rant, r «3 129 10713 9200 Q " ‘ ‘7 L- ..... i i "‘* " ‘-___ P ._L. ”Why n -_-- L 3.. ,‘::“;-..;a This is to certify that the dissertation entitled THE EFFECTS OF POVERTY, INCOME INEQUALITY, AND UNEMPLOYMENT ON CRIME RATES presentedby Joanne Belknap has been accepted towards fulfillment of the requirements for Doctoral degreein Criminal Justice 7/7 WW1 // Major professor Date ijlfi 307 / 9&2; MS U is an Affirmative Action/Equal Opportunity Institution 0-12771 IIV‘ESI_J RETURNING MATERIALS: PIace in book drop to LJBRARJES remove this checkout from ” your record. FINES wil] be charged if book is returned after the date stamped be10w. MAG'CZ mic 139,9 91 1999 4 d c \ H N ". m J ‘ v; .‘ ‘ — .9- w ' i 9-.7o eMEwww ”5 -, - -- 9- LL!’ ‘unq .r‘“ h. . - W0 2* «3m ' . , ar'u' . ' 3! he a V {A ‘ luv-"j? 91h: r r‘ ‘ FE§15 2833 ; 1:35; w *9 [3‘31 3 “1952 @AI ;1:5?:2§03 I L ’ I m3 999; 39691 " ‘ “a... 7.; V -j .4 ___ ._~ 1 - .ng U5? 'rtj’fé’zmamv 13 2997 _ ,, ‘3‘. _._, THE EFFECTS OF POVERTY, INCOME INEQUALITY, AND UNEMPLOYMENT ON CRIME RATES By Joanne Belknap A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY School of Criminal Justice 1986 Copyright by JOANNE ELIZABETH BELKNAP 1986 ABSTRACT THE EFFECTS OF POVERTY, INCOME INEQUALITY, AND UNEMPLOYMENT ON CRIME RATES BY Joanne Belknap This study used opportunity theory from sociology/criminology, and expected utility theory from economics to guide an empir- ical analysis of the relationship between the economic struc- ture and the crime rates. In addition to incorporating vari- ables measuring poverty, labor force participation, and income distribution, it also controlled for demographic variables (e.g., race, education, age, population size and density, ur- banization, population mobility, and region), and deterrence variables (police production and severity of punishment). All seven index crime rates (homicide, rape, robbery, aggra- vated assault, burglary, larceny, and auto theft) constituted the dependent variables. This cross-sectional analysis was conducted on 260 SMSAs in 1980. In addition to using an or- dinary regression model, a simultaneous equations model was tested, which allowed the crime rate, police expenditure, and the severity of punishment to all act as both dependent and independent variables. The explanatory ability of both models (ordinary and simultaneous) was moderately high, and the sig- nificance of income levels as important predictors of crime rates was supported. Another important predictor was the racial distribution, and of less importance were the popula- tion size and density, and the educational attainment. The simultaneous model was useful in showing that the police ex- penditure was significantly impacted by the violent crime rate and the tax rate, but that the crime rate was not affec- ted by police expenditure. Unlike the ordinary regression, the simultaneous model showed poverty to significantly affect the (property) crime rate, as did the percent urban. The labor force variables were never significant, nor were any of the deterrent variables. Thus, these findings lend more sup- port to opportunity theory than to expected utility theory (the latter focusing more on risk and deterrence), and suggest the importance of treating crime at a more fundamental level than strictly focusing on the "after the fact" approach of strengthening punishment and police expenditure. Dedicated with the deepest appreciation to Dr. Merry Morash for recognizing my potential and helping me achieve it How do you thank someone who has taken you from crayons to computer analysis? ACKNOWLEDGEMENTS Undoubtedly this process of acquiring a doctorate and writing a dissertation would not have gone as smoothly without the exemplary mentorship of Dr. Merry Morash. In addition to being extremely knowledgeable and patient, she makes a remarkable and rare commitment to her students. Dr. Larry Martin, another committee member, was helpful in numerous ways, but in particular I was grateful for his ability to help me find some humor in economics and the dissertation process. Dr. Jim Morrow was also extremely generous with his time (especially during the final stages), and I remain immensely impressed with his patience with my "identification problem" (of my simultaneous model, that is). Dr. Tim Bynum also contributed guidance and criticism to this dissertation. Dr. Bob Trojanowicz and Dennis Banas, vi although not actually on my committee, were very generous with positive feedback and faith in my ability to succeed. I am also grateful to the Charles S. Mott Foundation for funding most of my graduate career, and for providing me with incredible research opportunities. Equally important, has been the family of friends who have encouraged and loved me throughout (and in spite of) all of this. These are not in order of importance, because that is impossible, and some people have been left out due to space, but I have much gratitude to these people for the following reasons: Mary (“Sully") Sullivan, whose tea and politics warmed by heart and heated my activism; Patti Witte, who is probably the funniest human being I have ever met, and certainly a wonderful cosmic twin; Brian Mavis and Jim McClurken, who consistently found fun things to do to avoid, I mean take breaks from, working on our dissertations; E. Gail McGarry, for her life-long and long-distance support; vii Renee Villeneuve, who helped me escape reality when neces- sary; Donna Hale, for letting me go into rages when neces- sary; Karen Williams, who helped me develop a new way of looking at the world; Anne Kellogg, who taught me there really are choices in life; Donna and JoAnn, who convinced me that being stable and outrageous are not mutually exclu- sive; and Jean Matievich, for her friendship and postcards that always arrived when I needed them most. I am also in- debted to the Radical Feminist and Remedial Volleyball Players for laughing at my autobiographical sketches, and the Amethyst Softball Team for taking my mind off work. Finally, I want to thank my loving housemates, Megan and Koshka, who were always glad to see me. viii CHAPTER 1 TABLE OF CONTENTS THE THEORIES I. Introduction II. Expected Utility Theory A. B. F. G. Introduction The History of Expected Utility Theory Rational Behavior Applying Expeted Utility Theory to Crime Recent Application of Expected Utility Theory Criticisms and Limitations Summary III. Opportunity Theory A. B. Introduction Theoretical Criticisms IV. Using Aggregate Data to Test these Theories ix 14 18 22 23 23 26 38 CHAPTER 2 THE LITERATURE REVIEW I. The Introduction II. The Economic Variables A. Income Levels and Distribution B. Unemployment and Labor Force Participation III. Demographic Variables A. Racial Composition B. Region C. Population Size D. Population Density E. Urban-Rural Distribution F. Age Structure G. Educational Level H. Population Mobility IV. Deterrence Variables A. Police Production: B. Certainty and Severity V. Conclusions to the Literature Review CHAPTER 3 THE METHODOLOGY I. Introduction II. The Need for Current Research 45 45 46 46 61 79 79 85 87 89 92 93 96 98 100 101 108 115 120 120 120 III. The Sample IV. The Statistical Technique V. The Variables A. B. The Dependent Variables The Independent Variables VI. Limitations of the Study VII. Transforming the Data VIII. Contributions of This Research IX. Building the Model CHAPTER 4 THE FINDINGS I. The Introduction II. Results from the Ordinary Regression Model A. B. C. D. Introduction The Economic Variables The Demographic Variables The Deterrence Variables III. Results from the Simultaneous Equations Model A. B. Introduction The Economic Variables xi 121 122 126 126 129 130 132 133 134 137 137 138 138 141 142 145 147 147 147 C. The Demographic Variables D. The Deterrence Variables IV. Conclusion CHAPTER 5 THE DISCUSSION I. The Introduction II. Theoretical Implications III. Policy Implications A. Economic Variables B. Demographic Variables C. Deterrence Variables D. Conclusion IV. Implications for Future Research V. The Conclusion APPENDICES Appendix 1: Variables used in Empirical Analyses Appendix 2: The Data Appendix 3: Simultaneous Models Used in Prior Research Appendix 4: The Tables REFERENCES xii 150 151 151 151 155 155 158 160 161 162 166 168 169 175 181 184 192 CHAPTER ONE: THE THEORIES ‘. I. INTRODUCTION: COMBINING ECONOMIC AND CRIMINOLOGICAL THEORIES Gordon (1971:55-56) claims there are three approaches to analyzing the earnings-crime link: conservative, liberal, and economic. The conservatives view criminals as irrational, while the liberals view criminals as "pushed toward irrationality." Economists, on the other hand believe criminal behavior, ”like any other economic activity, is eminently rational" (1971:56). Model economic analysis "postulates that an individual chooses to commit a crime depending on expected benefits and costs" (Freeman,l983:90). This implies that crime rates may be affected through policies influencing unemployment, income distribution, certainty and severity of criminal sanctions. "The expected losses from criminal sanctions enter the criminal problem in much the same way as would, for example, the nonpecuniary aspects of a job" (Viscusi,mimeo:1). Block and Lind (1975:487) suggest that "if one were to stratify the population according to legal wealth, including both the value of assets and the present value of future earnings, one would expect to find that the percentage of individuals engaged in crime would decrease with each successive increase in wealth." CHAPTER ONE: THE THEORIES Sociologists, on the other hand, tend to see the "decision" to commit crime quite differently. "Criminologists approached the task of explaining illegal activity by attempting to determine those psychological and/or physiological factors that are unique to criminals" (Heineke,1978:1). Further, Heineke (l978:l) believes this inductive approach generally leads to untestable models. Cobb (1973:20-21) refers to this sociological approach as the "sickness hypothesis" which views criminals as sick rather than rational, and therefore see rehabilitation as the only solution to controling criminal behavior. ”Thus, the issue that divides the two schools [traditional criminologists and economists] concerns their view of man: whether his behavior is the result of natural causes or whther he freely chooses to behave as he does" (Orsagh,l983:392). Expected utility theory is a subset of the theory of choice under uncertainty, which implies risk-taking. ”The general picture of such a theory [of choice in risk-taking situations] is the following: There is a set of conceivable actions which an individual could take, each of which leads to certain consequences. The individual has in mind an ordering of all possible consequences of actions, saying, for each pair of consequences, either that he prefers one or that he is indifferent between them...” (Arrow,197l:l). For risks to be CHAPTER ONE: THE THEORIES taken, it is implied that there is some expectation of profit (l971:6). Both theories view the individual as discrete and autonomous in the decision-making process (see Monzingo,1977:262 regarding expected utility theory), and opportunity theory is more like economic theories and less like other criminological theories in its view of the decision to be criminal as a somewhat rational decision. However, these theories clearly have differences which will be elaborated on in the remainder of this chapter. An economist would postulate that, given his en- vironment, a criminal would act in a rational manner to maximize utility. Thus, two identi- cal individuals who face different environmental constraints would behave differently with respect to criminal intentions, not because their aims were different but because their environments differed. Note that the ecologist would say that the environment is causing the criminal to act, whereas the economist would say that the criminal is acting, taking his environment into account (Chapman,l976:50). Regarding the ecomomic approach, Orsagh (1983:393-394) claims: "In the economist's view, man is embedded in an environment which limits, and in a sense, determines his choices." Opportunity theory, then, suggests that variables such as class, race, gender, education, and skill level will all influence the optimal job opportunities, legitimate and CHAPTER ONE: THE THEORIES illegitimate, one has open to him or her. "Crime should be more attractive to nonwhites because there is less employment discrimination in illegal than legal occupations" (Swimmer,l974a:6l9). In the economic framework, the decision to become involved in crime is influenced both by the effectiveness of criminal justice deterrents, such as arrest and conviction rates (to be discussed further), and also by the limitations an ex-offender faces in the job market due to his/her criminal record. "Recidivism is thus not necessarily the result of an offender’s myopia, erratic behavior, or lack of self control, but rather may be the result of choice dictated by opportunities” (Ehrilich,1975b:3l9). Swimmer (1974a: 618) claims: "The crime rate of a city is negatively related to the probability of being caught which in turn depends upon police expenditure." Heineke (l978:2) views the criminal choice as a special case of the economic choice theory, and hypothesizes that "all individuals, criminal and non-criminal alike respond to incentives; and if the costs and benefits associated with an action change, the agent's choices are also likely to change." Opportunity theory contends that demographic constraints affect legitimate/illegitimate work decision making. Opportunity theory is in a sense a sociological theory that is a counterpart to the economic expected utility theory. CHAPTER ONE: THE THEORIES II. EXPECTED UTILITY THEORY A. Int d ° e e t Ut t eo Expected utility theory is an approach to describe how individuals order uncertain consequences, the point being to maximize their expected utilities.1 Kahneman and Tversky (1979:263) claim: "Expected utility theory has dominated the analysis of decision making under risk."2 According to expected 1 Savage (1954:69) describes a utility function: ”A function U that arithmetizes the relation of preference among acts will be called a utility. It will be shown that the multiplicity of utilities is not complicated, every utility being simply related to every other." 2 Decision making under certainty, uncertainty, and risk are commonly referred to in this literature. Luce and Raiffa (1957:13) adequately differentiate between the three: ”We shall say that we are in the realm of decision making under: (a) Certainty if each action is known to lead invariably to a specific outcome...(b) Risk if each action leads to one of a set of specific outcomes, each outcome occurring with a known probability. The probabilities are assumed to be known to the decision maker...(c)Uncertainty if either action or both has as its consequences a set of possible specific outcomes, but where the probabilities of these outcomes are completely unknown or are not even meaningful." CHAPTER ONE: THE THEORIES utility theory, choices made under uncertainty about "the state of the world" are made under "the valuation of the consequences and the relative strength of belief in occurrence of the different states of the world“ (Arrow,197l:45). Kahneman and Tversky (1979:265) claim: In expected utility theory, the utilities of outcomes are weighted by their probabilities." There is a method of assigning utilities to the individual outcomes so that the utility assigined to any probability distribution is the expected value under that distribution of the utility of the outcome. The numbers are said to be utilities because they are in the same order as the desirability of the distri- bution or outcome (desirability itself being merely an ordinal concept) (Arrow,19Slz425). B. The Histoty of Expected Utility Theoty Daniel Bernoulli is commonly thought to be the founder of expected utility theory. "The idea that choices among alternatives involving risk can be explained by the maximization of expected utility is ancient, dating back to D. Bernoulli's celebrated analysis of the St. Petersburg paradox" (Friedman and Savage,l948:281). Bernoulli’s paper on measurement and risk was first published in 1738, and much later translated from Latin into English in 1954. He "argued that the individual acts in such a way as to maximize the mathematical expectation of his CHAPTER ONE: THE THEORIES utility" (Arrow,1958:4). Bernoulli was interested in probability theory and found that pure monetary expectations did not tell one which bets were likely, or how the value of money decreases as one obtains more of it. Therefore, when analyzing betting, the focus should not be on the money per se, but rather on the yglug a person would gain. If the utility of each possible profit expectation is multiplied by the number of ways in which it can occur, and we then divide the sum of these products by the total number of possible cases, a mean utility (moral expectation) will be obtained, and the profit which corresponds to this utility will equal the value of the risk in question (Bernoulli,l954:24). This work led economists to looking at marginal utility, but Bernoulli's approach has since been discredited. Utility theory remained dormant from this point until 1944 with Von Neumann and Morgenstern's book titled Ibeoty gt Games and Eggngmit Behavior. They created game theory as a field and conducted a rigorous approach to create the modern field of utility; focusing on the individuals' risk in order to gain. CHAPTER ONE: THE THEORIES KW Inherent in an analysis of expected utility theory is the assumption of rationality. ”In its broadest sense, rational behavior simply means behavior in accordance with some ordering of alternatives in terms of relative desirability..." (Arrow,197l:3). Rational behavior implies actors selecting actions (means) which will best attain pre-determined goals (ends). "When people have: (1) nonprobabilistic preference, (2) transitive preference and indifference, (3) 'forced choice', (4) nonperverse selection rules, we refer to them as rational” (Frolich and Oppenheimer,l978:ll). Utility functions, thus, purport to project future behavior. Additionally, rationality assumes that behavior is consistent over time; that people do not radically change their goals. Riker and Ordeshook (l973z8) claim: "[P]eople behave as if they arranged their preferences in some logical pattern. Such behavior we describe as tatignal,..." Furthermore they state "[W]e believe it is usually possible to define rationality in terms of expected utility: rational behavior is, then, the choice of that alternative which maximizes expected utility" (1973:20). Arrow (1958:1-2) claims: "What may be termed a tgtigng1_mgggl_gt choice (or decision-making) has the following well-known form: an CHAPTER ONE: THE THEORIES individual is assumed to rank all alternative logically possible decisions in order of preference; in any situation, only some of the logically possible alternatives are in fact available, due to budgetary or other limitations, and the individual is assumed to choose among the alternatives available that one which is highest on his ranking." Frohlich and Oppenheimer (l978:5) note that the consequences of actions may be uncertain, however the perceived consequences still affect the choice determination. In choosing among alternatives open to it, whether or not these alternatives involve risk, a consumer unit (generally a family, sometimes an individual) behaves as if (a) it had a consistent set of preferences; (b) these preferences could be completely de- scribed by a function attaching a numerical value-- to be designated 'utility'-- to al- ternatives each of which is regarded as certain; (c) its objective here to make its expected utility as large as possible (Friedman and Savage,l948:287-288). D. A i cte Ut o o m eh v r The bulk of information on expected utility theory applies it to decision making concerning the buying of insurance, gambling, and betting. However, more recently it has been 10 CHAPTER ONE: THE THEORIES applied to the decision of whether to partake in criminal behavior. Expected utility theory does appear applicable to the decision of criminal behavior in that there is risk involved (i.e., being arrested, incarcerated, etc.), and that the benefits of committing crimes (even when allowing for the chance of being "caught") likely result in some people (i.e., those lacking in education, employment, and other legitimate opportunities) perceiving it as a "rational choice." For instance, Kau and Rubin (1975:75) state: "Nonwhites, because of inferior legitimate market opportunities and lower opportunity costs of imprisonment, may be expected to have a significantly greater incidence of crimes.” Becker (1968:172), who was probably the first to directly apply expected utility theory to criminal behavior, claims: "The approach taken here follows the economists' usual analysis of choice and assumes that a person commits an offense if the expected utility to him exceeds the utility he could get by using his time and other resources at other activities." Regarding punishment, Becker (1968:179) believes costs can be converted to monetary equivalents, which can be done directly only for fines. The cost of imprisonment varies depending on the discounted sum of earnings foregone. Clearly, the unemployed and poor have less to lose by going to prison in this instance. 11 CHAPTER ONE: THE THEORIES Palmer (l977:6) discusses four costs of criminal behavior: (1) money spent on resources to commit the crime, (2) the cost if the individual does not "enjoy the criminal act," (3) the opportunity cost of investing time in criminal behavior and away from lawful activities, and (4) the expected loss due to punishment for the crime. Sullivan (1973:14) similarly attempts to evaluate the costs and benefits of criminal behavior: According to the economic explanation of criminality, the individual calculates (1) all his practical opportunities of earning legitimate income, (2) the amounts of income offered by these opportunities, (3) the amounts of income offered by various illegal methods, (4) the probability of being arres- ted if he acts illegally, and (S) the prob- able punishment should he be caught. After making these calculations, he chooses the act or occupation with the highest discounted re- turn . Friedman and Savage (1948:283) claim: "The major economic decisions of an individual in which risk plays an important role concern the employment of the resources he controls: what occupation to follow, what entrepreneurial activity to engage in, how to invest (nonhuman) capital." Block and Heineke (1973:381) believe that even returns from the labor market are uncertain and labor allocation decisions depend on risk-taking behavior, but 12 CHAPTER ONE: THE THEORIES they also admit "the individual uses the labor market as a 'hedge' against uncertainty.” Markowitz (1952:151) states the following concerning the above Friedman-Savage hypothesis: “Each individual (or consumer unit) acts as if he (a) ascribed (real) number (called utility) to every level of wealth and (b) acts in the face of known odds so as to maximize expected utility.” Applicable to the offense decision is Arrow's (l971:6) statement: "It has been vaguely contended that 'profits are the reward of risk-taking,’ in the sense that the expectation of profits is a necessary inducement for risk bearing." Similarly, Knight's (1957:236) belief may also be applied to the crime decision: ”We shall assume, then, that if a man is undergoing a sacrifice for the sake of a future benefit, the expected reward must be larger in order to evoke the sacrifice if it is viewed as contingent than if it is considered certain, and that it will have to be larger in at least some general proportion to the degree of felt uncertainty in the anticipation." Luce and Raiffa (1957:21) believe we need to construct utility functions on an individual basis, which represents choices among gambles with expected values for the corresponding gambles. Assuming the large reward usually associated with successful completion of an undetected crime, Knight's (1957:237) statement can be adequately applied to a person with a small or 13 CHAPTER ONE: THE THEORIES no income: "Hence, it is correct to treat all instances of economic uncertainty as cases of choice between a smaller reward more confidently and a larger one less confidently anticipated." Similarly, in a discussion of the risks unskilled workers are willing to take concerning gambles, Friedman and Savage (1948:299) claim: "Men will and do take risks to distinguish themselves, even when they know what the risks are." In the same vein, Arrow (1971:45) claims: "[I]t is clear that any choice for actions must be synthesized out of two components, the valuation of the consequences and the relative strength of belief in the occurrence of the different states of the world." In addition, the ”risk" involved in a decision also plays an important part. Diamond and Stiglitz (1974:337) point out that not only are some situations riskier than others, but some people are more risk averse than others, thus increases in risk should affect the more risk averse individuals more than the less risk averse individuals. 14 CHAPTER ONE: THE THEORIES E. e n c ‘ ec ed Consistent with expected utility theory, Phillips et al. (1980:492) hypothesize that crime rate increases can be explained by ”deteriorating economic opportunities.” Danziger and Wheeler (1975:114) state: "According to the neoclassical view of crime, individuals calculate the returns from all possible activities-- illegal as well as legal-- and choose those that maximize expected utility.” This theory follows the contention of criminals as individual, rational decision makers "capable of scanning the set of occupational choices they feel are open to them and selecting that alternative which stands the highest in their own scale of preferences" (Krohm,l973:3l). Viscusi (mimeo:23) claims that individuals choose to participate in crime "if the expected net rewards from doing so are positive." Similarly, Ehrlich (1974:69) views ”the general theory of occupational choice by presenting the offender's decision problem as an optimal allocation of resources under uncertainty to competing activites both inside and outside the market sector, rather than as a choice between mutually exclusive activities.” Finally, Heineke (l978:2) claims: "In a sentence, the models of economic choice theory, of which the criminal choice is a special case, hypothesize that all individuals, 15 CHAPTER ONE: THE THEORIES criminal and non-criminal alike respond to incentives; and if the costs and benefits associated with an action change, the agent's choices are also likely to change." Freeman (1983:91) states: "Because crime is risky, attitude toward risk always plays an important role in the decision.” However, Block and Lind (1975:246) claim that criminal actions "are independent of attitudes toward risk.” In his discussion of expected utility theory, Markowitz (1952:157) refers to the "fun of participation" hypothesis, although he does not mention crime in its applicability, but rather as a function of behavior under uncertainty. Kau and Rubin (1975:69-70) take this a step further in their discussion on crimes against the person: "The model to explain non-economic crimes is more difficult to specify. The argument in the utility function is no longer measurable in money terms; nonetheless, costs of arrest and conviction would still affect crimes of this sort.” It has been pointed out that crime may increase one's ”psychic" income, benefits, returns, or gains (Ehrlich,l975:314; Avio and Clark,1976:27; Palmer,l977:5; Sjoquist,l973:439). Thus, “psychic” benefit may be especially pertinent in explaining violent crimes, as these generally result in no material gains, although the role it may play in the 16 CHAPTER ONE: THE THEORIES "gamble" of "getting caught” for property crimes is also feasible for "risk lovers."3 A primary focus in the economists’ individual choice model is the matter of time allocation (Avio and Clark,1976; Becher,l968:l76; Block and Lind,l975:242; Block and Heineke,l975:314; Ehrlich,l974; Krohm,l973:34; Sjoquist,l973:439; Witte,l980:59). This is basically viewed as the person's choice as to whether her or his time is more profitably spent in legal or illegal activities. However, as Ehrlich (1974:71) points out, this is not really an "either-or” case. That is, a combination of legal and illegal activities can be used at one time or be spread over a life time. A person's allocation of time-spending will depend on how "marketable” that person is in the legitimate market. For example, the assumption here is that a highly skilled, trained, or educated person would tend to find more benefits in the legitimate labor market, while the ex-convict, poorly trained, unskilled, and uneducated person would find more benefits in the illegitimate market, and therefore be more likely to (re)turn to 3Solberg (1982:457) claims: ”Not all persons are risk averse. A person is called a risk lover if the expected utility of a gamble is greater than the utility of an amount of money equal to the expected value of the gamble." 17 CHAPTER ONE: THE THEORIES crime. Just as expected increases in gains from criminal activity will likely result in increased criminal activity and increases in costs will reduce it, increases in the likelihood of failure or increases in the uncertainty surrounding the risks should decrease resources allocated to criminal activity (Heineke,1978:9). The costs of engaging in property crimes are dependent upon (1) the foregone income that could be obtained in a legitimate occupation, (2) the costs cf being apprehended and charged by police and of appearing in court, (3) the costs of serving a prison sentence, and (4) the reduction of legitimate opportunities for any- one with a criminal record (Avio and Clark, 1976:14). Orsagh (1983:394) believes that economists try to explain criminal behavior through "environmental stimuli mediated by individual preference." Traditional criminologists, on the other hand, place more emphasis on what economists call "tastes"-- attitudes, values, and personality. Stern (1978:139) claims that the economic arguments "singly or collectively, cannot provide a satisfactory answer and concludes that we have to search outside the economic models, simple or complicated." Future analysis requires both disciplines to examine the relationship of certain economic, demographic, and deterrence variables, with crime rates . 18 CHAPTER ONE: THE THEORIES F. Cr t eo Expected utility theory has been met with some cynicism and reluctance. Friedman and Savage (1948:282) state: "Whether a numerical utility function will in fact serve to rationalize any particular class of reactions to risk is an empirical question to be tested; there is no obvious contradiction such as was once thought to exist.” Similarly, Bernoulli (1954:24) originally put forth: ”However, it hardly seems plausible to make any precise gereralizations since the utility of an item may change with circumstances.” Savage (1954:98) insists that there do exist meaningful measures of utility. He claims: ”The confusion arises only because von Neumann and Morgenstern use the already pre-empted word 'utility' for what I here call the 'von Neumann- Morgenstern utility'", where ”every von Neumann-Morgenstern utility is an increasing function of every other” in order to satisfy their classical utility“ (1954:98). Von Neumann and Morgenstern (l964:9) believe that "there exists at present, no satisfactory treatment of the question of rational behavior. There, may, for example exist several ways by which to reach the optimum position..." In the same vein, Friedman and Savage (1948:292) believes the choices consumer units make involving risk are usually much more complicated than 19 CHAPTER ONE: THE THEORIES expected utility theory hypotheses portray. "There are two chief sources of complication: Any particular alternative typically offers an indefinitely large number of possible incomes, and 'other things' are generally not the same" (1948:292). Concerning the rational model of choice under uncertainty, Arrow (1958:4) states: "There are a number of possible states of nature and a number of possible actions we can take. We do not know which state of nature, among a certain class, is the true one." Kahneman and Tversky (1979:265) criticize expected utility theory in its weighting the utility of outcomes by their probabilities. Their findings refute this contention, in finding that "people overweight outcomes that are considered certain, relative to outcomes which are merely probable-- a phenomenon which we label the cettainty effect." However, I believe this supports the occurrence that most people, even the unemployed and poor, do ggt commit crimes. That is they feel more secure with a dependable small income than in taking the risk of being caught committing a crime. I suggest that there is some "adversity to risk" factor involved in addition to the poor-have-more-to-gain- and-less-to-lose argument in the commission of crimes. 20 CHAPTER ONE: THE THEORIES Kahneman and Tversky (1979:271) further criticize expected utility theory through a phenomenon they call thg_iggl§tign effett. This phenomenon claims that people focus on components distinguishing alternatives rather than what they share, thus resulting in inconsistent preferences ”because a pair of prospects can be decomposed into common and distinctive components in more than one way, and different decompositions sometimes lead to different preferences" (1979:271). Again, this does not necessarily impinge on the application of expected utility theory to criminal behavior, since people will differ in their adversity to risk concerning crime commission. Riker and Ordeshook (1973) defend the plausibility of rationality in expected utility theory. They believe that most researchers' findings of irrationality in decision making result from the researchers' own poor judgement in realizing the subjects' actual goals (1973:16). Riker and Ordeshook (1973) also defend against the focus on people being unable to maximize expected utility because they err. "Rather the question concerns the frequency of error. Is it so general and pervasive that the concept of rationality is useless as description because it is not parsimonious?" (1973:24). Instead, they claim that most error is the result of "misinformation and reasoned rejection of the cost or correction, not from simple stupidity" (1973:31). 21 CHAPTER ONE: THE THEORIES Similarly, Sullivan (1973:142) states: "The basic economic assumption does not maintain that people do not make mistakes, but rather that they do their best given their reading of present and future possibilities and given their resources." Markowitz (1952) criticizes the Friedman-Savage (1948) hypothesis elaborated on earlier. He believes their utility function implies that a "moderately wealthy person is willing to risk a large fraction of his wealth at actuarially unfair odds" which is probably highly unlikely (1952:153). Thus, he believes there needs to be an alternative utility analysis capable of explaining risk-taking among the rich and poor. He then proposes a utility function which distinguishes between what he calls "customary" and "present" wealth. To have an exact hypothesis-- the sort one finds in physics-- we should have to specify two things: (a) the conditions underwhich cus- tomary wealth is not equal to present wealth (i.e., the conditions referred to as recent windfall gains or losses) and (b) the value of customary wealth... when customary wealth is not equal to present wealth (1952:157). Finally, Markowitz (1952:157) claims that the 'fun of participation' (discussed in the last section) does not explain observable behavior. “It [the 'fun of participation' hypothesis] 22 CHAPTER ONE: THE THEORIES seems to say the utility of a gamble is the expected utility of the outcome plus the utility of playing the game (the latter utility is independent of the outcome of the game)" (1952:157). I do not believe this assertion of the ”fun of participation" hypothesis is inconsistent in the explanation of violent crimes, because in most cases interpersonal violence is not precipitated over acquiring an outcome such as property or money, but is more likely to be to gain a sensation, frequently that of power. The sensation of power, then, may be the ”fun of participation” associated with violent criminal acts. 69-min: This section has summarized expected utility theory regarding its definition, development, limitations, and applications to criminal behavior. Clearly more research needs to be conducted to test expected utility theory, especially as it pertains to the decision under risk, particularly that of criminal behavior. The following section will review opportunity theory, which can be viewed as a sociological counterpart to expected utility theory. 23 CHAPTER ONE: THE THEORIES III. OPPORTUNITY THEORY A. u ' ve o e o O ortu eo Studies by sociologists and criminologists on the relationship between economic factors and the crime rate are largely based on opportunity theory. Opportunity theory is a "strain theory.” Strain theory does not view delinquents as ”motivated by deviant values, but rather as by values of the "American Dream" held by all of American society.4 Therefore, lower-class youth violate the law over frustration at being unable to attain this dream (Empey,1982:233). Opportunity theory is most commonly identified as evolving from Cloward and Ohlin's book Dg1iggggnty_agg_gppgtttnity. Leonard (1982:122) claims that anomie theory suggested illegitimate opportunity was easily accessible or unimportant, Sutherland recognized the importance of illegitimate opportunity 4There will be a discussion in the literature review chapter that income distribution (relative poverty) appears to have a greater effect on crime rates than does the porportion of a population living in poverty (absolute poverty). Similarly, economists Block and Heineke (1973:383) claim: "[I]ncreases in the dispersion of the wage rate induce the individual to supply more effort in an attempt to compensate for the additional income uncertainty." 24 CHAPTER ONE: THE THEORIES but ignored legitimate, and Cloward and Ohlin combined the two. The objective of opportunity theory is to consolidate two traditional theories: (1) Durkheim's and Merton's anomie theory, and (2) Sutherland's and Shaw and McKay's "cultural transmission" and "differential association” theories (Cloward,l959:l64; Cloward and Ohlin,l960:151). "We can now look at the individual, not simply in relation to one or the other system of means, but in relation to both legitimate and illegitimate systems" (1960:151). Opportunity theory was developed by Cloward and Ohlin (1959,1960) as an explanation of delinquency of adolescent males in lower class sections of large urban areas. Empey (1978:290- 291) classifies this view of delinquents as more "rational and utilitarian" than most delinquency/criminological theories: "When channels for success are closed to them [adolescent, working-class, urban males], they simply turn to illegitimate ones if at all possible." In fact, Gibbons (1979:96-97) claims of this theory: "The raw material for delinquent gangs consisted of boys concerned about economic injustice rather than middle- class status.” These youths "internalize conventional goals, but are blocked by limited economic and educational opportunities" (Davis,l980:l4l). Opportunity theory has been expanded to attempt to explain the effects of unemployment, poverty, and income inequality on crime rates. 25 CHAPTER ONE: THE THEORIES An inherent theme in Cloward and Ohlin's (1960) opportunity theory, is the theme of a delinquent "subculture." Their theory claims that in addition to feeling the frustration of blocked legitimate access to culturally agreed upon goals, for some delinquent types to emerge, accessibility to illegitimate means of goal achievement may be necessary.5 Thus, they claim: "Illegitimate figures become no less obscure as role models to lower-class youth than the role-models of the legitimate world of finance and corporate enterprise" (1960:206). Cloward and Ohlin's use of such terms as 'opportunity structures', 'legitimate and il- legitimate means,’ 'role models,’ 'integration of age levels,’ etc. implies that somewhere within the institutional or informal social context of an adolescent's life there exist individuals who can offer or withhold keys to certain sectors of adult status... (Rivera and Short,1967:78). 5It is interesting to note that the expected utility theory literature also touched on this idea independently from opportunity theory. For instance, Friedman and Savage (1948:302) state: "Opportunities to assume risk may not exist." More vague, but still applicable is Von Neumann and Morgenstern's (l964:8) discussion of how there may be a number of actions capable of leading to the same desired goal, but the chosen action ”may depend upon the knowlege and understanding which the individual has and upon the paths of action open to him.” 26 CHAPTER ONE: THE THEORIES B-WW l. u ure Cloward and Ohlin's portrayal of the subculture has been questioned not only as to the extent in which it plays a major role in delinquency, but even in its existence. Matza (1961:632) criticizes Cloward and Ohlin for perceiving the ranges of access to both legitimate find illegitimate opportunities to adolescents in lower classes as much more limited than they actually are. He does not believe the organized rackets "virtually exhaust available criminal endeavors...Do not the great majority of criminals take the path of entrepreneurial petty crime?” (1961:632). Similarly, Bordua (1961:135) believes Cloward and Ohlin "may make too much of the necessity for systematic, organized criminal careers in order for the illegitimate opportunity structure to have an effect." Gibbons (1979:97) claims Cloward and Ohlin's definition of subculture "limits the applicability of their theory to a minority of delinquents. Critics contended that much gang delinquency in working class areas is more spontaneous and unstructured than acknowledged by Cloward and Ohlin." Kornhauser (1978:155) claims that Cloward and Ohlin's ”conception of working 27 CHAPTER ONE: THE THEORIES class subculture is so tortured as nearly to defy comprehension, and so bizarre as, when understood to defy belief.” Bordua (1961:133) believes Cloward and Ohlin confuse the "justifacatory function of delinquent subcultures with their causation." In short, there is a strong pressure here to- ward a simple isomorphism of culture content and problem of adjustment. The complex society of street groupings characteristic of slum life is given little or no culture creating potien- tial of its own. The group is simply the in- strument that translates the individual discon- tent into the collective solution (Bordua, 1962:255). Taylor et al. (1973:134) believe that Cloward and Ohlin do not grasp the diversity of subcultures in modern industrial societies. Toby (1961:283) suggests that in viewing delinquent behavior as required and a central group activity in delinquent subcultures, that Cloward and Ohlin fail to see that delinquency may emerge from personal pathology. "They also ignore delinquent repsonses to cultural or subcultural phenomena that do not involve a well-organized group with clear-cut prescriptions and interests" (1961:283). Empirical support for the existence of a delinquent subculture has been confirmed by some researchers (Vold,1958:l89; Moore et al.,l983; Weinber,l976; Yablonsky,l962; Thompson et al.,l984; Mooris,l965). However, Lermen (l967:7) warns against 28 CHAPTER ONE: THE THEORIES viewing ”gangs" and "subcultures” as synonymous, and claims we need to differentiate between "subculture" and "interaction.” Similarly, Morash (1983:328) found "negligible association between membership in a stereotypic gang and delinquency." Although Yablonsky (1962:143) found support for the existence of a subculture in the form of a gang, he believes Cloward and Ohlin "tend to minimize the importance of individual personality factors and characteristics." Rahav (1984:172) concluded that "age-sex norms are pervasive across the various sectors of the social structure, and differ only minimally among members of various social categories." There has also been research conducted on the role of adults and orgnization in subcultures as a means of creating legitimate or illegitimate opportunity, there has also been research conducted. Downes (1966:189) found that although there appeared to be a positive relationship between adult and youth crime, "the nature of the criminality suggests Skid Rows rather than professional crime...any organization is simply the minimum required for the commission of these offense types." For the most part, the studies found the role of significant adults in youths’ lives play an important part in influencing delinquency (Kobrin et al.,l967; Spergel,l964; Rivera and Short,1967). 29 CHAPTER ONE: THE THEORIES In summary, Cloward and Ohlin's belief in a delinquent subculture has been vigorously challenged both theoretically and empirically. Although some support has been found, the overlying belief is that this approach is too simplistic. For example it ignores individual personality traits, and the existence of upper- and middle-class delinquency, and the limited delinquency among girls in these deviant subcultures. However, the impact of adults with which youth have contact, do appear relevant in affecting delinquent behavior. Part of the logic behind opportunity theory rests on the assumption that all classes and walks of life hold the same values. Taylor et al. (1973:135) criticize opportunity theory's narrowness: "At no time is the cultural diversity of goals and means, and the multitude of graduations of acceptance and rejection of utilitarianism, existing in modern industrial societies, fully encompassed." They wonder how Cloward and Ohlin would have classified the Black Panthers and hippies in their typology. 30 CHAPTER ONE: THE THEORIES Empirical evidence overwhelmingly suggests that there is a common "American Dream" that cuts across social class, that there are universal goals to which everyone aspires, including delinquents ( Short et al.,l965; Gordon et al.,l963; Short and Strodtbeck,l965; Spergel,l963; Elliot,l962; Sherif and Sherif,l964; Lerman,l966). However, Thompson et al. (1984:20) found ”while relationships may appear to exist between attachments and conventional attitudes with delinquent acts, these relationships diminish to the point of being virtually non- existent when the effects of delinquent companions is controlled.” Perhaps the effect of common aspirations has been overemphasized in place of the influence of one's cohorts in research up to this point. 3. ce tu i c s The gap between aspirations and expectations that Cloward and Ohlin propose that the lower class experiences is in some cases hypothesized to lead to delinquency out of a sense of injustice or frustration. Bordua (1961:133) criticizes Cloward and Ohlin for "forcing" themselves into believing the feeling of injustice is objectively correct, thus injustice flourishes among those actually "fitted for success" without the legitimate means 31 CHAPTER ONE: THE THEORIES of achieving it. Hirshci (1969:184) claims that the commission of delinquency is unrelated to whther a boy blames himself or the social system for his potential failure. Bordua (1961:134) questions Cloward and Ohlin's belief in the gang boys actually possessing the requirements even without structural opportunity blocks. In fact, he seems to blame the delinquents, claiming they are often "the end product of a long history of their progressively cutting off opportunity and destroying their own capacities..." (1961:134). Matza (1961:632) criticizes Cloward and Ohlin for not seriously considering the actual range of legitimate opportunities open to lower class youth. Empirical support is evident for Cloward and Ohlin's contention of a perception of blocked opportunities by delinquents (Spergel,l963; Spergel,l964; Short et al.,l965; Short and Strodtbeck,l965; Elliot,l962; Cernkovich,l978; Cernkovich and Giordana,l979; Datesman et al.,l975; Segrave and Hastad; Sherif and Sherif,l978). Johnstone (1983:295) claims youths resisted gang recruitment as long as they believed they could achieve success in conventional manners. Hardman (1969:178) found little resentment or even awareness of opportunity deprivation, although it existed. Others agreed that occupational and middle-class aspirations did not pressure boys toward delinquency 32 CHAPTER ONE: THE THEORIES (Short,1964:115; Reiss and Rhodes,l96l:729; Hirschi,l969:l85; Simons et al.,l980:49). Thus the role of blocked opportunity does not appear entirely clear. The literature reviewed in this section suggests that perceptions of opportunities may work through other variables in affecting delinquency (Cernkovich,l978), and/or perceived blocked opportunity may cause stress, but this does not automatically lead to delinquency (Hardman,l969). At any rate, there is much to be learned concerning the role perception of blocked opportunities plays with delinquency. Kornhauser (1978:156) accuses Cloward and Ohlin of wrongly interpreting their data (finding higher income people more concerned with status and lower income pople more concerned with money) in a zero-sum manner. ”Now certainly it must be true that money is more eeliegt to those who have little of it...I think it is safe to assume that all people, of whatever rank, value most of all that which is a necessity for survival: money" (1978:156). Vold (1958), whose writings preceded Cloward and Ohlin's opportunity theory, was irritated with any preoccupatin of using economic variables to explain crime rates. ”If economic 33 CHAPTER ONE: THE THEORIES conditions are influences contributing to the structure and functioning of various social arrangements, including crime, it is equally true that the existing social and institutional structure molds and influences the nature of economic affairs. Economic conditions are then no longer 'determining influences'..." (1958:182). Bordua (1962:252) similarly criticizes Cloward and Ohlin's focus on class position "to the near exclusion of other social attributes" frequently associated with those in the lower economic strata such as rural migrants, immigrants, blacks, etc. Toby (1961:287) criticizes Cloward and Ohlin because "by assuming that feelings of eeenemie deprivation are of central motivational importance, they simplify their analysis of the impact of social structure on the alienated youngster." Furthermore, Vold (1958:172) claims the following in a discussion on middle-class and similar crimes: "If one assumes that there is a causal significance to the situation under which 30.7% of the population 'produces' 56% of the juvenile delinquents, one must be prepared to explain why the 'comfortable' and 'luxury' classes 'produce' the other 44%... It would be more logical to conclude that neither poverty nor wealth, as these are experienced in modern society, is a major 'determining' influence in crime and delinquency." Cloward and 34 CHAPTER ONE: THE THEORIES Ohlin's theory implies that the middle-class, in addition to having more legitimate opportunities to gain success, has less exposure to illegitimate means. I suspect that they have access to different methods of illegitimate means, or are able to "purchase" it from the lower class. Scott and Vaz (1963) believe the age of industry, post World War I, negatively affected the cohesiveness of the American family. "The adolescent conspicuously lacks an exact definition of the expectations and obligations attached to his role in society and he is left to define for himself what is 'right' conduct” (1963:327). Membership in a peer group implies the youth is socially adjusted (1963:330). They claim: "If we are correct, the opportuntities which exist for the middle-class adolescent to engage in legitimate, approved activities will greatly influence the probability of his becoming involved in illegitimate disapproved behavior" (1963:333). Matza (1961:632) questions the magnitude of class immobility that Cloward and Ohlin imply: "It seems important to investigate the relation between the delinquents' implicit model of the legitimate system of opportunities-- a closed system-- and the realities of social mobility in what many sociologists characterize as approximating an open system.” Hirshci (1969:226) is more direct: "Strain theories are class 35 CHAPTER ONE: THE THEORIES theories...The trouble for these theories thus begins with failure of social class to be importantly related to delinquency." Empirical research has supported an inverse relationship between social class and delinquency (Hardman,l969:l78; Segrave and Hastad,l983:450-451; Cernkovich,l978:344), however the reasoning and logic used to come to these findings I believe are frequently suspect. Others found no relationship, or a relationship too complex to pass judgement (Cernkovich and Giordana,l979:l48; Elliot,l962:226; Reiss and Rhodes,l96l:729). Tittle et al. (1978:653) analyzed 35 studies using individual- level data to compare class position and crime/delinquency, and found the overall mean association to be only -.09, and this relationship became increasingly weaker in more recent studies, through to studies since 1970 which found no relationship. They claim this may be because self-report data is more accurate and less class biased, or it may be that class has generally become less important (1978:654). Cloward and Ohlin have been criticized for their failure to explain middle-class delinquency. Chambliss (1973:29) found a community he studied reacted differently toward an upper-middle class white male gang than to a lower-class white male gang. Despite similar degrees in delinquency, the former were not seen 36 CHAPTER ONE: THE THEORIES as a problem, and the latter were. Thus, any association between delinquency and social class may be due to differential labeling. Cernkovich (1978) claims that structural theories (such as opportunity theory) have historically been applied to lower class delinquency, while control theories are applied to middle-class delinquency. Hirschi (1969:7-8), a control theorist and a popular critic of strain theory claims: If he ever feels called upon to explain middle-class delinquency, the strain theorist has two options: he can argue that apparently middle-class boys committing delinquent acts are 'really' lower class boys; or he can re- verse his original procedure and ask: What is it about the middle class situation that pro- duces middle-class delinquency? A review of evidence from empirical studies challenges Cloward and Ohlin's obsession with lower class delinquency. Opportunity theory fails in its inability to explain middle-class delinquency. Further research and theory building must proceed in attempting to determine if we must find separate theories for middle- and lower-class delinquency or if a single theory or combination of theories can depict delinquency etiology. 5. ac et Race and ethnicity are other variables which researchers have found problematic when considering Opportunity theory. 37 CHAPTER ONE: THE THEORIES Bordua (1962:252) states: "Cloward and Ohlin discuss racial and ethnic differences in the types of delinquent subcultures found in different neighborhoods but do not seem sensitive to the possibility that race, ethnicity or rural origin may be heavily involved in the 'problem of adjustment' to which the delinquent subcultures are solutions." Yablonsky (1962:184-185) points out: Prejudice first appears in the 'community' to force the family of the minority group gang member into the disorganized slum... Most of the newly arrived population are unskilled labor, and this too affects their economic and financial condition...on the more direct discriminatory level, minority groups, particularly Negroes and Puerto Ri- cans, do not have equal access or are simply barred from living in certain neighborhoods regardless of financial ability to pay their rent regularly. Cernkovich and Giordano (1979:150) found blocked opportunity more strongly associated with the delinquency involvement of whites than nonwhites, regardless of sex. Taylor et a1. (1973:134) state: "Negro adolescents may quickly become aware that it is their blackness, rather than than any individual inadequacy shared by their peer group, which prevents them from competing.” Yablonsky (1962:189) found "some individual victims of prejudice use the gang vehicle of violence as a direct weapon for 'getting even'." This brief discussion suggests opportunity 38 CHAPTER ONE: THE THEORIES theory needs to be examined more closely to determine the role that race and ethnicity play. Clearly, Cloward and Ohlin's opportunity theory was developed for the explanation of juvenile delinquency. More recently it has been generalized to understand adult criminality. Cloward and Ohlin's contention of the importance of the "subculture" has not been widely supported, however there is some evidence as to the importance of perceived blocked opportunities. Therefore, this aspece of opportunity theory, blocked opportunities, is now being applied to research on adult criminal behavior and earnings opportunities. IV. USING AGGREGATE DATA TO TEST THESE THEORIES Cloward and Ohlin's opportunity theory was clearly developed using and assuming individual-level data. However, more recent studies based on opportunity theory and expected utility theory, especially those concerned with the economics-crime link, have utilized aggregate level data. Thus, although opportunity theory was initially a theory of individual behavior, as the last section of this chapter indicated, it is not at all well supported. Some researchers borrowed the idea of opportunity-- separate from all of Cloward and Ohlin's ties to the 39 CHAPTER ONE: THE THEORIES "subculture"-- and used it to predict crime rates at the aggregate level. Due to the problems supporting opportunity theory in an inidividual-level analysis, there exists a great need to conduct more research on opportunity and crime at the aggregate level. Among others, Firebaugh (1978) and Hannan (1970) refute Robinson's (1950:357) contention that ecological correlations cannot "validly be used as substitutes for individual correlations." However, before elaborating on this it is important to review some of the limitations posed by the use of aggregate data. W.S. Robinson (1950) is believed to be the person who first recognized the problem of ecological fallacy.Babbie (1979:91) defines ecological fallacy as "the danger...of making assertions about one unit of analysis based on the examination of another." Robinson (1950) claims that in ecological correlations "[t]he variables are percentages, descriptive properties of groups, and not descriptive properties of individuals." Wagner (1964) refers to this as the homology thesis-- which is when researchers assume there is consistency across levels of analysis. Thus the problem comes in making insinuations about individuals from aggregate- level data, or vice versa. 40 CHAPTER ONE: THE THEORIES Lower status areas may have higher crime rates becuase a small proportion of people within those areas commit a lot of crimes or because outsiders come into these areas to do mischief. And, it is possible that people who live in low status areas and commit crimes are not necessarily themselves of low status, since the composition of urban units such as census tracts are often diverse (Tittle et al., 1978:644). Researchers since Robinson (1950) have argued that the use of aggregate-level data is not inherently poor methodology. Land and Felson (1976:566) claim that most researchers agree that social indicators, such as the unemployment rate, ”are measures of social conditions." Hannan (1970:23) points out that when individual measures are not available or realistic (and he uses delinquency research as an example) ”researchers often shift levels and employ tetee of certain behaviors." In fact, Riker and Ordeshook (1973:33-34) deal with this abstractly in their discussion of expected utility theory: "Individual perferences and the choices that reveal them have social consequences. Indeed these consequences together make up the soical environment and are in fact social institutions. As such they stand behind all individual action.” Furthermore, Hannan (1970:66) relates the aggregation problem to theory: 41 CHAPTER ONE: THE THEORIES But what of the situation when there is like- ly to be a substantively interesting causal re- lation between the aggregation critereon and one or more variables of interest? Without a theory specifying such relations, we would face an overwhelming range of choices of causal specifications of such relations in any fairly complex situation. This problem seems to piont to a real need in the face of aggregation- dissaggregation problems of etese-level theotiee. If we had such theories, we would be in a better position to uncover the various biasing aggre- gation problems discussed to this point. Better yet, we would be able to explicitly include the operative variable(s) under a surrogate variable (like region) in the model. Additionally, Mathieson and Passell (1976:83) believe: "Just as models of collective consumer behavior can have powerful analytic uses without benefit of a complete psychological theory of consumption, so too can testable models of 'rational' criminal behavior fruitfally bypass the fundamental psychological origins of crime." By using geographical areas as a units of analysis (as opposed to individual people), and then using percentages, rates or propensities as the variables, we are attempting to determine the degree to which "social indicators" (such as the unemployment rate, the percent poor, the police expenditure per capita, etc.) are related to social conditions such as suicide, mental illness, or in this case, the crime rate. We cannot assume if the findings are significant in the direction expected 42 CHAPTER ONE: THE THEORIES that, for instance, unemployed peeele commit more crimes, simply because we find the unemployment rate positively and significantly related to the crime rate. However, we can claim that our findings suggest that eteee with higher unemployment rates are more prone to crime, given what we know and allowing for what we have controlled. Of course, measuring the structure of dominant/subordinate relations within popu- lations requires research at the aggregate level, meaning that populations are the appropriate units of analysis, not individ- ual persons. Interestingly enough, aggre- gate empirical examinations of conflict criminology have been conspicuously rare, even though such examinations promise to be quite instructive (Williams and Drake, 1980:564). Babbie (1979:92) claims: "The appropriate unit of analysis for a given research question is not always clear, and it is often debated by social scientists, especially across disciplinary boundaries." In his conclusion of a review of delinquency and opportunity structures, Gibbons (1979:98) claims: The major thrust of this commentary has been that real-life social structure in a society such as the United States is ex- ceedingly complex, involving interwoven layers of social variables that are com- bined in varied ways and that produce be- havioral outcomes such as delinquency. In short, theories of gang delinquency are not yet rich enough or elaborate enough to 43 CHAPTER ONE: THE THEORIES encompass the varieties of real-life exper- ience. In summary, it is time that Cloward and Ohlin's theory be expanded to include aggregate levels of analysis, since findings at the individual level have not frequently substantiated their opportunity theory. By using geographical areas as the units of analysis, we can use opportunity theory as a basis for hypothesizing that eteee' unemployment and poverty rates will be negatively related to their crime rates, and the more equally distributed the income in an area, the lower the crime rates, cetetis petihee. Therefore, it now becomes pertinent to test which sorts of social conditions affect the level of crime rates. CHAPTER TWO: THE LITERATURE REVIEW I. THE INTRODUCTION The literature researching the link between economic factors and crime rates varies widely in the level of analysis (e.g., individual, city, nation, state, etc.) and the choice of predictor and dependent variables. The present research uses the most comprehensive combination of variables of which the author is aware, and focuses on the SMSAl, as the unit of analysis. The limited number of variables in prior research is partially due to the fact that economists and sociologists/criminologists tend to "stick to" their respective fields when conducting literature reviews for their research, whereas this research has combined both disciplines. This research includes as dependent variables the composite crime rates, as well as rates for the individual index crimes (homicide, rape, aggravated assault, robbery, 1 "The general concept of a standard metropolitan statistical area (SMSA)... is one of a large population nucleus together with adjacent communities that have a high degree of economic and social integration with that nucleus...The SMSAs are designated and defined by the Office of Information and Regulatory Affairs, U.S. Office of Managment and Budget (OMB), which follows a set of official published standards developed by the interagency Federal Committee on Standard Metropolitan Statistical Areas" (U.S. Bureau of Census, State and Metropolitan Data Book, 1982: xv). 45 CHAPTER TWO: THE LITERATURE REVIEW larceny, burglary, and auto theft), and both composite property and violent crime rates. The predictor variables can be grouped into three major categories: economic, demographic, and deterrence variables. The literature review consists of sections describing previous findings for each of the variables. Because the studies varied in the variables chosen for analysis, Appendix 1 shows which variables were included in which studies. The economic variables consist of income levels and distribution and unemployment and labor force participation rates. The income level variables include mean or median family income, and relative and absolute poverty.2 The demographic variables comprise racial distribution, geographical region, educational level, population size and density, urban-rural distribution, and age structure. The deterrence variables include the police expenditure and police employment per capita rates, and the certainty and severity of punishment. 2 Absolute poverty implies the proportion of peOple living at a subsistence or below level, or rather, the percent of people living in poverty. Relative poverty implies the distribution of income, or rather, income inequality. 46 CHAPTER TWO: THE LITERATURE REVIEW II. THE ECONOMIC VARIABLES This section will discuss the major economic variables thought to influence crime rates: absolute (subsistence) and relative (income inequality) poverty, mean or median income levels, and unemployment and labor force participation rates. The measurement of these variables has been widely disputed and thus a lengthy discussion is necessary. "No one would gainsay that there is some correlation between the labor market and crime, but the strength and magnitude of the link are more subtle and difficult to determine than one might expect" (Freeman, 1983:90). We might expect this contention to hold for the effect of poverty and income inequality as well. A. Income Levele and Distributioe It has been frequently asserted that decreasing a person's income, especially below the poverty line, should increase the propensity for that person to turn to crime. However, analyses of the poverty level have led researchers to believe that the degree of absolute poverty may not be nearly as threatening to 47 CHAPTER TWO: THE LITERATURE REVIEW crime rates as the degree of relative poverty. This implies that income inequality may be more influential on the crime rate than poverty, per se. Chester (1976) claims we cannot just focus on poverty, but rather we must look at "perceived relative deprivation.“ This should also enhance our understanding of middle-class and white- collar crimes. Danziger and Wheeler (1975:119) believe "the potential criminal is concerned not only with his own income, but with how this income compares to that of his reference group, and to the economic structure of the society, as expressed through the relative distribution of income." This, then, is linked to strain theory, and the concern becomes lagging rather than lower incomes (Messner,l982:104). These studies on poverty and income inequality have focused on units of analysis including individual data, and data aggregated by police districts, cities, SMSAs, states (or provinces), and nation-states. Discussions of each of the variables will be organized by units of analysis. 1. v d - eve ata Viscusi's (mimeo) individual-level analysis of self-reported data on black youths found: "The differential earning from crime and legitimate occupations are a significant, but by no means 48 CHAPTER TWO: THE LITERATURE REVIEW dominate determinant of criminal behavior." Witte's (1980) study used individual-level data to study arrest and conviction frequencies per month of released prisoners. She found that the legal wage had a significant, negative relationship with these frequencies, and the initial wealth variable had a negative (but insignificant) relationship. Thus, the individual level data tend to support the hypothesis. 2. at d a Most studies analyzing the link between income variables and crime rates from a national perspective have shown only partial or no support for the hypotheses concerning income levels and crime rates. Land and Felson (1976) conducted a time series analysis of U.S. national statistics from 1947 to 1972. They found that inflation (measured by the consumer price index) had the predicted positive, significant relationship with property crime rates, but a negative, insignificant relationship with violent crime rates. However, inconsistent with expected utility and opportunity theories, prosperity (GNP per capita) had a significant positive relationship with property crime rates. (These effects may be spurious: crime, inflation, and GNP per capita could have all gradually risen over time, but perhaps all for different reasons.) 49 CHAPTER TWO: THE LITERATURE REVIEW Similarly, other studies had less than conclusive findings concerning the income-crime link. Regarding international differences, Krohn (1976) examined homicide, property and total crime rates in nation-states. He measured income inequality with the Gini index3 and found it to be positively correlated with homicide, and negatively correlated with property and total crime rates. To measure economic development he used GNP per capita, and found it negatively correlated with homicide, and positively correlated with property and composite crime rates. Messner's (1980) analysis of 39 nations found only income inequality had a significant relationship with murder in the zero-order correlation, while GDP4 per capita was positive, but insignificant. Some studies have provided no support for this relationship between income variables and crime rates at the national level. Stack (1984) looked at 62 nations' property crime rates and 3 The Gini index is generally computed using the Lorenz Curve (Ray and Singer,l973), which consists of two axes representing (1) cumulative proportion of income, and (2) cumulative proportion of the population. "We may find, for example, that the poorest .25 of the population earns only .05 of the nation's income" (Allison,l978:872). A high Gini coefficient implies a high degree of income inequality. 4Gross Domestic Product (GDP) is similar to Gross National Product except that it includes only what is produced within the physical boundaries of a country, regardless of who owns them. 50 CHAPTER TWO: THE LITERATURE REVIEW measured inequality by (l) the Gini index of household income, and (2) the ratio of income share of the top decile to the bottom decile of households. He measured "economic development” as the GNP per capita. Surprisingly, he found the relationships between these variables and property crime rates to be in the opposite direction to what was hypothesized. He suggests that this may be because low development may cause heth high inequality and low property crime rates, thus ”making the development-property crime relationship spurious" (1984:241). Wellford's (1974a) cross- national analysis found that GNP per capita best explained the crime rate fluctuations (compared to his other "dimensions of nations" variables: population and political orientation), and this relationship was (again) poeitive, except for homicide and drug offenses. However, Wellford (l974a:5-7) claimed that none of his "dimensions of nations' variables "account for much of the variation in reported crime." 3. st e v n d ta Data collected at the state/province level was generally supportive of the expected relationship between income levels and crime rates. Avio and Clark (1976) studied Canadian provinces, and measured income distribution as the percent of families 51 CHAPTER TWO: THE LITERATURE REVIEW earning less than one-half of the median family income. The elasticityS estimates for this variable were generally positive and significant and more than one for theft, fraud, and breaking and entering, and insignificant for robbery. Ehrlich's (1974) analysis of states found that the percentage of families below half the median income had positive relationships with property crime rates. Ehrlich's (1975b) comparison of states measured income inequality as the percentage of families below the state's median and the median family income. These varied positively and significantly with the rates of robbery, burglary, larceny and auto theft in all regressions and for all census years. Loftin and Hill's (1974) analysis of states' homicide rates found the Structural Poverty Index6 was the best predictor of homicide rates although the Gini index was also a positive and significant predictor. Median income did not predict as well as the Gini Index (1974:720). Although Smith and Parker's (1980) analysis of states and two types of homicide rates found the 5 Elasticities represent the percent change in the crime rate divided by the percent change in income. An elasticity of greater than one, thus implies the percent increase in the crime rate is more than the percent decrease in income. 6 The Structural Poverty Index is based on infant mortality, literacy, absolute poverty, and educational rates, and the percent of children living with one parent and the percent of the population who failed the armed forces mental test. 52 CHAPTER TWO: THE LITERATURE REVIEW effect of the Gini coefficient was negative and insignificant for each category, the Structural Poverty Index was positive for all, and significant for composite homicide rate, and primary (non- stranger) homicide rates. Thus, most studies using individual states as the unit of analysis have found the predicted relationships between the income variables and crime rates. Some studies of metropolitan areas were overall, unsupportive of any of the income variables in their predicted effect on the crime rates. Sjoquist's (1973) analysis of 53 cities found absolute poverty tended to have a negetiye, insignificant effect on the composite property crime (robbery, larceny, and burglary) rate. Income had a poeitiye, insignificant effect. Thus, both of these "wealth" indicators were in the opposite direction to what was hypothesized. Messner (1982) looked at homicide rates in 204 SMSAs and measured relative poverty with the Gini coefficient of family income, and absolute poverty by the percent of the population below the Social Security Administration poverty line, and the percent below $1,000. Beth measures of absolute poverty were negetiyely and eignifieently related to the homicide rate. However, he found 53 CHAPTER TWO: THE LITERATURE REVIEW relative poverty to have an insignificant, positive coefficient when demographic variables were controlled. However, many studies of metropolitan areas have generally supported the link between some of the income variables and crime, while discounting others, within the same study. Blau and Blau (1982) looked at the 125 largest SMSAs to examine the composite violent crimes, as well as individual murder, rape, robbery, and assault rates. Absolute poverty was measured by the percent of the population below the Social Security Administration poverty index, and was positively (but insignificantly) related to rape and assault, but negetiyely related to the composite violent and robbery rates (significant for robbery). Income inequality was measured with the Gini coefficient for family income, and was (as predicted) consistently positively related to the crime rates, and signficantly so to murder and assault. Socioeconomic inequality between races was measured with Duncan's SEI7 scores, and was positively and significantly related to each of the crime rates. Additionally, relative deprivation produced more criminal violence than absolute poverty. (The coefficients for income inequality were much higher than those for poverty.) 7 Duncan's SEI (socioeconomic inequality) is used to measure the diferences in average socioeconomic status between whites and non-whites. 54 CHAPTER TWO: THE LITERATURE REVIEW Another study lacking clear-cut evidence for the income- crime connection, Jacobs (1981) looked at theft rates (burglary, grand larceny, and robbery) in SMSAs. He measured economic inequality using the Gini coefficients on family incomes, absolute poverty using the percent of families below the poverty line, and economic development using median income. Economic inequality was an especially strong and significant determinant for burglary and larceny, and not as strong, but still positive for robbery. However, median income had a consistently peeitiye, significant relationship with the theft rates, but they were very weak. Although this finding is in opposition to expected utility and opportunity theories, Jacobs (1981:23) contends that these relationships support the argument that thefts are more likely to occur in affluent communities where there is more to steal. There was 39 relationship between absolute poverty measures and the theft rates. Another study generally supportive, but not consistently so, to the income-crime relationship, Danziger and Wheeler (1975) conducted both time series analysis (from 1949 to 1970) and cross sectional analysis (on 57 SMSAs) for burglary, aggravated assaults, and robbery rates in the U.S. The time series analysis found both absolute and relative poverty had significant, positive relationships with the crime rates. However, while the 55 CHAPTER TWO: THE LITERATURE REVIEW cross-sectional analysis found income inequality significantly and positively related to the crime rates, the coefficient for absolute poverty was insignificant and negetiye. Other studies tended to find the effect was in the predicted direction, but was not significant. Bechdolt (1975) analyzed property and violent crime rates comparing census tracts in Los Angeles, and comparing police districts in Chicago. The results for Los Angeles found median family income as the only variable which was not significantly (although it was negatively) related to both property and violent crimes. In Chicago it was not found to be significant either. He claims: ”[G]iven unemployment, it does not matter how high or low the income of an area is in determining the rate of property crime; what does matter is the rate of unemployment” (1975:138). Allison's (1972) analysis of Chicago and its surrounding communities did not find per capita income to have a significant relationship with the crime rate. Fleisher (1966:130) analyzed 101 cities, Chicago communities, and Cook County suburbs and found ”there is a non-negligible negative effect of the income dispersion, or supply variable.“ However, Weicher (1970) replicated Fleisher's (1966) study replacing his definition of broken family and found the impact of income distribution less significant. 56 CHAPTER TWO: THE LITERATURE REVIEW Bailey (1984) replicated and extended Messner's (1982) research, claiming cities were better units of analysis than SMSAs to study homicide rates. His analysis was for 1970, and comparisons with 1950 and 1960 were made. His economic deprivation variables were absolute deprivation (percent of the population below the poverty level), relative deprivation (the Gini index) and low income (percent of families with an annual income of less than $1,000). His study consistently found absolute poverty and homicide rates were positively and significantly associated (unlike Messner), but found no evidence that relative deprivation contributed significantly to homicide rates (like Messner). In fact, he believes the slight and insignificant relationship between relative deprivation and homicide is merely a function of the association between income inequality and poverty (1984:544). In contrast to the review above, many metropolitan area studies have provided more clear-cut support of the relationship between income variables and the crime rates in metropolitan areas. DeFronzo (1983) used 39 SMSAs to replicate Messner's study by controlling for the cost of living and governmental assistance. He employed the following economic variables: (1) cost of living adjusted monthly AFDC per family member, (2) cost of living adjusted percent of families living in poverty, (3) the 57 CHAPTER TWO: THE LITERATURE REVIEW income inequality index, (4) cost of living adjusted median family income, and (5) the Household Activity Index.8 Both adjusted absolute poverty egg income inequality were positively related to all seven index crime rates, and these relationships were significant for burglary and larceny. The adjusted AFDC assistance per aided family member was negatively related to homicide, rape, burglary, aggravated assault, and auto theft rates, and the former three were significant. Thus, they all tended to be in the expected direction. Swimmer's (1974a) simultaneous equations9 approach, where the dependent variables were the logarithms of the seven index crime rates and police expenditurelo, found the median income variable was significantly related to the demand for police 8The Household Activity Index is proposed to measure high- risk victimization households, using the time activities are spent away from the home and/or the ownership of goods likely to be stolen. 9The use of the simultaneous equtions (or systems) model has become increasingly popular over the last decade in order to describe the relationship between crime and economic factors. This is largely due to (l) the increased awareness of econometric models, (2) the realization that deterrence variables probably affect the crime rate and vice versa, and (3) that similar independent variables affect both the crime rates and the deterrence variables. 10 Logarithms of variables are used to either adjust for the functional form or to decrease the range of values of a variable. 58 CHAPTER TWO: THE LITERATURE REVIEW expenditure. The percent of rich and poor families was positively related to all index crimes except rape, and was significantly related to all of the property crimes except larceny, but not to any of the violent crimes. Williams and Drake (1980) viewed the arrest rates for the seven major felonies in SMSAs of 500,000 and larger populations. They measured economic inequality using the Gini index, and found it significantly and positively related to all of the crime rates except auto theft. McCarthy et al. (1975:777) found: ”The relationship between percentage poor and homicide is strong outside the South, although the relationship between percentage poor and assault is weaker outside of the South and negative in the South.” Humphries and Wallace (1980) found effects of the Central City Hardship Index11 had a positive relationship with personal and property crime rates. 11The Central City Hardship Index (CCHI) was developed by the Brookings Institute and measures social and economic disadvantage using unemployment, age structure and dependency, educational level, income level adjusted regionally, crowded housing and poverty. 59 CHAPTER TWO: THE LITERATURE REVIEW 5. co u on Studies focusing on nation-states as the unit of analysis frequently included GNP 2e; eapite as a measure of income level or economic development. This variables was occasionally related to the property crime rate in the predicted (negative) direction (Land and Felson,l976), and was consistently related to the homicide rate as predicted (Krohn,l976; Messner,l980; Wellford,l974a). However, GNP per capita tended to have the opposite effect as predicted by expected utility and opportunity theories on property and composite crime rates (Krohn,l976; Stack,l984; Wellford,l974a). A variable comparable to GNP per capita is the meee 0t medien inceme. This variable was consistently insignficant, and was negative (as prededicted) in some studies (Bechdolt,l975; Allison,l972), but positive in others (Sjoquist,l973; Jacobs,l98l). The type of crime did not appear to affect the findings. An important possibility concerning these findings is that the relationship between (especially property) crime rates and income levels may be expected to be peeitiye in that affluent areas may be likely crime targets. 60 CHAPTER TWO: THE LITERATURE REVIEW Studies employing peyetty as a variable tended to find it positively related to the crime rate (Avio and Clark,1976; Ehrlich,l974; Ehrlich,l975b; Blau and Blau,l982; Deonzo,l983; Bailey,l984; Danziger and Wheeler,1975; McCarthy et al.,l975; Humphries and Wallace,l980). However, some studies found poverty had the opposite (negative) effect as predicted (Blau and Blau,l982; Messner,l982; Jacobs,l981; Danziger and Wheeler,l975: Sjoquist,l973). The unit of analysis did not appear to influence the findings, except that research using states consistently found poverty affected the crime rates in the predicted direction (Avio and Clark,1976; Ehrlich,l974; Ehrilch,l975b; Smith and Parker,l980: Loftin and Hill,l974). All three of the studies utilizing a simultaneous equation which included poverty found it to be a significant variable in the predicted direction (Avio and Clark,1976; Ehrlich,l974; Ehrlich,l975b). Although most studies found poverty affected the crime rate in the predicted direction, there were no tendencies for certain crimes to be more affected by poverty than others. Similar to poverty, most studies using ineeme_inegge11ty as a predictor variable found it affected the crime rate in the predicted (positive) direction (Krohn,l976; Messner,l980; Loftin and Hill,l974; Blau and Blau,l982; Messner,l982; DeFronzo,l983; Bailey,l984; Jacobs,l981; Danziger and Wheeler,l975; 61 CHAPTER TWO: THE LITERATURE REVIEW Swimmer,l974a; Williams and Drake,l980; Fleisher,l966). However, some research found income distribution affected the crime rates in the opposite direction (Krohn,l976; Stack,l984; Smith and Parker,l980). It is interesting to note that the findings on income inequality are less ambiguous and more likely to be in the predicted direction, the less encompassing the unit of analysis. That is, the findings for states and nations were inconsistent, while metropolitan areas consistently found income inequality to affect the crime rates in the predicted direction. This suggests that areas larger than states may be inappropriate to test the models. The sophistication of the analysis did not appear to influence the findings on income inequality. KW Previous studies have generally found that “rises in unemployment and/or declines in labor participation rates are connected with rises in the crime rate, but... the effect tends to be modest and insufficient to explain the general upward trend of crime" (Freeman,l982:96). Freeman's (1983) literature review suggests that labor force participation is linked more closely with the crime rate than unemployment, implying that those who leave the labor force are the most crime prone. Phillips et al. 62 CHAPTER TWO: THE LITERATURE REVIEW (1972:493) believe that one advantage of labor force participation data over unemployment data is that "while unemployment rates reflect cyclical and short run conditions in the labor market, participation rates capture secular changes, including the influence of past unemployment rates." Consistent with opportunity and expected utility theories, then, we hypothesize that unemployment will be positively related to crime rates, and labor force participation rates will be negatively related to crime rates. The remainder of this section will discuss empirical research testing this hypothesis. One issue that has been discussed is that of how age affects the relationship between unemployment and crime. More specifically, the question is why there is less empirical support for the unempoloyment-delinquency link, than for the unemployment and adult crime rate link. Gibbs (1975) accounts for this difference in the direction of age groups' relationships between unemployment and crime as being due to "status integration.” (He also uses this concept to explain low crime rates among older/retired persons.) "Status integration refers to the degree to which occupancy in a population conforms to a particular pattern" (1975:971). Gibbs (1975:98) thus hypothesizes that "unemployment in an age group varies inversely over time with the property crime rate to the extent that members of the age group 63 CHAPTER TWO: THE LITERATURE REVIEW are not employed."12 Gibbs (1975:100) also uses ”status integration" in an effort to explain the low rates of female criminality: "Further, since the labor force status of women varies a great deal from place to place and time to time, there is a basis for anticipating a differential influence of unemployment on the female crime rate." The majority of the individual-level research supports the hypothesis that unemployment and crime are related. Thornberry and Christenson (1984) studied 567 of Wolfgang's cohort subjects and predicted a reciprocal effect interaction between unemployment and crime rates. They found unemployment exerts rather immediate effects on criminality, while criminality exerts a more long-range effect on unemployment (1984:405). There were significant effects from unemployment to crime rates and from crime to unemployment rates (1984:408). Also, the model fits better for the more socially disadvantaged (e.g., ex-delinquents, 12 In the same vein, Phillips et a1. (1972:493) suggests that labor force participation rates are especially advantageous over unem-ployment rates regarding youth "since youth have low participa-tion rates, unemployment will have less weight because of the considerable fraction of youth outside the labor market." 64 CHAPTER TWO: THE LITERATURE REVIEW blacks, and persons whose fathers were blue- rather than white- collar workers). Other research supporting the unemployment-crime link included Krohm (1973:33), who analyzed burglars in Chicago and found "...unemployment is no small threat to persons having the sociodemographic characteristics of the typical burglar." He believes the income incentives to commit burglary are especially large for youths because legitimate opportunities are positively related to age, thus there is less reason for older persons to choose crime as an occupation (1973:34). Viscusi (mimeoz32) used individual-level data for his study on black youths, and found that "one is less likely to engage in crime if one is currently employed." Furthermore, those who were not employed or in school "were much more strongly driven by the economic incentive to commit crimes” (Viscusi,mimeo:6l). Cobb's (1973) analysis of thieves also found frequent periods of unemployment. However, some individual-level research was less supportive of the unemployment-crime link. Witte (1980) looked at 641 men released from prison and their post-release arrest and conviction rates. She measured "expected unemployment" as the number of months between the offender's release and his finding his first job. Expected unemployment had a negative but insignificant effect on both dependent variables. It implies that a longer job 65 CHAPTER TWO: THE LITERATURE REVIEW search leads to a decreased crime rate. Schmidt and Witte (1984) analyzed these same ex-inmates looking at (a) serious consumption crimes, (b) serious income offenses, (c) nonserious income offenses, and (d) other crimes. They found the expected unemployment rate was positively but insignificantly related to serious income offenses, negatively and insignificantly related to serious consumption crimes and other crimes, and negatively and significantly related to nonserious income offenses. They believe their "expected unemployment rate" variable may actually be measuring ”legitimate job search time" instead of the expected unemployment rate. Further, Holzman's (l982)research suggests that unemployment may net be related to crime He looked at 29,474 U.S. males with at least two convictions for robbery and/or larceny, and had some interesting findings and speculations regarding ex-inmates and unemployment and recidivism: The results of the present study suggest that known recidivists in robbery and/or burglary are very likely to be legitimately employed while continuing to pursue their criminal ca- reers. Furthermore, when one examines the oc- cupational status of these recidivists in light of the literature on moonlighting, it seems reasonable to view their continued involvement in property crime as secondary employment of an entrepreneurial genre (1982:1971). 66 CHAPTER TWO: THE LITERATURE REVIEW Overall, the individual-level data support the unemployment- crime relationship. The only studies (Schmidt and Witte,l984 and Witte,l980) which did not find support for this relationship questioned their measure of expected unemployment. 2. e o t a d data Many studies supported a significant, positive relationship between unemployment and crime. Sjoquist (1973) analyzed 53 municipalities and found the labor force unemployment rate was positively and significantly related to the property crime rate. Using a simultaneous model, Mathur (1978) looked at U.S. cities with populations over 100,000 in 1960 and 1970, and found unemployment positively and significantly related to buglary, larceny, and auto theft in 1960, and larceny in 1970. Unemployment was occaisionally negatively related to robbery, but not significantly so. Kau and Rubin (1975) analyzed sixty U.S. cities in 1960 and 1970 with a simultaneous model and found the unemployment rate was positively related to both violent and property crime rates in both years, and this relationship was significant for violent crime rates in 1960. Jacobs (1981) looked at 195 SMSAs and found the male unemployment rate had a positive, significant effect on burglary, grand larceny, and 67 CHAPTER TWO: THE LITERATURE REVIEW robbery, although the effect was not strong. DeFronzo's (1983) analysis of 39 SMSAs on the seven index crime rates found the unemployment rate was positively related for all the crimes, and significant for rape, burglary, and larceny. Other studies also supported the predicted effect of unemployment on crime. Bechdolt (1975) analyzed Los Angeles census tracts and Chicago police districts and both property and violent crime rate composites. For Los Angeles, male unemployment rate (as a percent of the labor force) had the most effect on both property and violent crime rates of all the independent variables. (It was a positive, significant relationship.) For the Chicago data, unemployment was the only significant variable and by far the most important. In fact, Bechdolt (1975:138) claims that "...given unemployment, it does not matter how high or low the income of an area is in determining the rate of property crimes; what matters is the rate of unemployment. Further showing a relationship of unemployment and crime, Phillips and Votey (1975) used a simultaneous systems approach to study fifty California counties, and found a variable comprising unemployment, labor force participation, and unemployment rates for males 18-24, and density had a significant positive effect on the crime rate. In fact ”...measures of economic conditions are 68 CHAPTER TWO: THE LITERATURE REVIEW more important than purely demographic measures" (1975:374). Chapman (1976) focused on 147 California cities (using a simultaneous approach) in 1960 and 1970 looking at arrest rates. He found the male labor force participation rate was negatively and significantly related to both property and violent crime rates, and the female labor force participation rate was positively and significantly related to property crime rates. A variable representing men not in the labor force or school was positively and significantly related to the property crime rate. Most metropolitan area and county studies on unemployment and crime found the predicted relationship, although these were not always significant. Using step-wise regression13, Allison (1972) analyzed communities within forty miles of Chicago using 14 independent variables and found unemployment was the variable most related to the composite crime rate, and in fact, claimed unemployment explained 57% of the variation in the crime rate. Similarly, Wellford (1974) researched 21 of the largest U.S. urban centers and combined the independent variables into 13 Step-wise regression ”picks” the best proxy variables, or the one(s) best representing the other independent variables,thus it is affected by the degree of multicollinearity the "chosen variables" have with the other independent variables. In addition to the possibility of ”picking" the wrong variables, it can bias the relationship estimates. Therefore, the meaningfulness of findings using this method are questionable. 69 CHAPTER TWO: THE LITERATURE REVIEW subgroups representing (1) demographics, (2) socioeconomic characteristics, and (3) crime control. He found the socioencomic variables (including male unemployment rate and male nonwhite unemployment rate) explained much more than social control (1974:200), and "a significant amount of the increase in the rate of crime is associated with changes in the demographic and socioeconomic characteristics of the society" (1974:207). Many studies found the predicted relationship between crime and unemployment, but this relationship was not significant. Wilson and Boland (1978) focused on 35 large U.S. cities (using simultaneous systems) to look at robbery rates and found unemployment was postively yet insignificantly related to robbery. Williams and Drake's (1980) anlaysis of 24 SMSAs found the unemployment rate was positively related to all seven major felonies, but none of the relationships were significant. Kvalseth's (1977) step-wise regression of 79 census tracts in Atlanta using six crime rates and twenty independent variables found the effects of unemployment were "inconclusive" except that male unemployment had a significant, positive relationship with robbery and rape, and female unemployment had a signficant, positive influence on rape. Similarly, Swimmer's (1974a) simultaneous equations studying cities found the unemployment rate postively related with the seven index crime rates, except 70 CHAPTER TWO: THE LITERATURE REVIEW rape (which was negatively related), and all were insignificant. He claimed the unemployment rate was "the most disappointing independent variable" (l974a2625). In a similar study on property and violent crime rates he claimed that unemployment had a positive, insignificant effect on both property and violent crime rates, and unemployment and schooling rates had the least explanatory power of all of the variables (l974b:309). However, several studies have provided little or no support for a relationship between unemployment and crime. Fleisher (1966) focused on 101 cities, Chicago communities, and Cook County suburbs and found no evidence of unemployment having a greater effect in the higher than the lower delinquency subgroups, as did the effect of income inequality. Weicher (1970) looked at 74 Chicago communities and found the male unemployment rate was negatively and significantly related to the delinquency rate. Carr-Hill and Stern (1973) analyzed British police districts using a simultaneous systems model and found unemployment unrelated to indictable offenses. Further studies discounting the predicted relationship between unemployment and crime include Spector's (1975) analysis of 103 SMSAs. He found unemployment was negatively, but insignificantly related to violent crime rates. Danziger and Wheeler (1975) conducted a cross-sectional analysis of 57 SMSAs, 71 CHAPTER TWO: THE LITERATURE REVIEW and found male unemployment rate was negatively, but insignificantly related to all three crime rates (burglary, robbery, and aggravated assault). Pogue (1975:33) looked at 163 SMSAs for three years using an simultaneous approach, analyzing the seven index crimes and found that unemployment had ”little systematic impact on the crime rates.” Both the unemployment rate variable and the change in unemployment had mixed, insignificant effects. 1W Some studies on unemployment and crime focusing on the state/province level, were supportive of the predicited relationship. Avio and Clark (1976) studied eight Canadian provinces using a simultaneous model and found the overall unemployment rate had a significant, positive relationship with both measures of theft and breaking and entering, as did unemployment of young males. The overall labor force participation rate was negatively and significantly related for some thefts, and the young male labor force participation rate was significantly and negatively related to robbery. 72 CHAPTER TWO: THE LITERATURE REVIEW Most studies focusing on the state as the unit of analysis did not show any clear-cut support for the effect of unemployment on the crime rate. Hemley and McPheters (1974) analyzed 32 states, focusing on robbery, burglary, and larceny rates. They found the percent employed in manufacturing had a negative and significant relationship with burglary and larceny, and a positive and insignificant relationship with robbery. The state unemployment rate had a negative relationship with robbery, and positive with burglary and larceny, and none of these were significant. The metropolitan unemployment rate had a negative, insignificant relationship with robbery, and a positive significant relationship with burglary and larceny. Another example of the unclear support of the unemployment- crime relationship was Ehrlich's (1974) analysis of states using a simultaneous approach, focusing on property and violent crime rates and certainty of punishment and police expenditure. He found the male unemployment rate for 14 to 24 year olds inconsistently and insignificantly related. Although the effects of the labor force participation rate on the same age group were inconlusive on the property crime rate, labor force participation had a consistently negative relationship with the violent crime rate (1974:105-6). 73 CHAPTER TWO: THE LITERATURE REVIEW knew The first major attempt to relate unemployment and crime rates was conducted by Glaser and Rice (1959) focusing on property, violent, and misdemeanor crime rates in the United states from 1932 to 1950 (using time-series analysis), and a cross-sectional analysis of Chicago, Boston, and Cincinnati for four census years. They found that unemployment and arrests for property crimes were positively correlated regardless of the age group, and especially for 20 to 45 year olds. However, juvenile crime rates varied inversely with unemployment. The time-series data for the United States showed a neative relationship between unemployment and crime rates for the older age groups, although the cross-sectional analysis did not (1959:684). ”It appears that unemployment is closely and directly related to criminality for males only when they are most strongly oriented to occupational stability or mobility" (1959:685). Fleisher (1963) re-analyzed the Glaser-Rice (1959) data on property crimes for the U.S. from 1932 to 1951, and found the effect of unemployment on juvenile delinquency to be positive and significant, especially for youths sixteen and older. "The principle difference between this analysis and that of Glaser and Rice, is that Glaser and Rice did not attempt to account for the effects of war or to include a trend variable in their regression 74 CHAPTER TWO: THE LITERATURE REVIEW equations” (1963:549). Fleisher (1963:551) also found unemployment related to property crime rate, regardless of age, in his cross-sectional analysis. More current time-series analysis of the U.S. has supported the earnings opportunity and crime link, but not always significantly so. Ehrlich (1975) conducted a simultaneous systems analysis of national U.S. data from 1933 to 1967 in his study on the effectiveness of capital punishment. The unemployment rate was positively and usually significantly related to the homicide rate. The labor force participation rate was negatively related, but not significantly. Ehrlich (1975:416) believes his findings suggest a systematic relationship between employment and earning opportunities and the frequency of murder. Passell and Taylor (1977) reconducted Ehrlich's (1975) data collection and analysis in an attempt to improve this research. They found unemployment positively but rarely significantly related to the homicide rate, and the labor force participation rate negatively and significantly related to the homicide rate. Krohn's step-wise regression analysis of nation-states found unemployment had a positive relationship with all three of the (homicide, property, and composite) crime rates, but was the least important of the 14 independent variables. 75 CHAPTER TWO: THE LITERATURE REVIEW Further research also reported inconsistent findings regarding the crime-unemployment relationship. Cohen et a1. (1980) focused on U.S. robbery, burglary, and auto theft rates from 1947 to 1972 and found the unemployment rate positively and significantly related to the burglary rate. However, unemployment was negatively and significantly related to the robbery and auto theft rates. They suggest the latter relationship is a result of people not working being less likely to "expose" their cars to theft (1980:108). Phillips et al. (1972:497) study on larceny, burglary, and robbery looked at 18- 19 year olds from 1964 to 1977 and found unemployment was positively and significantly related for nonwhites, but negative for whites. The not-in-the-labor-force rate variable had a significant and positive sign (as expected) for nonwhites, but insignificant for whites. Overall, labor force participation rate had more explanatory power than did the unemployment rate. The findings in some time-series studies have been inconsistent regarding crime and unemployment, depending on the type of crime. Land and Felson (1976) conducted a time series, simultaneous systems analysis of the U.S. from 1947 to 1972 and found unemployment to be negatively but insignificantly related to violent crime rates, and positive but insignificant for property crime rates. Danziger and Wheeler's (1975) time-series 76 CHAPTER TWO: THE LITERATURE REVIEW analysis of the U.S. from 1949 to 1970 found unemployment was not systematically related to the crime rates; unemployment was positively related with burglary and negatively related with assault and robbery. Similarly, Cook and Zarkin (1985) studied nine business cycles between 1933 and 1981 and crime data for this same period in the U.S., and found through both parametric and non-parametric analyses, that recessions appear to result in significant increases in robbery and burglary (and thus are counter-cyclical), and weaker evidence that auto theft is procyclical, and homicide is insensitive to the business cycle. Studies focusing on delinquency rates and unemployment rates had inconclusive findings. Fleisher (1966) claimed there was not relationship, Weicher (1970) claimed it was negative and significant, and Phillips et al.'s (1972) time-series study of 18 to 19 year olds claimed unemployment and labor force participation were usually positively related to three property crime rates, and this relationship was significant for nonwhites. Although not an empirical analysis, Calvin (1981:234) reviewed some research on black youth, delinquency, and unemployment, and claims that contrary to previous research (which has misused or not used facts on unemployment among black youth, according to Calvin), that crime by black youth over different time-periods neee show a close relationship with the concurrent economic conditions. 77 CHAPTER TWO: THE LITERATURE REVIEW In conclusion, regarding composite violent and composite property crime rates, the effects of unemployment and labor force participation are fairly consistently in the hypothesized directions (positive for unemployment, and negative for labor force participation), and frequently significant. However, these relationships are usually not very strong. Concerning homicide, rape, robbery, burglary, and larceny individually, again, the effects of unemployment and labor force participation are usually in the predicted direction, and frequently significant. However, regarding aggravated assault and auto theft, although the effects are usually in the expected directions, they do not tend to be significant. This review supports the previous contentions that the labor force participation rate is a more accurate measure of who has dropped out of the labor market than the unemployment rate is, and thus the labor force participation rate tends to be a more explanatory variable of crime rates than is the unemployment rate. It is also important to note the role these variables representing the labor market play in affecting both the yielent as well as the property crime rates. 78 CHAPTER TWO: THE LITERATURE REVIEW This review indicates that ordinary regression analysis supports the contention that legitimate earnings opportunities are more likely to affect property crime rates than violent crime rates, especially if one classifies robbery as a violent crime. It also suggests that perhaps (at least time-series) findings may be more substantiated in the hypothesized direction when the researcher looks at nonwhites. Positive, significant effects of unemployment on crime rates were more likely to occur when analyzing property crimes, than violent crimes, and among the property crimes, burglary and larceny were much more likely to have positive, significant findings than was robbery. In fact, the unemployment rate was frequently negatively related to robbery, and this relationship was sometimes significant. An initial reaction to the review of the unemployment/labor force participation studies using a simultaneous equations model is their increased amount of support for unemployment (labor force participation) rates having a positive (negative for labor force participation) and often significant relationship with violent crime rates. This is quite different than the ordinary regression findings. Additionally, the simultaneous systems approach still supported the existence of a relationship between unemployment (labor force participation) and the property crime rates. Overall, the findings using the simultaneous approach 79 CHAPTER TWO: THE LITERATURE REVIEW suggested a link between employment opportunities and the crime rate, although the findings were not consistent. This review also suggests that the labor force participation rate may be a better measure of employment opportunities than the unemployment rate, in so far as it more clearly affected the crime rates. III. DEMOGRAPHIC VARIABLES In addition to the economic variables, we must also examine and control for demographic variables thought to influence crime rates. This section will discuss key sociological variables that have been suggested by previous research to be significant. They include: race, geographical region, population size and density, age structure, and educational level. A. Rec 0 o It has been hypothesized that race and crime are related, with the implication that nonwhites are more crime prone, especially regarding violent crimes. For instance, Swimmer (1974b:299) suggests: "Crime should be more attractive to nonwhites because there is less employment discrimination in illegal than legal occupations.” Pogue (1975:40) claims that the 80 CHAPTER TWO: THE LITERATURE REVIEW significant effect of race on the crime rate ”suggests that standards of law obedience are different for whites and nonwhites, a reflection perhaps of racial discrimination and the alienation it entails.“ However, more recently this assertion has been challenged. At any rate, overwhelming support has been indicated for the significant and positive relationship between race and crime rates (usually measured as the percent nonwhite) in the literature (Bailey,l984; Chapman,l976; Danziger and Wheeler,l975; Ehrlich,l974:94; Jacobs,l981; Messner,l982; Mooris and Tweeten,l97l; Pogue,l975; Sjoquist,l973). Other studies have found this relationship generally positive, but not always significant. Swimmer (l974a) found the percent nonwhite positively related to murder, rape, aggravated assault, robbery, and larceny rates and significant for all but the larceny rates. Similarly, Kau and Rubin's (1975) research on 60 cities in 1960 and 1970 found the percent nonwhite significantly and postively related to both violent and property crime rates, except this relationship was not significant for property crime rates in 1960. Mathur's (1978) analysis of cities found the percent nonwhite positively related to all the seven index crime rates except rape, and these relationships were significant for murder, robbery, and assault. Williams and Drake 81 CHAPTER TWO: THE LITERATURE REVIEW (1980) found percent nonwhite positively related with rape, robbery, aggravated assault, burglary, larceny, and auto theft, but only significant with auto theft. (Also, percent nonwhite was significantly and positively related to arrest rate for robbery, consistent with conflict criminology (1980:567)). The studies tended to support a positive relationship between percent nonwhite and race, often finding this insignificant. In their analysis of homicides, Smith and Parker (1980) found race positive, but not significant. Similarly, Pressman and Carol (1971) found the percent nonwhite postively related to all seven index crime rates, but only significantly related to burglary and larceny, but suggested that race may be "hiding" another variable. Both Swimmer (l974b) and Greenwood and Wadycki (1973) found the effect of percent nonwhite was postive for both composite violent and composite property crime rates, but only significant for violent crime rates. Wilson and Boland (1978) found the percent nonwhite had a positive but insignificant relationship with robbery rates. Similarly, Phillips and Votey (1975) found that neither percent black nor percent with Spanish surnames were significant in affecting the crime rate. Allison's (1972) step-wise regression analysis showed race to be a less significant variable. Fleisher (1966) found race insignificant concerning delinquency rates, 82 CHAPTER TWO: THE LITERATURE REVIEW while Weicher's (1970) reanalysis of Fleisher's work claimed blacks were signficantly leee likely than whites to be delinquent. Other research has controlled for variables thought to exaggerate or imply a relationship between race and crime which may be non-existent. Blau and Blau's (1982:125) analysis of violent crimes found: ”When all other influences on crimes of violence are controlled, and SMSA's racial composition continues to exert some influence on it, though it is not great, and it is significant only for murder and assault (as well as all four types combined) and not for rape and robbery. Ehrlich (1975az42) found the percent nonwhite became insignificant when he introduced a time trend variable. DeFronzo (1983:130) drew the following conclusions: ’ Percentage of the population that is black-- hypothetically an indicator of the percentage of the population that shares a criminogenic sub- culture, especially with regard to violence-- had no independent positive effect on rates of burglary, larceny, or auto theft. Among violent crime, percentage black had significant positive effects, except with regard to rape. In the case of rates of homicide and aggravated assault, the standardized regression coefficients for per- centage black were equalled or exceeded in magni- tude by the corresponding coefficients for AFDC assistance, suggesting that the heretofore gener- ally neglected structural factor of level of gov- ernmental assistance to the poor may account for 83 CHAPTER TWO: THE LITERATURE REVIEW at least part of the supposed effect of crimino- genic subcultures on crime rates. Huff and Strahura (1980:465) suggest that the ”percentage of blacks exerts substantial indirect effects on both violent and property crime rates, especially through percentage of low income people and police employment rates...” and found only a moderate eiteet effect of percent black on the violent crime rate. Blau and Blau (1982) claim that when other variables influencing crimes of violence are controlled, that the impact of race is greatly reduced. "The hypothesis inferred is that socioeconomic inequalities are associated with ascribed positions, thereby consolidating and reinforcing ethnic and class differences, engender pervasive conflict in a democracy" (1982:118-119). Some research exploring the relationship between race and crime have discussed employment opportunities as having a potentially powerful likelihood of distorting this relationship. Thornberry and Christenson (1984:406) claim that the relationship between unemployment and crime is stronger for nonwhites than for whites. Phillips et al. (1972) found variables representing unemployment and "not in the labor force” (NLF) were significant for both racial groups. "Collinearity between the explanatory variables by color is such that one can only obtain significant 84 CHAPTER TWO: THE LITERATURE REVIEW estimates of the impact of changing labor market conditions for both colors jointly using either the white or the nonwhite variables” (1972:503). Similarly, Holzman (1982:1782) claims: "[T]he labor force participation of black offenders more closely resembles that of the general nonwhite population than the participation of white offenders resembles that of whites in conventional society.” Avio and Clark's (1976:41) research in Canada states: The significance of the variable representing the proportion of the population that is North American Indian depends largely upon whether labor force variables are included in the equa- tion. If so, the North American Indian coeffi- cient is generally significant, positive, and less than one in value for robbery, breaking and entering, and the theft classifications. [It also appeared that this variables was rela- ted to Indians on reservations.] Therefore, it appears that differential employment opportunities for whites and nonwhites could distort the actual relationship between race and crime, making it appear that nonwhites are more criminal than whites, while it is likely that nonwhites and whites respond similarly under the same employment limitations. In summary, the literature largely supports the contention that racial composition is related to the crime rates. Although this is substantiated in the research on property crime rates, racial composition seems to be more consistently related to the 85 CHAPTER TWO: THE LITERATURE REVIEW violent crimes (except rape). However, certain caveats were mentioned by numerous researchers (which were elaborated on above), warning against assuming too much from the significant, positive findings associated with the percent nonwhite and crime rates. Most frequently mentioned was the belief that the effects of racial composition may be confounded with with the effects of other variables that are strongly related to race. Thus, the current research, controlling for many of these confounding economic and demographic variables, will improve on previous research. B-Essisn Numerous studies predicting crime rates have included region as a dummy variable representing whether the area studied is in the South. This variable has been included because the literature "suggests a subculture of lethal violence in the South" (Messner,l982le7). This implies a Southern cultural "tradition" of violence, based on beliefs that oppression, conflict, broken homes, and exploitation in black ghettoes lead to condoning and acceptance of violence (Blau and Blau,l982:115). Some studies have found this variable to have a positive, significant effect (Danziger and Wheeler,1975; DeFronzo,l983; 86 CHAPTER TWO: THE LITERATURE REVIEW MCarthy et al.,l975; Messner,l982; Swimmer,l974). Greenwood and Wadycki (1973) found "region" was positively related to both property and violent crime rates, but only significant for property crime rates. In contrast, others have found no significant effect (Bailey,l984; Blau and Blau,l982; Humphries and Wallace,l980; Land and Felson,l976; Loftin and Hill,l974; Smith and Parker,l980). Bailey (1984:542) found no significant relationship between region and homicide when he controlled for poverty and inequality. Blau and Blau (1982) conducted a study of SMSAs, and claim their results "support the hypothesis advanced here that inequality-- especially ascriptive inequality-- engenders violence and can explain the higher incidence of most violent crimes in the South, obviating the need to interpret it with conjectures about a Southern tradition of violence" (1982:122). Furthermore, they claim "if there is a culture of violence, its roots are pronounced economic inequalities, especially if associated with ascribed position" (1982:114). Some research shows ne support for the ”tradition of Southern violence” hypothesis. Mooris and Tweeten (1971:42) found the highest crime rates in the West, and then the Northcentral, Southern, and Northeastern regions, respectively. 87 CHAPTER TWO: THE LITERATURE REVIEW Fleisher (1966:131) found delinquency tended to be lower in the South. Similarly, Swimmer (l974b:310) found a negative and significant effect concerning property crime, and a negative insignificant effect concerning violent crime, implying Southern cities have less property crime and about the same violent crime rate as indentical non-Southern cities. Certainly the hypothesis of a "Southern Culture of Violence" is not well-founded at this point, and the implications of assuming such a belief may be detrimental in influencing stereotypes and misguiding research. Previous research has supported the contention that the population size has a significant, positive impact on crime rates (Messner,l982; Mooris and Tweeten,l97l; Pogue,l975; Sjoquist, 1973; Spector,l975; Swimmer,l974b; Webb,l972; Weicher,l970). Kau and Rubin (1975) found population size positively related to violent and property crime rates in 1960 and 1970 (except negatively related to property crimes in 1970), and this relationship was significant with violent crime rates. Swimmer (1974a) found population size positively related to murder, rape, aggravated assault, robbery, burglary, and larceny, and significant for all but burglary and larceny. Similarly, Blau and 88 CHAPTER TWO: THE LITERATURE REVIEW Blau (1982) found the population size postively related to the composite violent crime rate, and to murder, rape, robbery, and aggravated assault rates, individually; and significant for all but murder and rape. Jacobs (1981) found the log of the population positively related to burglary, larceny, and robbery, and significant for burglary and robbery. Pogue (1975) found the population size was positively related to the composite crime rate and with each of the seven index crime rates, and these relationships were significant concerning the composite, robbery, and larceny crime rates. Hemley and McPheters (1974) focused on robbery, burglary, and larceny and found positive relationships for each of these with the population size, but the only significant relationship was with robbery. Others have found population size consistently positive, but insignificant (Allison,l972; Huff and Strahura,l980; Messner,l980; Williams and Drake,l980). Bailey's (1984:543) research on homicide found population size significant for only one of the three years he studied (1970), and believed this significance "may reflect the impact of some other (unknown) factor(s), and not the effect of population size per se." In summary, the literature overwhelmingly supports a positive relationship between population size and the crime rate, regardless of the type of crime. However, the findings are mixed 89 CHAPTER TWO: THE LITERATURE REVIEW concerning the significance levels. It does appear that composite violent crime rates are more likely to be significantly related to population size than composite property crime rates. Of the individual index crime rates, robbery is the most likely to have a significant, positive relationship with population size, while in none of the research was the rate of rape found significantly related to the population size. The remaining index crime rates (homicide, aggravated assault, burglary, and larceny) seemed as likely to have signficant relationships with the population size as they were to have insignificant relationships. kW Research on the relationship between economic variables and crime rates has frequently included the demographic variable population density. This variable represents, among other things, "crowding," or a sense of a less than ideal living environment. Population density has also been included "to reflect the fact that the closer the opportunities are, the greater the likelihood the individual can acquire the information necessary to perform a crime“ (Sjoquist,l973:443). 90 CHAPTER TWO: THE LITERATURE REVIEW Researchers looking at the composite crime rate found population density had a positive, insignificant effect (Pogue,l975; Carr-Hill and Stern,1973), while Weicher (1970) found density had a positive, significant effect on the delinquency rate. Regarding the composite property crime rate, most researchers found it was positively related to population density (Huff and Stahura,1980), many of them finding it a eignifiieent, postive relationship (Bechdolt,1975; Mooris and Tweeten,197l; Sjoquist;l973). However, some researchers found property crime rate was negatively and ineignifiieently related to the population density (Bechdolt,l975; Greenwood and Wadycki,l973), and one negatively and eignifiieently related to property crime rate (Humphries and Wallace,l980). Most of the research on the violent crime rate found it positively related to population density (Bechdolt,l975; Huff and Strahura,l980), with some finding it eignifieent and positive (Bechdolt,l975; Mooris and Tweeten,197l). 0f the studies finding a negative relationship between violent crime rate and population density, both reported insignificant findings (Green and Wadycki,l9731; Spector,l975). All of the research including population density and homicide rates as the crime rate found they were negatively related (Messner,l980; Pogue,l975; Pressman and Carol,l97l; 91 CHAPTER TWO: THE LITERATURE REVIEW Bailey,l984), but it was rarely significant (Messner,l982). Regarding rape and population density, Pogue (1975) found this relationship positive and insignificant, Pressman and Carol (1971) found it negative and insignificant, and Kvalseth (1977) found it negative and significant. All the research on population density and aggravated assault found this relationship negative (Pressman and Carol,l97l; Pogue,l975; Danziger and Wheeler,l975), but it was rarely significant (Kvalseth,l977). Research on the relationship between population density and robbery rates found the relationship to be generally positive (Pressman and Carol,l97l; Wilson and Boland,l978; Pogue,l975) and sometimes significant (Humphries and Wallace,l980; Cohen et al.,l980). However, population density was also found negatively and significantly related to the robbery rates (Kvalseth,l977; Danziger and Wheeler,l975). There were inconsistent findings regarding burglary and its relationship with population density. Cohen et a1. (1980) found it positive and significant, while others found it positive, but insignificant (Pressman and Carol,l97l; Pogue,l975). Others found it negative and significant (Humphries and Wallace,l980), and still others negative and insignificant (Danziger and Wheeler,l975). Larceny was always negatively related to papulation density (Pogue,l975; Pressman and Carol,l97l), and once this was significant 92 CHAPTER TWO: THE LITERATURE REVIEW (Humphries and Wallace,l980). Auto theft was always positively related to the popoulation density (Pogue,l975), and sometimes significantly so (Pressman and Caor1,l97l; Cohen et al.,l980). In summary, overall composite crime rates, and composite property and composite violent crime rates appear to be positively related to population density. Homicide, larceny, and aggravated assault are usually negatively related to poulation density, while auto theft and robbery rates tend to be positively related with population density. The relationships population density has with rape and burglary were inconclusive. The relationship between the urban-rural distribution (usually measured as "percent urban") and the crime rate was not consistently researched. Those who did analyze this variable tended to find it positive and insignificant (Smith and Parker,l980; Phillips and Votey,l975; Loftin and Hill,l974; Messner,l980). Although Smith and Parker (1980) found this relationship between percent urban and the crime rate positive and insignificant for "acquaintance" and the total homicde rate, it was positIve and eignitieent for the total homicide and stranger homicide, and positive and insignificant for 93 CHAPTER TWO: THE LITERATURE REVIEW acquaintance homicide. Hemley and McPheters (1974) found the percent metropolitan was positively and insignificantly related to the robbery rate, but positively and etgnifieently related to both the burglary and larceny rates. Carr-Hill and Stern (1973) found the percent urban postively and significantly related to the overall crime rate. From the little research conducted in this area, it appears that percent urban may be more related to property crimes. At any rate, most of the research including composite homicide rates and percent urban found this relationship insignificant (Smith and Parder,l980; Loftin and Hill,l974; Messner,l980). Clearly, the effects of this variable are not really understood. Many studies conducted in this area have included the percent of the population (usually male) between 15 and 24. This is because youths of this age account for almost 60% of all city arrests and these tend to be the ”peak" crime years (Viscusi, mimeo:6). Thompson and Cataldo (mimeoz9) claim that "income- oriented street crime is neth a very widespread phenomenon within poor populations and (within the individual life cycle) relatively short-lived." Danziger and Wheeler (1975:122) state: 94 CHAPTER TWO: THE LITERATURE REVIEW "Youth are considered to be more prone to crime either because they have not been completely socialized into accepting societal norms or because they have extra time on their hands." Ehrlich (l975b:104) suggests: "Possibly, then, not age per se, but the general opportunities available to offenders determine their participation in crimes against property. At any rate, the hypothesis is that the percentage of males aged 15 to 24 in the population is positively related to the crime rate. Most studies which have analyzed the age structure variables have found it insignificantly related to crime (Bailey,l984; DeFronzo,l983; Ehrlich,l974; Jacobs,l981; McPheters and Stronge,l976; Mooris and Tweeten,197l; Messner,l982; Viscusi,mimeo; Wilson and Boland,l978). Land and Felson (1976:592) found age structure had a significant relationship with property crime rates until they included unemployment in their equation. In this case, it was positively related to property crime rates and negatively related to violent crime rates, both relationships insignificant. Other research has found age structure positively and significantly related to gene crime rates. Carr-Hill and Stern (1973) found age structure was positively related to the serious offense rate in 1961 and 1966 in England, and was significant for the latter year. Danziger and Wheeler (1975:126) found a 95 CHAPTER TWO: THE LITERATURE REVIEW significant negative relationship with robbery, but not burglary and aggravated assault in his cross-sectional analysis, and in his time-series analysis found burglary, aggravated assault, and robbery all postively related with the age structure, and this relationship was significant regarding aggravated assault. Avio and Clark (1976:40) found the age structure variable had a significant, inyetee relationship with fraud and robbery, but not for burglary and different types of thefts. Huff and Stahura (1980) found age structure had a strong positive effect of property crime rates, and a weak positive effect on violent crime rates. Parker and Smith (1979) and Smith and Parker (1980) found the percentage of the population between 20 and 34 significant for acquaintance homicides, but not for stranger and composite homicide rates. In contrast, some studies provided consistent support between age structure and various crime rates. Cohen et al.(l980) found a significant, positive relationship with burglary, robbery, and auto theft rates. Pogue (1975) found age structure was postively and significantly related to all of the seven index crime rates and the composite crime rates. Others (Passell and Taylor,l977; Loftin and Hill;1974; Ehrlich,l975a) found age structure positively and significantly related to the homicide rate. Allison's (1972) step-wise regression analysis of 96 CHAPTER TWO: THE LITERATURE REVIEW delinquency rates found age structure to be a positive and ”more significant” variable. In summary, age structure was consistently positively related to both the composite crime rate and the composite property crime rate. However, these relationships were not consistently significant. Findings regarding age structure and the violent crime rate were never significant, and the direction was inconsistent. Homicide and aggravated assault were usually positively related to the age structure, and this relationship was frequently significant. Robbery was usually negatively related to the age structure, and this relationship was sometimes significant. Burglary was usually positively related to the age structure, and this was occiasionally significant. Separate analyses for rape, larceny, and auto theft with age structure were rarely conducted, but these relationships appeared to be positive, and sometimes significant. G. Edneational Level It is suggested that the educational level should be representative of earnings opportunities, and thus should be related to the crime rate. Most studies measure schooling as the median years of education for persons over 25. Consistent with 97 CHAPTER TWO: THE LITERATURE REVIEW the expected utility and opportunity theories, we would expect the relationship between educational level and the crime rate to be inverse. However, the findings using this variable were rarely significant, and frequently positive. Regarding the composite crime or delinquency rate, many studies found a eignifiieent and noeitive relationship with educational level (Allison,l972; Pogue,l975; McPheters and Stronge,l976). However, one study found the predicted negative, significant relationship between delinquency rates and educational level (Weicher,l970). and Viscusi's (mimeo:26) study of delinquents found no significant relationship. Relationships between homicide and aggravated assault and the educational level tended to be in the expected (negative) direction, but were never significant (Mathur,l978; Pogue,l975). Contrary to the theories, robbery, burglary, and larceny rates were usually neeitiyely related to the educational level (Danziger and Wheeler,l975; Jacobs,l981; Pogue,l975), sometimes significantly. The evidence concerning the relationship between murder, rape, auto theft, property and violent crime rates is too limited and inconsistent to draw any conclusions. In summary, the findings concerning the relationship between educational attainment and crime rates are at best disappointing. Ehrlich (l975bz333-335) posits a number of suggestions for this 98 CHAPTER TWO: THE LITERATURE REVIEW insignificant, inconsistent, and frequently positive relationship between crime rates and educational attainment: (1) education may be negatively related to the unreported crime rates, and (2) "It is essentially the inequalities in the distribution of schooling and training, not their mean levels, that appear to be strongly related to the incidence of many crimes.“ I suggest that since it is the various property crime rates (robbery, burglary, and larceny) which tend to most frequently be postively (and sometimes significantly) related to the educational level, that property crimes may occur in the areas where the more affluent (usually associated with the more educated) people live. Thus, it seems best to hypothesize educational level is negatively related to violent crime rates and positively related to the property crime rates, although the latter is inconsistent with the theories. Population stability or mobility has sometimes been used as an independent variable in previous studies exploring the relationship between the eocnomic structure and crime rates. THis is usually a measure of anomie, with the idea that the social fabric is more tightly woven (and thus, commitments to 99 CHAPTER TWO: THE LITERATURE REVIEW societal and moral norms are more binding) in more stable environments. ”Presumably new residents are less likely to observe the established codes of behavior of the community than the old residents” (Fleisher,l966:125). Also, a more stable population aids the criminal justice system, in particular, police officers, in keeping track of who and where the criminals are, and subsequently in locating them. (Thus, it would not be out of the question to have stability act as a predetermined variable in both the crime and police production equations within the simultaneous systems model.) Too few studies incorporated this variable to differentiate significant and direction of its affects between types of crime, statistical techniques, and units of analysis. For the most part, this variables was in the predicted direction, but largely insignificant (Weicher,l970; Fleisher,l966; Allison,l972; Phillips and Votey,l975). However, two research found it to be a signficant predictor of crime rates. McPheters and Stronge's (1976:392) (non-simultaneous) analysis found their measure for “urban decay," comprised of the percent change in the white population, had a significant, positive effect on the crime rate, and in fact, called it a ”major determinant of crime.” Hemley and McPheters' (1974) analysis of 32 states found the ”percent change in the population" end the "net migration” were eeth 100 CHAPTER TWO: THE LITERATURE REVIEW negatively related to robbery rates (significantly for the former). However, both measures of stability were positively and significantly related to burglary and larceny. Finally, although Mathieson and Passell (1976) did not include the stability variable in the crime equation of their simultaneous model, they found it had a positive, insignificant effect on the arrest rate in that equation. IV. DETERRENCE VARIABLES A study determining the relationship between economic variables (income distribution, poverty and unemployment) and crime rates, must also examine the relationship between criminal sanctions and crime rates. In fact, Stern (1978:139-140) suggests that the economic approach "cannot be the whole story" and that we must also examine "the effectiveness of deterrence damages from offenses and the costs of apprehension and punishment." Taylor (1978:38) claims: ”Viewed in conjunction with the behavior of criminals, therefore, the models suggest that crime and punishment are jointly determined. Neither can be said to depend on the other, but together depend on other variables which are exogenous to both.” This requires a simutaneous equation to model the interaction. Freeman's 101 CHAPTER TWO: THE LITERATURE REVIEW (1983:96) review of the literature suggests that studies that include deterrence variables find them to be more closely related to crime than the labor market variables. Tullock (1974:109) points out that for punishment to have a deterrent effect on crime, potential criminals must have some information on the certainty (or risk or likelihood) and severity of punishment. Similarly, Ehrlich (l975b) believes the probability of imprisonment and the average time spent in state prisons needs to be measured, and Becker (1968) believes that we need to include the number of convictions and arrests as well as the number of offenses in our analysis. However, Block and Lind (1975) criticize Becker (1968) for attempting to represent all effects of crime and punishment in monetary equaivalents. Block and Heineke (1975:323) state that previous researchers may be misguided, and that enforcement, punishment, and certainty have "no qualitative supply implications under traditional preference restrictions.” AWWW Police production is usually measured as the number of police per capita and/or the amount of money spent on police protection per capita. Frequently, these variables are included 102 CHAPTER TWO: THE LITERATURE REVIEW in models using simultaneous equations, where one of the equations has police production as the dependent variable. The first portion of this section will discuss how police production variables behave where a crime rate is the dependent variable. That is, how police production variables affect crime rates. The second section will examine how other variables (including the crime rate) affect police expenditure or police employment. That is, where the police production variables are the dependent variables. LEW The majority of studies researching the impact of police production on crime rates found no or inconclusive effects. Using step-wise regression, Allison's (1972) analysis of Chicago and its communities found the number of citizens per police department employee and police expenditure per capita to be less significant variables. Carr-Hill and Stern's (1973:313) analysis of British police departments found the effects of "creating" (increasing the number of) police officers and ”expenditure- increases” (more expenditure per officer) on ”the 'true' number of offenses were obscured by the creating and reporting effects," and thus were inconclusive effects. Humphries and Wallace's (1980:186-7) analysis of the seven index crimes in 23 cities 103 CHAPTER TWO: THE LITERATURE REVIEW found police force size insignificant, except that cities with large police forces in 1970 "experienced increases in their rate of homicide since 1950, and that police employment is negatively related to auto theft rates. Similarly, Wellford's (1974b:208) analysis of 21 large urban centers in the U.S. found that police budgets and employment did not account for much variation in the crime rate. Many findings were inconclusive. Mathur (1978) conducted a simultaneous analysis of the seven index crimes in large U.S. cities in 1960 and 1970 and found disappointing results for police expenditure: negative in 1960 and positive in 1970 for most crime equations. In fact, "the deterrent effect of police expenditure can be felt only through certainty of punishment where it had a significant, negative effect for robbery, burglary, and larceny in both years, and murder in 1970 (1978:460). Pogue (1975) conducted a simultaneous, cross- sectional analysis of SMSAs and found police employment and police expenditure did not significantly affect the probability of detecting criminals, and thus the crime rates. Avio and Clark's (1976) study of Canadian provinces had mixed findings concerning the impact of the police production function on property crime, and the total government spending on police protection. Ehrlich's (l975a:412) analysis of US crime rates 104 CHAPTER TWO: THE LITERATURE REVIEW from 1933 to 1969 found the relationship of police expenditure to crime was inconclusive. Some studies supported the hypothesis that police production and crime rates are inversely related. Mathieson and Passell (1976:94) conducted a simultaneous approach to research homicide and robbery in New York City precincts and found police employment significantly related to robberies: "...we would expect a 1% increase in police protection to reduce robbery arrests by 180." Swimmer (1974a:629) police expenditure had a significant, negative relationship with all of the crime rates except auto theft, and this relationship was sometimes significant for murder and rape. Swimmer's (1974b) analysis of large American cities found police expenditure had a negative effect on both property and violent crime rates, and this relationship was significant for violent crime rates. Ehrlich's (1974) analysis of the seven index crime rates across states for 1940, 1950, and 1960 found per capita expenditure on police was positively realted to the likelihood of apprehending and convicting felons, and negatively related to the crime rate (1974:108). Land and Felson's (1976) time series analysis of the U.S. reported property and violent crime rates from 1947 to 1972 found police employment was strongly and negatively related to the crime rates. McPheters and Stronge (1976) looked at large 105 CHAPTER TWO: THE LITERATURE REVIEW U.S. central cities with a simultaneous model and found police expenditure had a negative effect on crime rate. However, more studies supported a noeitive relationship between police production and crime rates, although this is not what the theories predict. Morris and Tweeten (1971) analyzed 754 cities in 1967 and 1968 and found police employment had positive, significant effects on both violent and property crime rates. Jacob and Rich (1981:121) critiqued Wilson and Boland (1978) in their analysis of nine U.S. cities from 1948 to 1978, and claimed increased police expenditure rates over 31 years generally related to higher robbery rates, probably because citizens were reporting it more. Greenwood and Wadycki's (1973) similar research on 199 SMSAs found police employment was positively and significantly related to both violent and property crime rates (especially violent crime rates), and that police expenditure was positively and significantly related to police employment. Pressman and Carol (1971) conducted a cross- sectional analysis on 95 SMSAs and found a positive and significant correlation between crime rates and police employment, probably because in high crime areas more police are employed. 106 CHAPTER TWO: THE LITERATURE REVIEW In conclusion, the findings on police employment per capita suggest its effects on the various crime rates are inconclusive as a whole. However, there does appear to be a tendency for police employment rates to have a positive, often significant effect on crime rates, particularly violent crime rates. It is suggested that this may be due to (l) hiring more police officers in high crime areas, and/or (2) citizens are more likely to report crimes where there are more police officers available to them. Police expenditure per capita is not included as frequently as police employment per capita in the various models, and the results from this variable also appear inconclusive. The effects of police expenditure rates on crime rates appear to be insignificant for the most part, and unlike police employment rates, its effects are frequently negative. As mentioned earlier, many of the simultaneous models used police production as both a dependent and independent variable. This section will discuss the effects of various crime rates and other variables on the police production variables. That is, where police production is the eeeeneent variable. For the most 107 CHAPTER TWO: THE LITERATURE REVIEW part, the crime rates were positively and significantly related to police employment (Philips and Votey,l975; Chapman,l976; Wilson and Boland,l978) and to police expenditure (McPheters and Stronge,l976; Greenwood and Wadycki,l973). However, Chapman (1976) found the violent crime rate was negatively and eignifiieently related to the police employment, and Wilson and Boland (1978) found the property crime rate was negatively (but insignificantly) related to police employment. Philips and Votey (1975) found the percent of non-violent crime positively (but insignificantly) related to the police employment. The median income was consistently positively related to the police production variables (Mooris and Tweeten,197l) and usually significantly so (Philips and Votey,l975; Greenwood and Wadycki,l973; Swimmer,l974a). In the same vein, the property tax per capital“ (or other measures of budgets affecting the expenditure on police protection) were consistently related to the police production variables (Chapman,l976; Greenwood and Wadycki,l973; Wilson and Boland,l978; McPheters and 1“Property tax per capita is usually included as a variable in a simultaneous model of this type in order to "identify” the system. That is, in order to have a variable which directly affects one of the dependent variables (in this case police expenditure or employment), while indirectly affects the other dependent variable(s) (in this case, the crime rate). 108 CHAPTER TWO: THE LITERATURE REVIEW Stronge,l976). None of the demographic variables were consistently related to the police production variables. In conclusion, models which included equations where police expenditure or police employment rates were one of the dependent variables usually found that the crime rate was positively and significantly related. Those including the property tax rates found it consistently positively and significantly related to these "police production" variables. Similarly, median family income always positively, and usually signficantly affected the police expenditure or police employment rates. There did not appear to be any pattern in the effect of various demographic variables on the police production variables. 9W Two other "deterrence" variables sometimes included in studies focusing on economic opportunities and crime are those of certainty and severity. Certainty measures the likelihood of being apprehended in some manner, while severity measures the degree of punishment, usually by sentence length. This section will review previous research inlcuding these variables. 109 CHAPTER TWO: THE LITERATURE REVIEW Schmidt and Witte's (1984:212) individual level data supported ”the contention that deterrence works and, indeed, that a percentage increase in the probability of being punished has a relatively larger effect on the number of arrests or convictions than does an equal percentage increase in the expected sentence." Certainty is a greater deterrent than severity. Witte (1980) had similar findings with the variables representing the probability of punishment and expected sentence. However, the deterrence effect varies with the type of crime (1980:79). Many studies on metropolitan areas have supported the contention of deterrence impacting on the crime rate. Carr-Hill and Stern's (1973:313) study on police departments in England and Wales found the probability of detection and severity of punishment if caught were significantly and negatively related to serious offenses. Philip and Votey's (1975:338) simultaneous analysis of the composite crime rate in 50 California cities found severity and certainty had a significant, negative (deterrent) effect on the crime rate. Mathur's (1978:462) simultaneous analysis of the seven index crimes across large American cities found: “Both severity and certainty of punishment retain the appropriate signs and are significant in the majority of the crime equations," but certainty is a greater deterrent than severity is to crime. "There is an inverse 110 CHAPTER TWO: THE LITERATURE REVIEW relationship between certainty and severity in all certainty equations in 1970 and in all certainty equations except murder in 1960" (1978:464). Other studies continuing to find the expected relationship between deterrence and crime, include Sjoquist's (1973) cross- sectional analysis of 530.8. municipalities in 1960, which found average sentence and convictions per arrest to have significant, negative relationships with property crime rates. In Swimmer's (1974a:622) analysis of large American cities, expected sentence (which was the average sentence times the probability of conviction) was negatively, related with all seven index crime rates, while this relationship was significant for burglary and aggravated assault. Swimmer (l974b) found the expected sentence negatively but insignificantly related to the crime rates. Similarly, Ehrlich's (1974:94) analysis of the seven index crimes across states in 1940, 1950, and 1960 found a negative, significant relationship between each crime category and the two independent variables representing deterrence: the probability of apprehension and imprisonment, and the average time served by offenders in state prisons. Avio and Clark's (1976:40) analysis of Canadian provinces found some evidence that apprehension and conviction rates are negatively related to property crime rates, and certainty of 111 CHAPTER TWO: THE LITERATURE REVIEW arrest/conviction deters property crimes more than severity does (1976:53). Ehrlich's (l975a:4l6) analysis of U.S. crimes from 1933 to 1969 found capital punishment was negatively related to murder rates. Arrest, conviction, and imprisonment rates were negatively related to the murder rate (l975a:416). Danziger and Wheeler's time-series analysis of the U.S. found the probabilities of being charged with a crime, convicted if charged, and of being sent to prison had a negative relationship with burglary, robbery, and aggravated assault rates, and was significant for imprisonment. McPheters (1976) looked at U.S. robbery rates from 1959 to 1971 and found arrest and certainty rates had significant, negative effects on the robbery rates. In contrast, other studies had inconsistent findings. Pogue (1975) conducted a cross-sectional analysis of the seven index crimes in SMSAs and found arrest rate was negatively and significantly related to the criminal activity. However, the severity of the sentence was significantly and neeitiyely related with the crime rates, except for burglary and grand larceny, which were insignificant. Danziger and Wheeler's (1975) cross- sectional analysis of burglary, robbery, and aggravated assault in 57 SMSAs found the probability of imprisonment was negatively related in all cases and significant for burglary and robbery. The average sentence had a negative relationship with burglary 112 CHAPTER TWO: THE LITERATURE REVIEW and robbery and was significant for burglary, while it had a positive, insignificant relationship with aggravated assault (1975:126). Passell and Taylor (1977) re-conducted Ehrlich's (1975) time-series analysis and found clearance rates per murder were negatively related to the murder rate, as was the conviction rate [certainty] related to the murder rate. The execution rate was negatively related to murder in some regressions and positive in others. Hemley and McPheters (1974) analyzed 32 states and found severity negatively related to robbery and burglary, and positively related to larceny rates. (These relationships were never significant.) However, certainty was consistently and negatively related to all three crime rates. In other cases, thefindings have also been inconsistent concering the effects of deterrence on crime. Kau and Rubin's (1975) simultaneous, cross-sectional analysis of 60 U.S. cities in 1960 and 1970 found certainty was consistently negatively related in both years and for both property and violent crime rates, and this relationship was significant, except for property ‘ crime rates in 1960. Severity was always neeitiyely related except for violent crime rates in 1960, but it was never significant. They claim that the "probability of conviction is by far the most important deterrent of crime" (1975:76). Chapman's (1976) simultaneous analysis of California cities found 113 CHAPTER TWO: THE LITERATURE REVIEW the percent nonwhite was negatively and significantly related to the arrest rate, while police employment was positively and significantly related to the arrest rate. Parker and Smith's (1979) analysis of 32 states found sentence certainty (proportion of homicides admitted to prison) significant and negative for stranger homicides and insignificant for acquaintance and composite homicides. Severity (median prison sentence length) was insignificant in all cases. Antunes and Hunt (1973) looked at the effects of certainty and severity as the only variables affecting the seven index crime rates in 49 states and found certainty had a slight to moderate negative relationship with each of the seven crime rates, while severity had a weak positive relationship, except with homicide and sometimes larceny. Black and Orsagh's (1978:626) simultaneous analysis across states in 1950 and 1960 claimed: ”We are unable to discover a consistent relation between sanctions and homicide." Futhermore, "the death penalty had no appreciable deterrent effect on homicide in the 1950 to 1960 period" (1978:629). Logan (1972:72) also used American states as units of analysis with a simultaneous approach and claimed: ”The findings for severity of imprisonment are not so impressive.” However, he suggests the role of certainty is probably important, but is likely curvilinear: "The relation between certainty and crime rate is 114 CHAPTER TWO: THE LITERATURE REVIEW increased in strength and slope under conditions of high severity, relative to conditions of low severity" (1972:71). Chiricos and Waldo's (1970) analysis of all 50 states on six crime rates (all index crime rates except rape) found the effects of certainty and severity, as well as the effects of percent change in certainty and severity had no significant, consistent results over three time periods. In summary, this section on the certainty and severity of punishment, strongly supports the contention that deterrence "works", and that deterrence variables are important in any model of the economcis-crime link. Certainty of punishment was consistently negatively related to the crime rates, and this relationship was usually significant, regardless of the type of crime. Severity of punishment was usually negatively related to the crime rates, but frequently insignificantly. There is undeniable support that certainty of punishment acts as a greater deterrent to criminal behavior than does the severity of punishment (Witte,l980; Schmidt and Witte,l984; Mathur,l978; Logan,l972; Hemley and McPheters,l974; Avio and Clark,1976). 115 CHAPTER TWO: THE LITERATURE REVIEW V. CONCLUSIONS TO THE REVIEW OF THE LITERATURE Almost two decades ago Tullock (1969:59) stated: "Among the various approaches to the study of crime, the economic perspective has been one of the least developed and utilized." Although sparse research on this topic existed previously, it has increased in magnitude and frequency since Becker's article in the Jentne1_ef_the_£e11t1ee1_fieeneny in 1968. In addition to an increasingly wider scope of variables included in the models, there has been an expansion of the units of analysis incorporated to test the expected utility and opportunity theories, relating the impact of the economic structure on the crime rate. These units of analysis have ranged in scope from individuals to nation-states. Additionally, improved statistical techniques (such as the simultaneous equations model) assist researchers in improving the determination of the nature of a system where demographic, economic, and deterrence variables all influence the crime rate, and an where an interactive relationship exists between the crime rate and the deterrence variables. Studies researching the link between income levels and crime rates have had mixed findings. (Income level measures include absolute poverty, relative poverty, median family income, and mean (or per capita) income.) Research on the mean and median 116 CHAPTER TWO: THE LITERATURE REVIEW income levels have been rarely signficant, and highly inconsistent in direction and strength. Many researchers have been surprised to find a positive and often significant relationship between the mean income and crime rates, and subsequently suggest it may be actually measuring the affluence of an area in the sense of that area being a reliable crime "target." Therefore, I suggest that future research should determine whether there is a distinction between median family income and the mean (or per capita) income. Since income distribution is not symmetric in the United States, the meany income is usually higher than the median income. Thus it is likely they are measuring quite different concepts, implying that future models need to include both. The absolute poverty has proved to be a more stable predictor of the crime rate than the mean or median income. However, not all studies found that absolute poverty was consistently related to the crime rate in the predicted (positive) direction. It does appear, however, that studies utilizing a simultaneous equations model were more likely to find that absolute poverty had a significant and positive impact on the crime rate (Avio and Clark,1976; Ehrlich,l974; Ehrlich, 1975b). Absolute poverty did not appear to influence certain crime rates more than others. Similar to the absolute poverty 117 CHAPTER TWO: THE LITERATURE REVIEW findings, analyses including the relative poverty (usually measured as the gini coefficient) usually found it affected the crime rate in the predicted direction, although not always. The sophistication of the analysis did not appear to affect these findings. Regarding variables representing the labor market, prior research suggests that the unemployment and labor force participation rates tend to influence the crime rates in the predicted directions, but not always significantly, and rarely strongly. An exception to this is that the unemployment rate was frequently negatively, and often significantly related to the robbery rate. Another important discovery from this literature review is that simultaneous equation models tend to increase support for the contention that employment rates affect violent as well as property crime rates. If the simultaneous equations model is indeed an improved manner of depicting the relationship between crime and the economic structure, then this impact on violent ene property crime makes this research especially pertinent regarding policy implications. The literature review suggests that the labor force participation rate may be a better measure of what is actually happening in the labor market than is the unemployment rate. This is not surprising since the latter does not include people who have quit looking for work, are on welfare, etc. 118 CHAPTER TWO: THE LITERATURE REVIEW Concerning demographic variables, previous research has widely supported a positive and often significant relationship between the percent nonwhite and the crime rate. It is possible that this relationship may not be as strong and as significant if the research controlled for many of the confounding demographic and economic variables associated with race and the crime rate. The literature review suggests that region (whether or not the area is in the South), may not significantly affect the crime rate, thus challenging the “Southern Culture of Violence" theory. The age structure was usually positively related to the composite and property crime rates, although these relationships were not always signficant. Age structure is never significantly related to the violent crime rates in the previous literature, and the directions were inconsistent. Unlike other types of crime rates, robbery rates were usually negatively related to the age structure. Findings concerning educational attainment and crime rates are often neeitjye and usually insignificant. Population size is usually positively related to the crime rate, but not always significantly so. Robbery is most likely to be signficantly related to the population size, while rape is never significantly related to the population size. The remaining crime rate types were inconsistently significantly related to the population size. Auto theft and robbery rates 119 CHAPTER TWO: THE LITERATURE REVIEW tended to be positively related with population density, while homicide, larceny, and assault, were negatively related to the population density. Rape and burglary had inconsistent relationships with population density. The effects of the percent urban on the crime rate are unclear, and usually insignificant. Population mobility was usually related to the crime rate in the predicted direction, but rarely significantly so. Police production variables suggest differential impacts of police employment per capita than police expenditure per capita, the former usually being positively and often significantly related to the crime rates, and the latter frequently negatively and insignificantly related to the crime rate. There is a need to collect data on both of these measures of police production in order to distinguish their effects. Simultaneous equations models including police production variables as one of the dependent variables, usually found median family income and the per capita tax positively and significantly related to the police production. Similarly, the crime rates in these equations were, as predicted, positively and significantly related to police production. Research incorporating certainty and severity of punishment usually found them negatively related to the crime rates, significantly so regarding certainty. CHAPTER THREE: THE METHODOLOGY I. INTRODUCTION Up to this point, the empirical findings and the theories backing the research on the relationship between economic variables and crime rates have been discussed in detail. This chapter will explain the methodology used in the present research to test this relationship, including the sample, the statistical techniques, the variables, and the models. The data for this study were collected from available data documents from a variety of sources. The source of each variable is included in Appendix 2. II. THE NEED FOR CURRENT RESEARCH Current research must take on the task of incorporating the numerous demographic, deterrent, and economic variables identified in the wide range of studies described in the literature review. The research to be described in this dissertation does this. Additionally, it is important that violent crime rates are included in the model, although neither expected utility nor Opportunity theories actually apply themselves to violent crimes. (The only exception to this would be opportunity theory's limited acknowledgement of the importance 121 CHAPTER THREE: THE METHODOLOGY of the criminal subcultures necessary for teaching criminal behavior, which might also be applicable to violent criminal behavior.) Prior research has suggested that violent crime rates may be affected by the same demographic, economic, and deterrence variables influencing property crime rates in models used to test the afore mentioned theories. Current research must also utilize a simultaneous equations model in order to capture the reciprocal effects between deterrence variables and the crime rates. The research described below meets all of these qualifications. III. THE SAMPLE The sample in this study includes 260 of the 318 Standard Metropolis Statistical Areas (SMSAs) in 1980.1 Two-hundred and sixty were chosen because this was the number of SMSAs for which crime rate data were available. As the literature review reported, a variety of units of analysis have been employed in the empirical testing of the earnings opportunities and crime rate relationship. These have included census tracts, police districts, urban areas, cities, SMSAs, counties, states, and nation-states. 1 A list of these 260 SMSAs constituting the sample is located in point one of Appendix 2. 122 CHAPTER THREE: THE METHODOLOGY SMSAs were chosen for this research because it was felt that they most adequately represented the conditions necessary for testing the relationship between economic variables and crime rates. For instance, states, and even counties, are too large to fully understand the dynamics economic fluctuations and demographic differences have on influencing the crime rates. On the other hand, census tracts, and even cities, are too small as units of analysis. More specifically, although there may not be much property crime within a ”poor” census tract, frustrated and "rational" persons from such a poor area would likely commit property crimes in more affluent census tracts, where the "earnings" from the crime would be greater. SMSAs are cohesive enough to include affluent, as well as rural and outlying areas, in addition to the inner cities. However, they are not so large as to present the generalization problems of a cross-sectional study where states or counties were the unit of analysis. IV. THE STATISTICAL TECHNIQUE The literature review chapter referred to prior studies which incorporated a simultaneous systems (or equations) model. The method was described briefly. At this point, I would like to elaborate on the need for such a model in research on the 123 CHAPTER THREE: THE METHODOLOGY relationship between economic and crime variables. In a discussion on the interaction between economic, deterrent, and crime variables, Palmer (1977:14) claims: ”What we need is some method which allows the concurrent estimation of systems of two or more equations where the dependent variable in one equation can serve as an explanatory variable in the other.” The simultaneous equations model is such a method: The distinguishing feature of simultaneous models is that the causal structure is no longer hierarchical, meaning that at least one equation contains a higher ordered endogenous variable as an explanatory variable. This gen- eral form, where each of the endogenous var- iables can interact with all the other endogen- ous variables, is much more useful...In cross- sectional studies, where data on many individ- uals are collected at only one point in time, the observations on any variables subject to reciprocal influences will appear simultaneous because no time-ordering is observed (Hanushek and Jackson,1977:246). In order to more accurately capture the relationship between crime rates and economic variables, it is necessary to include the influences that deterrence variables also have on the crime rates. However, these variables have separate variables affecting them (thus indirectly affecting the crime rates), and also, the crime rates might eimelteneenely influence these deterrence variables. For instance, areas with high crime rates may decide to increase police expenditure and/or increase the 124 CHAPTER THREE: THE METHODOLOGY severity of sentences, and this, in turn, should affect the crime rate. Greenwood and Wadycki (1973:141) explain why a simultaneous equations model is appropriate concerning crime rates and police production variables: The model that has been specified above [in their model] is circular in the sense that crime rates are viewed as determining the amount of expenditures that society chooses to devote to police protection. Given the cost of police protection, the amount of expenditures decided upon by society determines the quantity of police protection that can be employed. The quantity of police protection is in turn a determinant of crime rates. Unfortunately, inherent in the constructing of a simultaneous equations model is the problem of adequately "identifying" the model. Wilson and Boland (1978:373) explain this problem in their study on the effect of police on crime: In estimating models of this sort where two variables, X and Y, are thought to affect each other mutually, it becomes necessary to make defensible assumptions that there are addi- tional factors that will affect one of the vari- ables but not the other. These assumptions, called 'identification restrictions,’ ensure that estimating techniques can separate the effect of X on Y from the effect of Y on X. If the model is not properly "identified“ statistically, there are no results. If the model is not properly identified according to the theory, then the results are meaningless. This 125 CHAPTER THREE: THE METHODOLOGY point cannot by over-emphasized: e od m e l ic- . 'o a r ,a . 1 1‘0 - z. ~ , z to. Regarding these restrictions, Fisher and Nagin (1978:371-372) claim the following: “If an equation is not indentified, one cannot estimate it. If one tries to do so using false restrictions to identify the equation, one can draw completely erroneous conclusions from the estimated relationship.” One method commonly used to determine if the necessary requirements are achieved in order to identify a model is called the ”order condition." The first step of the order condition is, for each equation in the model, to count the number of dependent variables and subtract one. The equation must exclude at least that number of predetermined (independent) variables. For example, if an equation has three dependent variables (one on the left side of the equal sign and two on the right), then it must exclude at least two of the predetermined variables. If it excludes exactly that number (two in this case) of predetermined variables, then it is considered "just identified." If it excludes more (three or more in this case), then it is ”over identified." And if it has excludes less (one or zero in this case), then it considered "un-" or ”under-identified.” Each equation in the model (or "system") must be just or over identified in order for the entire model (or ”system") to be 126 CHAPTER THREE: THE METHODOLOGY identified. It is fine if the model is either just or over identified, but it cannot be computed if the entire model (thus, even one equation) is unidentified. While the statistical techniques in this research will include ordinary regressions, additionally it will employ the simultaneous equations model to further test the hypotheses, with the expectation that the latter is a more appropriate test of the empirical questions, for the reasons stated earlier in this section. This section was necessary to precede the section on dependent variables, in order to explain why this model will contain more than one dependent variable. At the conclusion of the discussion on the various dependent and independent variables incorporated in this study, sections will follow explaining the use of cost of living adjustments, and finally, the specific models to be tested. V. THE VARIABLES The literature review suggests the importance of including violent as well as property crime rates in analyzing the link between the economic structure and crime rates. Although both 127 CHAPTER THREE: THE METHODOLOGY the expected utility and opportunity theories generally pertain to criminal behavior being "rational” in an effort to achieve the "American Dream,” empirical research also suggests a link between yielent crimes and the earnings opportunities variables. This is understandable if one considers such issues as that unemployment and low incomes likely result in low self-esteem, high tension, and high frustration. In addition to the likelihood of these "victims" of unemployment and poverty taking their anger out on those around them, is the plausibility of taking these frustrations out on those who appear to be living the "American Dream." However, Webb (1972:646) makes the following differentiation between violent and property crimes in these models: The violent or expressive crime is gener- ally spontaneous or an act of passion, commit- ted without relfection or premeditation. The property or instrumental offense, on the other hand, is generally the result of rational and motivated behavior, and as such seems a more valid index of the offender's normative orien- tations, which are in turn a reflection of the underlying social structure. My argument is that unemployment, poverty, and a high level of income inequality, while directly negatively affecting a person's self-perception, may indirectly negatively affect the "social fabric" of society (i.e., by increasing tension, 128 CHAPTER THREE: THE METHODOLOGY frustration, self-hate,depression, etc.). For these reasons, the current research incorporates each of the seven individual index crime rates for 1980 (homicide, rape, aggravated assault, robbery, burglary, larceny—theft, and motor vehicle theft), as well as the composite crime rate (all seven index crime rates combined), the composite violent crime rate (homicide, rape, aggravated assault, and robbery rates combined), and the composite property crime rate (burglary, larceny-theft, and motor vehicle theft). This is useful to help determine how the various independent variables behave differently in impacting on specific types of crime rates. The crime rates are measured as the crime rate per 100,000 inhabitants within each SMSA. In addition to the crime rates, there will be other variables in the final model which will behave as both independent and dependent variables. Police employment was measured as the number of police officers per capita, while the police expenditure variable was measured as the amount of money spent per capita on police protection. The severity of the sentence was measured as the median time served for first releases in 1981. This was available as a composite score as well as individual scores for robbery, burglary, and larceny. 129 CHAPTER THREE: THE METHODOLOGY B. The Independent Varieblee The demographic variables in the current research are the percent nonwhite, the population size, the population density (measured as the number of persons per square mile), the geographic region (a dummy variable representing whether the state was Southern), the percent young males (also called "age structure," and measured as the percent of males between 15 and 24 years old), the percent urban, and the educational level (measured as the percent of the population over 25 years old who have graduated from high school or, alternatively, the percent with four or more years of college). The economic variables included the percent in poverty,the average monthly family afdc payment, the median family income, the mean income, and the income distribution, as well as the unemployment and labor force participation rates. For the current research, those variables acting as both dependent and dependent variables (in addition to the crime rates) include police expenditure (or police employment, and the severity of the sentence. (Ideally, some measure of certainty (i.e., the number of offenses cleared through arrest) would also have behaved as both a dependent and independent variable. Unfortunately, no such data were available for SMSAs for 1980.) 130 CHAPTER THREE: THE METHODOLOGY Furthermore, in order to help identify the simultaneous model and explain the changes in the police production variables, the variable entitled "tax" was callected. This variable was measured as the general revenue from local taxes per capita. It is proposed to directly affect the police production variables (police employment and expenditure), while only indirectly affecting the crime rates. Prison capacity was used for similar purposes in the severity equation, with the idea that the fullness of the prisons should influence how severely the offenders are sentenced. For instance, states with prisons filled to or above capacity will probably result in having their inmates serve less time in order to deal with the overcrowding. VI. LIMITATIONS OF THE STUDY Inherent in almost any analysis using secondary data is the problem that the variables are not always collected/measured in a manner considered ideal to the researcher. This research was not exempt from these problems. Probably the most serious limitation the current research faced was that there was no measure of the certainty of punishment available for 1980. If this variable had been available it would have either been multiplied by severity to create one certainty-severity variable which would be the 131 CHAPTER THREE: THE METHODOLOGY dependent variable in one of the simultaneous equations in the model, or have been the sole dependent variable in its own equation (with certainty also being a dependent variable in one equation). Furthermore, not all of the data were collected for the "right” years or unit of analysis. For instance, the severity of the punishment was only available by etetee for lgel, and even then, not all of the states. Similarly, poverty variables were not collected for 1980 for the 1980 census, but rather they were collected for 1979 rates. Additionally, the police production variables were only available for SMSAs for the fiscal year 1981- 1982. Regarding employment data, Ehrlich (1975:411) claims that time-series data are better than cross-sectional, since time- series ”measures the variations in both involuntary unemployment and the duration of such unemployment more effectively than do variations in U [unemployment] across states at a given point in time.“ One caveat is that of problems with the Uniform Crime Reports (the source of the crime data for this study). In particular is the problem of the collection process. ”If police departments have their own definition of crime or desire to suppress some information, no one will be any the wiser" 132 CHAPTER THREE: THE METHODOLOGY (Swimmer,l974b,303).2 Furthermore, Swimmer (l974b:304) goes on to point out: "The thorniest data problem concerns differential underreporting of crime by victims and police departments across cities." However: "If all underreporting were uniform, it would not affect the proposed study" (l974b:304). Ennis (1967:13) found in a 1966 survey of 10,000 households in all parts of the Continental United States that "at least twice as much major crime as is reported occurs..." I can only hope that underreporting is fairly consistent across the police departments in this sample, as this is the only available data on crime rates for SMSAs for 1980. The literature suggests that there is systematic rather than random bias in the Uniform Crime Reports. VII. TRANSFORMING THE DATA This section will discuss the alleged value of transforming the data by making cost of living adjustments. Consistent with James DeFronzo's (1983) suggestions regarding cost of living 2 However, this may not just be a function of police departments cross-sectionally, but also over time. For instance, Mathieson and Passell (1976:84) believe time-series crime data "are particularly suspect, especially for recent years in which public awareness of crime has been great and law enforcement officials have been under pressure to lower crime rates and raise both police inputs and police efficiency. 133 CHAPTER THREE: THE METHODOLOGY adjustments and the inclusion of governmental assistance, this study conducted analyses with and without these adjustments. DeFronzo found that manipulating the data in this manner improved on previous research, suggesting that these cost of living adjustments and governmental assistance are important variables in determining the link between the economic structure and the crime rate. It stands to reason that such adjustments and aid should affect the income level, possibly dramatically, when making cross-sectional comparisons. For a more detailed account of how this research followed DeFronzo's (1983) calculations, see point 10 in Appendix 2. VIII. CONTRIBUTIONS OF THIS RESEARCH This research expands previous research in a number of ways. First it includes the most comprehensive and wide range of economic and demographic variables of which the author is aware. Additionally, the analysis is conducted on all seven index crime rates, as well as the composite crime rate, and both the composite violent and composite prOperty crime rates. Frequently these studies are limited by a small sample size. This study includes 260 SMSAs, which is one of the largest sample sizes for this type of study. 134 CHAPTER THREE: THE METHODOLOGY In addition to a large number of cases and variables, this study expands others by its advanced statistical techniques. It incorporates a simultaneous equations systems model, in addition to ordinary regressions with the crime rates. That is, it models the criminal justice system as interactive, where deterrence variables affect and are affected by the crime rates simultaneously. Further, the data are manipulated to account for cost of living adjustments. (The only other person to make cost of living adjustments, DeFronzo (1983) did not include deterrence variables nor use a simultaneous systems approach.) Finally, analysis of 1980 data has not been conducted thus far, at least in publication. IX. BUILDING THE MODEL Various models to test the earnings opportunity and crime rate link have been constructed. (Some of these models are displayed in Appendix 3.) A prior section of this chapter discussed how the current research uses both ordinary regression models and a simultaneous systems model. Some "tips" suggested by previous researcher identifying the model include McPheters and Stronge's warning against overusing demographic variables in the police production equation: "We would argue that such 135 CHAPTER THREE: THE METHODOLOGY variables [as population density, percent nonwhite, age, or housing] are important determinants of police expenditures enly WWW. Their influence is thus captured in a proper specification of the crime function." Additionally, many researchers have included a measure of income as an exogenous variable in the police production equation (Greenwood and Wadycki,l973zl44; Swimmer,l974bz301). For instance Greenwood and Wadycki (1973:144) claim: ”Communities with higher income levels tend to make greater expenditures for public goods of various sorts, including police protection, and a positive sign is expected for the income variable." (Symbols for the variables are listed in Appendix 2, point 11.) This research will start by testing the following ordinary regression models for each of the seven index crime rates, in addition to the composite, composite violent, and composite property crime rates. ("CR' in these models implies these various compositions and types of crimes will all be run for each equation.): dina es on e CR - f(POLX(or POLM),SEV,UE,LFP,AFDC,POV,MEDYF,MEANY(or YDIS), AGE,POP,URB,DENS,MOB,ED,NW,SOUTH) 136 CHAPTER THREE: THE METHODOLOGY equation 1: CR - f(MEANY (or YDIS), MEDYF,AFDC,POV,UE,LFP,POP,DENS,URB,MOB, AGE,ED,NW,SOUTH) equation 2: POLX (or POLM) - g(CR,URB,MOB,MEDYF,TAX) equation 3: SEV - h(CR,NW,CAPAC) The results from these models will be discussed in the following chapter. Chapter four will elaborate on the strength and significance of the variables, as well as the importance of conducting cost of living adjustments. Chapter Five will discuss the implications of this research on future research and concerning policy implications. CHAPTER FOUR: THE FINDINGS I. THE INTRODUCTION The findings discussed in this chapter result from ordinary regression techniques (where there is only one dependent variable and that is the crime rate), and from two-stage least squares techniques (where there are three dependent variables: the crime rate, the severity of the sentence, and the police production). Thus, the two-stage least squares approach is used to compute the simultaneous equations model. Tables 1 through 8 are included in Appendix 4, and show the results of the analyses conducted in this research. Tables 1 and 2 are included to inform the reader of the zero-order correlations between the variables used in this study. Table 1 focuses on correlations between the predictor variables, while Table 2 looks at correlations between the predictor variables and the crime rates. It was noted that there were high correlations between mean income and the median family income (r-.88,p$,01 without cost of living adjustments, r-.83,p5,01 with cost of living adjustments), which is not surprising since it seems they would be similar measures of the same phenomenon, the "normal" income level. The income distribution variable, on the other hand, was believed to measure income inequality. 138 CHAPTER FOUR: THE FINDINGS II. RESULTS FROM THE ORDINARY REGRESSION MODEL A. I t o uc 0 When conducting the analysis, it was kept in mind that the median family income and the mean income may net be measuring the same phenomenon, in spite of their high correlation. As mentioned previously, since income is not symmetrically distributed, mean income is a more skewed measure of income, thus "overrepresenting" the outlying higher extremes. The median family income, on the other hand, is more a measure of of "central tendency” incomes, more adequately representing "the norm.” For these reasons, regressions included both of these variables ene the income distribution. In the course of the analysis it was discovered that two variables did net significantly increase the R2”.1 Surprisingly, income distribution was one such variable. It is believed by the 1 This was determined through the following formula: Fcb,n-p“7b" - 1 - 2 (1-R‘p/n-p-q-1 Rl- R with everything in the equation Rfl- R with the questionable variable(s) left out q - number of variables left out n - sample size p - number of variables in the complete equation 139 CHAPTER FOUR: THE FINDINGS author that the inclusion of both median family income and the mean income, negate the impact of income distribution on the crime rate. The other variable found not to contribute significantly to the Rx'was police employment per capita. It should be noted that police enneneitete per capita did significantly increase the R1} in addition to frequently being significant in the ordinary regression equations (although it was not in the hypothesized direction). Another important finding was that the regressions had significantly higher st when the cost of living adjustments were conducted on the variables representing monetary amounts. For this reason, all such economic variables were adjusted for the cost of living in the findings reported in the ordinary regression and two-stage least squares (simultaneous equations model) findings. Tables 3 and 4 represent results from the final ordinary regression models and are differentiated only by the former measuring educational attainment as the percent of the population 25 and older graduated from high school (EDl), while Table 4 measures educational attainment as the percent 25 and older with four or more years college (ED2). Because previous research had inconclusive findings concerning the variable "educational attainment," it was felt that since there has been a paucity of analyses including this variable in any form, that two different 140 CHAPTER FOUR: THE FINDINGS measures might aid in determining the effects of this variable. Thus, in an "exploratory” manner, analyses were conducted with both of these measures of educational attainment in the ordinary regressions. Consistent throughout the regressions reported in Tables 3 and 4, is the moderately high RL and signficance levels (RLranges from .49 to .68,ps.01 in all cases). This suggests that these models have achieved a fairly high status of predictive power. The R tended to be slightly higher when using ED2 (Table 4), suggesting that it is an improvement over EDl on predicting the crime rate. Additionally, the regressions incorporating ED2 tended to have more significant findings. However, the R for rape was higher using EDl (R -.68,p$,01), than for E02 (R1 -.58,p$,01). At any rate, there was little difference between the models based on the measurement of educational attainment, especially concerning violent crime rates. It is interesting to note that the crime rates best explained through ordinary regression models are the homicide and robbery rates (RI-.68,ps.01 in all cases). This is of particular interest since the economic structure was expected to have a greater impact on the property crime rates. Assault and burglary rates were the most poorly predicted of the crime rates when using EDl (R‘-.49,p5.01), while larceny was also relatively low 141 CHAPTER FOUR: THE FINDINGS (RI-.55,p$.01). Using ED2, assault was the most poorly explained crime rate (R3-.50,p$,01), with burglary and larceny being relatively low (RP-.57,p§;01 in both cases). This is somewhat of a surprise since the theories used to guide this research, expected utility and opportunity theories were designed largely to account for property crime rates, and for the most part have ignored the repercussions of the economic structure on the violent crime rate. Concerning the labor market variables, the findings are unexpected. Neither the unemployment nor the labor force participation rates are significant in any of the ordinary regression models. Similarly, the absolute poverty level also remains insignificant throughout the ordinary regression models. However, the mean income and the median family income proved to be important predictor variables of the crime rates. With the exception of larceny rates, the mean income had a significant, positive influence on all of the crime rates, in both Tables 3 and 4. Conversely, the median family income was negetiyely and significantly (again, except in the case of larceny rates) related to the crime rates in all of the regressions. This 142 CHAPTER FOUR: THE FINDINGS suggests that, holding the median family income constant and allowing the mean income to increase, results in higher crime rates. Alternatively, holding the mean income constant and allowing the median family income to increase, results in lenet crime rates. This finding suggests why the variable measuring income distribution did not significantly add to the model: including both the mean income and the median family income accurately portrays differences in income distribution. That is to say, these findings analyzing the interactions between these different measures of income level claim that as income distribution becomes more spread out, the crime rates increase. The tables suggest that ED2 has a greater impact on the crime rate than ED1. Not only do the models including ED2 have higher R for the various and composite property crime rates, education is more likely to be significant when ED2 is included. A surprising result from this variable is that in every case that it is significant, it is not in the hypothesized direction. Table 3 claims that the mete people that have graduated from high school, the highet the rape rate (p$.05). Table 4 claims that the nighet the percentage of the population with four or more 143 CHAPTER FOUR: THE FINDINGS years of college, the bignet are the composite, larceny, and property crime rates (ps.01 in all three casesl These findings concerning ED2 in Table 4 suggest that areas where more people have college educations may be tetgete of property crimes, thus measuring the affluence of an area. Additionally, especially concerning the significant positive effect of education on the rape rate suggests (consistent with findings from victimization research), that the more~educated are the more likely to repett crimes against them. It does not seem unlikely that more educated women would be more likely to report rapes. The percent of the population that were males between 15 and 24 years old proved to be weak, usually positive, and insignificant variables in all cases. Whether the SMSA was located in the South was only significantly related to auto theft rates, and there were signficantly more auto thefts enteiee of the South. However, the percent of nonwhites was one of the most powerful and consistent predictors of the crime rate, particularly for violent crime rates. It was positively and significantly related to the composite, homicide, rape, robbery, burglary, and composite violent crime rates in both Tables 3 and 4. 144 CHAPTER FOUR: THE FINDINGS We cannot assume from these findings that nonwhites are simply "more criminal" than whites. A number of explanations have been offered to explain the frequently positive and significant relationship between the percent nonwhite and the crime rate. Pogue (1975:40) believes the significant effect of race on the crime rate "suggests that standards of law obedience are different for whites and nonwhites, a reflection perchaps of racial discrimination and the alienation it entails. In further reference to the variable “percent nonwhite“, Swimmer (1974b:300) states: “certain groups in society may become involved in crime as well as other antisocial behavior) as an outlet of frustration." Kau and Rubin (1975:75) point out that the inferior legitimate market opportunities and lower opportunity costs of imprisonment for nonwhites may lead to a "significantly greater incidence of crimes" among nonwhites. Furthermore: ”If the property rights of nonwhites are not protected by the established law enforcement authority, then nonwhites may use self-enforcement policies. This self-enforcement policy may lead to more murders and assaults among nonwhites" (1975:75). Population size was positively and significantly related to homicide and robbery rates and negatively and significantly related to the larceny rates in both Tables 3 and 4. However, in the model including ED2 (Table 4) population size was 145 CHAPTER FOUR: THE FINDINGS significantly and negatively related to the property crime rate. The population density was significantly and negatively related to rape and larceny and positively related to auto theft in both Tables 3 and 4. Additionally, where ED2 was included (Table 4) population density was negatively and significantly related to the overall composite, and composite property crime rates. The results concerning population density suggest that "crowding" does not negatively affect the crime rate (with the exception of auto theft rates), but rather, that more "crowding" may represent more ”protection" by other people from both violent and property crimes. The percent urban was only signficantly (and positively so) to the auto theft rate, and this was only when EDl was used. In no cases was population mobility significantly related to the crime rate in the ordinary regression equations. D. e ence Va bl s The cost of living adjusted police expenditure per capita was a peeitiye, significant predictor, except for homicide and burglary in both of the ordinary regression models and auto theft in Table 3, and rape in Table 4. Prior research has suggested why the police production variables may be peeitiyely related to 146 CHAPTER FOUR: THE FINDINGS the crime rates, as opposed to the predicted negative direction. For example, Greenwood and Wadycki (1973:141) claim that we are ”unable" to specify this relationship since the "police perform two major functions-- crime prevention and crime detection.” Regarding the positive relationship between police expenditure and the robbery rates, Jacob and Rich (1981:117) claim: “Thus it is more likely that increased expenditures led more to increased citizen reporting and police recording of robberies than to deterrence of them." It is difficult, particularly without the use of a simultaneous equations model, to determine the causal order. That is, police expenditure per capita is likely positively, strongly, and significantly related to the crime rates in the ordinary regression models because areas with high crime rates are spending more on the police. It is probably not that a higher police expenditure results in a higher crime rate, unless reporting and/or detection rates are increasing due to the increased police production. The severity of the sentence was consistently negatively related to the crime rate, but never significantly so. 147 CHAPTER FOUR: THE FINDINGS III. RESULTS FROM THE SIMULTANEOUS EQUATIONS MODELS Because ED2 was deemed a more important variable (if only slightly) in the ordinary regressions, the simultaneous models included ED2 (instead of EDl) in every case. The results from the simultaneous equations models are represented in Tables 5,6,7, and 8 in Appendix 4. Table 5 represents results when using the composite crime rate, Table 6 uses the violent crime rate, and Table 7 models the property crime rate. Table 8 includes both the property and violent crime rates separately (as opposed to Table 5). It should be noted that each of the four models had moderately high R‘s, which were significant (pSIOI), with the exception of the severity equations, which had extremely low and insignificant st. As a whole, the economic variables behaved similarly in the simultaneous equations models as in the ordinary regressions. However, there were some differences. For instance, absolute poverty had a positive, significant (pg,05) effect on the 148 CHAPTER FOUR: THE FINDINGS property crime rate in the simultaneous models (Tables 7 and 8). The labor market variables (the unemployment and labor force participation rates) continued to be insignificant, although their coefficients were in the predicted directions. Again, consistent with the ordinary regression models, the median family income was negatively and significantly related to the crime rates, and the mean income was positively and significantly related to the crime rates in all of the models. Government aid (a cost of living adjusted and family size adjusted computation of Aid for Dependent Children payment) was again insignificant. Another economic variable, ”tax”, represented the budget affecting the police expenditure, and had the predicted positive, significant effect on the police expenditure in all four of the simultaneous models. The median family income did not have a signficant impact on the police expenditure, although it was in the predicted, positive direction. Police expenditure had a positive but insignificant effect on the crime rate. This is interesting since this variable was strongly and significantly related to the crime rate in the ordinary regressions. This suggests that the ordinary regression is not properly capturing the actual causal ordering. In fact, the models showed that the violent crime rate had a significant, positive impact on the police expenditure, thus police expenditure increases with increases in the violent crime rate (p$.01). 149 CHAPTER FOUR: THE FINDINGS C. emo ra c Va es The findings concerning the demographic variables were also consistent with the afore mentioned ordinary regression results. However, the educational attainment (ED2) had no significant impact on the crime rate (although it was positive). Similarly, population size and density, region, age, and mobility rates had insignificant effects on the crime rates. However, the percent urban had a significant, positive effect on the overall composite and composite property crime rates (pS.01). The mobility variable (the percent change in the population) had a positive and significant effect on the police expenditure in all four models (ps.01), as did the percent urban (p$"01). In the simultaneous equations model, the percent nonwhite was again a positive and significant predictor of the crime rate (p$,01). However, it was a negetiye, but insignificant predictor of the sentence severity. D. e ete Va a e The results of the deterrence variables were unexpected. In no instances did the deterrence variables significantly affect the crime rates. Furthermore, prison capacity was not a significant predictor of the severity of the sentence. 150 CHAPTER FOUR: THE FINDINGS IV. CONCLUSION This analysis suggests that the findings are importantly differentiated depending on the statistical technique used. Although the mean income, the median family income, and the percent nonwhite remained important predictors of the crime rate regardless of whether ordinary regression of a simultaneous model were used (psIOl), educational attainment and police expenditure ceased to be significant predictor variables of the crime rate. This simultaneous model helped us determine that the percent urban, the percent change in the population (mobility) and the tax revenue per capita, all positively, and significantly influence the police expenditure (p$.01). The simultaneous model also resulted in absolute poverty becoming a significant, positive predictor of the property crime rate (p5.05). Thus, the key findings of this study are the importance of income distribution and the percent nonwhite in influencing the crime rate. Many suggestions for the reason why the percent nonwhite may lead to increasing crime rates have already been discussed. The significance of income distribution as a potential for policy change in affecting the crime rate cannot be underplayed. The following chapter will discuss these findings, guidance for future research, and policy implications as potentially influenced by the findings in this research. CHAPTER FIVE: THE DISCUSSION 1. THE INTRODUCTION This dissertation has evaluated the theories and prior research linking economic and demographic variables with crime rates, and has developed a simultaneous equations model to test a refined version of prior theories. It has increased the existing knowledge on the relationship between the economic structure and the crime rate by building a comprehensive model and applying advanced statistical techniques. In addition to testing the model in a new manner on a large sample, it uses the most recent (1980) data available, to continue research conducted on 19703 data. This chapter will discuss the potential changes in policy which are suggested by the findings, as well as the implications for future research. II. THEORETICAL IMPLICATIONS Research in the area of economic factors and crime rates has been guided by one of two theories: (1) from the economic perspective, expected utility theory was incorporated, and (2) from the sociological/criminological perspective, opportunity 152 CHAPTER FIVE: THE DISCUSSION theory was used. Although these theories are similar in many respects, such as focusing on the impacts of environmental constraints on the crime rate, they also differ. Particularly, expected utility theory suggests that deterrence and risk variables play a larger role in the crime rate than does opportunity theory. A fine distinction between the two theories is that expected utility theory looks at the criminal as a "rational" person, while opportunity theory views the criminal as "driven" by outside pressures. Additionally, when these theories are applied to the violent crime rate, ferreting out differences between the two theories becomes even more difficult. The findings of this research tend to lend more support to opportunity theory than to expected utility theory. Specifically, the deterrence variables (e.g., police production and sentence severity) did not prove to affect the crime rate. Theories do not "own" variables, and the remaining (non- deterrence) variables seemed equally feasible in representing the conditions required by both expected utility and opportunity theories. Overall, opportunity theory appears to be more substantiated by the current research than does expected utility theory. 153 CHAPTER FIVE: THE DISCUSSION The first chapter discussed the potential use and appropriateness of incorporating aggregate level data to test the theories. If theories are true, they should leave traces that support them at the aggregate level. Although this macro approach is appropriate as stated, for determining what types of conditions about an area predict the crime rate; the usefulness of individual level data cannot be ignored. What is needed are cross-level theories that apply to aggregate ene individual level data. Individual level data would most likely prove useful in testing some aspects of the theories not really feasible through aggregate level data. For example, individual level data could help measure ”psychic” income and "fun of participation” concepts, while this would be impossible through aggregate level data. Additionally, individual level data would be more appropriate than aggregate level data for determining individuals' perceptions of the certainty and severity of punishments, and individual level data could help determine opportunity theory's ”socialization to crime," such as Cloward and Ohlin's subculture. Obviously, individual level data would be better for predicting individual criminal behavior decisions than would aggregate level data. Determining the interaction between risk, environment, and income distribution with 154 CHAPTER FIVE: THE DISCUSSION individual reasoning in the prediction of individual criminal behavior decisions would help advance these theories. It would be interesting to determine the role race would play in the crime rate using individual level data. Another problem with these theories is applying them to women. What we know about female economic and criminal levels is in complete contrast to what these theories claim: women tend to be poorer than men ene have lower crime rates. Thus, the model should predict that women commit more crime, while in fact, they commit less. The impact of gender is also important in this research as women, on the average, have more flexibility, wider economic opportunities, thus there is a wider range of differences between women. Individual level data may help future research determine the role of gender in these theories. However, it is also necessary to make gender distinctions at the aggregate level. Unfortunately, the data collections process does not always make this gender distinction, specifically regarding the crime rates for SMSAs. Additionally, it would be useful to look at the degree of violent crimes that occur within families and how the economic conditions affect this. Clearly, there is a lot of potential for the incorporation of gender in the research leading to advancing these theories. 155 CHAPTER FIVE: THE DISCUSSION III. POLICY IMPLICATIONS A. o Var a e A major finding in this analysis was the consistent negative, significant effect of the median family income on the crime rate. This suggests the importance of income level as a potential contributor for policy change in effecting change on the crime rate. While the absolute poverty was usually insignificant, it was significantly and positively related to the property crime rate in the simultaneous equations model. In the same vein, mean income had a significant, positive effect on the crime rate, implying that as the median family income is held constant, and the pet_eepite (or mean) income increases, the crime rate will increase. Thus, in addition to low median family incomes and the poverty levels significantly influencing the crime rate, this findings of mean income points out that income eiettientien is also an important predictor of the crime rate. Of extreme importance in this analysis is the degree to which these economic variables (median family income and mean income) were found to affect the yielent crime rate in addition to the property crime rate. The one exception was that absolute poverty 156 CHAPTER FIVE: THE DISCUSSION was not significantly related to the violent crime rate in the simultaneous equations model, while it was significantly (and positively) related to the property crime rate in this model. This study therefore supports bringing about change in the crime rate at a very fundamental level, that is, by improving income distribution. This is of particular importance when one realizes that the deterrence variables, police production and severity, had no real impact on the crime rate.1 Thus, the root causes of crime are probably more fundamental and basic than simply ”beefing up" the social control mechanisms. This is congruent with Viscusi's (mimeo:62) statement: In terms of the relative efficacy of the 'carrot' of improved economic opportunities and the 'stick' of stricter criminal en- forcement, there is strong evidence that the 'carrot' is more effective. The results concerning the labor market were unexpected regarding what was hypothesized from the expected utility and opportunity theories. In no cases were unemployment and labor force participation rate significantly related to the crime rates. It is likely that since these measures do not adequately 1 The significant, positive effect of police expenditure on the crime rate in the ordinary regressions was ”flushed out" in the simultaneous equations models, suggesting that the ordinary regressions do not adequately model the causal ordering. 157 CHAPTER FIVE: THE DISCUSSION account for the length of the unemployment, reasons for unemployment, and the interplay of unemployment with other variables. Unemployment itself may be voluntary (i.e., through retirement, or due to inheritance, etc.), temporary, etc., but the measures used in this research may not account for these issues. Also, while being unemployed may not in itself be enough to convince a person to turn to crime, having a low income relative to others probably does, but the measures used in this research may not account for these issues. Congruent with the labor market findings, the results for the variable measuring government aid were consistently insignificant. We cannot conclude from these findings that unemployment is not a contributor to the crime rate. It is likely that unemployment, especially over time, is related to some of the economic variables found to be significantly related to the crime rate. For instance, it is very likely that unemployment affects the median family income and the mean income levels. Thus, what is needed is a time-series approach to studying the economic structure as it relates to the crime rate, where it can be determined how unemployment over time affects the income levels. 158 CHAPTER FIVE: THE DISCUSSION B. D r V r 3 Another important predictor of the crime rate identified in this (and much of the previous) research, is the effect of racial distribution on the crime rate. The percent nonwhite was fairly consistently significantly and positively related to the crime rates, espeically the violent crime rates. As discussed in Chapter Four, there are a number of explanations for this. for example, Swimmer (l974b:299) discusses how the percent of the population that is nonwhite is ”indicative of the size of that population subgroup facing discrimination." Thus, the results of these findings are commensurate with Pogue's (1975: 40) findings and conclusions: With respect to public policy, the findings of this study support a tentative conclusion that criminal activity may be better con- trolled by reducing poverty and racial dis- crimination than by increasing police spending and employment and the severity of criminal punishment. Other findings from this research include very little support for the importance of most of the demographic variables (other than racial distribution) in influencing the crime rate. The positive, significant effect of four or more years of college' on the crime rate, particularly the property crime rates, implies that this variable measures affluence and "target” crime areas 159 CHAPTER FIVE: THE DISCUSSION more than the lack of educational opportunities. At any rate, this significant effect of education disappears in the simultaneous equations model. The age structure (percent males between 15 and 24) was insignificant in all models. This study did not find any evidence for the "Southern Culture of Violence" theory. In fact, the only crime rate significantly related to region was the auto theft rate, and this claimed that auto thefts were significantly more likely enteiee of the South. The ordinary regression models showed that larger populations are significantly more likely to have higher robbery and homicide rate, and significantly lower larceny rates and overall property crime rates. Furthermore, the ordinary regression models showed areas which were less densely populated were more likely to have higher rates of rape, larceny, and auto theft, in addition to the overall composite and the composite property crime rates. Thus, population density, as an over- crowding variable, did not behave in the predicted manner, but rather the results suggest the sparseness of the population allows for mete criminal behavior, perhaps because of fewer people around to "protect" against crimes. Only in one ordinary regression model was the percent urban significantly related, and this was a positive relationship with the auto theft rate. Population mobility was not significantly related to the crime rate in any of the ordinary regressions. 160 CHAPTER FIVE: THE DISCUSSION The simultaneous equations models refuted some of the prior findings on the demographic variables. In particular, they never found the percent of the population with four or more years of college to be significantly related to the crime rates. Similarly, the population size and density, region, age, and mobility rates had insignificant effects on the crime rates. However, the percent urban (unlike in the ordinary regression models) was found to have a significant, positive effect on the overall composite and the composite property crime rates. CW As pointed out earlier, the deterrence variables did not have a significant effect on the crime rates. In none of the models was the sentence severity related to the crime rate. Furthermore, the unpredicted peeitiye, significant effect of police expenditure on the crime rates was ”washed out" in the simultaneous equations models, and showed that instead, the violent crime rate had a positive, significant effect on the police expenditure. Unlike most previous research, this research did not find the median family income to be significantly related to the police expenditure. However, the tax rate, population mobility, and the percent urban all had significant, positive 161 CHAPTER FIVE: THE DISCUSSION influences on the police expenditure. Regarding the equation for severity in the simultaneous equations model, neither the percent nonwhite nor the prison capacity significantly affected the sentence severity. In fact, this was clearly the weakest equation of the model with an extremely low explanation of variation (the R2 ranged from .01 to .03). This research has pointed new directions concerning policy implications. Most important is the need to achieve more equitably distributed income levels and to decrease racial discrimination. Unfortunately, in the real world, prescriptions are easier to formulate than to implement into policy. Miller (1974) discusses the diversity in ideologies concerning criminal justice policies, and the subsequent implications for change. It does not appear too far-fetched to propose the major kinds of necessary social adaptation in the U.S. can occur only through the medium of ardently ideological social movements-- and that the costs of such a process must be borne in order to achieve the benefits it ultimately will confer (1974:34). Achieving consensus on ideologies and the subsequent policies is extremely difficult, therefore the probability of seeing them come to fruition is even more unlikely. 162 CHAPTER FIVE: THE DISCUSSION Additionally, Morash's (1983) findings on adolescents in community programs are evidence of the problem of why it is so difficult to "get at” the root causes of crime. In her explanation of why delinquency treatment may be ineffective, she claims "individual treatment will be overshadowed by the inadequacy of communities and the larger society to establish bonds with and meaningful roles for adolescents” (1983:63). Thus, identifying the need for social, political, and/or economic change is not sufficient for ensuring its implementation, nor the desired results. Unfortunately, to too many, the reactive approach of decreasing crime through strengthening deterrence (i.e., through increasing incapacitation, capital punishment, etc.) appears to be viewed as a solution over the proactive approach of attempting to deter crime petete it occurs, such as through major social, economic, and political changes. IV. IMPLICATIONS FOR FUTURE RESEARCH This study improved on previous work by including a more comprehensive set of variables, by building a new model, and incorporating advanced statistical techniques. Additionally, it had a large sample size and adjusted for the cost of living. This adjustment in the cost of living proved important in 163 CHAPTER FIVE: THE DISCUSSION increasing the explanatory power of the model, and suggests it should be included in future research on the economic structure and the crime rate. The second section of this chapter, "Theorectical Implications," discussed many theoretical issues that should help guide future research. This section will describe future research implications that surfaced specifically through the current data analysis findings. The simultaneous equations model was useful in differentiating the effects of police expenditure on the crime rate, from the effects of the crime rate on the police expenditure. That is, while the ordinary regression techniques implied increased police expenditure leads to increased crime rates, the simultaneous equations model claims that police expenditure is not significantly related to the crime rate, but rather the violent crime rate is positively and significantly related to the police expenditure. In addition, while the ordinary regression models did not find the percent urban an influential variable on the crime rate, the simultaneous equations model found it had a significant, positive effect on the overall composite and the composite property crime rates. These findings suggest that the simultaneous equations model more adequately represents the true relationship between police expenditure and the crime rate. 164 CHAPTER FIVE: THE DISCUSSION Some of the unexpected findings in this research, such as the insignificant effect of the labor market variables, may be due to not accounting for lagged effects. For instance, unemployment may not lead to crime immediately due to temporary unemployment payments, optimism of returning to work, etc. However, it is likely that over an extended time period that unemployment could lead to crime. Similarly, the "socialization" to crime probably odes not happen overnight. For these reasons, the usefulness of time-series data to test these theories is clear. Another problem inherent in this and similar research is the operationalization of the concept of ”average income." This variable, frequently measured as GNP or income per capita, was found in this and frequently in previous research to be in the "wrong” direction, often time significant. The problem is that we cannot be exactly sure what we are measuring: these "mean income” variables may be measuring (a) legal opportunities, or (b) potential loot. Since we would expect this variable to impact on the crime rate in different directions depending on which one of these concepts you believed the variable was measuring, the results are clearly problematic. Individual level data may be more appropriate for distinguishing between these concepts. 165 CHAPTER FIVE: THE DISCUSSION Finally, the major weakness of this research is that no measure of the certainty of punishment was available. It is believed that such inclusion in the model would allow an improved depiction of the actual "picture," and a greater understanding through the simultaneous equations model. As the literature review explained, the certainty of punishment has consistently proven to be a better predictor of the crime rate than the severity of the punishment. Due to the termination of the Law Enforcement Assistance Administration and a slack in time before the Bureau of Justice Statistics took up collecting some of the data (such as certainty of punishment and police production), they were not collected in a fashion timely for this project. (The police expenditure and employment data used were for the 1981-82 fiscal year). However, certainty is currently being measured by the Bureau of Justice Statistics, and the potential for replicating this and other models including the certainty of punishment should aid in our understanding of the relationship between the economic structure and the crime rate. 166 CHAPTER FIVE: THE DISCUSSION V. THE CONCLUSION This dissertation has carefully evaluated expected utility and opportunity theories, and reviewed the literature. Some valuable issues surfaced in this research: (1) the importance of the economic structure in influencing the violent as well as the property crime rates, (2) the importance of conducting cost of living adjustments on the variables representing monetary amounts, (3)the realization that economic structures affect the crime rates more significantly and powerfully than the deterrence measures (thus implying that increasing sanctions are more a manner of treating the symptom instead of the cause), and (4) the need for including a simultaneous equations model in order to adequately capture interactons between the crime rate and the deterrence variables in the analysis. This research points out the importance of examining crime causality from a very fundamental level: the economic structure. In particular is the potential to influence the crime rate through improved income distribution. Furthermore, this analysis supported the use of a simultaneous equations model in order to adequately represent causality between deterrence variables and the crime rate. Additionally, the inclusion of a large number of demographic and economic variables helped improve the explanatory power of the model. In conclusion, the role of the economic 167 CHAPTER FIVE: THE DISCUSSION structure in affecting the crime rate is becoming increasingly understood, however the need for future research remains. 168 o>_ucouc_ u_Eonouo .m:_m> >ucmaota .co_umuL0amcmLu o__nsa mc_m: deuce Loom. N .ama m_mEme .cm__ou-ws_n N Amkmpv cmEamcu mmmcmeeo Amam_v acm_o_> w .mmm_u m_vv_E ncm mc_xL03N cumum ucm ___:uccmu momcomeo Ammmfiv EOLL mc_mm .mmm_o o_vn_E vcm mc_xL03 N ccmum ucm ___I-Lcmu uoLOUm _mm m_:muc:o .eouLo>_v N Amwmfiv :m_m pcm :m_m omumc o_mEow\o_mE .m__ :o_um_:aoa N .co_u:ooxm mo >u___nmnota Awmm—v Lawmuo vcm xumg mo_am_cm> mc_vzotu Locuo N Ammm_p un_OLoom Aswm_v >o__mm >< Ammm_v Jam—u Ucm o_>< Ammm_v ucsz ocm moczuc< >u___noe .co_umotuoc w mxcmq co otgu_vcmaxo .mm_mEN .otou >u_u vcm >u_c:EEOU coozumn oocmum_u Ammm_v com__~< lasso axcwm meweu-mmcwccmm we» we mmmz_mc< quwLTQEm cw umm: manmem> 169 Xmu >utoaota mu_amu Lou .mu_c: mc_m:o: um_a:uuo Loczo m:_m> cm_coe Amam_v _xo>um3 w UOOZCooLo Ammm_v mu_m w cmmm_u woumtmaom to vmoLo>mu a? co>o c0503 N .>ub__noe Aeem_v Lm£m_o_l >EE3v ecotunoe_u .moo_>com voscm c_ mcomcoa do LopE3: Amem_v cw;m_m_m Anmkm_v ;u__czm Ammmm—v co__ccm m_mE N Aamm_v £o__tcm >__Emm Lou mocmum_mmm uod< .vxoec_ >u_>_uo< v_o;om:oI fimwm_v cwcocmoo mam—v Lm_om;3 w cmm_~cmo Aomm_v ._m so emcee oE_u Lo>o >uc_mutou c. omcmcu N .>u_co>om c_ omcmco N Aomm_v OU_m3 w m0u_c_£u Lasso a/w nosc_ucoo F x_ucmeo< 170 Ammm_v ammo; Ammm_v __m30ooz w c_um04 ooo.oo_\mbmn _mB_amo; .exobc_ >uto>ou _mL:uo:Lum x Aqmm_v ___I m c_ue04 co_um_ec_ pflmmm_v com_om w been x nAommpv czoLx x x AmNm_v c_n:m a gag co_um~__m_tum3ec_ x x A_wm_v mnoomfi A_mm_v co_a s poems mote» acnm_ :_ mc_csuumeacme & .o_qu Aowm_v oum__m3 w mo_ccQE:: >< Momm_v mcncmum a was: mEoo:_ .maccm .Amumv _o>o~-_m3n_>_vc_v cowumazuuo Lo oa>u >< Amwmpv cmEN_ox co_umcm_s Doc .>u___noe :mm_v mcouocmoz w >o_EmI segue po3c_ucou F x_vcmoe< o_mE N x x x .x x x x x A_mmfiv commosk w m_tcoz x x x x x x x x sflmwm_v Lwcmmmz x x x x Aowm_v Loammoz onco>oc xmu mu_amu Lou .Lm__ou-oum53 N x x x x x x x x x x Awmm_v Legumz ooo.m~m Lm>o moEou:_ cu_3 mo___Emm .mutoat_m to mxema oatm— cmoc w_ to» >EE:v .muuc_uota :muumzcmz col >E53u .oE_Lu toe mow.0a .ueoaum ”mum—V mo mo__E .uo_tum_v mmoc_m:n to» >EE3v x x x x __mmmmm w como_zumz oEouc_ >__Emw cm_ooE u_o _mcucou Lo o_umc .motsu_vcoaxm cu_mo; mu_qmo toe .m:_m> >utoqoLa mu_amo Lou .>u___noe .ncmbcmum Imam mc_maoc N .moumzpmtm mmo__ou N .pmm Meme—o3 co mmm__Emm N .mwcnu_vcoq Ixo co_umu:po new co_ummtuoL.mc_U30Lu Aomm_v .mco_um_:aoa sum—n w ou_;3 c_ omcmco N x x x x x x x omcotum w mcmuocmuz wo_uonnot EOLL mc_mm x Amxm_v mcoumcauz mo_nm_tm> mcmbzoLu Locuo 03“ x x x x x x Ammm_v ._m we >cucmuoz geese om3c_ucou F x_ucmoo< 172 Ammm_v Louuoam xovc_ >uco>oa _mL:uu:Lum w .ooo.oo_\mpon _mu_amo£ Roma—V Loxtmm w Lu_Em ucoEme_nmumo Lon mo_mm __muoc.>u_>_uum _mmo_ Eotm mc_mm Ammmfiv um_svohm uomeE Lepm_ ecu :_ segue; mo oumc vouuoaxm Aqwm—V ouu_3 w uv_Ecum ooo.o_m Lo>o oEouc_ >__Emm N.co_umtm_euc_ um: .mumE__o A_Nm_v _0cmu w cmEmmoLm moxmu .vwm _.u>0m .>u___noe .m_me N Ammm_v mam0a moEmcLsm ;m_cmam £u_3 :o_um_:aoa N .ucmcma o: to o_mc_m ;u_3 mc_>__ cocv__cu N .>u_ -_mnoE .muaoaoLe _oo£om Lm_z N .moLoL Lone. c_ :Num_ N.Ecmmncoc-_mtst N “mkm_v 3zo> w ma____;a ANAm_V ._m so ma____£a o_nm_tm> ncmcuuoe_u .ogmc co_u:omxo Amam_v L0:8 w __mmmma exocc_ >uto>oa _mcauu:cum “mum—V cu_Em w Loxcmm wozcmucoo — x_UdUdd¢ 173 mumL mmmz _mmw_ x x Aowm_v muu_3 mmxmu.>u_mcmv mc_m:o;.mLmu Lmu_mmo -N :_ mLmu_mmo N.u_:3 _oLumu Lma me_Lu.u_c: _0Luma\mco_um_o_> mc_>oe x fimxm_v vcm_0m w cow__3 x x Aommpv mxmLo w mEm____3 ummvsn >u_u Ou ummnan mu__0a mo o_umL .umumLmamm\vmuLo>_v c0503 N .muEOL vwmzoLUILw>o.me0; vm_a:uuo mezo N.u:mE>0_aEwc: uu_53ucoc 0—mE x x An:mm_v ULOm__03 co_umucmmLo _mu_u__oa x nfimqmm_v vL0m__m3 mmc___wzv nmumv_am__u N .mm___Emm acmLmauozu N .vmuLo>_n N .co_umgocmm UCN w um_ N .>u___noe x x x AONm_V Lm;u_m3 >u_mcmu _m_u0m.Lonm_ mo :o_m_>_v Ammm_v @302 Fave Tam—v ucmzcc__mv .co_umqsuuo m_meumw x comcoum_L;u w >LLwncLozk .mm__E mLmacm :_ m~_m >u_u .moxmu._oocum c_ mv_o me> m_-:_ N x x x Anqmm_ w mamm_v LoEE_3m ./ 95% ngpo vmscmucoU _ x_vcwoo< 17h .mumg AumLmHWFPP new .pmmu poucme mmugom nmsg< may umpmm$ on: mmmuuzucw a ._m>wp chowpmuaum .mmw_w5m% acmgma-mpmcwm a .xugm>oa manpomnm .XHmepLos ucmmcm "manmem> mcmzoppom mg» mmmwgasou xmccH Augm>om ngzpuagpm mch .ngm>on new .mcwmaon umuzogu .xmecowmmL umpmzwcm ~m>m_ msoucw .Fm>mp chowpmuanm .>ucmu:mamu vcm mgauuagum mam .pcmEXopquca "mm_nm_gm> mcwzoPPom mzp smaogzp mmmucm>cmmwv u_Eocoum new meuom mmgammme .musuwpmcmmmc_xoogm as» an umao—m>mv AHIQQV xmncH awcmugm: pru Fmgucmu mch .cmpopm mg on z_mx_— mvoom $0 a_:mgmczo may L0\u:m use; mg“ sogw xmzn mgm mmwuw>wuum me_u mcvm: .mupocmmzo; co_um~wewuu_> xmwg-cmw; mgzmmms o“ ammoaoga mw xmncH zuw>wuu< upocmmzo: ash .xu_Pm:cm:w u_eocoumowuom Fm_umg mgammme ow umm: mgm mmgoum Hum m.cmu::o .m—nmu mwcu cw meoucw came mm umecmoumL mw pH .ucmsqo~m>mu quocoom mgzmmme on mu_QQU\azu umm: mmwuzum pmcowumc msom .gmcuocm Lo; uoc pan .mpmg mswgu we max» mco go» pcmo_uwcm_m mgmz mm_nm_gm> mEom auspm m cw;p_3 mmamumn umpmuwucw go: mgm mmm can mFm>mp mucmuwkwcmwm .cowuwuum :H .Emcu cwcng mars: can mmwgommgmu ucmgmww_u mo Lm353: m mpmgoogoucw mmwuzum mmmcu mo msom mmamumn umua~ucw uo: mgm mwmxpmcm mo muwca new mmymg memgu $o mmaxh r APPENDIX 2 THE DATA The following SMSAs were included in this research: Abilene,TX;Akron,OH;A1bany,GA,Albany-Schnectady-Troy,NY; Albuquerque,NM;A1exandr1a,LA;A11entown-Beth1ehem—Easton,PA- NJ;A1toona,PA;Amarillo,TX;Anahe1m-Santa Ana-Garden Grove, CA;Ann Arbor,MI;Anniston,AL,App1eton-Oshkosh,WI;Ashev111e, NC;At1anta,GA;At1ant1c City,NJ;Augusta,GA-SC;Aust1n,TX; Bakersfield,CA;Ba1t1more,MD;Baton Rouge,LA;Batt1e Creek,MI; Bay C1ty,MI;Beaumont-Port Arthur-Orange,TX;Billings,MT; Biloxi-Gulfport,MS;Binghamton,NY-PA;Birmingham,AL;Bismark, ND;Bloom1ngton,IN;Bloomington-Norma1,IL;Boise C1ty,ID;Bos- ton,MA;Bradenton,FL;Bridgeport,CT;Brockton,MA;Brownsv111e- Harlingen-San Benito,TX;Bryan-College Station,TX;Buffalo,NY; Burlington,NC;Canton,OH;Cedar Rapids,IA; Champaign-Urbana- Rantoul,IL;Char1eston-North Charleston,SC;Char1eston,WV; Charlotte-Gastonia,NC;Chattanooga,TN-GA;Chicago,IL;Cincin- nati,OH-KY-IN;Clarksville-Hopkinsville,TN-KY;Cleve1and,0H; Colorado Springs,CO;Columbia,M0;Columbia,SC;Columbus,GA-AL; Columbus,OH;Corpus Christi,TX;Da11as-Fort Worth,Tx;Daven- port-Rock Island-Moline,IA-IL,Dayton,OH;Daytona Beach,FL; Decatur,IL;Denver-Bou1der,CO;Des Moines,IA;Detroit,MI; Dubuque,IA;Du1uth-Superior,MN-WI;Eau Claire,WI;El Paso,TX; Elkhart,IN;E1m1ra,NY;Erie,PA;Eugene-Springfield,OR;Evans- ville,IN-KY;Fa11 R1ver,MA-RI;Fargo-Moorhead,ND-MN;Fayette- ville,NC;Fayetteville-Springdale,AR;F11nt,MI;Florence,AL; Fort Lauderdale-Hollywood,FL;Fort Myers-Cape Cora1,FL;Fort Smith,AR-OK;Fort Wayne,IN;Fresno,CA;Gadsden,AL;Gainesville, FL;Ga1veston-Texas City,TX;Gary-Hammond-East Chicago,IN; Grand Forks,ND-MN;Grand Rapids,MI;Great Fa113,MT;Green Bay, WI;Greensboro-Winston-Salem-High Point,NC;Greenv111e- Spartanburg,SC;Ham11ton-Midd1etown,OH;Harrisburg,PA;Hart- ford,CT;Honolulu,HI;Houston,TX;Huntington-Ashlund,WV-KY-OH; Huntsville,AL;Indianapolis,IN;Iowa City,IA;Jacson,MI;Jack- son,MS;Jacksonv111e,FL;Janesv111e-Beloit,WI;Jersey City,NJ; Johnson City-Kingsport-Bristol,TN-VA;Johnstown,PA; Kankakee, IL;Kansas City,MO-KS;Kenosha,WI;K111een-Temp1e,TX;Knoxv111e, TN;Kokomo,IN;La Cross,WI;Lafayette,LA;Lafayette-West Lafey- ette,IN;Lake Charles,LA;Lake1and-W1nter Haven,FL;Lancaster, PA;Lansing-East Lansing,MI;Laredo,TX;Las Vegas,NV;Lawrence, KS;Lawton,OK;Lewistown-Auburn,ME;Lexington-Fayette,KY;Lima, 0H;L1ncoln,NB;L1tt1e Rock-North Little Rock,AR;Long Branch- Asbury Park,NJ;Longv1ew-Marsha11,TX;Lorain-E1yr1a,0H;Los Angeles-Long Beach,CA;Louisv111e,KY-IN;Lubbock,TX;Lynchburg, 176 APPENDIX 2 VA;Macon,GA;Madison,WI;Manchester,NH;Mansfie1d,OH; McAllen- Pharr-Edinburg,TX; Melbourne-Titusville-Cocoa,FL;Memphis,TN- AR-MS;Miami,FL;Midland,TX;Milwaukee,WI;Minnea-polis-St.Paul, MN-WI;Mobile,AL;Modesto,CA;Monroe,LA;Montgomery,AL;Muskegon- Norton Shores-Muskegon Heights,MI;Nashville-Davidson,TN; Nassau-Suffolk,NY;New Brunswick-Perth Amboy-Sayerville,NJ; New Haven-West Haven,CT;New London-Norwich,CT—RI;New Orleans, LA;New York,NY-NJ;Newark,NJ;New-port News-Hampton, VA;Norfolk-Virginia Beach-Portsmouth,VA-NC;Northeast,PA; Odessa,TX;Ok1ahoma City,0K;0maha,NB-IA;Orlando,FL;Owensboro, KY;Oxnard-Simi Valley-Ventura,CA;Panama City,FL;Parkersburg- Marietta,WV-OH;Paterson-Clifton Passaic,NJ;Pensacola,FL; Peoria,IL;Peterburg-Colonial Heights-Hopewell,VA; Philadel- phia,PA-NJ;Phoenix,AZ;Pinebluff,AR;Pittsburgh,PA; Pitts- field,MA;Portland,ME;Portland,OR-WA;Poughkeepsie,NY; Provi- dence-Warwick-Pawtucket,RI-MA;Provo-Orem,UT;Pueblo,CO; Racine,WI;Raleigh-Durham,NC;Reading,PA;Reno,NV;Richland- Kennewick-Pasco,WA;Richmond,VA;Riverside-San Bernardino- Ontario,CA;Roanoke,VA;Rochester,MN;Rochester,NY;Rockford,IL; Sacramento,CA;Saginaw,MI;St.Cloud,MN;St.Joseph,MO;St.Louis, MO-IL;Salem,OR;Salinas-Seaside-Monterey,CA;Salt Lake City- Ogden,UT;San Angelo,TX;San Antonio,TX;San Diego,CA;San Francisco-Oakland,CA;San Jose,CA;Santa Barbara-Santa Maria- Lompoc,CA;Santa Cruz,CA;Santa Rosa,CA;Sarasota,FL;Savannah, GA;Seattle-Everett,WA;Sherman-Denison,TX;Shreveport,LA;Sioux City,IA—NB;Sioux Falls,SD;South Bend,IN;Spokane,WA;Spring- field,IL;Springfield,MO;Springfield,0fl;Steubenville-Wierton, OH-WV;Stockton,CA;Syracuse,NY;Tacoma,WA;Tallahassee,FL; Tampa-St.Petersburg,FL;Terre Haute,IN;Texarkana,TX-AR; Toledo,OH-MI;Topeka,KS;Trenton,NJ;Tuscon,AZ;Tulsa,OK;Tusca- loosa,AL;Tyler,TX;Utica-Rome,NY;Vallejo-Fairfield-Napa,CA; Vineland-Millville-Bridgeton,NJ;Waco,TX;Washington,DC-MD-VA; West Palm Beach-Boca Raton,FL;Wheeling,WV-OH;Wichita,KS; Wichita Falls,TX;Williamsport,PA;Wilmington,DE-NJ-MD;W11- mington,NC;Worcester,MA;Yakima,WA;York,PA;Youngstown- Warren,OH. Regarding geographic region, the following states were coded as Southern: Texas, Oklahoma, Arkansas, Louisiana, Mississippi, Alabama, Georgia, Florida, South Carolina, North Carolina, Tennessee, Kentucky, Virginia, West Virginia, Maryland, Deleware, and Washington,D.C. The remaining states were coded as non-Southern. 177 APPENDIX 2 The 12§Q_g;1mg_1§;g§ were obtained from the Federal Bureau of Investigation Uniform Crime Reports published in 1981 by the U.S. Department of Justice, entitled §11Qg_1n_§hg_nniggg Stages. Similarly, the 1212_g11mg_;§§g§ were collected from the 1980 publication of Qg1mg_in_ghg_yni£§g_§§§§gfi. Police production variables (police employment and police expenditure) as well as the variable "budget" were gathered from the publication: ve e t act e , Census of Governments, U.S. Department of Commerce. Volume 5. G082(S). Issued March 1985, by the Bureau of Census. Police employment is measured as the number of police officers per capita (Table 4). Police expenditure is the per capita amount of dollars spent on police protection (Table 5). "Budget" is measured as the general revenue from local taxes per capita (Table 5). The data on §§3§11§1_gf_ngnighmgn§ were secured from the U.S. Department of Justice, Bureau of Justice Statistics, Special Report. entitled WWW 12g; (Table 8). Unfortunately, severity data were not available for 1980, and were only available by state, and not for all states. SMSAs falling in more than one state had to be coded as missing data regarding severity. Additionally, data were not available for all of the individual crime types. Composite scores as well as individual scores for robbery, burglary, and larceny were reported for the median time served (in months) for first releases. The W. Maize. was: (measured as the number of inhabitants per square mile), populatign_§§§b111§1 (measured as the percent change in the population from 1970 to 1980), pxiggn_g§p§g1;y (measured as the state correctional facilities capacity in 1979), and average per month aid; payment data were collected from the State—WWW for date on 1980 (except prison capacity data were for 1979). This document is published by the Bureau of Census, U.S. Department of Commerce. 178 APPENDIX 2 The l2§Q_Sen§u§_2f_£he_fonglation Volume 1 (Characteristics of the Population) Chapter C (General Social and Economic Characteristics) publication by the Bureau of the Census (PC80-l-C4) was used to collect pge gttpctutg (or the percent of males between 15 and 24 years old) (Table 115), the pgptptipppl_lggp1 (one measured as the percent of the population 25 years and older with high school graduates and another measured as the percent with four or more years of college) (Table 56). the lab2r_force_22r£icination_rate for all persons sixteen years old and older (Table 120). The l2§Q_Qan§2§_2f_£he_ronalation Volume 1 (Characteristics of the Population) Chapter C (General Social and Economic Characteristics) Part 1 (United States Summary) published by the Bureau of Census, Issued December 1983 (PC80- 1- Cl) was used to collect data on mtgipp_fi_m11x_ip_pm§ for 1979 (Table 247). 2_rcent_families_xiEh_inc2me§_bel2__the_1212_22_ertx 1ev pl (pt pp§plptg poverty) (Table 247) and the pgttppt pttpp (Table 56). (Income and poverty data are only available for years ending in "9" because that is how the census bureau collects them for the census years.) Unfortunately, the Census Bureau did not conduct a collection of data on income distribution for SMSAs in the 1980 census. In fact, the last available data were for the 1970 census. However, 1980 figures were found by SMSA in the 19§1 Sutxex of pripg Eowgt, a sales and marketing management magazine published July 27,1981. The "Effective Buying Income" (also called "EBI," which is measured as the personal income less personal tax and non-tax payments) was reported by percentages of the population falling into each of five categories: the percent of the households with EBI (1) below $8,000, (2) $8,000-$9,999, (3) $10,000-$lh,999, (4) $15,000-$24,999, and (5) $25,000 and over. The author would like to thank Dr. Paul Menchik, of the Economics Department at Michigan State University for developing the following formula to determine the income distribution: 10. 179 APPENDIX 2 X - the weighted average -£p:x; pL- proportion of the population in each income group xL- group mean (or 2 times the limit value) (Thus, in each case, x,- $4,000, x;— $9,000, x3- $12,500, X4- $20,000, and x;- $50,000.) weight - pix 100 5‘ — w '- z 100 -o-" fi-o— The Coefficient of Variance - - income 9' 3': distribution Thus, a high coefficient of variance indicates high income inequality. Consistent with suggestions from James DeFronzo in a recent article in Qtimipplpgx, this research included variables representing government payments to the poor and tp§t_pf 1121pg_pdlp§tmpptfi. The Cost of Living Index was acquired through the American Chamber of Commerce Researchers Association (ACCRA) W Liyipg_lng1tptpt§. Census SMSA cost of living assessments are not conducted for more than a small number of SMSAs each year (too small for the purposes of this research), so I was forced to look at this quarterly publication. Unfortun- ately, this document reported a lot of variation, even between consecutive quarters. Therefore, I collected all of the data available for each SMSA for each quarter between 1979-1981 (thus, 12 quarters), and took the mean. I was, thus, assuming that relatives do not change much and that the high variation between quarters was due to sampling variations. Where the data were available for SMSAs composed of more than one city, I weighted the cost of living index appropriately, depending on what percentage of the SMSA the city constituted. Therefore, this cost of living adjustment is by no means perfect, but it was the best I was able to find and compute. Over half of the SMSAs had data that were complete for 11 or 12 of the quarters. Using DeFronzo's (1983) formulas, a vi was calculated for each SMSA. This was done by taking the cost of living index for the individual SMSA and dividing this by 11. 180 APPENDIX 2 the mean cost of living index for the sample (100.9 in this sample). The variables involving monetary values were then divided by the cost of living ratio for each SMSA in order to devise new variables adjusted for the cost of living. Those variables for which cost of living adjustments were made, include the mean income, median family income, police expenditure, and budget. Furthermore, DeFronzo's (1983) example for determining the cost of living afdc payments for families for each SMSA was followed by dividing the average per month family payment by the cost of living ratio, and then dividing this quocient by the average size of families living in poverty. Variable Codes (Unless otherwise stated, the crime rates are for 1980): COMP-composite crime rate,VIOL-composite violent crime rate, PROP-composite property crime rate, ROM-homicide rate,RAPE-rape rate,ASSLT-assault rate, ROB- robbery rate, BURG-burglary rate,LARC-larceny-theft rate, AUTO-motor vehicle theft rate,NWhpercent nonwhite,POP- population size, DENS-population density, EDl-percent population 25 or over with high school degree, ED2-percent population 25 or over with four or more years of college,AGE- the percent of the population which is male and between 15 and 24 years old, URB-percent urban,SOUTH-dummy variable representing if SMSA is in the South (not in the South-0, South-l),MOB—percent change in the population from 1970 to 1980),UE-percent unemployed,LFP-1abor force participation rate,POV-percent of families below the 1979 poverty level (absolute poverty), MEDYF-median family income, MEANY-mean income, YDIS-income distribution, TAXPbudget,or the general revenue from local taxes per capita, AFDC-average monthly afdc family payment, CAFDC- afdc payments adjusted for cost of living and family size using DeFronzo's (1983) formula, POLXF police expenditure per capita, POLMh police employment per capita, SEVC- median time served for first offense for all offenses, CAPAC-State Correctional Facilities' Capacity rates for 1979). A ”C" before any of these variables represents that cost of living adjustments have been made. APPENDIX 3 SIMULIANEOUS MODELS USED IN PRIOR RESEARCH Previous researchers have identified the simultaneous systems model regarding the relationship between the economic structure and the crime rate in a number of ways. This appendix lists some of the models used. Variable symbols will be the same as those used in this research, as listed in point 11 in appendix 2. If the variable is one that was not used in the current research, it is spelled out in the models below. Blatk and QI§§Eh (122§: HOM - f(certainty,SEV,probability of execution,MEDYF,NW) Certainty - g(HOM,MEDYF,POV,NW,POLX,HOM in t-1,UE,AGE, % living in SMSA,male/female ratio,POP,REG,ED) SEV - h(same variables as in Certainty equation) Probability of Execution - i(same variables as in Certainty) Car - 9 3 COMP - f(certainty,POLM,POLX,URB,%working class) Certainty - g(AGE,POP,POLX,%violent offenses,%working class) POLM - h(%violent offenses,%middle class,URB) Catr-Hill and Stern (1222) COMP - f(Certainty,%working class,AGE,URB,POLX) Certainty - g(AGE,POP,POLX,URB,%working class,%violent) POLM - h(%midd1e class,%violent,URB) Chapman (19Z§) PROP - f(NW,LFP,Certainty,female LFP,total losses from felony crimes per capita) VIOL - g(NW,LFP,MEDY,PROP,Certainty) Certainty - h(NW,POLM,%using public transportation to work) POLM - i(PROP,VIOL,%taking public transportation to work, per capita property value,%blue-collar,police wage) Ehrlith (192g) CR - f(Certainty,MEDYF,SEV,NW,%families below 1/2 MEDYF) Certainty - g(SEV,POLX,CR,MEDYF,UE,NW,AGE,REG,POP,ED,%families below 1/2 MEDYF,male/female ratio,%population in SMSAs) Gre wo an Wad k 9 3 PROP - f(POLM,POV,DENS,NW,REG,median value of owner occupied housing units) VIOL - g(POLM,POV,DENS,NW,REG) POLX - h(VIOL,PROP,MEDYF,TAX) POLM - i (POLX) 182 APPENDIX 3 BMW PROP - f(POP,NW,UE,YDIS,SEV,MEANY,Certainty) VIOL - g(POP,NW,UE,YDIS,SEV,MEANY,Certainty) Certainty - h(POP,NW,AGE,POLM,MEANY,ED,REGR,PROP,VIOL) Lan e PROP - f(PROPT,AGE,POLX,UE,inflation,GNP/POP) VIOL - g(VIOLT,PROP,AGE,UE,POLX,inflation) POLX - h(PROPT,VIOLT,GNP/POP,inflation) Mc he nd t on e 9 COMP - f(central city decay,central city affluence,NW,ED,housing quality,youth presense,COMPT,POLX) POLX - g(COMP,TAX) Mathittpn and E§§§§ll (l9Zé) COMP - f(Certainty,MEDYF in precinct,MEDYF in adjacent precincts, %families with incomes more than $25,000, dummy for business district,dummy for if in presencs of large park or airport) Certainty - g(POLM/reported crime,MOB) POLM - h(POP,business dummy,large park or airport dummy,COMP, dummy for if in Manhattan,miles of street) Mat u 9 COMP - f(Certainty,SEV,MEDY,YDIS,NW,ED,UE,REG,%white collar) Certainty - g(SEV,POLX,COMP,POP,POP squared,NW) POLK - h(COMP,TAX) Ph V te 9 CR - f(Certainty,SEV,and combinations of various socioeconomic variables) Certainty - g(CR,POLM) POLM - h(CR,MEDYF,%violent offenses,police wage) Police wage - i(MEDYF,DENS,POLM) Eogue (1225) CR - f(MEANY,NW,POV,AGE,POP,UE,DENS,ED,Certainty,%male,change in POP,change in UE) Certainty - g(POP,DENS,POLX,CR) POLX - h(MEANY,NW,POV,AGE,POP,UE,DENS,ED,POLX,%male,change in POP,change in UE) 183 APPENDIX 3 Swimmet (1924;) CR - f(SEV,Certainty,gain/offense,income from illegal activities) POLX - g(MEANY,TAX,safety rate (which is the inverse of the crime rate) Sw 9 4 CR - f(NW,ED,MEDY,UE,POLX,REG,POP,POP squared,YDIS,SEV x Certainty) POLX - g(MEDY,TAX,POP,DENS,PROP,VIOL,REG,POP squared) Wilso an a 9 ROB - f(Certainty,AGE,NW,UE,DENS) Certainty - g(NW,crimes/patrol unit,moving violations/patrol unit) POLM - h(PROP,VIOL,TAX,REG,police officer salary) Patrol units on the street/100,000 populuation - i(POLM,DENS, POP,%officers assigned to two-officer cars) APPENDIX 4 THE TABLES 184 3. w. a... 3. w a . 00.. .33. team. e.~¢. eegn. H0. Conn. eenn. 00. .ena. n0. e.nn. econ. 00.. 00. .00. en“. ..n~. c.0n. 40.. .00. ..m~. ..0d.. 00mm 00.. «ego. econ. ..qn. c.0n. .enu. c.0H. .300. ..mu. e.~4. .0a.. 00. n0.. s.q~.. 30.... ..0~.. #0. 00. e.-. do. ..n~. n0.. 00.. 40C; 3.. ..nn. 2.3. :00. :2”. 00.. 8.. :3. .13. :3. :3- :3. «0. «0.. 2.3. «0.. 2.2. :0“. :2. 2.2.. no. .2. ”0.. 0.2? «0.. :39 ..n.~. «.2... .63. 00.. ..an. :3. 0a.. 00.. .3. :2... ..an. .3. .3. .3. .2. :2. ..n~. 00.. oo. .13. :2: 0.5..— .3.. :2... ..0n. :3. 2...}. :3. ..nn. ..~n. :2. .Ln. «0.. :3. .0... ..o«.. 0». .15 00. .3. :2. no. :3. no. :3; case .3; :64. scan. e.an. e.n.~. eenn. 0.0“. ..Hn. no. it}. ”0.. n0. 00.. C.-.. ..nN.. ..n~.. ..na.. ~0.. No. 0.0H. ..Hn. .13.. 00.. H4mm< #0.. 2.60. .500. some. 0.00. n0. «0.. 9.2:. :34. e.~w. ..n~.. .HH. n0. 0A.. «.0. 00.. c.0N. .od. ..un. ..NN.. 0a. 00. 00.. no“ «0. :3. «0.. 3.. No. .1". 40. 00.. 3.. a0. 3.. #0.. 40.. ..nu.. 00. .3.. do. .2. no. «0.. no. 00. 00.. ”:5. .2: 23. 2.3. :3. 2.3. :3. .1". 2.2. .3. ..0n. 3.. 3.. ..on.. 23.. .3... 2.2.. 2.2.. 3. no. .3. :3. 3. .3: to: 00.. .350. .50.... e.o.~. .52”. 00. item. ..on. .NH. 0.0N. H0. ..0n. .50". 00.. no. 00.. .0a. ..nn. ..0n. n0.. 00. .32. ..u~.. mrou >mm xuqomu 50m r400 3.2 FHDOM no: man 930 000 u0< «am an“ Don—<0 Uom< 9:0wa From: Card tax Mug >00 mm.” m: mmaHdem<> mbozmoova mam. 02¢ mmH2z mbozmuoxm mxH zmeHmm mZOHH¢Ammmoo Mmomo-ommN N mAm