CRIME, THE POLICE AND CRIMINAL STATISTICS An Analysis of Official Statistics for England and Wales using Econometric Methods R. A. CARR-HILL St. Andre de Majencoules 30570, Gard, France and N. H. STERN Department of Economics University of Warwick, Coventry, England 1979 ACADEMIC PRESS London New York San Francisco A Subsidiary of Harcourt Brace Jovanovich, Publishers ACADEMIC PRESS INC. (LONDON) LTD. 24/28 Oval Road London NW1 United States Edition published by ACADEMIC PRESS INC. 111 Fifth Avenue New York, New York 10003 Copyright© 1979 by ACADEMIC PRESS INC. (LONDON) LTD. All Rights Reserved No part of this book may be reproduced in any form by photostat, microfilm, or any other means, without written permission from the publishers Library of Congress Catalog Card Number: 78-66678 ISBN: 0-12-160350-4 Printed in Great Britain by THE LAVENHAM PRESS LIMITED Lavenham, Suffolk List of Tables 2.1 Expected Chances of "Getting Away" with Each of Nine Prompted Offences by Offence Score 24 2.2 Confidence About Not Getting Caught by the Police 25 2.3 Ranking of Deterrents to Crime 28 2.4 A Typology of Modes of Individual Adaptation 37 2.5 Incidence of Delinquency and Crowding in the Home 45 3.1 Aggregate Shortfall in Police Numbers 72 3.2 Shortfall in Police Numbers in Detail in 1966 72 3.3 Extent of Violations and Per Cent Undetected 82 3.4 Willingness of the Public to Contact the Police in Different Situations 85 3.5 Willingness to Contact Anyone/the Police in Different Situations: Youth Sample 86 3.6 Differences between Official and Self-Reported Delinquents 90 3.7 Effects of Extra Police on Criminal Statistics 98 4.A.1 Indictable Offences 1966 118 4.A.2 Crimes Known and Cleared-Up in England and Wales for the Main Classes of Indictable Offences 121 4. A.3 Means of Each Variable 121 5.1 Variables in the Full Model 138 6.1 Likelihood Values for Chi-Square Tests on the Role of Unemployment 167 6.2 Chi-Square Values for Tests on the Role of Unemployment 169 6.3 Likelihood Values from Restricting the "Full Model" 170 6.4 Testing the Hypothesis that e = ij 172 p 6.5 Main Estimated Equations 174 6.6 Tests of Over-Identifying Restrictions in the Models of Table 6.5 182 6.7 A Model with Severity of Punishment Endogenous 184 6.8 A Model with Breaking-and-Entering Offences 187 6.9 Derived Reduced Forms 192 6.10 Variance-Covariance Matrices 196 6.11 Ordinary Least Squares Estimates 200 6.12 Two-Stage Least Squares Estimates 204 6.13 Overcrowding in the Household 208 6.14 Quantiles of the Distribution Function of x2 209 6.A.1 The "Full Model" 211 6.A.2 Correlation Matrices 217 7.1 Standard Deviations of Variables 223 7.2 Inter-Variable Correlations Over 0-7 in the Data Sets 225 7.3 Coefficient of Certain Variables in the Second Equation 226 7.4 Coefficients of All Variables in the Third Equation 229 ix X List of Tables 7.5 Coefficients of "Economic" Variables in the First Equation 232 7.6 Average Value of F—The Proportion of Those Convicted who are Incarcerated 233 7.7 Coefficients of Police Activity Variables in the First and Second Equations 238 7.8 Coefficients of Age and Social Class in the First and Second Equations 243 7.9 Coefficients of Police Activity, Age and Social Class Variables in the Model using Breaking-and-Entering Offences 245 7.10 Proportion of Variance "Explained" in the Second Equation of the Derived Reduced Form 248 7.11 Correlations Amongst Residuals for the Structural Equations which are Above 0-7 in Absolute Value 251 8.1 Changes in the Value of Property Crime Relative to Income and Ownership 270 8.2 Convictions and Imprisonment in Scotland in 1954-5 and 1961-2 272 8.3 Deaths and Discharges from Non-Psychiatric Hospitals in England and Wales broken down by Cause 276 Preface Crime fascinates many people. The study of crime and the reaction of society to it, is an important feature of several academic disciplines: politics, sociology, the law and economics as well as criminology. We believe that this study has benefitted from the fact that each of us has a very different view of the world which is, at least in part, due to the different disciplines to which we are attached. R.A.C-H. was first a criminologist and then a social statistician and N.H.S. is an economist. Criminal statistics are important. They are frequently used in popular discussion of crime and they form the basic raw material for the academic and the policy maker. The relationships between crime, the police and the criminal statistics are the central issue of our study. But as well as being fascinating and important the study of these relationships is complex. The complexity is an unavoidable consequence of theories we have had to consider and in the techniques required when using the criminal statistics to study those theories. The complexity derives mainly from the fact that we have to examine many things at once rather than any overwhelming difficulty of particular aspects of our work. The resulting broad spread of our subject matter has led us to explain the theories and statistical procedures under discussion in some detail. Accordingly any one reader will find that some parts of the book appear laboured whereas others require extra concentration. We hope that our work will be of interest to a variety of readers. It will be of special interest to criminologists, economists concerned with social problems in general and the economics of crime in particular, sociologists of deviance and of the law and to statisticians interested in the problems of social data. The only requirements are the willingness to become involved in the development of a formal model of the processes involved and in the discussion of the quantitative results from testing this model statistically. The book can be considered a monograph in the sense that it is a report on a single study, and it is therefore best read as a whole. We think, however, that some readers may wish to concentrate on certain chapters. xi xii Preface For example, Chapters 2 and 3 contain a review of theories of crime, criminal statistics and some aspects of the police. Chapter 5 provides a description of the simultaneous equations techniques of econometrics which are used extensively in our analysis. An examination of the use of criminal statistics in public discussion is offered in Chapter 8, and in Chapter 9 we provide a critical review of recent attempts to provide a purely economic theory of punishment. Whilst each of the five chapters just mentioned is to some extent a review of an area none of them is intended to be exhaustive. The study was planned and executed jointly and our work has been in 50- 50 partnership. However, R.A.C-H. takes particular responsibility for Chapters 2, 3, 7, 8 and N.H.S. for Chapters 5, 6, 9 and the Appendices to Chapters 2 and 9. Responsibility for Chapters 1, 4 and 10 is equal. The Data Appendix was compiled by Jocelyn Kynch. Our work on this project began in 1969 when we were graduate students together at Nuffield College, Oxford. During the course of writing this book we have published three reports: Carr-Hill and Stern (1973), Carr-Hill and Stern (1976), Carr-Hill, Hope and Stern (1972). We have presented various aspects of our work to a large number of seminars involving groups from a wide range of academic disciplines in universities and research institutes beginning with our first performance in Nuffield College in 1969. The first presentation at a public conference was at the Barcelona meeting of the Econometric Society in 1971 (see the summary of the papers published by the Organising Committee). We intend this book to be the final statement of our results. It has several features which cannot be found in our earlier published work. Thus there is a thorough discussion of the criminological, economic and sociological assumptions (in Chapters 2 and 3) underlying the model which is developed in Chapter 4. This last chapter provides an extensive discussion of the measures used. We begin in Chapter 4, and develop at length in Chapter 5, the problems of identification and unobserved variables, which occur in our analysis. There are two particular aspects of the model itself which do not appear in our earlier work. The first is the discussion of unemployment as a possibly important factor (see Chapter 2) and the second concerns the question whether or not there is such an entity as the "real" or "true" offence rate. The relevance or existence of such an entity would be denied by most sociologists of deviance. Indeed, the academic debate in this area has been fruitful for our study. But for more of this see Chapter 3. Finally, this monograph contains, in Chapter 7, a thorough discussion of the implication of our results for substantive criminological theory. We made one last recheck of our data before performing the calculations for this final report and picked up some minor errors. We were able to use data on socioeconomic groups from the 1971 census which had not been available for our previous studies. We have presented a very wide range of results, Preface xiii not only to cover the range of arguments under examination but also to provide help to the reader who may have wished to pursue a line slightly different from our own. For all these reasons we hope that quotes will be made from, and issue taken with, this book rather than our earlier articles. We have adopted certain conventions in the book which we should mention here. Equations are numbered from (1) for each chapter. Cross references to equations are to the same chapter unless otherwise stated. There is a list of works cited at the end of the book and we refer to works in the text by name of author, with the year of publication in brackets. "He" is used when referring to a person simply for convenience without connotation of gender. Standard data sources are not included in the list of works cited. A full description of data sources and formal definitions is contained in the appendix on "Definitions and Sources". Our main econometric model and notation are summarized in Table 5.1 on p. 138. Our data have been deposited at the S.S.R.C. Survey Archive at Essex University. There are many individuals who have been both kind and helpful to us during this study; of most importance are the members of our families. We have been fortunate with many institutions as well. We mention below those whose help was of special importance and apologise now to those whom our forgetfulness has led us to omit. We received financial support from the S.S.R.C. (Grant Number HR3733) and Oxford University. Academic Press have shown consideration and patience in waiting for this book for so long. We are grateful to all three. We have been supported throughout our work by the respective institutions to which we have been attached. R.A.C-H. was at Nuffield College, Oxford from 1968 to 1971, at the University of Sussex from 1971 to 1974, at the OECD in Paris from 1974 to 1977, and from January 1978 has been unemployed in the South of France. N.H.S. was at Nuffield College, Oxford from 1968 to 1969, The Queen's College, Oxford from 1969 to 1970, St. Catherine's College, Oxford from 1970 to 1977 and from January 1978 has been at the University of Warwick. We are grateful to St. Catherine's College, Oxford and Oxford University for granting leave in the autumn 1977 thus hastening completion of the book, and special thanks are due to Claude Henry and the Laboratoire d'Econometrie of the Ecole Polytechnique in Paris for hospitality and to the French Government C.N.R.S. for providing finance. Jocelyn Kynch and David Deans have worked with us on this project from time to time for several years and we are especially grateful to them. Jocelyn Kynch collected some and checked all of the data which are the basis of our empirical estimates. She kindly wrote an account of the data which has been included under her name as an Appendix. She has also compiled the subject and author indexes, and located many references. We are greatly indebted xiv Preface to her. David Deans carried out most of the programming with considerable expertise and good humour. Oliver Morgan did the remainder and thanks are due to both of them. We owe an important debt to Cliff Wymer, then of the L.S.E., who made available his SIMUL package programme. The options available in this programme were a major influence on the econometric work we were able to do. Jerry Hausman adapted the programme for the Oxford University ICL 1906A computer, on which the estimates reported here were produced. Clive Payne provided patient advice over the years. We were also fortunate to be able to use the FAKAD package developed by Emil van Broekhoven and Kenneth MacDonald. There are many individuals who have had an intellectual influence on the development of our work. We were particularly helped by the detailed comments of Art Goldberger on an early unpublished paper presented at the 1971 Barcelona meeting of the Econometric Society. Grayham Mizon was very helpful too on the econometric side. We benefitted from comments at various stages of our work by David Downes, John Flemming, Jerry Hausman, David Hendry, Al Klevorick, B. S. van der Laan, Gordon Wasserman and Nigel Walker. Mary Mcintosh was kind, enough to read and comment on Chapter 3, and Chapter 5 benefitted greatly from the comments of David Begg, Jerry Hausman and Kenneth MacDonald. Peter Rossi provided useful comments on most of the manuscript. Seminar audiences have responded with vigorous and often constructive comments at a number of university and research gatherings too numerous to name. We must mention in particular, however, the comments by the group working on the econometrics of crime at the Hoover Institution at Stanford University and we are grateful to Michael Block for arranging a visit there in 1976. We have been particularly fortunate in the kindness shown to us by the secretaries who have been involved in this work. These are Audrey Hiscock, Vera Kastner and Betty Wilbery at St. Catherine's College, Oxford; Marie- Louise Pouderous at the Ecole Polytechnique, Hazel Coleman, Lesley Mcintosh, Irene Sinha at the OECD, Brenda Jeater and Kathy Thorpe at the University of Sussex, Ann Sampson, Kerrie Beale, Shirley Patterson and Yvonne Slater at the University of Warwick, and Yvonne Scragg. Finally, and most important, is our debt to our friend Andrew Cornford. He commented in great detail on the whole manuscript and hunted down a large number of references for us. In view of the long list above we must conclude by stressing that all errors are our own. March 1978 R.C-H. N.H.S. 1 Introduction The focus of our study is the relation between official criminal statistics and the activities which they are supposed to reflect. Theories of crime should be able to tell us something about the generation of criminal statistics and the analysis of criminal statistics should improve our assess ment of theories of crime. Accordingly, we shall be interested in testing theories of the determination of the number of offences. Our concern is, however, much wider than this. One of the central interests of the criminal statistics is their relation with real events. The statistics are produced by official bodies, in particular the police force, and are to a large extent a function of the choices made by these bodies. Thus an examination of criminal statistics requires at the same time analysis of police activities. In other words, we shall be investigating theories of the generation of criminal statistics as well as of crime. It is unfortunate that many commentaries on, and academic studies of crime and deterrence, have been marred by a failure to appreciate and take account of the difference between offences as recorded and the number of illegal events. We shall see that this difference will be crucial to an understanding of criminal statistics and in the interpre tation of our results. But our interests do not, and we shall argue, cannot stop with theories of the level of offences, as officially recorded, and crime. Broadly speaking the problem is as follows. Simple deterrence theories suggest that the level of offences will depend on the probability of being caught. It is also reasonable to suppose that the number of offences which are recorded depends on the number of police. In turn one can argue that the probability of being caught, or the proportion of offences solved, depends on the number of offences to be solved and the strength of the police force; and again the allocation or recruitment of police officers would be determined in part by the recorded l 2 Crime, the Police and Criminal Statistics level of crime and the "success rate" or proportion solved. We are faced, therefore, with a problem of simultaneous causation: a given factor or variable will simultaneously contribute to the determination of another and be determined by it (for further details see Chapters 4 and 5). One cannot ignore, although it is all too common, these simultaneous relationships in a statistical investigation without producing biased or misleading results. Here, we are interested in the determination of all three variables—the offence rate, the proportion of offences solved and the number of police. But even if we are interested in just one of these, and the offence rate is a popular candidate, we could not omit some study of the other two. Our first task is to present the theories we are to examine and to place them in the form of a particular model. There are, again broadly speaking, three groups of theories concerned with the determination of the level of recorded offences: deterrence theories which concentrate on the con sequences, possible pay-offs and penalties of acts, in the determination of whether or not they are committed; the theories of traditional criminology which specify certain groups as being more prone to offending; and theories of some sociologists which concentrate on the recording of offences and the labelling of offenders. These three groups of theories are not necessarily competitive. Few commentators would argue that just one is relevant and that the other two can be excluded, but analysts differ considerably in their emphasis. We shall be examining all three groups of theories. In modelling the determination of the second variable, the proportion of recorded offences solved, we shall, in addition to the number of recorded offences and the number of policemen, be including variables intended to capture various aspects of the difficulties in solving an offence. In the process of modelling the determination of the third variable, the number of policemen, we shall be discussing factors affecting both allocation and recruitment as well as the number of recorded offences and the proportion solved, as already mentioned. Our data are for the cross-section of police districts in England and Wales and we examine this cross-section for the years 1961, 1966 and 1971. The police district is the lowest level at which the offence data are reported. Our measure of recorded offences is "all indictable offences known to the police" (see the annual Home Office publication Criminal Statistics). Indictable offences are, loosely speaking, those serious enough to allow the possibility of trial in front of a jury (for a full list of offence types included, see the appendix to Chapter 4). We must emphasize at the outset, particularly for readers from the U.S.A. who may associate indictable offences with their "felonies", that there is no monetary cut-off for an