A MODEL FOR EXPLAINING THE DISCREPANCY BETWEEN OFFICIAL CRIME RATES AND VICTIMIZATION SURVEY CRIME RATES Dissertation for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY HERBERT L. TYSON, JR. 1977 IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII 293 00085 0507 This is to certify that the thesis entitled A MODEL FOR EXPLAINING THE DISCREPANCY BETWEEN OFFICIAL CRIME RATES AND VICTIMIZATION SURVEY CRIME RATES presented by HERBERT L. TYSON, JR. has been accepted towards fulfillment of the requirements for PH . D . degree in Sociology Major profes r > C! C DateW}? 0-7639 ABSTRACT A MODEL FOR EXPLAINING THE DISCREPANCY BETWEEN OFFICIAL CRIME RATES AND VICTIMIZATION SURVEY CRIME RATES By Herbert L. Tyson, Jr. Some researchers have investigated the discrepancy between official crime and survey crime, and have found that the two rates covary, but are not perfectly correlated. This thesis sought to explain the discrepancy by hypothesizing that the ratio of the two rates can be used as a social indicator, utilizing a framework derived from the social theories of Emile Durkheim. This study specified three regression equations using combinations of ten independent social and economic variables to explain each of nine different categories of crime. The equations used operationalize factors specific to the popula- tion of large cities, large cities' police departments, and a combination of both types of factors. It was found that factors relating to pcpulation size, density, and age composition were not significantly related to the discrepancy, while the proportion of blacks, the aver- age earnings from manufacturing, police per capita, and police civilian employment tended to explain most of the variance in the dependent variable. This dissertation concluded that factors such as police bias and data manipulation are not the primary causes of the discrepancy. The most important finding was that the Herbert L. Tyson, Jr. discrepancy is caused by perceptual factors which influence the attitudes of the population, such as the visibility of racial differences and the visibility of the police. The study suggests that the best way to deal with inaccuracy of police crime statistics is through public education and policies which promote police-community based approaches to crime. A MODEL FOR EXPLAINING THE DISCREPANCY BETWEEN OFFICIAL CRIME RATES AND VICTIMIZATION SURVEY CRIME RATES By Herbert L. Tyson, Jr. A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Sociology 1977 This work, as is everything I do, is dedicated to my wife, Karen, whose love, patience, intelligence, and un- derstanding make every thing and every idea I value worthwhile. ii ACKNOWLEDGMENTS I gratefully acknowledge the guidance, assistance, and encouragement of my doctoral committee: Jim Zuiches, Chris vanderpool, and Tom Conner. Their thoughtful advice has provided me with more help than they can realize. I would especially like to acknowledge the apprecia— tion I feel for the time, guidance, editing, advice, and insight given me by Peter Manning. Without his help, I have serious doubts that this dissertation would have been undertaken, let alone completed. I would also like to thank Larry Hoover, Mike Gins, Bernie Gilman, Harry Greenspan, and Karen Tyson, who have all contributed in various ways to this dissertation. iii TABLE OF CONTENTS List of Tables . . . . . . . . . . . . . . . . . . . . Vi List of Figures . . . . . . . . . . . . . . . . . . . . Vii Chapter 1 Introduction . . . . . . . . . . . . . . . . 1 Chapter 2 Literature and Theory Bearing on the UCR and Other Measures of Crime . . . . . . 13 Introduction to the Problem . . . . . . 13 Theory and Literature Bearing on the UCR and Other Measurs of Crime . . 2O Victimization Surveys . . . . . . . . . 27 A Question of Measurement . . . . . . . 28 Reconciling the UCR and N03 Crime Rates 31 Production of the Discrepancy-~Ctizens and the POI-ice O O O O O O O O O O O O 37 Hypotheses, variables, and Propositions 41 Chapter 3 Data, Method, and Analysis . . . . . . . . . 50 Data and variables . . . . . . . . . . 50 Analysis . . . . . . . . . . . . . . . 54 Citizen Sector variables . . . . . . . 59 Police Sector . . . . . . . . . . . . . 63 General Findings . . . . . . . . . . . 65 Path Analysis Decomposition of U/v . . 68 Analysis of Propositions . . . . . . . 73 Analysis of Separate Crime Equations . 87 Chapter 4 Summary and Conclusions . . . . . . . . . . 101 The Problem . . . . . . . . . . . . . . 101 The Theory . . . . . . . . . . . . . . 103 Results . . . . . . . . . . . . . . . . 105 Discussion . . . . . . . . . . . . . . 105 Conclusion . . . . . . . . . . . . . . 110 iv/t/ Table Table Table Table Table Table Table Table Tabl e 3.1 3.2 3.3 3.4 3.5 3.6 3.8 4.1 LIST OF TABLES a. Zero Order Correlatin.Matrix for u/v Discrepancy by Level of Crime and Social and Police Indicators . . . . 55 b. Zero Order Correlation Matrix for Social and Police Indicators . . . . . 55 U/V Ratios and Frequencies for Official and Victimization Offenses . . . . . . . . . 58 R—Squared for Regression Equations and Significance by Category of Crime . . . 67 Independent and Common variation EXplained by Civilian and Police Sectors and F Tests of Significance for Each Sector in Whole Model by Category of Crime . . . . . . . . . . . . 69 Zero Order Correlation Matrix of Selected In- dependent and Dependent variables with Logarithmic Transformations . . . . . . . . 72 Means and Standard Deviations for Selected Dependent variables 0 o o o o o o o o o o o 72 Estimated Coefficients and T Statistics for Citizen Sector Regressions by Category 0f crime 0 O O O O O O O O O O O O 0 O O O O 74 Estimated Coefficients and T Statistics for Police Sector Regressions by Category Of Crime 0 O O O O O O O O O O O O O O O O O 84 Summary of PrOposition Results . . . . . . . 106 Figure Figure Figure Figure Figure 3.1 3.2 3.3 3-4 LIST OF FIGURES Flow Chart Showing Processing of Potential Criminal Event . . . Adjustment Procedure for Making UCR and NCS Comparable within each Category of Crime . . . . . . . . Data Sources and Interpolation Method Used in this Study . . . . . Method Used to Create variables and Symbols . . . . . . . . . . . . Path Decomposition of u/v . . . . . vii 36 51 53 53 70 CHAPTER 1 INTRODUCTION The use of quantification procedures in attempting to analyze the workings of society is an established practice. The roots of thought concerning the deveIOpment of knowledge about cause and effect regarding social relations owe a cer- tain partcf their heritage to scientific investigation. Such must have enticed the ancients to consider questions of society in a similar way. It was not, however, until comparatively recently that such thought became socially acceptable. The acceptability of theorizing about society in such terms, cause and effect, has, after all, been challenged at many points in time by conservative influences which considered such questions as not only beyond man's reach, but beyond his jurisdiction as well, bordering on sacrilege and heresy. The division between traditional beliefs and the ideas preposed by sociology's founders and the challenge presented by empiri- cism, have been discussed by Martindale (1960) and others (Barnes, 1940; Barnes and Becker, 1952). Much of the question has revolved about the possibility, utility and apprOpriateness of applying scientific questions to social phenomena. Many have considered, as some still do, that human social behavior is beyond the realm of measurement and generalization, that each being is unique, and that to attempt to prove otherwise is fallacious. It is as if the very posing of such questions were an insult or a threat to man's uniqueness, individuality, and his distinctive posture in the world. To attempt to find patterns in the way people organize 1 2 represents a very real attempt to destroy the mystique in which mankind has basked for so long. Thus, it has not been willingly that many have come to accept the science of sociology as a valid and justified dis— cipline. Many still challenge its most fundamental postulates. Perhaps it was fitting then, that sociology itself began as a challenge: not as an academic discipline, but as a practical step in which certain men sought to use such knowledge to make society "better"-—to reform-or to use it for their own gain. Michael Cullen (1974) put forth the thesis that man first began to collect information about social aggregates during the 1600's, and perhaps earlier. The first applications oc- curred in Europe, primarily as a means of determining how much tax was due land owners and governments, based on the enumeration of those who occupied it. It was at about the same time that some began to question the humanity of particu- lar social conditions. Social statistics, or "political arith— metic", was then becoming useful for reformers, who sought to prove that certain conditions or relations existed which were contrary to the moral tenets by which they were governed. Thus, such statistics were a political instrument of challenge, in that the conditions to which they testified were a recom- mendation for change. Through many phases, the practice of collecting data a- bout groups and social conditions develOped, having been for- malized in some societies through a regular census, and having 3 been incorporated into the decision making apparatus of many governments. Still, the study of social relations and human behavior were only recent arrivals, and were, in and of them- selves, seldom the object of study. The introduction of human behavior as something which could be measured and ex- plained by aggregate data was still not in vogue into the 19th century.1 Jack Douglas (1971) has provided a context which is es— pecially relevant to this development. Durkheim, he says, set about to specifically engineer a discipline which would deal with the science of human behavior. A principal vehi- cle for this effort was the classic study Suicide. Into the 1800's, as even today, subide was considered by some to be the single most individualistic act that a person could per- form-the ultimate demonstration of free will and self—deter- mination. Little of anything else that man could do could so clearly articulate man's independence. While many had gathered statistics on suicide, it was not until the early 1800's that enough data had been gathered to allow trends to become visi- ble. Until then, few took such statistics at more than face value. In the years preceding Durkheim, students of moral stat- istics, as they were called, began to notice regularities in suicide rates. Quetelet, Morselli, Buckle, and others noticed that some groups were consistently undergoing the same suicide rates, (see Douglas, 1967: pp. 7—12). Announcing the birth of social indicators, Durkheim's predecessors claimed that these rates were all indications of specific conditions in 4 society, or certain underlying causes. Joining them in this denial of the individuality of suicide, Durkheim chose to launch the science of sociology, claiming as did they that certain moral relations may be indicated by aggregate rates, and that a certain dimension of underlying meaning produces the connection between different social rates. There are, Durkheim implied, several layers of meaning 2 There are the meanings that peOple which must be understood. themselves give to the various forms of interaction and be- havior with which they are occupied. These are the ostensible rationalizationsthat one encounters. As such, these reasons apply to suicide to the extent that one might cite them as those which compel one to take his own life-~money, love, re— venge, etc. Aside from the meanings which individuals assign and re- cognize, there are meanings within society as a whole-relevant to the systems of relations that exist in socidies. That is, why are so many individuals-—for whatever personal justifica- tions-~at the same time and with the same frequency within given societies, compelled to commit suicide? Why is it that so many peOple simultaneously begin to steal, murder, and rape? If such acts were truly individualistic, then would not one expect them to occur randomly? Durkheim posited several underlying phenomena. He theor- ized that one special general dimension was the rise of modern societies—-industrialization and urbanization—-and the gra- dtuil breakdown of socializing structures which promote some lexrel of equilibrium or social health (see Divisbn of Labor 5 in Society, pp. 353-373). On one level he analyzed variations from group to group, with very specific (and perhaps ad hoc) rationales. On the broad level, however, Durkheim posed several different theories which relate to people,and why the behavior of groups and societies change over time. To analyze the change, oneimnrpursue either of two courses. One may analyze the individual groups and investi- gate how the composition of society changes when some groups wax and wane. 0n the other hand, one may analyze society to determine what kinds of broad structural changes are occurring. Either the changing composition or broad structural changes may account for changes in human behavior as indicated by aggre— gate rates, but the broad level deals especially with the cases where the general society is in flux. Thus, on a particular level, we may determine that anomie is becoming widespread, or that one group is more diSposed to egoism than the rest of the society-~upsetting the equilibria, however defined, and causing rate changes for that group. On another level, however, a different kind of balance is dis- turbed—-that of socialization routines and the social en- vironment. More specifically, we may determine, as did Durkheim, that certain forms of organization are apprOpriate for particular sizes of societies. In contrasting mechanical with organic solidarity, Durkheim describes a scenario in which all social behavior is determined through internaliza- tion, with small close-knit societies wherein solidarity is insured by virtue of sameness (Division..., pp. 70-110 and 111-132). On the other hand, large groups with diverse 6 employment and interests can not rely on sameness. The equil- ibrium of interdependence is more tenuous. One might further hypothesize that socialization practices have not evolved quickly enough to c0pe with certain ecological changes, e.g. those in technology and population, resulting in an imbalance. Hence, anomie, a condition which is related to the failure to communicate what is socially acceptable and/or eXpected, can be seen as more prevalent in a rapidly changing society, in which many individuals are alienated and are unaware of what society wants of them. Anomie is used in a loose sense throughout this dis- cussion, in no way intending to incorporate or assume all of the conceptual dimensions implied by Merton, nor confined just to that range of meaning used in Social Theory and Socid Structure. There, Merton uses a more specific conceptualiza- tion of anomie, in his discussion of "means and ends", in which anomie is the "imperfect coordination of the two," (p. 159). In the present discussion, anomie is presented as a general state of normative disorder in which social disorganization is prevalent as manifested by high crime rates and other social indicators, such as divorce, suicide, racial strife, and juvenile delinquency. This is similar to the way Lander (1954), Chilton (1964), and Slaten (1969) view society, each of whom makes reference to a more general conceptualization. Here, especially at times when society is in flux (which may be the way many sociologists view modern society), what ig and ig ngt normative becomes problematic. Thus, the useage is in- tended to be taken as the general disorganization resulting 7 from and/or attending a problematic or unclearly defined normative structure. Such changes may be seen as occurring over time, but additionally, if we concede the universality of the mechanism which creates anomie, we may measure the difference between particular societies or groups within the same time period. Thus, we might observe different suicide rates in different countries and arrive at certain inferences about them. We may decide that one is closer to equilibrium than the other. Then again, we may decide that one is composed of more in- dividuals who are "well-adjusted" or properly balanced, owing to special characteristics of their groups. In either case, we are using a rate as a measure of some underlying dimension or meaning. At the same time, it should be clear that the social health or equilibrium may be measured in other ways as well: in terms of crime rates, mortality and accident rates, economic levels, and demographic composition. The de- cision to proceed along these lines, however, is subject to several criticisms. Douglas criticized Durkheim for precisely this reason-— that he usually assumes the underlying moral relations or meanings, and assigns them his interpretation by using "com- mon sense," rather than empirical validation. This applies specifically to assumptions about what is considered right and wrong in a given society, but as well to the existence of underlying dimensions. Validation of right and wrong as seen by the pOpulation is certainly obtainable, within certain limits, through empirical research. Validation of a theoretical concept, 8 however, is subject to numerous problems which should be obvious from the history of science and sociology. Even with empirical validation, one must be aware that whatever underlying dimensions are "discovered"-—through factor analysis or whatever—~are still subject to more detailed interpretation, and in the last analysis are still theoretical. Thus, all of Durkheim's constructs are simply tools for or- ganizing reality as he sees it, just as any theory uses and invents concepts as are necessary to "explain" a phenomenon. Whether or not such theories are t£3g_is beyond practical proof. They may simply remain "valid" until disproven, but they can never be said to be true. Hence, even well taken, Douglas' criticism may never be fully heeded. Another caveat relevant to the summarization and analysis of aggregate data is that of the "ecological fallacy" (Robinson, 1950). When one says that suicide rates increase due to anomie, another wonders why, then, everyone does not commit suicide. It is here that the task is arduous, and one must become increasingly specific. It is not that anomie creates or causes suicide, but rather that it increases the suanpti- bility of some who may be in special states of disequilibrium related to certain specific circumstances. The effect is simply that, by the laws of probability, a certain constant prOportion of a large population will be in such a state, and that increased anomie will activate the susceptible.fraction to a uniform degree. Even so, not everyone in such circum- stances will commit suicide—-the decision of any one individual is still highly unpredictable. Stochastically, however, we 9 may know that a particular proportion will commit suicide with a given change in certain social indicators which we say measure anomie. Rather, it is this confusion, that of concernwith specific individuals, which creates the problem, in that the economic and social health of a society is in a real sense the composite health of its members. The idea then, is not to relate to the behavior as individual per se, but to provide broad outlines of expectations about large groups, and in the apprOpriate setting to allow individuals to utilize thatinformation accordingly. By analogy, we can not tell if an individual will die from smoking tobacco. However, we can apprise him of the risk to his age group, sex, race, and so on. Thus, the ecological fallacy is well taken as a caveat, but should not deter the science of generalizations provided that we do not make pre- dictions about individuals. These cautions in mind, then one may turn to the ques- tion of the general state of society. We have theorized that suicide and crime rates may be indicators of certain underly- ing societal dimensions. That of crime, particularly, may be seen as relevant to a failure to adhere to socially approved behaviors. Thus, in an anomie prone society, we may ex- pect relatively larger prOportions to fail to identity with whatever socially approved means are present, and to take short cuts to perceived goals-—i.e. commit crimes. But, what about those against whom crimes are committed? In his discussion of the "collective conscience," Durkheim posits that certain acts are crimes because they 10 shock the collective conscience, and not vice-versa. There are also established routines for dealing with such offenses, once juridically defined. There are additionally, in some societies, laws which require members of those societies to report such offenses when they are observed, and which re- quire the utilization of those estafidshed routines. Durkheim effectively defines lawseuscodified norms, with a potential for strong negative sanctioning of certain behav- ior by society. Thus, by a Durkheimian definition, reporting crime is the apprOpriate response to crime which society con— dones, and which it requires in order to mete out even justice. What then, of those do not report crime? Is this a form of anomie? Within the Durkheimian conceptualization, the failure to report a crime must certainly be viewed as an indicator of certain dysfunctions of the society. As something which shocks the collective conscience, the prescribed demonstration of that shock is crime reporting. That failing to occur, may be taken as a social indicator, a measure of some underlying meaning. The failure of a large proportion of peOple to report, consistently, may then be taken as a measure of that society's health, and possibly as an indication of anomie. It may also be that, as a parallel with increasing com- plexity, the trend towards increasing formalization increases the collective obligation-i.e. at the formal level-while decreasing that of the individual. That is, a decline in the focus on individual and informal control may provide for the shrinking role of individuals in the control process. Hence, 11 as a parallel to formalization, individuals may report crime less often, especially when it does not involve them per— sonally, leaving such matters to the formal process. To attempt to incorporate this dimension, formalization (as Opposed to anomie, as an additional underlying meaning), into this study, however, might be less than useful. As will be shown, this study concerns a group of large U.S. cities. By their very nature, such cities may have achieved the same level of formalization-a consequence of being part of the same large industrialized country. Rather, such a dimension would be more appropriate to a cross-national study, the s00pe of which would require considerably more sophisticated and ex- tensive data than is available. Within the U.S., the degree of formalization within large cities is therefore, assumed to be constant, and not very amenable to empirical study due to measuring problems and insignificant variation. Thus, the central question, anomie, concerns the various levels of meaning involved in the reporting phenomena. There are, as before, several levels. First, we have the ostensible reasons for failing to re- port a crime—~memories of being called a "tattletale", fear of reprisal, fear that one's own culpability might be revealed, or sympathy with the offender. The fact that these, however, are allowed to outweigh the influence of the "codified norm," or the collective need to have all such crimes reported is a definitecflue about the society. The fact that so many peOple, in particular places, are subject to such circumstances is as much a measure of the general state of society as crime itself. 12 We may eXpect that differences which occur are related, as is crime, both to the general conditions of a place-~the economic or developmental level—~as well as the composition of a place-~age, race, sex, etc. We may expect that such factors will be related as well to the indicators of that dim- ension. Additionally, we may expect that a composition which contains more members which are susceptible or predisposed to the given malady will be correlated with higher rates for the indicators. The missing link, then is the theory , that which bridges the gap between the several indicators and explains whatever relationhip exists. After that, then one may ex- plore those relationships, explicating them both in theoreti- cal and practical (common sense) terms. Hence, derived from Durkheim, the basis of this disserta- tion utilizes the theory that certain.moral relations may be indicated by the way in which members of a society report crime. Greater reporting indicates a more universal incor- poration of the norms governing the reaction to crime and norms governing crime itself (e.g. see Rossi and Berk, 1975; Erickson, Gibbs and Jensen, 1977).3 Such behavior may be seen as corresponding to various levels of economic and social integration, however Operationalized or measured, as well as the composition of particular places, all of which are in some way determined or made problematic by the underlying dimen- sions relating to anomie. 0n the basis of these broad theoretical concerns, we would now like to look at the general problems of measurement and at a conceptualization of our problem. CHAPTER 2 LITERATURE AND THEORY CONCERNING CRIME REPORTING Introduction to the Problem Historically, there has been a need for accurate crime statistics for use by law enforcement officials, administra- tors, scholars and the public. Accurate measures of crime are an invaluable tool in planning policy, developing theor- etical perspectives on criminality, and giving the public an idea about personal risk and the effectiveness of their police departments. However, the measurement one applies will de- termine the extent of its usefulness, as will its specific applicability for the problem under study. This concern for obtaining measures of crime has, paradoxically, resulted in there being two such measures, with one being very extensive, yet with much question about its validity, and the other be- ing rather limited in scope, but believed by some to be super— ior in measuring the magnitude of crime. In 1929, primarily in response to the administrative needs of police departments and law enforcement agencies, the first of these measures, the Uniform Crime Reports (UCR) "crimes known to the police" was initiated by the Interna- tional Association of Chiefs of Police. This reporting sys- tem was delegated, by Congress, to the FBI in 1931, and con- ‘tinues to this day. Prior to that, statistics had been kept on the local JJBXrel, but had never been fully collected on a national basis. JFEIITther, such statistics were likely to be of much better (leéaJlity in the large cities in the northeastern parts of the 13 14 U.S., where policing had become a formal and collective pro- blem (Lane, 1971). The establishment of the FBI and the creation of the UCR, in fact, mark the beginning of the re- cognition of crime as a national phenomenon. Increasing mo- bility and the common problems attending large cities empha- sized the inability to locate lpggl culpability, and the need to focus national resources and attention on the problem. It was in this context that crime became a national concern, as recognized by Congress. The FBI's Uniform Crime Reports collects data on "offen- ses known to the police" from over 11,000 rural, state, and urban law enforcement agencies, along with other information, in monthly and annual reports or returns. The annual returns are used to summarize and correct monthly returns, and are filed on forms provided by the FBI. The returns are volun— tarily filled out and mailed to the FBI each January, and cover the previous calendar year. The offenses of primary interest in this study are those used by the FBI to construct the "Total Crime Index" (TCI). They are known as "Part I Offenses" and consist of seven crimes which are common and subject to "frequent" reporting. They are: 1. Criminal homicide--"The killing of one human being by another" 2. Rape--"The carnal knowledge of a female forcibly and against her will" 3. Robbery-~"A special and vicious type of theft. It takes place in the presence of the victim. To obtain the prOperty or thing of value the robber uses force or violence on the victim or puts the victim in fear by use of threats, violence, etc." 4. Aggravated assault-—"An unlawful attack by one person upon another for the purpose of inflicting severe 15 bodily injury usually accompanied by the use of a weapon or other means likely to produce death or great bodily harm. Attempts are included." 5. Burglary--"Housebreaking, safecracking, or any un— lawful entry of a structure to commit a felony or a theft, even though no force was used to gain entrance; includes attempts" 6. Larceny--Theft-"Felonious stealing, taking and carrying, leading or driving away of the personal prOperty of another without claim of right, with in- tent to deprive him of his ownership or to convert such prOperty to the use of the taker or another" 7. Auto theft-~"All cases where automobiles, trucks, busses, motorcycles, motor scooters, or other self- propelled vehicles that run on a non—rail surface, are taken by persons not having lawful access thereto and are later abandoned" The total crime index consists of all of these (in which homicide excludes negligent manslaughter). The FBI gives ex- plicit instructions for the enumeration of these index crimes in the Uniform Crime Reporting Handbook which defines each crime (the above definitions are paraphrased from the Handbook) and provides procedures for deciding which crimes to include. The Handbook provides numerous detailed examples. For example, with burglary, actually at least two crimes are usually committed-—burglary and larceny. However, larceny would apt, be recorded, since it is lower on the list-~i.e. less "serious." They also detail what p33 to include, e.g. "taking for temporary use when actually returned by the taker, that is when prior authority has been granted or can be assumed, such as a family situation, or unauthorized use by chauffeurs and others having lawful access to the vehicle, are not counted for auto theft" is provided as a guide for auto theft (pp. 39- 40). Situational examples are also provided, such as that .Presented in the appendix,for auto theft. 16 As indicated, the UCR uses a statistical convenience for its classifications which effectively reduces the volume of crime. For example, in the case wherein a number of crimes are committed by an individual in the pursuit of one parti- cular criminal epidode, the FBI instructs reporting stations to include only the crime which is highest on the list of index crimes (see pages 14-15). In a case wherein burglary was the intent, then, a "secondary consequence" such as rape, aggravated assault, robbery, or murder may be recorded. While such crimes are certainly serious, where such combinations occur frequently, one could not begin to estimate the risk of being burglarized by looking at the UCR data. A large number of the aggravated assaults and robberies are, in fact, the result of burglary, in Which the burglar did not anticipate a personal confrontation. To a large extent, of course, the UCR treatment of crime does not influence either the local police department's hand— ling of such a case nor the legal system's disposition of the case. The individual may still be charged with each crime actually committed, and the actual final disposition of the case is the product of a complex interaction of several fac— tors, the analysis of which is beyond the intent of this study. Such factors may include plea bargaining and other local forms of negotiation contingent upon local practices and traditions. An additional problem may be found to exist inasmuch as the UCR contributors are not under any effective monetary or ‘administrative compulsion to participate, nor to follow 17 instructions correctly. The lack of any effective control or supervision in this area results in the aggregation of a wide variety of data which may have been collected in as many dif- 4 Here, the large ferent ways as there are police departments. potential for differences renders the entire data methodology suSpect. The fact is that it is practically impossible to sort out this dimension, even at a basic level. The aggre- gation employed by the FBI denies this very real problem. When assembled, these data are supposed to be "uniform", but uniform only in the final assembly process. Although the FBI cautions against using the data for comparisons, the fact that they, themselves, aggregate the data for states, regions, and the nation as a whole betrays this caution. This is true of the mass—media as well, which was recently illustrated by the release of the second quarter crime figures for 1977 (Detroit Free Press, July 8, 1977). The newspapers and the national broadcast networks readily compared the data without any mention of the lack of comparability, touting the "pro— gress" of some cities, which, as illustrated herein, may be ambiguous progress at best. Important also is the fact that the UCR total crime index, which is frequently cited as the principal measure of crime, excludes a number of economically and socially important offen- ses such as arson, forgery, fraud, embezzlement, and kid— napping. There is no attempt to incorporate this dimension, Which means the effective exclusion of many "white collar" and federal offenses. uh.» Ev. 18 The changing way in which crime is defined has made many hesitant to make comparisons over time. Gould (1971) expressed this concern in trying to assess whether or not crime has been increasing in the U.S. Because of citizen underreporting, ex— clusions mentioned above, and other features which change from year to year, it is difficult to interpret and generalize appar- ent changes, when attempting time series comparisons. For example, for every U.S. city, between 1972 and 1973, the T01 increased by between 20 and 120 percent. This, not because of a crime wave, but because the FBI "suddenly" started inv eluding larceny below $50, which it had previously excluded. Larceny under $50, it should be pointed out, is typically as large if not larger than any other single category of index crimes. The potential for gross error is further increased by various opportunities for establishing policies of as- sessing the value of stolen goods in particular ways (see below). Additionally, because of the greater availability of well-identified police organizations at the city level, and the presence of modern data processing equipment, there is a tendency for rural crime to be underrepresented. This, combined with the other features, and numerous objections related to specific police depertments (see below), has left the UCR Open to many levels of criticism. In 1967, in response to the criticism, and in response to the wish to obtain an alternative way of assessing police .Performance, the Justice Department (OLEA), in connection 'Mith a President's Task Force, began a series of victimization 19 surveys which have continued to the present. The various goals of the surveys are enumerated by Reiss in that task force's report (Presidents Commision on Law Enforcement and Administration of Justice, 1967). The surveys, with the ongoing sponsorship of the LEAA, have produced several nation- wide summary publications, as well as surveys of individual U.S. cities, the latter of which are the focus of this study. These surveys have provided an alternative picture of crime for the nation, individual states, regions, and selected large cities, which contrasts in a number of respects with the picture provided by the UCR. The most general and important conclusion has been that there is at least twice as much crime in the U.S. as is reported in the UCR (Hood and Sparks, 1970: p. 23). Additionally, Skogan (1974) compared victimi- zation (NOS) and UCR rates for robbery, showing NCS estimates to be as much as 5.6 times as great as UCR figures. While the UCR rates were higher in some cases, in the vast majority they were not. (This is also dealt with in Chapter 3, see table 3.2.) Indeed, a large part of the problem is manifested by these two measures, in that, historically, there has been a question of what is the best way to measure crime. 0n the one hand, we have a measure which reflects the general citi- zenry's picture of crime, as reported to us by them. On the other hand, we have a measure which reflects the organizational structure of one component of the legal system «he police) and the way it deals with crime. While either measure is likely to contain a certain amount of measurement error, each 20 measure is useful for specific applications, depending on what is being studied. It is a primary contention in this dissertation that, While the applicability of any measure depends upon the specific context of the problem studied, most of the histori- cal applications relevant to the points raised in the intro- duction and most of the current research on the causes of crime, can be benefitted best by a measure which contains minimal institutional filtering, and which comes directly from the population under study. One objective of this study, therefore, is to attempt to reconcile the differences between.NCS and UCR data, by showing the theoretical significance of those differences. The specific aim, described below, is to develop a model for explaining the discrepancies, showing that they are neither sporadic nor random, and thereby also to allow a degree of predictability about the level of victimization that occurs based on UCR figures. Theogy and Literature Bearing on the UCR and Other Measures of Crime In their 46 year life, the UCR have been criticized on a number of grounds. These include differential enforcement policies (wherein some departments enforce laws which others ignore) accidental and deliberate misclassification (see be- low), record manipulation, and individual and departmental discretion (e.g., choosing which calls to answer, which crimes to report, etc.). Beattie (1960) warned against using the UCR because of their compilation in different areas in ways which make them 21 unreliable, inconsistent, and incompatible for various re- search related comparisons. For example, different states have different systems of criminal law and procedures. Different departmental emphasis can create other differences. Additionally, large urban police departments have access to a much more sophisticated level of record-keeping hardware, such as computers, while many small cities and rural areas may have no specific statistical personnel at all. To then combine such diversely gathered data under one label may require making assumptions about their consistency and com- parability which are unwarranted. Limitations based on the UCR method itself are explored by Wolfgang (1963) and Pittman and Handy (1965) concerning the classification system, inappropriate indexing of crimes, poor reporting, and the fact that the system is voluntary rather than mandatory. Pittman and Randy say that the UCR inr dexing of Part I Offenses should be subdivided into two categories-crimes against peeple and crimes against prOperty, as being distinct in their origins and social implications. While many theorists do in fact subdidde them for their own purposes, the FBI‘ggggflggt, Additionally, they point out that specific "crackdowns", direct underrecording by police officers, and downgrading of certain offenses are facts of life which seriously damage the UCR's useability. In elucidating her assertion that the UCR are not "worth the paper they are written on," Robinson (1964) attacked the official crime rates on the grounds that large city data are 'fiuggled", limited by quota systems, biased in favor of insurance 22 related complaints (i.e. wherein the complaint is required by the insurance company to validate the loss), and otherwise manipulated for reasons of the specific departments, as re— ferred to below. Within this context, there are a number of potential sources of bias that come from the police which may be inter— nal (from within the department), external (related to polit- ical and citizen pressures), formal (specified department policy) and informal (tacit agreement about the way it should be done). One of the prOblems is that the Specific factors are seldom easy to identify, but frequently present in com- binations which make it difficult to sort out the varied effects. Black (1970), in a study based on police-citizen en— counters in which arrests did and did not take place, found several sources of bias. There was a tendency to over-repre— sent "white collar" complaints, and to underrepresent cases in which the victim and the offender are closely related. There was also a tendency for police to consider more thorough- ly the complaints of citizens who were especially deferen- tial towards the police. To a large extent, this tends to bias the kinds of data that show up in official reports, thus dis- torting the actual distribution of crime. Conklin (1972) ec- hoed the criticism of police officer discretion, not only in taking the complaint seriously, but in classifying it correct- 1y. There are, of course, several varieties of discretion to Consider. Goldstein (1964) for example, says that certain 23 types may be beneficial. These exist because of public opinion (wherein a law is obsolete but remains on the books), limitations on manpower, a need to establish priorities, or to achieve a social good (as in ushering a drunk home rather than incarcerating him). Goldstein sees much of the problem as political, in that the police must consistently publicly deny that they use discretion-i.e. politically, discretion is seen as an "evillh He suggests, in fact, that the police and the public could be better served by admitting the fact of discretion, and then making the amount of discretion a matter of publicly determined policy—~i.e. setting the level of enforcement (see also J. Goldstein, 1960, and Davis, 1969 and 1975). The problem of deliberate manipulation of data is addressed by Siedman and Couzens (1974) in noting the great amount of discretion available in assessing the value of stolen preperty. They identify several ways in which the police manipulate data to "reduce crime." They focus here on discretion available which allows the police to juggle the property crime rate by undervaluing the reported amounts of preperty stolen. By undervaluing, certain crimes are downgraded, thus reducing the volume of "grand larceny" recorded for the city. They demonstrated that such was actually being done in.Washington, D.C., as evidenced by the large reduction in larceny following the placement of Jerry Wilson as police chief there.5 In effect, Siedman and Couzens concluded that, since crime statis- tics are a police department's report card, the very oppor- tunity for manipulation renders them suspect, and "useless 24 as a tool for evaluation of social policy," since they distort the picture of crime. Arguments that such variations and biases are random and tend to cancel each other out, in general, were challenged by De Fleur (1975) who found that, in a study of Chicago drug arrest data "biases that influence official data...distort the validity of drug rates as measures of drug use rates to unknown degrees," i.e. they have distinct patterns of bias and are £23 random. Thus, much of the criticism focuses on the sources of bias available to the police as well as the amount of dis— cretion they have. In effect, it is the interaction of discretin and bias which creates much of the problem, in that the discretion allows the bias to surface, and to show up in the data which is reported in the UCR. On the other hand, several studies have indicated a fair amount of correlation between UCR crime rates and substitute measures, which tend to show that such measures are indicative of the same underldng phenomena-actual crime. Price (1966) used insurance rates relevant to certain types of crime as an alternative measure, and found that the UCR rates correlated moderately with the alternative measures. However, the use of insurance rates, which is itself a biased measure, is ques- tionable. In light of Robinson's findings, such a correlatin is explained in terms other than crima.per se. Given the low rate at which the average person purchases such insurance, the use of this measure as a crime indicator at all, let alone Price's assumption that it is a better measure than the UCR 25 is unwarranted. Thus, his conclusion, even while critical of the UCR, is of questionable validity. Skogan (1974) provides evidence that the UCR may have specific utility. Using robbery and burglary, he paired NOS andUCR rates for correlation analysis. He found, for example, that while the discrepancies between the rates were quite high, the two rates have almost identical rank orders. This was true both when analyzing by city and by type of crime (e.g. if the UCR rape rate was highest for New York, then so was the NOS rape rate; if rape is the least frequent UCR crime, then it is the least frequent NCS crime). Hindelang (1975) compared UCR homicide rates with data from the Center for Health Statistics. While not exactly comparable because of different inclusions and exclusions (e.g. CHS includes all deaths caused by other persons, in- cluding those brought about by "due process" while the UCR rate includes only criminal homicide: due process includes self-defense, police line of duty, and executions), they show a very close correlation over time. Using 1967 N.O.R.C. victimizatin data, he found correlations with the UCR rates for other crimes as well, while comparing across national regions and metrOpolitan area subclassifications. His findings held even when applying verious weighting schemes to account for the relative "seriousness" of crime (see Rossi and Berk, op. cit.). As did Skogan, Hindelang concluded that official crime rates do have selective utility, provided that we are mindful of their limitations. 26 In a purely rhetorical vein, Lejins (1966) argues that the UCR are deficient only in that criminologists make un- reasonable demands of theme-that they attempt to employ them for purposes for which they were not specifically designed. He claims that the UCR remains useful as a tool for evaluating, improving, and comparing in police work and policy formula- tion. Kituse and Ciccourel (1963) make the radical claim that crime is uniquely defined by each organizational segment which deals with it, and that it is not unidimensional. Thus, accuracy is not a relevant point since crime is only that which a given organization says it is. Of course, the problem of this dissertation would be omitted by such a claim, though, to a certain extent, one would have to conclude that or— ganizational segments are utilizing an incomplete and biased data set. Additionally, one must question the extent to which statistics such as the UCR, given the criticisms ex- pressed throughout the literature, can be used in either case. They do not really represent all of that with which the police deal, nor are they accurate enough for evaluation pur— poses. To use them in such a way leaves out the bulk of crhme, distorting the needs and effectiveness of the various branches within the criminal justice system. Studies on juvenile delinquency can also shed some light on this debate, in that many of them have used "self-report" data. Chambliss and Nagasawa (1966) attempted to test the validity of official statistics by comparing them with self- report data, which asks the potential deviant actors them— selves the extent of their delinquent behavior, if any.6 27 They found that official data are biased in that they lead to conclusions about the distribution of actual black—white- Japanese delinquency which are contradicted by self-reports of juvenile delinquency. Reiss and Black (1970) also in- dicated the possibility of situational bias in that arrests are related to the amount of deference which a suspect shows to the police. This may result in an artifact of racial bias in that the deference shown towards the police was re- lated to race (less shown by blacks), even though racism was not identified as a factor. Victimization Surveys Although welcomed as long overdue, victimization surveys have not been without their critics. The problem of tele- scopipg (bringing in events from beyond the reference period ) and forggtting, have been discussed in the victimization reports themselves and by Skogan (1974b) and Hood and Sparks (1970). Hood and Sparks and Skogan detail the various ways in which a crime may be missed or misrepresentedly surveys. For example, Hood and Sparks report (pp. 28-30) that many peOple (over 25%) refused to be interviewed. Additionally, "conspiratorial" crimes are not likely to be reported, and people under 18 were not interviewed (although persons over 18 were asked about the experience of the entire household, there are likely to be many events that they are not aware of). Older persons and women who tend to be at home more were also more likely to be sampled-—and are overrepresented. 28 Levine (1974) outlines a number of ways in which the vic- timization surveys may result in overreporting. He warns against a large potential for lying, misperceptions, experi- menter expectancy, bad memories, improper coding, and over- zealous interviewing. He does not, however, offer documentation to demonstrate that such actually occurs. In NEWS REPORT (April, 1977) it was reported that a panel of the National Research Council Assembly of mathemati— cal and Physical Sciences Committee on National Statistics concluded that the victimizatin surveys do not, in fact, provide a "picture of actual crime nor of any regglar dis- crepancies between" victimization and UCR rates. In effect, that report (Surveying Crime , Penick, ed.) concludes that neither NCS nor UCR data are indicative of the real rate of crime, however they might define that. While concluding this, however, they report that certain adjustments to the NOS pro- cess could produce a series which could be useful for assessing the risk of crime. Additionally, they leave open the door for further question about the UCR and NCS relationship in that their conclusion that the victimization rates can not be "easily" determined from the UCR is not really demonstrated. By not exploring the possibility, they leave unanswered the central problem of this dissertation. A Question of Measurement All of the above illustrates an ongoing debate about the best way to measure crime. As indicated by Biderman and Reiss (1967), any given technique will contain some Specific 29 selective preperties which make it more useful for some purposes and less so for others. The existence of two mea- sures, both of which contain error and fact, may even make it possible to estimate the amount of error in each, and thus to provide more meaningful measurement. Indicative of this debate is the question of proactive versus reactive behavior of the police. Black (1968) says that there are two ways by which police become involved-— they react to a request for assistance (reactive) or they ins tervene on their own authority (proactive). The problem.is that only a relatively small amount of crime can be proactively discovered, due to realistic constraints such as the number of policemen, the budget, administrative control and guid- ance, and public discensus (see Goldstein, 1964). Conse- quently, much more crime is not discovered both because the police can not ferret out each offense and because much crime is not reported-—the latter being quite significant because most crime is dealt with reactively. Pittman and Handy (1964) provide evidence of this in a study of a random sample of aggravated assault in St. Louis. They randomly (method not specified) chose 241 of the 965 recorded aggravated assault cases for 1961. Examining the police files, they found that less than.7% of the 241 cases were actually "proactively" discovered, and that the rest were reported to the police by victims, bystanders, and the offenders. Pittman and Randy point other problems of data quality in that the error from citizen reporting is compounded with that from police recording. The simuhaneous inclusion 30 of both types of error would then serve only to confound the problem of measurement, making it even more difficult to sort out. The reasons given by households and commercial estab- lishments for nonvreporting are detailed throughout the literature (Skogan, 1974; Hood and Sparks, 1970; Criminal Victimization in the U.S. 1973, 1975). They include such reasons as fear of reprisal, victim precipitated crimes, sympathy with the offender, personal capability, and a lack of faith in the police. Further, the distribution of these reasons has been shown to vary with the type of crime-i.e. the public may have more faith in the police's ablity to handle a murder than to deal with a case of domestic aggra- vated assault. This has been shown to be the case by Lynn Curtis (1974). Thus, there is a host of problems associated with mea- suring crime and criminal victimization in the U.S. These problems are helped very little by the varying interpreta- tions offered in the literature concerning their useability, and advocating the use of one versus the other. For example, in his work on urban crime and public policy, Y.H. Cho (1974) comments about measuring the extent of crime in the U.S.: Although the accuracy has been questioned, the crime statistics compiled and published by the Federal Bureau of Investigation in its Uniform Crime Reports are un- doubtedly the best source available to measure the mag- nitude of crime in the United States (p. 10) This he states, even while citing problems concerning the UCR and the discrepancy of magnitude revealed by the NOS sur- veys. While it is certainly clear that the extent of coverage 31 by the FBI reports is unmatched, there is a very real ques— tion about whether they can be used as a measure of magnitude. Doubt mpgp exist because of the large gap between the level of crime that is officially reported and the level to which the population claims to have been victimized. Rather, as shown by Skogan, what the UCR can be used for is to measure the relative magnitude of crime. He shows that using either measure will result in similar conclusions about the effects of certain variables relating to the social and economic structure of particular urgan areas and their police departments ( in Social Science Quarterly, op. cit.). However, Skogan points out that the magnitude of the coefficients which are obtained ( in regression analysis, for example) will be very different, depending on what measure is used. Thus, the absolute effects are not the same. However, it should be stressed that Skogan found that the NOS and UCR measures covary closely, and appear to be measuring parts of the same phenomena. Thus, while one may not be able to judge the exact probability of getting assaulted in New York, one can say Whether or not assault is more likely there than in Boston, and whether or not assault is more likely than robbery. Reconciling the UCR and NCS Crime Statistics There are fundamental contrasts between what the UCR and $08 rates are measuring, even given a certain level of similar- ities. At the broadest conceptual level, the UCR would be intended to measure crime. The most pOpular impressions of the UCR, especiallywith the public, is that its rates reflect 32 the actual level of crime in applicable jurisdictions, and that such rates provide an effective way of gauging the risk of crime in a given city. This is not surprising, considering the way in which such statistics are publicized and used. What the UCR is actually measuring, however, is somewhat different. To see this clearly, involves tracing each "crime" back to its inception. There is a continuous stream of events occurring in society, many of which may eventually be classified as crime. What happens is that the UCR taps this stream of events, but only after much filtering has reduced its magnitude. When a person is victimized, a perceptual filter Operates so as to determine whether or not that person will invoke the aid of the police. .I£_the police are notified, then additional fil- ters, institutional and perceptual, intercede between the issuing of the report and the recording of the crime. These filters have been illustrated in the preceding section. Fur- ther, such processes act in different ways on different crimes-—because of varying opportunities and motives. Thus, an official crime rate must be viewed, not as that full amount of crime that occurs, and not even the amount of crime which is reported or known to the police, but rather, the volume of crime which succeeds in making it through the various levels of filtering. Victimization rates, on the other hand, attempt to deter— mine the level of crime before filtering occurs. As has been illustrated, however, filtering‘gg involved, even though less than that which is present when measurement occurs at 33 the UCR stage. Thus, victimization surveys and the UCR are both attempting to measure the same victimizationqhenomena, from the same stream of perceived offenses. The difference lies in the fact that the measurements are being taken at different points-—NCS near the beginning and UCR nearer the end. Hence, it follows that the UCR and NCS will be different, since much filtering and processing has occurred between the two measurements (e.g. see Sellin and Wolfgang, 1964). Conceptually, there are in fact other measurements which could be taken which can illustrate this process. For ex- ample, the individual himself may choose to ignore an offense, and not recall it for the victimization survey. So, if one were able to monitor an individuars life and interactions, they might determine that many "offenses" go unnoticed or ignored, and are filtered out even before the victimization survey level. At some later stage, still before the survey, the individual may discuss the event, and decide not to in- clude it in the report to the interviewer, for whatever rea- sons. Thus, we can not say_that victimization surveys are without filtering, but just that the amount of filtering is cumulatively less than that at the UCR stage. On the other extreme, filtering may occur after the UCR stage. Only a certain proportion of events generates any arrests. Of those arrests, only a lower proportion are ever charged and arraigned. Following that, fewer still are tried and convicticsi with a small fraction ending up in prison. Estimates of the preportions which "survive" to each subsequent level, however, are not easy to obtain. 34 The President's Commission on Crime and Law Enforcement attempted to demonstrate the filtering effect in the gpallenge of Crime in a Free Society (1967), but concluded that their attempt is "only suggestive since no nationwide data exists" (pp. 8-9). Thus, if we trace an offense from its inception, we find that the level of crime may appear different at different points, depending on where it is measured, and the amount that is subject to formal punishment is only a small fraction of the original stream. The question then, is one concerning not only what kind of filtering occurs, but additionally, at what stage we wish to measure the flow of crime. Much of this may be dealt with by asking what the original intent of measurement was. Clear- ly, any measurement will contain filtering, both by virtue of the point at which it is observed, and biases contained in the measuring device itself. Then one must ask how much filtering is apprOpriate or allowable for whatever is being studied. If one is studying police effectiveness, then one would measure at two points, depending on how effectiveness is defined. For example, we might compare the number of com- plaints with the volume of resolved cases, to determine a clearance rate. Alternatively, one might measure the effec— tiveness by comparing the number whih goes to trial and those convicted. Additionally, one must be aware that filtering can even occur in response to the fact that monitoring is taking place (e.g. Siedman and Couzens). If one is studying recividism, the appropriate points of measure might be the 35 proportion of persons who are in prison who have been there before. Thus, the means of measurement determines the level of what can be studied by its use. If one is interested in reconciling the UCR and the NOS victimization rates, the apprOpriate way of going about it is to view these two measurements as being taken on the same stream of events, but at different stages. To reconcile the two, then one must measure the difference or discrepancy and determine what events or conditions have transpired between the two measurements which have affected the filtering. The problem then, is that there are many possible ways and types of filtering which occur. Basically these can occur during either of two phases: 1) between the time that the victim perceives the eventas "important enough" to in- clude in a victimization report and the time that he/she reports it to the police and 2) between the time it is reported to the police and the time they record it for the UCR. This conceptualization may be aided by Figure 2.1, as constructed by Willmer (Crime and Information Theopy, p. 2). From the flow chart, it is clear that there are many points at which the "event" may be diverted from the stream. The event may be witnessed or not witnessed. Innumerable events are not witnessed at all. Even if it is observed, it may be quickly dismissed as trivial or not worthy of further attention. Additionally, at any point new informa~ tion may alter an individual's decision to report or not to report-i.e. time consumption, involvement of friends, etc. 36 FIGURE 2.1 FLOW CHART SHOWING PROCESSING OF POTENTIAL CRIMINAL EVENT event occurs < yes V I police < I await new procedures information A V / yes availabI p \ no no recorded in official not recorded in criminal statistics official criminal statistics From M. A. P. Willmer, Crime and Information Theory, p. 2. 37 Once the event is reported, additional factors may intercede to prevent it from being recorded. The police may not consi- der it a crime, or as pointed out earlier, a "more serious" crime may have been committed which will keep the actual complaint out of the UCR report. It is, however, a practical impossibility to monitor each event and determine what decisions affect its completed or aborted journey. Even if we were to trace a single event, we would miss others that occur during our study of the first. Further, we have no guarantees that the events are perceived by the potential victims, let alone that they are perceived by the victim in the same way as they are by us. Even between our two points of focus, many things happen which are too complicated to thoroughly sort out. How then do we measure what has taken place between the two points? Production of the Discrppancy—-Citizens and the Policq We are aware, from Skogan, Levine, and others, of the various reasons for underreporting and the recording biases that are thought to occur. Thus, the factors that attend the discrepancy must be located in two social subsets-the citizens and the police. Of course, this narrows the focus of events substantially, in that they are only those which are dealt with by both sources of information (i.e. NCS and UCR data). As has been argued in Chapter 1, the difference between the two measures can itself be a measure of some set of moral relations in a given society, or parts of that society. Spe- cifically, the conditions of the society may in some way be 38 gauged by how much of the stream of events is diverted be- fore being recorded as "crime". To a large extent, then, this represents the degree to which an event is "important" enough to be recorded in both places. Effectively, this would be a measure of the rate at which the given society or group translates the events from one form of seriousness to another. If a translative failure occurs, then it is a feature of the area under study, and its measurement, while a rough aggregate, is an indicator of the social and moral relations in that society. To the extent that such aggregate rates dif- fer from place to place, it isihrther possible to make in— ferences about the relative conditions of the two places, based on the inferred meaning of the measure. For many purposes, those events which are intercepted are not simply ignored, but are thought to be disposed of without involving the police. They may be dealt with in many ways. Within this conceptualization, however, as ex- plained earler, the social system has prescribed ways of dealing with events such as those covered by the UCR and NCS data. As evidenced by the existence of laws requiring the following of such prescriptions, the failure to do so may be viewed as deviance from some formally defined norm. While it is possible and perhaps valid, to argue that reporting is not really a norm (Gibbs, 1966), but rather is actually a violation of a popular norm which negatively sanctions those who "squeal" or inform, even that may be taken as a form of relative measure. That is, to the extent that underreporting 39 is due to the operation Of such popular norms, the comparative to which such norms operate may be partially addressed by this measure. Hence, while such contrary norms may exist, they Operate to different degrees in different places. In that they have replaced the formal procedures, and may not contain a definition of appropriate behavior, a gap is, in fact, produced. This is clearly a form of anomie, as used here, in which a complete definition of appropriate conduct is replaced by or challenged by an incomplete one. Thus, an individual being acted upon by contrasting rules may be effec— tively alienated from the social system, and if this con~ dition is socially prevalent, we may properly call it anomie, the breakdown of normative structures. It is in this context, then, that one may view nonrreporting as a manifestation Of the state of social relations. Where cities vary on this measured dimension, the we assume that they vary as well on some underlying dimension that accounts for or explains the variation. Following this line, we read from Durkheim that other features of society may similarly be affected by or be a reflection of the same underlying dimenv sions. That is, as this phenomenon, however we define it, occurs, there occurs also a corresponding set Of phenomena that is associated with it. Where the dimension is strong, that corresponding set may be more predominant, and less so when the dimension is weak. Thus, by taking the two measures of the stream of Offenses, we may form a new social indicator. This indicator tells us how much filtering occurs between the 40 two measurements. The UCR measure divided by the NOS estimate becomes U/V, and is then an indicator Of what proportion of the Offenses survives to the second measurement point. There is a possible Objection to this conceptualization, as one might tell from an examination of auto theft. In some instances, U can be higher than V, and there is no underrepor- ting. Thus, the measure, U/V, will also reflect those events whose seriousness or relevance was negated following an Official report. The two effects are practically impossible to sort out—-that is, overreaction versus underreaction. We may note, however, there is a net underreaction (i.e. U/V less than 1.00) for most crimes (this will be demon- strated in table 3.2). That the net effect of a community's reaction is less than 100% reporting may be taken as a mea- sure Of that community. For example, one can not always tell an election's results by one candidates share of the votes. However, by noting the net surplus or deficit, the result is Obvious. Thus, the net effect, even if it allows for both types Of influence, can be a telling factor. It can tell us the extent to which the society, on balance, overreports or underreports. The next logical step in this inquiry, then, is to find out whether or not other indicators of the state Of society covary with the measure defined here. variation which is common may indeed indicate the Operation of an underlying dimension relating to anomie, as conceptualized here. For this, we now set forth the factors and relationships which are relevant to the phenomena. 41 HypothesesJ variables and Propositions From Skogan and Hindelang, we know that victimization rates and the UCR rates are measuring a similar phenomenon, by virtue of their strong intercorrelations. Given that there are large differences between the two, the question is whe- ther or not those differences represent a useful or informa- tive indicator about the social relations. If they are ran- dom with respect to other parts of the social system,. then one would conclude that they are not useful in this way, and that whatever conclusions one draws from one measure must be the same as those drawn from the other. If, however, these differences occur systematically, then the variations which result are important. Given that different cities are de— fined by different demographic, social and economic varia— bles, variations in crime rates which result from or are correlated with such variables are worthy of examination. From the nature of the problem, it follows that the largest part of the gap between the two measures may be seen as underreporting. This is a central hypothesis in this study. If everyone were to report everything, the UCR and NCS would be much closer. However, because Of the extra gap created by underrecording, some discrepancy would remain. Thus, by adding in both sides of the discrepancy, the UCR rates would be "restored to that of the victimization level. Thus: V = U + b1(citizen variables) + b2(police variables), where the vector B is the functional delineation Of the coef- ficients. 42 It is hypothesized that certain relationships exist, and that a measure of underreporting and underrecording, U/V, will be "explained" in large part by cities' characteristics which relate to the underlying dimension. Based on the findings discussed earlhr, and based on.the following discussion, the following propositions have been derived: Reporting 1. Reporting is negatively correlated with population size. 2. Reporting is negatively correlated with population density. 3. Reporting is negatively correlated with the proportion of the population under 20. 4. Reporting of prOperty crime is positively correlated with property crime. 5. Reporting is positively related with the suburban/metro— politan population for violent crimes, and negatively for property crimes. 6. Reporting is negatively correlated with the proportion of the populatin which is black. Recording 1. Recording is positively correlated with the amount of money a police department Spends in connection with each Officer. 2. Recording is positively correlated with the ratio of police to alternative income, i.e. relative pay. 3. Recording is positively correlated with the percentage of police personnel which is civilian. 4. Recording is positively correlated with the number Of police per capita. From the underreporting and reporting side, most of these relationships have an empirical basis, as spelled out below. Additionally, factors such as pOpulation size, the presence of a large number of young people, and the urban concentration of minority groups are also correlates of in— dustrialization and urbanization, factors which Durkheim associated with the onset of normative breakdown. Under- reporting may be seen as related to such factors. These, in 44 are the ostensible meanings which peOple assign to their actions. In a Durkheimian sense, the underlying meaning may be wholly apart from the common sense rationalizations that people use to justify their actions. Rather, the ap- proach would be to look in terms of other factors, and to leave such reasons for a more practical level of analysis. That is, the process is basically an inferential one. Vhile we may have a collection of reasons for not reporting crimes tO the police, as revealed by the victims, those reasons are more appropriately left to some other facet of analysis, since they are not useful in a Durkheimian context. In the second place, one of the desired consequences of this study was to find a method of estimating the level of victimization in cities by reference to factors that tend to explain it. Even if such reasons provide us with reliable regression estimates for the cities we have which are covered by the surveys, this would be of no use in cities for which such data have not been collected. Thus, it is to our theo- retical and practical advantage to utilize only those data which are generally available, and not to use those generat- ed by the survey, except Of course for the construction of the dependent variable. much Of the literature has indicated factors which tend to be associated with reporting discrepancies. The survey results themselves have provided summaries of reporting behavior by race, age, sex, and income, along with other factors. Empirically, reporting was lower for persons under 20, and varied for particular cases of income and race. 45 Whether or not the victim was female, as Opposed to male, did not appear to have significant bearing on the question. Income has several possible relationships which have been indicated by the survey data. From a theoretical stand- point, one expects that higher income will be associated with higher levels of personal efficacy and commitment to the social system and its rules. On the practical level, it has been shown that very high income groups tend to report prOperty crime with greater frequency than the general pOpulation. This may be a reflectix1of the theoretical relationship which is additionally indicated by the higher amount of reliance on insurance among high income groups, and the consequent "need" to involve the police to collect on their insurance piicies. There is some indication, which we do not attempt to explain, that the very high income groups to report violent crime with less frequency. Racial composition, similarly may be a confounding fac- tor. Some empirical investigations have shown that more blacks are associated with greater reporting (which we at- tempt to explain later), while the theoretical expectation is that more blacks means lower reporting. By virtue of their exclusion from many facets of the social and economic mainstream, blacks are expected to have less normative commitment to the system, and to be somewhat disaffected. Consequently, one expects that a large prOportion of blacks would reduce the U/V ratio by adding in a component with a lower reporting probability. 46 Population density and magnitude are expected to be re- lated in that they indicate the potential for urban.complexity and diversity. While not necessarily the case in an under- deveIOped country, in an industrialized nation, higher pOpula- tion and density is expected to be associated with more come plex levels of development. Within that greater complexity, coordination and the balance of interdependence is harder to manage, yielding a breakdown in various means of social control. As this occurs, various forms of deviance, as herein defined, come to be more prevalent, including none reporting. From a practical standpoint, larger and denser populations are thought to make it more difficult for an individual to receive aid from the police, and to make the police's job harder. As suggested by Goldstein (1964), the reduction in reporting may well be a reflection of the "level of enforcement" which, because of largeness, com- plexity, and changing norms, is appropriately lower in larger cities. The ratio of suburban to metropolitan population is also included. In this case, the adjacent population is likely to move in and out (commute) of the urban centervdth great fre- quency. The larger this adjacent population, the greater the "exposed" pOpulation within the city, which may not be re- flected in the urban population itself. Thus, this proportion is intended to correct for that gap. Additionally, a large adjacent population may be a further indicator of the moral relation in the area under study. The fact that so large a pOpulation has formed outside the city, but not in it, 47 indicates different styles of living, and perhaps differing amounts of normative commitment. Thus, its inclusion will add a distinct component into the total, in that a pOpulation with potentially different ways of dealing with crime is con- tributing to the enumeration of events which are bang measured at the UCR stage. This may additionally be important in that this group may contribute at the UCR stage, but not at the victimization level, since the NOS surveys give only urban estimates. ~Thus, a large suburban balance may have practical as well as theoretical impact on the set of’relationships. Within the police sector, Skogan, Robinson, and Hood and Sparks provide a rationale for eXpecting the influence of several variables which might affect the ability and motivation of the police in .andling cases at various levels. Each of these is expected to have a cumulative impact on recording. That is, the combined features which might contribute to effecting the greater normative commitment of officers are expected to effectively increase the pro- portion of complains which are translated into official reports. Civilianization has been posited as a possible factor. The including of clerks whose duties are specifically related to the processing of crime rate information may enhance the chances for accurde data production. Further, the use of such clerks may break the utilitarian linkage between two conflicting facets of police work--i.e. that of preventing crime on the one hand, and that of telling the population how much crime is occuring on the other. Thus, the linkage 48 which might be automatically made by an officer who is re- quired to do his own paperwork is not so automatic here. Of course, even civilians may be requested to "fudge" the statistics, however, the probability that civilians will be motivated to do so on their own is reduced. Additionally, removing the police officers from such paperwork duties lessens their resentment and may enable them to spend more time in ways which they feel are more productive-— crime prevention. This may provide them with less motivation to reject the normative system and as well reduce the problem of specific police-victim-offender interactions affecting the recording. This may be related to another variable as well, that of police per capita, in that both translate into more potential time to spend doing actual enforcement activity. This has the effect of increasing the effectiveness of each officer, on the one hand, by reducing his workload, and of in- creasing the visibility of the police on the other. Police department expenditures may also be related. By taking the amount of money a city spends per police officer, salry included, one may obtain a relative measure of each urban area's commitment to crime prevention and control. This type of factor has been used by others(Wellford, 1974), but not in this way. What was expected to be important is the degree to which a police department utilizes its man— power. One may hypothesize that more money spent per officer indicates greater utilization of that officer and more re- sources for that officer to use. Thus, a department which 49 only spends enough for stationhouse upkeep and salaries may be making less effort than a department with an additional thousand dollars per officer-—providing better equipment, benefits, training, etc. (Since data for the specific categories are not available, this would appear to be a rea- sonable alternative.) One other economic factor is included. Similar to the factor just discussed, the relative pay of police officers (relative to what they might make in another form of employment) is expected to play a role in that it indicates the relatively higher or lower value the community assigns to police functions. Additionally, this factor has a relation- ship, potentially, to the individual police officer's self esteem, being an indication that he/she is of higher status than some around him/her. This should contribute to his commitment to his job, in that for his occupation, the system is providing a relative advantage over others. Such higher status would be expected to increase a U/V ratio by producing better morale and job performance. The main purpose of this study, then is to test whether or not the theoretical propositions spelled out above can explain the discrepancy between the victimization rates and official rates. USing the factors just outlined, the next section explains the methods which were used to Operationalize the dependent and independent variables, the mode of analysis, and the results of that analysis. The next step, then is to provide a close examination of data and sources which are available to effect this study. CHAPTER 3 DATA, METHOD, AND ANALYSIS Data and variables This study utilized ten independent variables and nine dependent variables. The dependent variables were formed by taking the ratio of UCR to NCS crime for each category of offense included in the study: total crime, violent crime, property crime, rape, robbery, larceny, aggravated assault, burglary, and auto theft. The UCR data was obtained from 1971-3 volumes of the Uniform Crime Reports, while the NCS data came from three victimization reports (Criminal Victimi- zation in the Nation's 5 Largest Cities, Criminal Victimiza- tion in 13 American Cities, and Criminal Victimization in 8 American Cities). Those three reports covered periods which were equivalent to January-December of 1972, January-December of 1973, and July of 1971 to June of 1972 respectively. For each group of 5, 13, and 8 cities surveyed by the NCS, corre- sponding crime figures were obtained from the UCR by matching each NCS estimate with UCR tabultations for the same year or period. This method provides a dependent variable with a sample size of 26, which is internally consistent. This pro- cedure has also been applied by Penick (1976), in a recent appraisal of the victimization surveys. Some further adaptation of the data was necessary before the dependent variables could be formed. The UCR reports its crime figures in seven categories (murder, rape, robbery, aggravated assault, larceny, burglary, and auto theft) while 50 51 the NCS reported the victimization survey estimates in three sectors (Household, Personal, and Commercial). In the same way as done by Penick, the data from the survey was tallied in such a way that the groups included were equivalent to those covered by the UCR. That adjustment procedure is summarized in Figure 3.1. FIGURE 3.1 ADJUSTMENT PROCEDURE FOR MAKING UCR AND NCS COMPARABLE WITHIN EACH CATEGORY OF CRIME Generic Label UCR Label NCS Lable Rape Rape Rape Robbery Robbery Personal and Commercial Robbery Aggravated Assault Aggravated Assault Aggravated Assault Larceny Larceny/Theft Household and Personal Theft Burglary Burglary/Breaking Household and and Entering Commercial Burglary Auto Theft Auto Theft Completed Motor Vehicle Theft Thus, this study concerns six crimes, and three crime categories, total, violent (rape, robbery, aggravated assault), and prOperty (larceny, burglary). Homicide is excluded be- cause the NCS does not include survey estimates for homicide. The actual number of homicides revealed in the surveys was too low to allow a statistically reliable estimate. Hence, there are nine distinct categories of crime to be analyzed 52 with respect to ten independent variables, which are dis- cussed below. The "adjustment" procedure just outlined provides a uniform sample with respect to time and type of crime. Similarly, all of the independent variables were derived from data sets in the apprOpriate time period, for each observa- tion. That is, due to the fragmented sample, data for the 26 cities were obtained for the apprOpriate year or were estimated using linear interpolation based on the most recent two or three observations available. So, for the first 5 cities (Chicago, Detroit, Los Angeles, New York, and Philadelphia), all independent and dependent variables are for 1972. For the next 13 (Boston, Buffalo, Cincinnati, Houston, Miami, Milwaukee, Minneapolis, New Orleans, Oakland, Pittsburgh, San Diego, San Francisco, and Washington, D.C.), all data are for 1973. Similarly, for the remaining 8 cities (Atlanta, Baltimore, Cleveland, Dallas, Denver, Newark, Portland, and St. Louis) all data were for the average of 1971 and 1972, except when otherwise Specified. In this last case, monthly data were not generally available, and, as did Penick (above) this study averaged the two years which the study overlapped. The exact way in which each variable was projected, where applicable, is described in Figures 3.2 and 3.3. Factors used in this study included some which were not part of the regression analysis presented later in this chap- ter. Basically, these were included to help explicate certain relationships, as described below, and as alternatives, which 53 FIGURE 3.2 DATA SOURCES AND INTERPOLATION METHOD USED IN THIS STUDY Label Used in this study Source Interpolation Method Percent of Pepulation U.S. Census Linear interpolation under 20 of 1960 and 1970 Percent black Statistical Linear interpolation Abstract of 1960 and 1970 Median Income Percent below poverty level, families Percent above $15,000, families U.S. Census U.S. Census U.S. Census Actual (1969) for families Actual (1969) Actual (1969) Urban Population Statistical Regression estimate Abstract from 1960, 1970,and 1973 S.M.S.A. population " " Area of City in square miles " Actual (1977 ed.) Average weekly manufacturing Employment Monthly reports, Jana- wage and Earnings Dec. for 1971 and 72, July-June for 1971/2 Total Police Employment U.C.R. Actual (1971-3 reports) 1971/2 - average of 1971 and 1972 Civilian Police U.C.R. Actual " Employment Entrance and Maximum municipal Actual (1972-4 report) police salaries Yearbook Police capital municipal " Outlays Yearbook Total Police Municipal Expenditures Yearbook " FIGURE 3.3 METHOD USED TO CREATE VARIABLES AND SYMBOLS Symbol variable Method and Components SALE Average annual wage in Weekly average wage X 52 manufacturing UPOP Urban population Method specified in Figure 3.2 ELK Percentage of blacks " in population AGE Percentage of population " under 20 Density Urban density UPOP/area PPOP Suburban proportion of [(S.M.S.A. pOpulation) - (UPOP)]/[S.M.S.A. S.H.S.A. population PRD Relative pay (Entrance + Eaximum police salaries)/2 SALE PCIV Civilianization Civilian Police Employment Total Police Employment EXP Expenditures per officer Expenditures/(Total - Civilian Employment) COPS Police Concentration (Total - Civilian Employment)/UPOP To minimize the possibility of error, the above calculations were all done by computer. 54 for various reasons were ruled out. Those factors included in the Zero-Order Correlation Matrix but not in the multivariate analysis were: the percentage of families below the 1970 poverty level (POV), percentage of families above $15,000 (RICH), median family income (MY), and police department capital outlays (PCAP) as a percentage of total police ex- penditures. All factors which were included in the analysis or which were used to form the independent variables are shown in Figure 3.2. From the factors shown in Figure 3.2, ten variables were formed to attempt to operationalize the factors outlined as Propositions in Chapter 2. These are: Urban Population Suburban/MetrOpolitan POpulation Percentage Black Percentage Under 20 Annual Manufacturing Wages Police/Manufacturing Income Ratio Police Per Capta Urban POpulation Density Police Expenditures Per Officer Civilian Police Employment Each of these was computed as shown in Figure 3.3. Analysis Data for the 26 cities included in this study were analyzed using a zero-order cormiation matrix (Table 3.1) and multiple regression. The latter form of analysis pro- vides several types of useful information. Foremost, it allows the specification of a set of factors whose contri- butions comprise an explanatin of the phenomenon under study. It provides a useful way of summarizing the relation 555 TABLE 3.1a Zero Order Correlation matrix for U/V Discrepancy By Level of Crime and Social and Police Indicators variable Crime Category 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Total Crime -.32 .26 .21 .15 .22 -.12 .50 -.08 .37 -.29 .50 -.51 .26 .29 Violent Crime -.29 .21 .19 .13 .23 -.15 .49 -.11 .36 -.26 .48 -.48 .37 .14 Property Crime -.33 .28 .22 .16 .21 -.14 .45 -.04 .34 -.28 .48 -.49 .20 .34 Rape -.18 .09 .38 .07 .43 .13 .63 .11 .29 -.11 .32 -.24 .70 -.02 Robbery -.16 .09 .23 .14 .33 -.08 .54 .04 .27 -.24 .43 -.42 .49 -.03 Aggravated Assault -.34 .28 .19 .08 .21 -.11 .45 -.18 .38 -.29 .52 -.51 .28 .18 Larceny .-.34 .30 .22 .18 .23 -.10 .48 -.04 .30 -.31 .48 -.49 .23 .31 Burglary -.27 .17 .12 .04 .10 -.21 .28 -.08 .40 -.20 .40 -.45 .04 .33 Auto Theft .05 -.09 .15 .23 .08 -.38 .05 .09 .34 -.25 .54 -.12 .23 -.02 numbered variables are: mac-11o» manufacturing wages Relative pay ratio Expenditures per officer Percent capital outlays Urban density Percent below 20 11 10 Percent above $15,000 6 7 Percent black 12 Median Income 8 urban Population 13 Police per capita 9 Suburban percentage 14 civilianization Percent below poverty level TABLE 3.1b Zero Order Correlation Matrix for Social and Police Indicators Variable Label 1 2 3 4 5 6 7 8 9 1o 11 12 13 14 1 SALE 1.00 -.52 .19 -.4o -.04 -.12 -.04 -.23 .26 -.28 .15 -.31 .35 .08 2 m - 1.00 .43 .13 .40 -.40 -.22 .51 -.08 .14 .33 -.16 .21 -.03 3 m * — -— 1.00 -.04 .51 -.43 .06 .39 .08 .04 .48 -.15 .35 -.03 4 PCAP -- -- — 1.00 .00 .21 .19 .28 -.55 .00 .07 .15-.01 .03 5 DENS -- - -— -- 1.00 -.26 .17 .64 -.04 -.53 -.06 .06 -.02 .64 6 AGE -— — -- —- - 1.00 .42 -.12 -.28 -.08 -.24 .31-.19 .06 7 an: — - — —- -- — 1.00 -.06 .17 -.05 -.21 .66 -.42 .62 8 UPOP - -- - — — - -— 1.00 -.60 -.28 .36 -.13 .29 .24 9 PPOP — -— -— -- — -— -— -- 1.00 .09 -.22 .08-.21 .12 10 PCIV -- -— —- -- —- -- -— -— —- 1.00 .16 .20-.10 -.48 11 nice” -— - - - - - —- - -— -— 1.00 -.58 .86 -.05 12 Dov” -- - - - -— —. - -— — -— - 1.00-.86 .13 13 a“ —— -— —- — — — — — -- -- — - 1.00 -.04 14 cops” — -- — — -- — -- — — — — — - 1.00 “For all variables except PCAP, RICH, POV, and HI, see text for full description. These are: PCAP Percent of Police Expenditures for Capital Outlays POV Percent of families below poverty level, 1969 HIGH Percent of families above $15,000 income, 1969 NY Hedian family income, 1969 56 between a set of independent or explanatory variables and a dependent variable. Further, multiple regression allows us to control for each of the other factors while examining par- ticular variables. Additionally, this mode of analysis provides a method of determining the extent to which the relationships defined by the regression equations-are statistically signi- ficant-whether or not confidence in the various "models" or particular variables is warranted, with reference to some specified level of conidence (here, p = .05, a 5% chance that the relationship estimated could be due to pure chance). The dependent variable was operationalized by forming the ratio of crime known to the police (U) to victimization survey crime (V). U/V represents the proportion of the total which is both reported by citizens and recorded by the police. Hence a ratio of 1 would mean that one of the following is true: 1. All crime is reported and recorded accurately. 2. Crime is underreported, but overrecorded such that the latter compensates for the former. 3. Crime is overreported, but underrecorded such that the latter compensates for the former. Any ratio less than or greater than 1 may involve either of the latter two arrangements in which compensation is not perfect, or cases in which both under or both over reporting and recording are occurring simultaneously. As illustrated previously, however, there are compelling reasons to expect that most crimes dealt with in this study and reported in the crime index are both underreported and“ underrecorded. The single exception (average) is auto theft. Auto theft repre- sents a particular case in which more is reported than 57 actually occurs. Because of the immediacy of the loss, and the fact that the loss is generally substantial, along with the fact that auto theft contains few of the specific per- sonal motivations for nonreporting that are discussed else- where in this study, people tend to report it quickly after the car disappears. If the car reappears, without police help, the proportion who cancel the theft report is some— what lower than 100%, thus producing a certain amount of inaccuracy in the official reports. Once the car is re- covered, many pe0ple feel embarassed at having involved the police and fail to report the recovery of the car. Others may entirely forget that they even called the police. So, when a victimization survey interviewer asks about auto theft, the respondent may say "No," while a police report says otherwise. Thus, a comparison of the two sources shows, on the average, that the police record 26% more auto theft than the victimization surveys, indicating that a sizeable propor— tion of the total may be in error. For this reason, auto theft is not included as a property crime--there is no underreporting which needs to be explained- nor in the total crime figures. It is, however, analyzed separately to determine whether or not any of the variables used in the model explain any of the U/V variation. The U/V ratios produced are presented in Table 3.2, along with the average relative frequency of each type of crime as reported in the UCR and NCS, for the 26 cities being investi- gated. £523 ovum an. oopsoemdm 909.828ro p.93 mm: “Comp 93.5 How ommpsoohoa D map odds; .mpmHmHSQ + knocked + pasmmmm + hamnpoa + mama ma poemsflaosmc n.0mmpsooama b mama one .msne .pxov any we cofiprSO msommoa Mom .popsasoolma mm: Hopop on» news; new .pmogp opsm paooxo mowaomovmo Ham mom phone cams paospas popsaaoo mm: Hence one: 58 0.009 0.00— mmm. mafiho Havoe *F.m— *>.¢ mmN.F amass 09:4 m.em o.mm mmm. sausages w.¢e >.me mom. ammoamq m.mb >.Pm ham. cease hpaoaoam F.b N.b NOm. Pasmmmd poPm>mawm4 w.mw ¢.oe mom. humanom m. a. Poe. some ae.om em.mP pom. mango pomaoa> mo oWMwMWomwmt> mo MWMNmewwm D ofipmm >\5 anoMmpmo cameo mmmzmmmo ZOHadNHEHBOH> 924 HdHUHmho mom mMHozmbommm 92¢ mOHsdm >\D N.m mummy 59 For both sets of data (UCR and NCS) the different types of crime have similar relative frequencies and rank orders, except for auto theft. It should be pointed out also that those crimes with the least frequency--rape, robbery, auto theft, and aggravated assault-are also the most completely reported and recorded, though not in perfect inverse corre— spondence. This would lend initial support to the idea that people react more thoroughly to crimes of rarer occurrence, and are perhaps desensitized or less responsive to those Of greater frequency. (This relatinship is additionally expli- cated by the whole question of normative agreement and report- ing and crime frequency, discussed earlier) For each of the nine crime categories listed in Table 3.2, multiple regressions were performed to attempt to identify the structure of the relatinship. Additionally, a correlation matrix summarizing each of the pairs of rela- tionships was created to guide the analysis. Before proceed— ing to the results, however, a discussion of the independent variables is in order. Citizen Sector Variables The theorectical expectations outlined earDer were trans- lated in each case into some Operational form. The Opera- tional form for each variable is used to attempt to grasp specific conceptual formulations, as social indicators for phenomena which may not be directly measurable. The average annual earnings in manufacturing (SALE) was used as a measure Of the economic position that the average 60 person might obtain. This figure may serve both as a compara- tive base for police income (see below) and as an indicator Of the city's general state of economic health. While this variable does not incorporate all of the dimensions of wealth and poverty which the given city experiences, it does pro- vide information about the relative economic climate, and the implications about factors which are necessary to generate such a pay rate (e.g. highly paid alternatives, unionization, regional differences, etc.). As shown in the correlatin matrix in Table 3.1, the manufacturing wage was tested for its relationship to other factors as well. Particularly, other economic dimensions including the percentage of families below the poverty level (POV) and the percentage above $15,000 median family income, along with the median family income (RICH and MY) were used alternately to determine which was the most useful, and to explicate the exact effect Of SALR (while people in manufacturing do not really receive salaries, which the label might imply, SALR is an annual average which is used to convey what a manufacturing salary might look like). As illustrated by the patterns of intercorrelations in Table 3.1, SALR is not as sensitive to the extremes (POV and RICH) as median family income.‘ The r's for SALR and POV and SALR and RICH were -.31 and .15 respectively, while the r's for median income were —.86 and .86 (POV and RICH). Thus, when a combination of SALR, POV, and RICH was used, excluding median income, in the multivariate model, three different concepts could be incorporated, instead Of one. In this way, 61 it was determined that manufacturing earnings was a much more useful concept, and that the extremes play little or no ex- planatory role. Thus, POV, RICH, and MY were excluded in the final analysis. Indeed, it was determined that the inclusion of any Of these three variables introduced multicollinearity in the multivariate analysis (a problem associated with ins tercorrelation among the independent variables, see Frank, 1971). Urban population (UPOP) was selected as a measure of demographic complexity. Contrasted with the s.m.s.a. popur lation, urban pOpulation was seen as more apprOpriate, even given.a large commuting population. The most effective and relevant method would have been to adjust the urban population for the number of inrcommuters, out-commuters, vacation net gain or loss, and other factors relating to time Of day and year, to arrive at the actual "exposed" population. Such data were not available. Rather, urban population was seen as the best average pOpulation for the area under study. The size of the suburban pgpulation, however, was viewed as a potential factor. Instead Of including just that, however, the suburban contribution to the s.m.s.a. was used, i.e. the ratio of pOpulation in the smsa.but not in the city under study, to the population of the entire s.m.s.a. Thus, this represents the percentage Of the s.m.s.a. which is not in the central urban area (although such pOpulation may be in the central areas Of other cities included in the s.m.s.a.), and was labelled PPOP. The larger the ratio, the more the exposed population and the greater their contribution to whatever 62 reporting behavior is observed-—i.e. the more their particular reporting characteristics will influence the U/V ratio. Urban density (DENS) was used as an additional attempt to grasp the concept of urban complexity. One of the concomi- tants of industrialization and urbanization, population density increases the Opportunity for conflict and the opportunity for conflictive interaction with someone whom one might not wish to report. An increase in density, without a corre— sponding increase in the mechanisms which promote OOOpera- tion and coordination, is seen as increasing the likelihood that certain formscf deviance (one of which is non-reporting) will occur. While a denser population, on a practical level, may increase the probabifity that a criminal will be Observed in the commission, various concomitants of this increase also increase the chance that certain acts (e.g. auto theft and burglary) may be observed but effectively go unnoticed. That is, in a dense pOpulation, each individual is acquainted with a smaller proportion of the total pOpulation, and the proba— bility that any given individual knows "who owns which car" or which peOple are authorized to enter which houses or businesses is correspondingly lower than in a town without such population dimensions. Whether this particular relationship is one re- lating to largeness or density is a question upon which the simultaneous inclusion was intended to shed light. Two additional factors included to incorporate the characteristics of the city were racial and age composition. Based on the theoretical reasoning outlined earlier, and based on NCS findings with respect to age, the age variable was 63 created by calculating the proportion Of the urban pOpulation Which is under 20. This group was viewed as particularly susceptible to the consequences of normative breakdown. The percentage of blacks in the urban pOpulation was included as an additional measure Of urban complexity. The initial theoretical expectation was that blacks, as a dis- affected and alienated portion of the pOpulation, would rely less upon institutional or formal channels for crime control, which would result in lower reporting and recording rates where blacks are a large proportion of the city under study. Police Sector In this sector, an attempt was made to incorporate variables which would be indicative Of several dimensions of police influence, police effort, and police susceptibility to bias. These include economic and compositional character- istics, just as the citizen sector. The relative pay_of police officers (PRD) with respect to those in manufacturing was used to reflect the relative position of the average patrolman to the average person in manufacturing-—i.e. a measure of relative advantage. It was hypothesized that this measure would yield the relative sta- tus of the patrolman and his relative value to the community, compared to that of a "typical" other. One problem with this variable was the numerator, which is the simple average Of minimum and maximum pay for police officers. That is, it is not known how many Officers are at each level, nor how much Officers are receiving at intermediate levels or grades. 64 Since additional information was not available, however, this was taken as a compromise. Actually, the assumption about equality at both ends may not be too presumptuous, since an entry and exit balance would be necessary to maintain the overall level Of police Officers, especially in a large city in which a certain proportion reach retirement age each year. Thus, the most problematic assumption is that the increments Of pay are uniform and that the distribution Of officers is normal about the mean. A second variable was also economic. The ratio Of expen~ ditures to the number Of police officers was used to reflect the city's police effort. Alternatively, total expenditures and expenditures per capita were analyzed, but were subse- quently seen as less useful, and conveying less information than that Which was used (EXP). Total expenditures, of course are not meaningful except as related to other city variables-total budget, pOpulation, etc.—-and was altered accordingly. Per capita expenditures, while providing more information than expenditures per se, still was not very useful in this model. Greater expenditures per capita can only indirectly be translated into better reporting, and only if money is spend in certain ways. Greater expenditures per capita may only reflect the greater difficulty of management in a large city. EXpenditures per officer allowed a measure of how much effort per unit of enforcement is being exerted. Here, greater expenditures per officer can more easily be translated into modernization Of recordkeeping (i.e. pur- chasing Of computers, clerical personnel) and perhaps greater 65 professionalization, having a direct impact on recording while the effect of per capita data would be more indirect, involv- ing more perceptual filters. The inclusion of non—Officer personnel, civilianization (PCIV), provided a variable which is the ratio of civilians employed by the police department to total police department employment. This is a measure of the extent to which the police department relieves the "paperwork" burden Of police officers-—which should result in better overall correspondence between recorded and reported crime. To clarify this point, if one's only job is recordkeeping, we may expect that job to be done more efficiently (benefits of specialization) than if ones primary function (law enforcement) is "interrupt- ed" by the relatively menial task of recordkeeping. Thus, less resentment of the task is anticipated on the part Of civilians who are hired specifically for that job. A final factor included was the number of police per capita (COPS). This variable is a direct way of measuring the potential visibility of police in the community. Of course, we have no present method Of determining whether the police are concentrated in diverse pockets or if they are dispersed evenly with respect to the exposed pOpulation. However, it is assumed that the greater the number of police per capita, the greater their visibility, and the more the Opportunity for reporting. Additionally, that greater visi- bility may produce greater confidence that the police may be effective (i.e. if the police are 3331 visible, one may be more likely to report a crime, in that, by virtue Of that 66 presence, the citizen expects that the police will be closer to the problem and available to solve the crime or alleviate the situation). General Findings For each Of the nine categories of crime, the following regressions were performed: (1) U/V = f (constant, citizen variables, police varie- bles (2) U/V = f (constant, citizen variables) (3) U/V In.all cases, due to the large number of variables, the f (constant, police variables) "whole" model (equation 1) explained more variation than either sector alone. However, because of the large number Of variables and the small number Of Observations, many of the variables which are significant in equations 2 and 3 are not significant in equation 1. Whether this is due to the sample size or due to a true null hypothesis, however, is not possible to say without a larger sample. The general results and F- statistics are shown in Table 3.3. In general, the whole model explains significantly more of the variation than either partial model. F-tests were per- formed to determine whether or not the results of equation 1 were significantly better than those for equations 2 and 3, for each category of crime. That is, starting with equation 2, for example, this tested whether or not the inclusion of the variables from equation 3 improves the results Obtained in equation 1, i.e. whether U/V'= f(citizen + police) is better than u/v = f(citizen) or u/v = f(police). 67 He>oa me. one we eoeeameemamx or. mm. we. geese case we. ow. em. hsmamasm *mm *Nm *mm. hamoamq soc *mm. sew. pasmmm< copm>oamm¢ om. ems. we. anonnom *mm :06 xme. seem em. *Nm. *mm. oafiho hphoaOhm xem. *em. mm. cease eeeaee> *mm *bm sow. cameo Hmpoa soapmscm soapmsum soapmdqm Oswao mo moaaom semapflo Odom; haemopmo mfiHmo ha Mmowmado Mm moz mowaom pom OP Qmpoavcoamzdmlm Scam Anomapfiovpoamzdmtm wswpomQPQSm an pom .QOmeaho> souwvwo pom OP Aampoev pmamodmlm Eosm AOOHHomVeoamsvmlm msHpOmAppsm an m.m manna Scam co>aamp mm; manna . mama .em.m m>onm some Ho>oa no. em psmowhwsmHm ma nopoom . moaaoaw.now m mam can mm .0 969m Gog; H964“ mo. an meOfimwsmHm m...” Hopoom nouflvfio ho.“ v mfm .haocrwpooamoh shepoom mowaoa can smufipfio map an po>oaaeH ma Admcoa Odomsvmm pomp mammApoahs on» poo» mowvmflpspm a 111 mm.m ea. mm.o mm. we. pawns case so. mm.w ea. em.e om. em. seeamwem Fm. mo.m ea. m..e am. we. assumes mm. mo.m mo. so.m em. em. pasemme eeeeeeemma mm. mm.» Po. me.m we. we. absence be. em.m ma. em.m mp. me. mamm om. mo.m ea. mm.» mm. mm. mambo seemeowm mm. mm.e mo. m~.m em. mm. eases eceaoa> em. m¢.m mo. ma.h om. mm. mango Hopes m newsmahm> h soapmwam> soapmaam> sowpmfiam> sossoo sowpmzum oowaom sowvmsvm :oewpwo sowpmScm macs; anomopoo mafiao mEHmo mo MmowmgHU Mm QMZHdeNm ZOHB ZOEEOO Qz¢ azmnzmmeZH ¢.m mamas 70 by Figure 3.4, in which X is an independent variable (such as COPS or BLK), D is u/v, and U and v are UCR and NCS figures respectively. In this analysis, X is related to both U and V. The extent to whichimese effects are reflected in D is deter- mined by the relative effects Of U and V on D. FIGURE 3.4 PATH DECOMPOSITION 0F u/v 1903 ’9) U w D I .p V :D V ‘l‘ «“1 Borrowing from Brown and Fuguitt (1972), who applied this type of analysis for black and white differences, their D = w — B becomes D = u/v. They show that rDX = p(W,D)’r(W,X) - p(B,D)‘r(B,X), where W and B are analogous to U and V, and p is: 6 6 p(U.D) = 3%- and p(V.D) = 3;— . the path coefficients for our model. Notice that u/v e U - v, but rather log(U/V) = log(U) - log(V). Thus the U, V, and U/V data are computed in logarithms for this analysis. Hence, it is clear that, in our analysis: 6 0 r(logD,X) = ._l9§£!l ‘ r(logU,X) --3£95L!l ‘ r(logV,X . Olog(D) log(D) 71 Using the information provided in Table 3.5, we find that the correlation of the log of U/V for total crime and BLK is .50. This .50 may be understood because an increase in BLK raises U while decreasing V, thus increasing U/V. The extent of this effect (rBLK,log(U) = .06, rBLK,log(V) = -.15) will be mitigated by the path coefficients, which are plog(U),log(U/V)= 2.16 and Plog(V),log(U/V)= 2.29. Applying the equation for r(logD,X) on page 70, we find that .50 = 2.16(.06) - 2.29(-.15), which is the correlation coefficient Obtained from the direct path. Thus, the effect Of the proportion of blacks may be seen as occurring because of its differential effects on U and V. Of course, there has not been postulated a logarithmic relation- ship between U/V and the independent variables. From a theoretical standpoint, however, logarithmic transformations are necessary in this analysis to effect the identity Of U and V with U/V while retaining the fundamental equation for path analysis. Additionally, the logarithmic relationship may be somewhat justified in that the zero order correlations between the unlogged independent variables and the logged dependent variables are very similar to the correlation coefficients Ob- tained in Table 3.1, in which nothing is logged; e.g. rBLK,U/V from Table 3.1 for total crime is also .50. Hence, while the discussion presented below focuses on U/V as a measure, it is implicit that the u/v behavior is de— termined by the effects of both U and V. In fact, for practi- cal purposes, much of the following discussion takes V as 72 sowpmfi>om wmm. mmh. mon. omm. omr. mwm. hue. New. mmm. camosmpm mop.a1 emm.PF mo>.oe mom.F1 mo>.PP oom.oF oso.a1 000.0, omm.w coca >\m eon > eon a con >\m eon >eoq D eon >\D eon >eou D eon Hopes apnoaohm . psOHOfiD oapmapmpm mm9m azmnzmmmn Dmaomqmm mom m20H9mD DmF.1 no.1 mm. mF.1 no. we. Fm.1 mm.1 >Hoa om. mF.1 Do. me. mp.1 Fo. mm. moo. mm. mam >\D com > woq D wog > D cog > cog D mon >.\.D Dog > mon D as moapmaamar Oswao Hmpoa mafiao apaoaoam mafiao vsoaoaD psopsoaoosH mZOHadfim0hmZ¢ma OHzmeHmdDOA DBHB mmndeM4> azmnzmmmn QZd azmnzmmeZH amaomflmm DO NHmBoa mo. esp pm Pamowmwswfim mas nOHpmsvo vmnp mopmoaanx so psooauwomfio ow oweuweopm a .osam> opsHomnm .mo.m some scammam ma .m.m ohDMwm ma powwmop mm was moanmfihm> oo.1 mm.1 om.~1 om.e mem. oma. mo.m oeeoflpopm a omme. .oooo.1 ewmo.1 Peoo. mooooo. Poooo. use.m phone osse oe.a me.1 mm.¢. mm.a mam. mm.1 mm.P enemaeoem 9 Dow. mooooo.1 whoo.1 waoo. Poooo. moooo.1 emo. shoamssm om. emm. mm.e1 ms.m mem.1 mm.~1 om.m casuaeoem a memo. eooooo. Pmpo.1 eaeoo. moooo.1 moooo.1 comm. sasoosoa wF. P¢.F mw.P1 em.m bm.~1 mo.m1 >N.N oaamfipmpm a Pmo. moooo. mmeo.1 eMNPQ. meooo.1 mpooo.1 moa.m seasomoe eoes>eumm< one. emo. we.a1 me.m oo.1 NF.P1 me.. caemweoem 9 came. mooooo. eaao.1 meoo. Nooooo.1 «0060. .me. essences Fm.P mm.F mom. me.m mmo. mN.P1 mum.1 oHpmeopm 9 ohms. Foooo. Feoo. emoo. moooo. moooo.1 memm.1 *oeom es. om. Po.e1 mo.m mme.1 .m.m1 em.m oaemaesem a ommF. Fooooo. oeao.1 mmoo. mooooo.1 eoooo.1 anew. *oewuo sebocosm as. ea. mm.e1 mm.m ems.1 me.m1 em.m canoeuopm a oeme. sooooo. ammo.1 memo. moooo.1 sooo.1 ommm.F season psoaoa> es. em. mo.e1 om.m emm.1 so.m1 ee.m cheesespm 9 Fame. moooco. mepo.1 mmoo. .oooo.1 eoooo.1 ooom. Newewbo Hosea moms mzma mes sum mom: anew essences showcase mEHmo mo Nmommado Hm mZOHmmmmwmm moaomm szHEHO mom mOHBmHadam a 924 mazmHOHmDmoo nmamfiHamm b.m mgmda 75 are all on the order of magnitude of 10"4 and 10-5, very simdlar. For total crime, the coefficient is approximately -000041. This means that for every $1 increase in manufac— turing wages, the u/v ratio decreases by .000041. Thus, if SALR is 10,000 and u/v is .329 (total from Table 3.2), then a 1000 increase in SALR will decrease U/V from .329 to .288. What this means is that manufacturing earnings is an effective variable whose one or two thousand dollar inter- city variations may account for a large amount of variation in U/V. The better the average economic position, the lower the rate of translation of V to U, i.e. U/V. The effect is especially noted for aggravated assault, whose coefficient is -.OOO18. What does this mean? Basically, especially for aggra- vated assault, greater income corresponds with less tendency to translate crimes reported in the victimization survey into "crimes known to the police." This may be eXplained in several ways. The NCS reports found thm higher income groups typically are more apt to report property crime, but are less likely to report violent crime. The fact that manufacturing wages' greatest effect is on aggravated assault-~i.e. that with a large potential for victim culpability and offender sympathy-- supports this. However, the fact that a negative coefficient exists for prOperty crime contradicts the expectation conveyed in pro- position 1. It was expected that higher income would be as- sociated with higher U/V ratios. What may in fact explain this is something which the data used here can not reveal. 76 That is, rather than higher income being associated with greater confidence in the police and higher average losses (i.e. the more you have, the more you lose), it may be that average losses are not higher for such groups. If, for example, the average larceny loss is $45 for all_income groups, then that $45 represents a smaller fraction of a high income than a low one. Consequently, the higher the income, the less salient such losses. Unfortunately, the data do not allow the testing of this idea. Reporting and Population Size-~UPOP Urban population has a negative coefficient for six of the crime categories, as shown.in Table 3.7, but is not significant in any case. If the sample were larger, it might be possible to make some inferences based on the fact that most of the coefficients are negative (the expected direction); however, at the .05 level, one can not be confident that the coefficient for urban pOpulation is significantly different from.zero. It was considered that the inclusion of urban density (see below) may have "washed out" the effect of UPOP, due to their moderate correlation (.64). However, when each was dr0pped from the model, the coefficient of the other did not significantly improve. One note of interest is the fact that, while insignifi— cant at the .05 level, UPOP was significant at the .15 level for aggravated assault. For this crime, more than others, urban population was expected to have an effect on the U/V ratio. Increased numbers of people increases the number of 77 conflictive opportunities, which was expected to decrease the likelihood that victimizations would be viewed as "important enough" to be translated into official crime reports. Such an effect can not be ruled out by this analysis, and given an expanded data set, warrants further examination. Reporting and Racial Composition--BLK The proportion of blacks in the population had a result which was contrary to the theoretical expectations of propo- sition 6, citizen sector. The coefficient for BLK was decid- edly positive in every case, and was significant in seven of the nine crime categories (all except burglary and auto theft). From Table 3.7, we see that the total crime coefficient for BLK was .0038. BLK ranged from 5.8 (Portland) to 76.3 (Washington, D.C.), and the U/V ratios for Porfiand and Wash- ington were .32 and .52, respectively. With a 76.3 - 5.8 or 70.5 difference, we would expect, based on Portland's .32 u/v ratio, for Washington to have a U/V ratio of .32 + 70.5 x .0038, or .58. As evidenced from their actual values, such an effect is present. Thus, a population with a higher proportion of blacks experiences a higher level of victimization to official report translation. Given the wide variations in the proportion of blacks, this effect can be significant, as illustrated by Port- land and Washington, in which the BLK variable would explain a U/V increase from less than 33% to nearly 60%. However, the theoretical expectations were that blacks would be asso- ciated with lower U/V ratios. That expectation focused on 78 blacks as an alienated or disaffected group which distrusts the police and does not rely on them. Does this mean that blacks have mggg active reporting habits than whites? Not necessarily. What is happening may be explained in part by two issues. The theoretical expectations focused on blacks primarily as a reporting group. However, it is possible that blacks involved more as a perceptual factor. First, it is possible that a larger proportion of blacks results in greaer enforcement-- i.e. a white police department expects more crime from a heavily black population and is more likely to record such crimes than crimes committed by whfles. This effects a higher rate of translation from V to U. Since the statistical pro- bability is that in a population with_more blacks, more crime is committed by blacks, it follows from previous findings that black crime is reported more than white crime, and that more blacks mean more crime in the official reports. This does ‘ngt mean that more blacks commit crime than whites. It sim— ply means that black crime is subject to a higher rate of V to U translation than white crime, and that, even if blacks and Whites commit crime at the same rate, more of the black crime will be transformed from victimization to official re- ports. A second issue bears on the influence of reporting habits. That is, the presence of blacks may encourage whites and blacks to report crime more-~i.e. a crime is more salient and less likely to evoke sympathy or empathy with the offender if the offender is of a different race than the victim. With a 79 larger prOportion of blacks, the probability of interracial crime increases. In the case of interracial aggravated as- sault, for example, the typical situation of family member- provoked assault and victim-offender relationships if effec- tively eliminated in most cases. Conversely, in all—white or all-black cities, there is a greater probability that assault is intraracial and dogs involve someone related to the victim. Thus, the effect detected in this analysis may be that whatever reporting habits the blacks may have as a group (presumably lower than whites), that effect is negated by their perceptual impact on the community and the police in increasing the probability that victimizations will be officially reported, both through the police and through the citizens. Reporting and Age Compositions-AGE The proportin of persons below 20 was expected to be one of the more significant variables in the analysis, having been indicated as a factor by the NOS reports. However, in this analysis, AGE was shown to be insignificant at the .05 level in every case except auto theft. Age should not be dismissed too quickly, however. The coefficients for AGE were uniformly negative-—the expected direction. .Additionally, as in the case of UPOP, several coefficients were significant at the .15 level (total, vilent, property, assault, larceny). A larger sample could very well provide us with confidence about these coefficients. Even so, the effect would be very difficult to detect both because of the low value of the 8O coefficient (.0118 for total crime) and because of the small amount of variation in age from city to city: the extremes were 27.2% for San Francisco, and 43.6% for Newark. From NCS findings, it was suggested that young people would have lower reporting rates, which would help to effect a lower U/V ratio. This may have been due to several factors. The most obvious is the stereotypical view of the young as a group whichxfistrusts the police and has a lower stake in society's rules. There is also, perhaps, greater fear among the young that dealing with crime at the legal level may get them into trouble with peers or the police, or that their age will mitigate their believability. Additionally, offen— ses against young people may be more likely to be committed by friends or family-wherein the fear of reprisal may play a large part. 0n the macro level, then, AGE was expected to be assodated with factors which reduce the probability that an even will be translated from victimization status into "official" crime. The consistently negative coefficients in- dicates that such a relationship can not be ruled out, but must wait for testing with a larger sample. Reporting and Urban Concentration--DENS Urban density was insignificant in every equation, and had essentially the same results as urban populatxn. As pointed out earlier, the dropping of either variable did not improve the results of the other. This might be explained in several ways. 81 The blurring of urban boundaries, the existence of high rise apartment buildings, and the general variance between neighborhoods may be too complex to allow density to be a clear cut factor, from city to city. That is, urban density may be quite low in some parts of the city, and quite high in others. Similarly, the U/V ratio in those parts of the city may be different from that in contrasting sectors. Thus, if density has any independent role, it is possible that it is simply not being measured adequately in this analysis. Rather a whole city density of 9,000/square mile may indicate a city with uniform density throughout, as well as a city with poc- kets whose density approaches 50,000/square mile but with large unpopulated parks. This possibility warrants con- sideration, but is too complex for the relatively less re— fined data which are currently available. Reporting and the Suburban Contribution--PPOP In this analysis, the suburban element was found to be an insignificant factor. While its coefficients are positive for all cases except auto theft, it is not possible to con- fidently draw any conclusions. Only in the case of rape was PPOP significant at even the .15 level. Here, this result might be due to the fact that a larger proportion of none residents reduces the probability that the victim is ac- quainted with the rapist, and consequently increases the probabilfiy that the victimizationwwill become part of the official crime records. Beyond this hedged inference, however, any conclusion would be tenuous. 82 Recording and Relative Pay--PRD The ratio of police to manufacturing earnings was positively correlated with the U/V ratio in every case ex— cept auto theft, but Was significant only for aggravated assault. Aggravated assault, however, does not appear to contain any special characteristics which would single it out for significance. Particularly, larceny was expected to be more clearly related, since the relative economic posi~ tion of the police was expected to influence the U/V in cases in which recording was expected to be especially susceptible to bias. However, larceny was not even sig- nificant for PRD at the .15 level. A better relative economic position of police officers was expected to be related to conditions leading to more professional conduct-including their processing of criminal statistics. Perhaps a higher relative wage does indicate greater value of the police to the community, indicated by the community's willingness to pay a higher wage to them than to the manufacturing worker. In the case of aggravated assault, then, the greater perceived status of the police may increase the V to U translation, by increasing the probability that the public will call in the police when such crimes occur. There may also be some effect in that the perceived higher status is perceived by the police as well as the public, af- fecting both sectors. 83 Recording and Civilianizations-PCIV Civilianization provided positive coefficients for each category of crime, which were significant for total, property, rape and larceny. The percentage of civilians employed by police departments varied from a low of 1.9% for Pittsburgh to a high of 27.9% for Oakland. The estimated coefficient for total crime was .773, as shown in.Table 3.8. Thus, given Pittsburgh's 1.9% PCIV, and its U/V of .255, and Oakland's PCIV of 27.9%, one would expect the Oakland U/V to be .255 + (.773)x(.279 - .019), or .455. This compares favor- ably with Oakland's actual U/V of .408. Conceptually, PCIV was expected to have a positive ef— fect on U/VA-more civilians reduce the burden on police officers and also represent a department's greater commitment to devéoping their data collectin methods. Clearly, such an effect is possible here. The use of civilians instead of police for recording purposes in effect reduces the effect of bias, Which is one of the perceptual filters which mitigates the VaU transformation. Thus, the U/V ratio increases as civilianization becomes more prevalent. More detailed ex- pectation might provide that PCIV's greatest effect would be to increase the translative processing of those crimes which are ordinarily most subject to reclassification-~i.e. those winding up in categories other than what the victim reported-- such as aggravated assault. A crime such as burglary, the nature of which may be considerably less ambiguous, has a significantly lower PCIV coefficient than assault and larceny. 84 50:» con? Hm>oa mo. .Ho>oa mo. can pm PamOfiMflQMam mm; soapmzdo soap moPmoHoGHx Po psmoamasmam ohm moapmwpmpm a .odam> .m.m oasmfia ca opsaomnc .wo.m woooxo dmswmou who moanoaam> em. em. ea. am.n em.m oaenapoem a moo. mac. was. mmm.: mm.F among ooze sp.aa m..n om.. 0mm. mo. oaenapsem a emo.s Noo.: man. one. mow. asoawssm we.m no.1 om.m ms.a mm.Fx oaenaensm a moo. ooo.u meo.e mosm. Pom.s sasoossa em.m mp..n mP.P ms.m m>.~u oaemapoem a mam. mmo.u mom.m new. mom.au seasonoa eopo>samw< mm.m se.: mo.F am. mm.n oaenaposn a ewe. soo.n com. was. no.1 anopnom Fo.m se.: oe.m mm. m¢.F: oaenapoem a mama. mooo.n oom.. omo. mm.u roams mm.m ms.u mm.m mN.P mmm.: oapoapspm a Omo. ¢oo.01 wmm. our. PmF.I mafiao Annmaohm sp.m om.eu em.a o>.. m>.Pu oaenaespm a mma. mac.: emm.a omm. «Ph.: soaaao esoaa> ~>.m sm.n ma.m ma." >>..n oasmaesen a ammo. mmoo.t new. .mm. mem.u ‘sosaso Hosea mmoo mam >Hom 9mm psosnsoo asomopoo mEHmo ho Mmoumadv Mm mZOHmmmmwmm moaomm EOHHOQ mom mOHamHsdam 8 92¢ mezmHonmmoo nmadfiHamm ®.m mqm Scam mouzaoxo ma oawso Hmpoa .pmosp opus mocsaosfi mafiao aphomoamx .m.m can m.m moans» ea macaw mm Ho>oa me. on» ad psoowmwsmam ohm moan; momdo mo hopes: on» opmowcsa monompsonma ma mnonasz m on F Ahv m o>apfimom ovfimmo hon oofidom .09 m on o Avv m o>apamom soapcuwsmwaa>ao .m oaoosw m on P AFV m o>flvfimom Mdfihfiwommsnm5\o0HHom .m m on w on w o>wpwmom yoowmwo hog mosdpwcsoaxm .w m on o Ahv m o>wpmMoz Medan psooaom .m e Mow P Mow m *aphmmonalobapmMoz cofipoasaoa s o o o e esoaow>uoewewccm we sowpwoeowa scnwsnsm .m m any h on m o>wwwmom momma wsahapomwsnos .d m AFV m on F o>wpmmoz om gobs: Psoohom .m m on m on b o>fiPcMoz apfimson .N m on m on m o>HPmMoz soapmasmom .— conco Aceoasoeaeassv Abscesseaeassv >\a so no nonssz o>wpomoz momma o>apwmom momma sconce coeooaxm mandaum> sofivamoaonm meHDmmm ZOHBHmomomm mo dezzbm P.¢ mamde H37 most of the analysis. What conclusions can be drawn from this? Previously, the literature has concentrated in large part on police bias and data manipulation as being respon- sible for the inaccuracy of crime statistics, and the failure of the UCR to fully measure crime. The findings here, however, indicate that such is ppt generally the case. The factors relating to the perceptual mechanisms and police interaction with the general population explain more of the variation, and with greater significance than factors relat- ing purely to police activity and potential bias. Only in the case of rape were the police variables more signi- ficant (and this may be "significant" in that the police side contains fewer variables than the citizen sector), and only by a marginal amount. Even there, a strong interaction effect may be evidenced by the common variance of the two sectors, which would be further indicative of the inability of the police factors to Operate in isolation (see Chapter 3, pages 73-88). The most significant factors in explaining the U/V variation were racial composition and police concentration. Both of these are seen as having strong perceptual links to the general population, such that their primary effect is in increased reporting-~i.e. from the citizen sector. In the case of racial composition, a larger concentration of blacks tends to heighten the awareness in the pOpulation, and effect more reporting, and hence a higher U/V ratio. 108 When there are more police per capita, it was argued, it is likely that proactive discovery will increase which increases reporting. There is also greater visibility. The effect which was inferred tend to provide evidence for both types of influence, indicating a further perceptual link between the variable (police concentration) and the U/V ratio. Bringing this back to a broader theoretical level, the U/V discrepancy occurs because of a translative failure-i.e. those victimizations have not been translated into official crime reports. At the broadest level, this failure occurs because of a failure to communicate or to internalize the reporting routines which society prescribes. Thus, as a normative indicator, U/V represents the degree to which the social aggregate fails or succeeds in doing what is theoretically socially appropriate-i.e. report crime. The larger this failure, the smaller the U/V ratio. The central hypothesis of this dissertation sought to examine the extent to which the u/v ratio covaries with certain other indicators which were related to anomie in a Durkheimian sense. Urbanization and industrialization being the central processes attending normative breakdown, in recent times, features which were related to both were incorporated- urban pOpulation, density, racial composition, etc. In fact however, those variables which were argued to be most re- presentative of Durkheim's thesis-p0pulation and density-- were not at all significant in this study, given the specific criteria for rejection of the null hypothesis. 109 What was found was that variables such as the percentage of blacks and manufacturing wages were related to U/V, but in an unexpectaidirection. Thus, as conceptualized herein, they were incorrectly assumed to be related to anomie, as defined, or they incorporated enough of alternative under- lying meanings that the expected relationships did not surface. Clearly, they are still indicators of some underlying meaning, but part of the basic assumptions about anomie necessarily attending urbanization and industrialization must be inappro- priate. The percentage of blacks, while increasing, does not result in a U/V decrease. This, given that historically, the increasing concentration of minorities in the central city has been a concomitant of urbanization. Rather, the perceptual linkedness is seen as important. That is, given a particular reporting norm, the degree to which it is acted upon will in many ways be mitigated by certain perceptual factors. While reporting may be appropriate behavior, enough normative variation apparently exists that the norm will be more closely adhered to when there is a larger proportion of blacks, more police per capita, and lower manufacturing wages (economic level). Better V to U translation is further facilitated by a large concentration of civilians within the police department. What is notable about this finding, however, is that three of the factors which were found to be related to the U/V ratio have current trends which would increase U as a fraction of V. Civilianization, the black proportion, and police dens— ity are all increasing. Thus, while we may still view large 110 cities as being in flux, as adjustment to modernization con- tinues, one measure of anomie (U/V) is decreasing. This fact is clearly spelled out in a recent nationwide victimization report (1977), which showed that 1975 victimization rates were closer the the UCR figure than in 1973 and 1974. Addi- tionally, the NOS reported reporting rates were increasing. Conclusion Two of the consequences of urbanization are increasing moral density and increasing individuation. The apparent contradiction is mediated by other factors in the social system. The mass media, for example, promotes the homo- genization of the population in a way which helps establish the collective consciousness. This intervening or mediating factor helps to decrease the discrepancy between the general sets of norms and the individual, increasing the U/V ratio, in this case. The media-promoted prominence of victimization has, then, a large (and apparently real) potential for increas- ing U as a prOportion of V. Additionally, we find that the level of integration of the population is increasing. This may be indicated by the U/V ratio in that increasing it means a merging of two general levels of normative reaction. That is, a police officer may attempt to apply a set of formal rules in his police work. However, the greater the integration of the community (moral integration), the closer will the set of rules used by the police be to that with which the population identifies. Thus, the more integrated population reduces the discrepancy. 111 While it is clear that a set of perceptual filters oper- ate to reduce crime between the two measuring points, it is not clear that such is due to anomie. Rather, the variables tend to take on a more active role-rahter than being asso- ciated with increasing anomie, they are associated with increasing reporting, as in the case of racial composition, police concentration and civilianization. In fact, if this were a time series analysis, each of these trends appears to be on the rise, which would indicate increasing sensitivity. The presence of a potentially alienated group, blacks, in high- er concentrations does a 3 reduce the reporting-any effect which they might have by virtue of lower reporting habits is more than cancelled out by the effect they have on reducing perceptual obstructions to reporting. The failure to find support for an anomie dimension, as defined abovr, may be viewed in several ways. First, it is possible that anomie is not present. Second, it is possible that by virtue of their similar positions within an industrial- ized nation, each of the large cities is at the same level with respect to anomie, and does not vary on the anomic di- mension. That is, perhaps anomie is a more subtle process which can not be measured in minute gradations, and a more appropriate comparison would be cross-national, or between large and small cities. A third possibility is that anomie is present, but can not be measured adequately in this framework (some support for this might be found in the con- sistent, if insignificant, effects of age and pOpulation). 1‘31u”.' (W4! IA. 112 Thus, the discrepancy between the UCR and NCS is not randome-it occurs as the product of perceptual filters which operate to different degrees depending on the characteristics of the individual cities. Further, those filters are more ac- tive at the citizen level, providing little support for the idea that the police are the primary source of the discrepaccy. Future Directions and Limits of this Study One problem of this dissertadon lies in the small sample of cities which was available. The confidence one may have in the results of empirical analysis is often determined by the sample size. Such was not possible here. That significant results were obtained in spite of the small sample provides convincing support for the conclusions offered above. The fact that many variables were not significant, however, demonstrates the problem of obtaining confidence when using such a small sample. Thus, one need in the area is that the victimization stu- dies be expanded to include more cities, and perhaps a wider vaiety of city sizes to provide a better basis for comparison. Perhaps more subtle variation could be detected if the data were more voluminous and more refined. Many of the relationr ships which appear insignificant in this study may in fact be significant in an alternative framewoek. Future surveys might also benefit by conducting their sample such that all 26 cities are surveyed for the same time frame. From a practical standpoint, certain economies might result from such an approach, given thatrmunrthings which are 113 replicated for each phase of each survey would only have to be done once. This would allow a sample which is more amenable to comparison, unlike that used here in which time series pooling was necessary. Thus, most of the future directions one might suggest are contingent upon an expanding data base. Comparisons across city sizes, time series comparisons, and analysis of variance can only be accomplished with a larger group of observations. For the latter, if one group of cities with a pOpulation above 250,000 has less or more U/V variation than a group of cities between 25 and 50,000, then we might make further tests about the notion that the smaller cities have less normative variation than larger cities. This would also release us from the assumption that normative conformity is only indicated in reference to U/V's distance from 1. Policy Recommendations There are, of course the obvious policy recommendations which accompany the expressed need for a larger sample. This need is clear. One of the most enlightening findings from this study, however, is that the power to increase the "accuracy" of crime statistics lies primarily in the hands of the general population. While civilianization is a very important factor in improving recording, most of the U/V discrepancy exists because crime is not reported to the police to begin with. Most important is the fact that the problem is a perceptual one o 114 Where the sensitivity of the populationis heightened, ei- ther by racial fear or greater police visibility, reporting is greater. That is not to say that racial mistrust and a high police/citizen ratio is good. It does, however, point out the importance of attitudes and perceptions in effecting a community-based approach to crime. This would suggest, for example, that the current trends in improving police-community relations, and the publicization of police emergency reporting numbers is in the appropriate direction. This lends support to the adea that such efforts will not be ineffective. Further, the more the public is made award of crime, it appears the more they will attempt to deal with it, rather than ignore it. This, in turn,has the the effect of increasing the effectiveness of the police- they have more information to work with-which may eventually have some impact on reducing crime itself. Additionally, greater community awareness may have the effect of improving normative control over crime, having a direct impact on its reduction. In line with this, a further policy recommendation would suggest that victimization survey results be more heavily publicized. If the victimization results were publicized as heavily as the UCR rates, perhaps greater awareness that a problem exists could be generated, leading to some alleviation of the problem. In general then, this study suggests a great potential for imporved reporting even without the touted com— puterization of information processing. The greatest potential for accuracy lies in the c00peration and education of the public. FOOTNOTES 2. 3. 4. 5. 7. FOOTNOTES For a discussion of the develOpment of empirical method- ology see, for example, Jack Douglas, American Social Order, 1971; especially pages 7-12. This is a recurring themem in Durkheim's approach, used here in the early chapters of Suicide, and repeated in The Rules of Sociological Method (1938: pp. 1-13; and The Division of Labor in Society (1933: pp. 49-63 . In many ways,'Durkheim's "underlying meaning" is a uni- fying concept, serving a purpose similar to that of function for the functionalists. This is a general concern of labelling theory, i.e. the norms governing reaction. For an overview, see Lemert (1951) or Becker (1963). Although there is a very large literature concerning reaction to "deviance", very few have been concerned with the particular issues discussed 'hereo This is implied by others who have noted the "local traditions" in policing such as Wilson (1968) and Although such incidents are thought to occur elsewhere, it is rare that documentation is so clear. In Washington, political pressures were quite strong, and were part of a larger context of public deceit. This is part of a large body of literature concerning self-reporting. See also Akers (1964). Wenninger (1962), and Reiss and Rhodes (1961). “ Using V as a standard can be justified by reference to NCS surveys which show that victim responses to their own reporting behavior are on the order of 90% correct or true. If we extrapolate that accuracy to the broader set of data being dealt with, even if not 100% accurate, the victimization rates are closer to "real crime" than the UCR. 115 .vl‘ 4 ’..".'l‘ .q APPENDICES .aoeoa no. so .mp.m o>ond esooawasmwo e «a esooecasmen noeoowose a .ohoa on assoc season page sconce oHSOB oompflmmca mo hmoho non: paooxo mpwmflo psmowmfismflm oohnv op stone and mossmfim mop. mooooo. mmm.e moo. moo. moooo. mmo. ooo. mooo. soooo. hem.e sowed oncosoem mop. soooooo. oNF.n ooo.w.mmop.moooo.n oso.u moo. moooo. >oooo.n Fee.m goose coda coo. Pooooo. mmm. see. now. .oooo. coo. moo. ooooo. moooo. was. gonna owsossem emo.u sooooo. mam. o>P.I mom. moooo. ooo.a moo. Foooo.u moooo.n one. anoamasm mmo. Pooooo. one. omm. 5mm. ooooo. woo. woo. ooooo. moooo. moo. tonne ossosswm mmo.: Pooooo. moo. map. mmo. .oooo. moo.n soo. mooo.u moooo.n nor. showcase amp. mooooo. mem.e sou. mam. moooo. mmo. poo. mooo. ooooo. mmm.m scene oscosoem mmo.u Pooooo.n Pom. wmo. oomp.soooo. moo.a meo. Fooo.u Pooo.: mam.e seasscca oso. .ooooo. oom. mom. mom. Foooo. meo. moo. mooo. moooo. moo. nouns oucoscem mao. moooooo.n Foo. oo..n mop. moooo. meo.n eoo. moooo. moooo.: oao. essence seo. Fooooo. mam. sow. sen. .oooo. mpo. moo. .ooo. moooo. Fem. sense oscocsem Noe. Pooooo. ePN.. ooo.a ohm. Foooo. meo. .ooo. ooooo.: moooo.c mmm.u woeom mao. Pooooo. mom. Foe. Fem. eoooo. moo. moo. ooooo. moooo. mam. costs oncosoem mmo.: oooooo. mmm. mmo. se.. Foooo. soo.: soo. soooo.u moooo.s Fem. *osweo aescoowm woo. mooooo. ems. 4mm. one. Poooo. oao. moo. Fooo. soooo. me..e sound accessem mao. Pooooo.n mam. Foo. eeo. Poooo. emo.u soo. moooo. ooooo.u eem.e ceaso esoaow> Pmo. Pooooo. mom. mop. mom. Foooo. ooo. moo. soooo. Noooo. mom. nosed cabochon mmo.n Pooooo. ems. moo. one. Poooo. ooo.n soo. ooooo.: moooo.u was. *oswno Hosea mmoo mam >Hom mam moms mama moo mum mom: mean azasmzoo assessem ZOHB¢Dcm EROS; mom mmommm QMdQZde 92¢ mezmHUHmmmoo QHB