooperation between local or state and federal authorities is never complete. Police discretion at the scene and prosecutorial discretion affect the totals. There will, then, continue to be an element of indeterminacy in crime figures. These can be minimized in much the same way that differences in interpretation of activities have been minimized in the NCVS. The NIBRS, when it is completely developed, should substantially expand the amount of crime data available to scholars and law enforcement officials. Even now, the fact that the UCR figures include many estimates for nonreporting agencies does not appear to bias estimates of national trends (Maltz 1999).

The NCVSs have profited from focused attempts to draw information from the population. The professional staff within the Department of Justice, the National Academy of Sciences, and academic criminologists believe that the combination of random selection, regular contact over a period of years, and better interview protocols have improved the quality of these data. These data can help deal with the problem of unreported or unrecorded crime, but of course they still have the problems of sampling and voluntarism. The sample never fully represents the entire population. The people it misses - those who are homeless, in transit, in hiding, or uncooperative - are those most likely to be both victims and offenders. The data show the victimization experiences of the broad middle part of the population.

Even though self-report data are the weakest link in efforts to estimate personal crime, they have an important role to play, especially with measuring victimless crime. There are limits to the degree of honesty one can expect in the population, and self-reported delinquency always should be considered to understate the true amount of crime. We should not expect people to report delinquent or criminal actions more accurately than they report, say, church attendance. In principle, ways of estimating the degree of underreporting of delinquency can be developed, thereby allowing the derivation of correction factors. The work of Fendrich and Vaughn is an excellent example of what might be done along these lines (Fendrich and Vaughn 1994). They compared self-reports of "lifetime”  drug use for specific cohorts across time and discovered a progressive tendency to underreport youthful drug use as people age. Such evidence could be used to adjust the self-reported data and develop more accurate measures.

Efforts to improve the quality of the historical crime record face greater challenges. Take the example of the reformulated NCVS. Recall that almost all forms of crime were found to be more common under the new survey protocols than under the old ones. The estimated rate of rape was 157 percent higher. So what shall be done with findings from the years 1973–1991? The decision was to take the rate (and count) of rape for each of those years and multiply it by the ratio estimator (correction factor) of 2.57 (Rand, Lynch, and Cantor 1997, pp. 6–8). The corrected figures are now published for the earlier years and are included in Table Ec160–168. Although this is not the only approach one could take, it is straightforward and should be reasonably accurate for nearby years. Yet, the further in time one moves from the year in which the figure was calculated, the greater is the likelihood that the figure is inappropriate. Relationships change; consequently, there is a great deal of uncertainty about the estimates for early years in the series. This kind of uncertainty exists whenever data collection procedures change. Because of the desire to implement improved procedures, we can expect further changes in data collection procedures in the future. Whenever this happens, it will introduce problems for comparing figures from different years.

The problem of comparing data collected using different methods complicates all historical work on crime trends. New York, Boston, and Philadelphia have the longest sets of criminal records. They began recording crimes in the eighteenth century.³ Eric Monkkonen has developed a two-century homicide data series for New York City (Monkkonen 1989). Midwestern and Northeastern states started to publish crime statistics in the nineteenth century. In the South and West, especially in rural districts, there is little good-quality data before the Second World War. Consider statistics on homicide, the most easily counted crime. Texas did not collect death data that met even the minimal criteria needed for admission to the U.S. death registration area until 1933. South Carolina's death data were considered admissible in 1916; however, years later registration officials published extended tables showing how poorly its data collection compared with that of Vermont (Census Bureau 1924, pp. 97–108). Indeed a recent detailed study of crime in post-Reconstruction South Carolina - hardly the most remote or decentralized of the Old South states - finds that its main newspaper missed probably one third of its murders and that records in the state archives are missing over half (Eckberg 2001). Fragmentary evidence suggests that the further south and west one traveled and the more rural the district was, the higher was the rate of murder (see Montell 1986; Vandal 1991, 1994; McKanna 1995; McGrath 1989). The record is even less clear for other crimes. Overall, the severe limitations in crime data mean that it is very difficult to generalize about crime in America as a whole before the middle of the twentieth century.

One should mention an exception to such a lack of data. In the case of lynching, usually the most public form of homicide, records drawn mainly from news accounts, and now housed at Tuskegee University, have long been the main data source for scholars. It has been long known that the Tuskegee data are somewhat unreliable, about when or where an incident took place and about whether it was indeed a "lynching”  (see the discussion in Tolnay and Beck 1995). Stewart Tolnay and E. M. Beck have gathered the most comprehensive state-by-state lynching data series, using modern methods, not for the entire nation but for ten Southern states. W. Fitzhugh Brundage has added lynchings for Virginia (Brundage 1993). In Figure Ec-G are plotted year-by-year totals from Table Ec251–253 and Table Ec254–289. It is clear that they parallel each other very closely. The new data series, which excludes Texas and all Northern and Western states, has slightly fewer lynchings, year-by-year, across the entire period.

Roger Lane attributes our difficulties in counting crimes, even in the rare "settled late nineteenth-century American cities, in the Northeast or West, with established police forces and recording procedures”  to two main causes (Lane 1992, p. 39). The first of these he calls "prejudice,”  referring to the prejudices that directed the passing of laws, arrests, and record keeping of nineteenth-century record keepers, but also referring to the ideological positions of modern scholars who take sides on the meaning of the old records. He does not find the first cause unsolvable, but the second one almost causes him to throw up his hands. This is that dark figure, the "holes in the historical record,”  caused by differences in categorization, lacunae in recording, indifference by authorities, changes in laws, and uncertainties about the sizes of populations and subpopulations, to name some factors (Lane 1992, p. 30). His complaint reads much like Sutherland's, and he believes we cannot in principle know with certainty if there was proportionately more, or less, of most types of crime in the past than now. He is willing to hazard guesses (his term) about broad quantitative changes in some offenses - murder of adults, theft, assault, perhaps robbery. The general trend of Lane's and Eric Monkkonen's findings on urban homicide, the best documented crime, is that it declined from the mid-nineteenth century to roughly the end of the century and then grew until the 1930s (Lane 1992; Monkkonen 1980, 1989). Lane is unwilling to discuss rape, except to speculate that it may have been common among working girls and step-children. He is similarly disinclined to discuss infanticide or abortion (Lane 1992, pp. 45–7). Infanticide, which he calls a "wild card,”  was often hard to detect, rarely acknowledged, and seldom prosecuted. It almost never led to a conviction. Many infant murders were a form of ex post facto abortions, replacing the ordinary ones that had been legal until the 1860s and that were mostly hidden afterward. Infanticides and abortions thus coexisted in a legal shadow land, with no clear line separating them. To develop estimates, one needs some kind of record with a known basis. Here the record is both mostly blank and murky. The other crimes - the ones that make up the bulk of the crime record - are much more difficult and much less reliable, though there is some historical record even for them. It remains to be seen what can be done with them.

Back To Top

[TABLE Ec-A Important events in crime and criminal statistics: 1850–2001]

Back To Top

Series Ec311–312 and Series Ec314 expressed as a share of series Aa113, series Aa116, and series Aa119, respectively. Beginning 1986, the number of nonwhite prisoners is calculated as the sum of series Ec312–313.

Figures cover sentenced prisoners admitted to state and federal institutions for whom racial data are available.

Back To Top

Series Ec191–192.

Back To Top

Series Ec12 and Series Ec17.

For display purposes, the two lines are plotted on different scales because the rate of property crime is so much higher than the rate of violent crime.

Back To Top

Series Ec969 and Series Ec1029.

Back To Top

Series Ec14 and Series Ec161.

Note that the National Crime Victimization Survey covers the population age 12 and older, while crimes known to police are expressed relative to the entire population.

Back To Top

Series Ec251 and Series Ec254.

The figures in series Ec251 present national totals for the reported white and black victims of lynching. Those in series Ec254 (Tolnay and Beck 1995; Brundage 1993) present totals of known victims of lynching (all races) in eleven Southern states.

Back To Top

Adler, Jeffrey. 2001. "Halting the Slaughter of the Innocents”: The Civilizing Process and the Surge in Violence in Turn-of-Century Chicago.”  Social Science History 25: 29–52.

Allport, Gordon, and Leo Postman. 1975 [1948]. The Psychology of Rumor. Russell and Russell.

Arrestee Drug Abuse Monitoring Program. 1999. 1998 Annual Report on Drug Use among Adult and Juvenile Arrestees. National Institute of Justice.

Barker, Tom, and David Carter. 1990. "Fluffing Up the Evidence and Covering Your Ass: Some Conceptual Notes on Police Lying.”  Deviant Behavior 11: 61–73.

Best, Joel, and Gerald Horiuchi. 1985. "The Razor Blade in the Apple.”  Social Problems 32: 488–99.

Bishop, G. F., and B. S. Fisher. 1995. "Secret Ballots and Self-Reports in an Exit-Poll Experiment.”  Public Opinion Quarterly 59: 568–88.

Bolton, Richard, editor. 1992. Culture Wars: Documents from the Recent Controversies in the Arts. New Press.

Bradburn, N., L. J. Rips, and S. K. Shevell. 1987. "Answering Autobiographical Questions: The Impact of Memory and Inferences on Surveys.”  Science 236: 157–61.

Brundage, W. Fitzhugh. 1993. Lynching in the New South: Georgia and Virginia, 1880–1930. University of Illinois Press.

Bureau of Justice Statistics. 1998. Sourcebook of Criminal Justice Statistics 1997. U.S. Government Printing Office.

Cahalan, Margaret Werner. 1986. Historical Corrections Statistics in the United States, 1850–1984. NCJ-102529. Bureau of Justice Statistics.

Carter, Dan T. 1979. Scottsboro: A Tragedy of the American South. Revised edition. Louisiana State University Press.

Census Bureau. 1906. Mortality Statistics: 1900–1904. U.S. Government Printing Office.

Census Bureau. 1924. Mortality Statistics: 1921. U.S. Government Printing Office.

Children's Bureau. 1958. Statistics on Public Institutions for Delinquent Children: 1956. Statistical series number 48. U.S. Department of Health, Education and Welfare.

Eckberg, Douglas Lee. 1995. "Estimates of Early Twentieth-Century U.S. Homicide Rates: An Econometric Forecasting Approach.”  Demography 32: 1–16.

Eckberg, Douglas Lee. 2001. "Stalking the Elusive Homicide: A Capture-Recapture Approach to the Estimation of Post-Reconstruction South Carolina Killings.”  Social Science History 25: 67–92.

Elliot, Delbert S., Suzanne S. Ageton, et al. 1983. The Prevalence and Incidence of Delinquent Behavior: 1976–1980. The National Youth Survey Report, number 26. Behavioral Research Institute.

Federal Bureau of Investigation. 1998. Crime in the United States, 1997. Uniform Crime Reports. U.S. Government Printing Office.

Fendrich, M., and C. M. Vaughn. 1994. "Diminished Lifetime Substance Abuse over Time: An Inquiry into Differential Underreporting.”  Public Opinion Quarterly 58: 96–123.

Goldstone, R. L., and C. Chin. 1993. "Dishonesty in Self-Report of Copies Made: Moral Relativity and the Copy Machine.”  Basic and Applied Social Psychology 14: 19–32.

Goodman, James. 1994. Stories of Scottsboro. Pantheon.

Gove, Walter R., Michael Hughes, and Michael Geerken. 1985. "Are Uniform Crime Reports a Valid Indicator of Index Crimes? An Affirmative Answer with Minor Qualifications.”  Criminology 23: 451–501.

Gurr, Ted Robert, editor. 1989. Violence in America, volume 1, The History of Crime. Sage.

Hindelang, Michael J., Travis Hirschi, and Joseph G. Weis. 1981. Measuring Delinquency. Sage.

Hindelang, Michael. 1978. "Race and Involvement in Common Law Personal Crimes.”  American Sociological Review 43: 93–109.

Hindus, Michael S. 1980. Prison and Plantation: Crime, Justice and Authority in Massachusetts and South Carolina, 1767–1878. University of North Carolina Press.

Hunter, James D. 1991. Culture Wars: The Struggle to Define America. Basic Books.

Kindermann, Charles, James A. Lynch, and David Cantor. 1997. Effects of the Redesign on Victimization Estimates. National Crime Victimization Survey. Bureau of Justice Statistics.

Kleck, Gary. 1991. Point Blank: Guns and Violence in America. Aldine de Gruyter.

Lane, Roger. 1968. "Crime and Criminal Statistics in Nineteenth Century Massachusetts.”  Journal of Social History 2: 156–63.

Lane, Roger. 1979. Violent Death in the City: Suicide, Accident, and Murder in Nineteenth Century Philadelphia. Harvard University Press.

Lane, Roger. 1989. "On the Social Meaning of Homicide Trends in America.”  In Ted Robert Gurr, editor. Violence in America, volume 1, The History of Crime. Sage.

Lane, Roger. 1992. "Urban Police and Crime in Nineteenth-Century America.”  In Michael Tonry and Norval Morris, editors. Modern Policing: Crime and Justice, A Review of Research, volume 15. University of Chicago Press.

Langan, Patrick A. 1991. Race of Prisoners Admitted to State and Federal Institutions, 1926–86. NCJ-125618. Bureau of Justice Statistics.

Leen, Jeff. 1998. "Number Jumble Clouds Judgment of Drug War: Differing Surveys, Analyses Yield Unreliable Data.”  Washington Post, January 2, p. A01.

Loftus, Elizabeth. 1979. Eye Witness Testimony. Harvard University Press.

Loftus, Elizabeth, and Katherine Ketcham. 1994. The Myth of Repressed Memory: False Memories and Allegations of Sexual Abuse. St. Martin's Press.

Maltz, Michael D. 1999. Bridging Gaps in Police Crime Data. NCJ-176365. Bureau of Justice Statistics.

Marler, Penny Long, and C. Kirk Hadaway. 1999. "Testing the Attendance Gap in a Conservative Church.”  Sociology of Religion 60: 175–86.

McGrath, Roger D. 1989. "Violence and Lawlessness on the Western Frontier.”  In Ted Robert Gurr, editor. Violence in America, volume 1, The History of Crime. Sage.

McKanna, Clare V., Jr. 1995. "Alcohol, Handguns, and Homicide in the American West.”  Western Historical Quarterly 26: 455–82.

Monkkonen, Eric H. 1980. "The Quantitative Historical Study of Crime and Criminal Justice.”  In James A. Inciardi and Charles E. Faupel, editors. History and Crime: Implication for Criminal Justice Policy. Sage.

Monkkonen, Eric H. 1981. Police in Urban America: 1860–1920. Cambridge University Press.

Monkkonen, Eric H. 1989. "Diverging Homicide Rates: England and the United States.”  In Ted Robert Gurr, editor. Violence in America, volume 1, The History of Crime. Sage.

Montell, William Lynwood. 1986. Killings: Folk Justice in the Upper South. University Press of Kentucky.

Porterfield, Austin L. 1943. "Delinquency and Its Outcome in Court and College.”  American Journal of Sociology 49: 199–208.

Presser, Stanley, and M. Traugott. 1992. "Little White Lies and Social Science Models.”  Public Opinion Quarterly 56: 77–86.

Rand, Michael, James A. Lynch, and David Cantor. 1997. Criminal Victimization, 1973–95. National Crime Victimization Surveys. Bureau of Justice Statistics.

Rand, Michael. 1998. Criminal Victimization 1997: Changes 1996–97, with Trends 1993–97. A National Crime Victimization Survey Report. Bureau of Justice Statistics.

Richardson, James T., Joel Best, and David G. Bromley, editors. 1991. The Satanism Scare. Aldine de Gruyter.

Skogan, Welsey. 1981. "Assessing the Behavioral Context of Victimization.”  Journal of Criminal Law and Criminology 72: 727–42.

Sutherland, Edwin H. 1924. Criminology. J. B. Lippincott.

Tolnay, Stewart E., and E. M. Beck. 1995. A Festival of Violence: An Analysis of Southern Lynchings, 1882–1930. University of Illinois Press

Vandal, Gilles. 1991. "‘Bloody Caddo’: White Violence against Blacks in a Louisiana Parish, 1865–1876.”  Journal of Social History 25: 373–88.

Vandal, Gilles. 1994. "Black Violence in Post-Civil War Louisiana.”  Journal of Interdisciplinary History 25: 45–64.

Weiner, Neil Alan, and Margaret A. Zahn. 1989. "Violence Arrests in the City: The Philadelphia Story, 1857–1980.”  In Ted Robert Gurr, editor. Violence in America, volume 1, The History of Crime. Sage.

Weis, Joseph G. 1986. "Issues in the Measurement of Criminal Careers.”  In Alfred Blumstein, J. Cohen, et al., editors. Criminal Careers and "Career Criminals.”  National Academy Press.

Wells, Gary L., and Elizabeth Loftus, editors. 1984. Eye Witness Testimony: Psychological Perspectives. Harvard University Press.

Wilbur, Cressy L. 1916. The Federal Registration Service of the United States. U.S. Government Printing Office.

Wright, James D., and Peter H. Rossi. 1986. Armed and Considered Dangerous: A Survey of Felons and Their Firearms. Aldine de Gruyter.

Back To Top