Factors Influencing Repeat & Multiple Victimization in Hierarchical Modeling, Slides of Criminology

Download Factors Influencing Repeat & Multiple Victimization in Hierarchical Modeling and more Slides Criminology in PDF only on Docsity! The author(s) shown below used Federal funds provided by the U.S. Department of Justice and prepared the following final report: Document Title: Repeat and Multiple Victimizations: The Role of Individual and Contextual Factors Author(s): Maureen Outlaw ; Barry Ruback ; Chester Britt Document No.: 194055 Date Received: March 2002 Award Number: 98-IJ-CX-0034 This report has not been published by the U.S. Department of Justice. To provide better customer service, NCJRS has made this Federally- funded grant final report available electronically in addition to traditional paper copies. Opinions or points of view expressed are those of the author(s) and do not necessarily reflect the official position or policies of the U.S. Department of Justice. / PROPERVOF . Natio&l Criminal Justice Reference Service (NCJRS) Repeat and Multiple Victimizations: The Role of Individual and Contextual Factors* 3 Maureen Outlaw Pennsylvania State University Bany Ruback Pennsylvania State University Chester Britt Arizona State University-West * This research was slipported by NIJ Grant 98-IJ-CX-0034 and the Center for Research on Crime and Jw?ice at Pennsvlvania State University. I position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. Repeat and Multiple Victimizations: The Role of Individual and Contextual Factors Criminal victimizations raise attributional questions of causation, especially for crimes involving strangers. Observers are likely to wonder why a particular individual was victimized: Was the victim merely unlucky?; Was the area unsafe?; Was the victim negligent in failing to avoid known risks? For any single victimization, observers are likely to be unsure about the reason for the crime, because several factors about the person and the area are plausible causal explanations. For repeated victimizations, however, this uncertainty is likely to be reduced, as some factors can be eliminated if they do not covary across the victimizations (Kelley, 1972). Thus, observers should be more confident about their causal attributions for repeat victims than for individuals who suffered only one victimization. . -- A common lay explanation Qr repeated victimizations is bad luck. The problem with this explanation, however, is that repeated victimization is not a random process, as would be expected if only bad luck were involved (Hindelang, Gottfredson, & Garofalo, 1978). That is, some individuals appear to be “victimization prone” (Hindelang et al., 1978, p. 130); they suffer significantly more victimizations than would be expected by chance. But characterizing individuals as victimization prone is merely the first step; it does not explain why they are victimized repeatedly. In contrast to prior research, which has investigated how person or contextual factors relate to repeat victimization, the present research uses hierarchical modeling to examine the relative contributions of factors about the person, factors about the context, and, most importantly, the interaction of individual and contextual factors. Moreover, we use this analytical framework to determine whether these main effects and interactions are the same for e 4 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. 0 those who are particularly susceptible to a certain type of criminal victimization as they are for those who are susceptible to crime generally. That is, we draw a distinction between repeat victimizations (i.e., more than one of the same type of criminal victimization) and multiple victimizations (i.e., two or more different types of criminal victimizations) and test whether these two different patterns of victimization are affected in the same way by individual factors, contextual factors, and the interaction of individual and contextual factors. An understanding of how the relative contribution of these three groups of factors affect repeat and multiple victimizations has implications for research, theory, and policy. Research on Repeat Victimization Repeat victimizations tend to be clustered among a few individuals and among a few places. Over a one-year period, most individuals do not suffer a victimization, and most of those who are victimized are victimized oply once. However, some individuals are victimized more than once, and this small percentage of people accounts for a disproportionately large number of criminal victimizations (Farrell, 1995; Pease & Laycock, 1996). For example, data from the National Youth Survey (NYS) indicate that about 6% of the sample accounted for 38% of the larceny victimizations, 5% accounted for 63% of the assault victimizations, and 3% accounted for 43% of the robbery victimizations (Lauritsen & Quinet, 1995). Similarly, data from the 1992 British Crime Survey indicate that 6% of the sample accounted for 63% of all property crimes and 3% of the sample accounted for 77% of all violent crimes (Ellingworth et al., 1995). Although most work in the area of repeat victimization focuses on this clustering among individuals and places within types of crime (what we are calling repeat victimization), there is also evidence of clustering among individuals and places across types of crime (what we are calling multiple victimization). In their analysis of victimization data from 26 cities, Hindelang 0 5 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. et al. (1978) found a positive relationship between personal and household crimes, such that individuals who lived in victimized households were significantly more likely to be victims of personal crime than were individuals who lived in nonvictimized households. Likewise, household victimization was significantly higher when a member of the household had been a victim of a personal crime than when no member had been a victim of a personal crime. Knowing that both repeat and multiple victimizations tend to be clustered among a few individuals and in a few places does not address the larger question about why some individuals or households are at greater risk than others. In answer to this question, research and theory have focused on the characteristics of the places where these individuals are likely to be and on those individuals’ demographic and lifestyle characteristics. Characteristics of Places That some areas have higher,crime rates than others is well documented. Because of disruptions in families and in communities, some areas have weak formal and informal social control and therefore higher rates of deviance. In particular, according to social disorganization theory (Shaw & McKay, 1942), areas that are characterized by high population density, ethnic heterogeneity, and residential mobility are also likely to lack the resources to fight off crime and decay. Places characterized by such social disorganization lack community investment (Bursik & Grasmick, 1993) and are typified by higher levels of neighborhood incivilities and high proportions of single-parent families (Sampson & Groves, 1989), all of which increase criminal victimization rates. These same factors are also generally predictive of repeat victimization. Areas with high rates of unemployment and deprivation also have high rates of repeat burglary victimization (Johnson et al., 1997) and, more generally, areas with the highest rates of victimization also have the highest rates of repeat victimization (Trickett, Osborn, Seymour, & 0 6 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. 0 crime. In cases of burglary, individuals in neighborhoods characterized by many incivilities were at significantly greater risk of victimization if they used fewer safety precautions, whereas for those in neighborhoods characterized by few incivilities, the difference in risk was not as dependent on the use of safety precautions. For violent crimes, individuals in ethnically homogeneous neighborhoods were at significantly greater risk of victimization if they were nonwhite than white, whereas in ethnically heterogeneous neighborhoods, the difference in risk was not as great between nonwhite and white residents. The current study applies the multi-level approach and measures used by Rountree et al. (1 994) to models for repeat and multiple victimization. We use the same victimization survey data from Seattle that Rountree et al. (1 994) used because it is one of the few available data sets that has both a large sample size and contextual information. Other data sets with large sample sizes, contextual information, and pgssibly clearer information regarding repeat and multiple victimization (i.e., the NCVS) are currently restricted and contain little or no information . -:- a‘-:‘’ relevant to the theoretically important factors that put people at higher or lower risk of victimization (e.g., routine activity measures). Based on prior research showing that areas with high rates of victimization also have high rates of repeat victimization, we expected that both person-level factors and place-level factors would be significant predictors of repeat and multiple victimization (Trickett et al., 1992). We also had two specific hypotheses with regard to the different types of repeat victimization. First, we expected that repeat property victimizations, compared to multiple victimizations, would be more context dependent. In the data set we used, property crimes were tied to the victim’s neighborhood, whereas violent crimes could occur either in the victim’s neighborhood or somewhere else. Thus, for repeat property crimes, factors about the neighborhood would e 9 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. covary with the victimizations. In contrast, the only factor that covaries for certain with the multiple victims (i.e., victims of both a property crime and a violent crime) is the person. Thus, we expected that contextual factors would play a smaller role for multiple than for repeat victimizations. In addition to these hypothesized main effects, we expected that there would be significant interactions of person-level and place-level factors. Generally, we expected that in low-risk neighborhoods, individuals at high risk of victimization would be significantly more likely to be repeat and multiple victims than would individuals at low risk of victimization, whereas in high-risk neighborhoods the difference between high-risk and low-risk individuals in terms of number of repeat and multiple victimizations would not be as great. That is, we expected person-level differences primarily in neighborhoods where the overall risk of victimization was low. 1 Aside from examining the role of person-level, place-level, and person-level by place- level factors in repeat and multiple victimizations for the entire sample (which included nonvictims), we also conducted more focused analyses in which we compared single victims to victims who have been victimized more than once. In the case of repeat property victimization, this comparison was between individuals who had been victims of one property crime in the prior two years and those who had been property victims two or more times in the prior two years. For multiple victimization, the comparison was between those who had been victims of either property or violent crime once in the prior two years and those who were victims of at least one property and one violent crime in the prior two years. These more focused comparisons of individuals who had been victimized at least once provide an additional test of the importance of person-level factors in repeat and multiple victimization. If the same factors that predicted repeat 0 10 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. 0 and multiple victimization when nonvictims were included in the models also predicted repeat and multiple victimization when nonvictims were excluded, we should have more confidence in our conclusions regarding the relationships between person-level factors and the risk of repeat and multiple victimization. METHODS Data The data used in the present study are telephone survey data from 5,302 adults who lived on one of 600 city blocks contained in 100 of the 12 1 census tracts in the city of Seattle in 1990 (Miethe, 1997). The sampling procedure involved the selection of three pairs of city blocks from each of the selected census tracts. One of the blocks in each selected pair contained a street address at which there had been a burglary in 1989, and the other block bordered this first block. In the present study, respondents frym each pair of blocks were aggregated, for a total of 300 local neighborhoods, distributed across the entire city of Seattle (Rountree et al., 1994). Housing units on each block were selected via a reverse telephone directory. Details of the telephone interview procedures and specific limitations of the sampling design are described in Miethe and McDowall (1 993) and Rountree et al. (1 994). This data set contains information on the number of several specific types of victimizations respondents had experienced in their lifetime as well as information regarding the recency of victimization. Further, the data include the rich detail, not available in most other data sets, that is needed to estimate multivariate models of individual- level risk factors and contextual variable5 on repeat and multiple victimization. The total sample in the data set consisted of 5,302 adults, but because data for some observations were missing, the final sample size for the present analyses was 5,049 individuals. Measures a position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. 0 last month respondents carried $50 or more in cash or wore jewelry worth more than $100 in a public place. Family income was used as an additional indicator of target attractiveness for both property and violent victimization. Guardianship represents the ability of individuals or households to prevent crime. Following Rountree et al. (1 994), we used an index reflecting the number of safety precautions the resident employed. These safety precautions included having door locks, leaving lights on, belonging to a crime prevention program, owning a burglar alarm, owning a dog, having neighbors watch the home, and owning a weapon. Additionally, the social dimension of guardianship was captured by whether the respondent lived alone or with other adults (Rountree et al., 1994). Both of these guardianship measures were used in the analyses of both property and violent victimizations. Aside from the demographic, measures and the individual-level indicators from routine activityAifestyle exposure theory, the explanatory variables included neighborhood-level contextual variables. These variables consisted of three factors related to social disorganization theory: neighborhood incivilities, ethnic heterogeneity, and population density. To measure incivilities, using Rountree et al.’s (1 994) method, we computed the number of neighborhood problems that existed within four blocks of the respondents’ homes by averaging responses within each neighborhood. These problems included teenagers “hanging out” on the street, litter and garbage on the street, abandoned houses and buildings, poor street lighting, and vandalism. Higher scores on this variable indicate more disorder. The measure of ethnic heterogeneity was the product of the percentage of white residents and the number of nonwhite residents in each neighborhood. Maximum heterogeneity, therefore, is indicated by a score of .25 (50% white and 50% nonwhite). To measure neighborhood density, for each neighborhood we averaged the 14 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. number of places available for public activity within three blocks of respondents’ homes. Called busy places by Rountree et al. (1 994), such places included schools, convenience stores, bars, fast food restaurants, ofice buildings, parks, shopping malls, hotels, and bus stops. Higher values on this variable indicate greater density. Descriptive information for all individual level and neighborhood level variables is provided in Table 1. Table 1 about here ANALYSIS: HIERARCHICAL MODELS Our test for neighborhood effects on repeat victimization and on multiple victimization proceeded in two general steps. First, we used random coefficient models to establish whether the individual predictors of victimization varied by neighborhood, and, where significant variation existed, to determine the degree of variation in the effects of individual characteristics across neighborhood. Second, we extended our random coefficient models from the first step in the analysis to include the measures of neighborhood context discussed above (incivilities, ethnic heterogeneity, residential mobility, neighborhood income and population density) as predictors of .” T the individual-level coefficients. The results from the second step in our analysis provide information on how individual characteristics interact with contextual characteristics to affect a person’s chances of being repeatedly or multiply victimized. Because we measured repeat property victimization as the number of property victimizations a person had experienced in the prior two years, we estimated a random coefficient Poisson regression model, which is the most appropriate technique for the analysis of count data (see, e.g., Maddala, 1983; McCullagh and Nelder, 1989). Although count data are e position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. often overdispersed (Le., the variance is greater than the mean), unless the overdispersion is severe, the consequences of overdispersion for the estimated parameters are slight (McCullagh 0 and Nelder, 1989). In these data, however, initial results indicated a significant binomial error term, so the error term was included in all of the poisson models to reduce bias in the estimates. Bryk et al. (1 996) have shown how the general poisson regression model can be extended to allow for random coefficients. The general form for this model is In equation (l) , lnhd represents the natural log of the respondent's expected rate of victimization, the (Xkij 'P Xkj) represent values of the independent variables for respondent in neighborhood j centered on the neighborhood mean (Xk,j), the Pkj represent the coefficient estimates for the effect of variable k on the dependent variable for each neighborhood j included in the analysis. Multiple victimization is measured as a dichotomous variable (multiple victim vs. non- multiple victim) which required the use of a random coefficient logit model to estimate the effect of neighborhood variation on the odds of being a victim of both violent and property crimes. The general form for this model is - - logit (Multiple Victim) = Poj + plj(X1ij - X1.j) + P*j(XZiJ - X2j) In equation (2), the coefficients and the measures of the independent variables are defined in the same way as in equation (1). For both the poisson and the logit model, random coefficient models necessary to establish neighborhood variation in the effects of individual characteristics on number and on type of victimization are specified through the following constraints 16 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. account for the neighborhood-level variation detected. Adding the relevant neighborhood characteristics into the model and retaining those that attained standard levels of significance individually produced several significant interactions between individual and neighborhood level factors (see Table 3). a First, the neighborhood differences in the mean number of property victimizations is largely attributable to neighborhood incivilities. In fact, once all three characteristics are entered simultaneously, incivilities is the only one that remains statistically significant. This effect indicates that those who live in neighborhoods with higher levels of incivilities experience more repeat property victimizations than do those who live in neighborhoods with lower levels of incivilities. This finding is consistent with Rountree et al.’s (1994) finding :.:$ -% with regard to single victimization incidents. -2& e+-- .;Jt‘ ? Table 3 about here The effect of sex on property victimizations is also dependent on the level of neighborhood incivilities, although the interaction is not significant in the full model. The sex difference in the number of property victimizations experienced appears only in areas of low or medium incivilities, where males experience a larger number. In areas with high levels of incivilities, there is no substantive sex difference in the number of property victimizations experienced. This pattern is generally consistent with our expectation that , . person-ievel differences in repeat victimization would be strongest in neighborhoods where the overall risk was low. There was also a significant interaction between race and neighborhood ethnic 19 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. heterogeneity. Specifically, nonwhites experience a larger number of property victimizations as the ethnic heterogeneity of an area increases. In areas with low heterogeneity, the effect of race is negligible. Further, once the neighborhood variation in ethnic heterogeneity is taken into account, there is no longer a significant individual effect of race. This effect, although somewhat stronger among repeat victims, is also consistent with the findings of Rountree et 0 al. (1994). Finally, there was also a significant interaction between the use of safety precautions and the number of busy places in the neighborhood. The best way, we think, to interpret this interaction is in terms of guardianship. That is, the highest numbers of property victimizations occur where there are the fewest spaces for public activity (i.e., more private residences, poorer lighting, fewer people out on the street to witness crime). In such places, the number of safety precautions u’sed has a clear negative linear relationship to the number ._ -.. -.-.. :-..- - a-$ of property victimizations experienced. In areas with more busy places and therefore higher guardianship, the opportunity for property victimizations is lower. In such places, the additional benefit from multiple safety precautions becomes negligible. That is, the risk of property victimization in such places is rather low, so that the use of more than one or two safety precautions is relatively unnecessary. Although the effects of income, expensive goods, and age also significantly varied by neighborhood context, none of the social disorganization variables in the present analysis attained statistical significance in the iargcr model. In the hll modei, however, the varimce components for both age and expensive goods become nonsignificant, although the variance components still approach significance (p = .09 and E = .l 1 , respectively). This pattern 0 20 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. indicates that although the effects of these factors are context dependent, the neighborhood- level variation is not the result of incivilities, ethnic heterogeneity, or the number of busy places. a Excludiny nonvictims. In addition to modeling repeat property victimization within the prior two years for the full sample, we also modeled repeat property victimization within the prior two years for only those individuals who had been victims of at least one property crime. That is, we compared one-time property victims (n = 143 1) to repeat property victims (n = 672) using random coefficient models comparable to those we used with the full s z q l e (full results . I available from the authors). This direct comparison indicated that, consistent with the model for the full sample, those who had higher incomes and owned more expensive goods were especially likely to be repeat property victims. Both of these effects were significant. The effect of race also approached statistical significapce (r! = .052), suggesting that those who were white were more likely to be repeat property victims. The effects for age and guardianship (home unoccupied), although not significant, were in the same direction as those in model with the full sample. The difference in the significance for these two variables across the two equations could indicate that age and the amount of time one spends away from home do not distinguish between single and repeat victims as well as they distinguish between victims and nonvictims. However, the lack of significance may also be at least partially attributable to the fact that the model excluding nonvictims had less statistical power to detect effects. The lack of significant interactions in this model may also be attributable to lower statistical power. The size and direction of the coefficients for these effects were similar in the two models, suggesting that low power (only 64 of the neighborhoods had sufficient numbers) 21 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. Multiple Victimization: Prior Two Years 0 As with repeat property victimization within the prior two years, our analyses of multiple victimization within the prior two years tested both the entire sample (including nonvictims) and the subsample of those who had been victimized at least once. We also examined lifetime multiple victimization. Full Sample. In contrast to our results for repeat victimization, the results for multiple victimization indicated no neighborhood-level variation (see Table 4). That is, the mean risk of ' . multiple victimization and the factors that increase a person's risk of being a multiple victir;; are constant across neighborhood context.' While we find no evidence of neighborhood variation in multiple victimization, the analyses indicate important individual characteristics related to the individual risk of multiple victimization. Specifically, those most at risk of being a multiple victim are young people, generally Tales, who participate in a number of dangerous activities. In fact, the odds of being a multiple victim are 1.42 times greater for each additional dangerous activity in which respondents participated. In addition, for each unit increase in the age scale, the odds of being a multiple victim are .24 times lower. Although not quite significant, the results also indicated that the odds of a male being a multiple victim were 1.35 times higher than they were for a female. . A. --- .> *. _.- . ......................... Table 4 about here Excludinn nonvictims. In addition to modeling multiple victimization within the prior two years for the full sample, we also modeled multiple victimization within the prior two years for only those individuals who had been victims of at least one property crime or one violent crime. That is, we compared one-time victims (n = 2041) to victims of both a property and a 0 ' 24 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. violent crime (n = 205) using a random coefficient model comparable to that we used with the full sample (full results available from the authors). This direct comparison indicated, a completely consistent with the model for the full sample, that those who were younger and those who participated in more dangerous activities were especially likely to be multiple victims. The direction of the coefficients for all other effects, which were nonsignificant in both models, were the same and the size of the comparable coefficients was very close in the two models. In the model with the full sample, none of the contextual effects was significant, and this absence of significant effects was also true for the sample excluding nonvictims.2 Multitde Victimization: Lifetime Measures We also examined the full sample using lifetime measures of multiple victimization. The results of this analysis are completely consistent with the results from the full sample examination of multiple victimization within the prior two years. That is, it indicated that individuals who are young, male, and participate in more dangerous activities have a higher risk , ..I of being a multiple victim. It also provided no evidence that multiple victimizations are dependent on neighborhood context. DISCUSSION The results presented here suggest that single victimization, repeat victimization, and multiple victimization are distinct phenomena that must be modeled separately. Consistent with other findings (Reiss, 1980), repeat property victimization was more common than multiple victimization (more than three times as common in our sample). Thus, it seems important to distinguish repeat from multiple victims. Further, based on our findings, it seems clear that demographic and routine activity factors influence each of these phenomena in different ways and that each is differentially dependent on neighborhood context. Moreover, the performance of 0 25 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. the social disorganization factors in the models of repeat victimization suggests that other neighborhood-level factors need to be considered. Finally, the findings reported here suggest that a reducing repeat victimization may not be the key to crime reduction. We elaborate these points below. First, OUT models of repeat victimization indicated that there are significant individual- level predictors for repeat property victimization. Whether analyzed with the full sample or only with those who had been victimized, being white, having higher incomes, and owning more expensive goods were associated with experiencing more victimizations. In other words, we can be fairly confident that repeat property victimizations are partially due to factors about individuals. Second, the models estimated here indicate that there is substantial neighborhood-level . " a:"- ?-.. ., - .I . - L --..: e-, variation in repeat property victimizftion and suggest that such variation is even more pronounced in the case of repeat victimization than is the case with individual victimization incidents. This large amount of neighborhood-level variation suggests that, although victimization is somewhat dependent on context, repeat victimization may be even more heavily dependent on neighborhood context. Unfortunately, because of limitations in the data set, we were unable to test more precise models of repeat victimization of different types. Future research should seek to provide more carefully bounded estimates of repeated incidents of exactly the same type as well as potentially important details regarding the temporal proximity of repeat victimization experiences of the same and different types. Further, it is important to continue to refine our understanding of routine activities and lifestyle factors by collecting more detailed information about individuals' activities during the day and the types of individuals with whom they come into contact. Such 0 26 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. e REFERENCES Bryk, A., Raudenbush, S., & Congdon, R. (1996). HLM: Hierarchical linear and nonlinear modeling with the HLM/2L and HLW3L programs. Chicago: Scientific Software International. Bursick , R., & Grasmick, H. (1993). Neighborhoods and crime. New York: Lexington Books. Buss, T. F., & Abdu, R. (1 995). Repeat victims of violence in an urban trauma center. Violence and Victims, l0, 183- 194. Cohen, L. E., & Felson, M. (1 979). Social change and crime rates: A routine activities approach. American Sociological Review, 44, 588-608. Ellingworth, D., Farrell, G., & Pease, K. (1 995). A victim is a victim is a victim. British .. .. . I Journal of Criminology, 35,360-365. Farrell, G. (1 995). Preventing repeat victimization. In M. Tonry and D.P. Fanington (Eds.) Buildinv a safer society: Strategic approaches to crime prevention. Vol. 19 (pp. 469-534) of Crime and justice: A review ofjustice. Chicago: University of Chicago Press. Hindelang, M., Gottfredson, M., & Garafalo, J. (1978). Victims of Personal crime. Cambridge, MA: Ballinger. Johnson, S . D., Bowers, K., & Hirschfield, A. (1997). New insights into the spatial and temporal distribution of repeat victimization. British Journal of Criminology, 37, 224-24 1. Kelley, H. H. (1972). Attribution in social interaction. In E. E. Jones, D. E. Kanouse, H. H. Kelley, R. E. Nisbett, S . Valins, & B. Weiner (Eds.), Attribution: Perceiving the causes of behavior (pp. 1-26). Morristown, NJ: General Learning Press. 29 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. buritsen, J., Quinet, K. (1 995). Repeat victimization among adolescents and young adults.” Journal of Ouantitative Criminology, 11, 143- 166. 0 Maddala, G. S. (1 983). Limited dependent and qualitative variables in econometrics. New York: Cambridge University Press. McCullagh, P., & Nelder, J. A. (1989). Generalized linear models (2”d ed). New York: Chapman-Hall. Miethe, T. (1 997). Testing theories of criminality and victimization in Seattle, 1960- 1990. [Computer file]. Blacksburg, VA: Virginia Polytechnic Institute [producer], 1991. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. Miethe, T., & McDowall, D. (1 993). Contextual effects in models of criminal victimization. Social Forces, 71,243-266. . - Nelson, J. F. (1 980). Multiple victimization in American cities: A statistical analysis of I rare events. American Journal of Sociology, 85,871-891. Pease, K., & Laycock, G. (1996). Revictimization: Reducing the heat on hot victims. NIJ Research in Action. Washington, DC: US Department of Justice. Polvi, N., Looman, T., Humphries, C., & Pease, K. (1991). The time course of repeat victimization. British Journal of Criminology, & 4 1 1-4 14. Reiss, A. J., Jr. (1 980). Victim proneness in repeat victimization by type of crime.” In S. E. Fienberg & A. J. Reiss, Jr. (Eds.) Indicators of crime and criminal justice: Ouantitative studies. Washington, D.C.: Bureau of Justice Statistics. Robinson, M. (1 998). Burglary revictimization: ‘The time period of heightened risk. British Journal of Criminology, 38, 78-87. 30 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. Rountree, P. W., Land, K. C., & Miethe, T. D. (1 994). Macro-micro integration in the study of victimization: A hierarchical logistic model analysis across Seattle neighborhoods. Criminology, 32,387-414. e Sampson, R. J., & Groves, W. B. (1989). Community structure and crime: Testing social-disorganization theory. American Journal of Sociology, 94,774-802. Sampson, R.J., and J. L. Lauritsen 1990. “Deviant Lifestyles, Proximity to Crime, and the Offender-Victim Link in Personal Violence.” Journal of Research in Crime and Delinauencv 27A 10-139. Shaw, C., & McKay, H. (1 969). Juvenile delinquency and urban areas: Revised edition. Chicago: University of Chicago Press. South, S., & Crowder, K. (1 997). Escaping distressed neighborhoods: Individual, community, and metropolitan influences. American Journal of Sociology, 102, 1040-1 084. Sparks, R. F. (198 1). Multiple victimization: Evidence, theory, and future research. Journal of Criminal Law and Criminology, 72,762-778. Spelman, W. (1 995). Once bitten, then what?: Cross-sectional and time-course explanations of repeat victimization. British Journal of Criminolog, 35,366-383. Trickett, A., Osborn, D. R., Seymour, J., & Pease, K. (1992). What is different about high crime areas?’’ British Journal of Criminolony, 32, 8 1-89. 31 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. e Fixed Effects Neighborhood Mean Number of Victimizations BASE INCNILITIES BUSY PLACES ETHNIC HETERO Age Sex BASE BASE INCIVILITIES Race BASE INCIVILITIES ETHNIC HETERO Home Unoccupied Family Income BASE Expensive Goods BASE Safety Precautions BASE BUSY PLACES Live Alone Random Effects Coefficient Standard Error p-value -.972 .072 .ooo .3 17 .040 .ooo -.O 18 .020 .3 84 -.285 .322 .377 -.145 .015 .ooo -.121 .093 .I97 .120 .294 .182 .lo5 -.085 .096 -378 -2.241 .990 .023 .024 .010 .016 .063 .018 .oo 1 .093 .016 .ooo -.123 .044 .006 .036 .012 .003 -.029 .047 -541 .090 .058 9 Variance p-va I u e Neighborhood Mean Component .079 .ooo TABLE 4: Random Coefficient Model for Multiple Victimization within 2 prior years Number of Victimizations Age Sex Race Family Income Safety Precautions ;eve1 1 Binomial Expensive Goods -. 3Toc 34 .009 .086 .122 .ooo .09 1 .044 .020 .047 .009 .111 .013 .077 .902 --- --- position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. -Fixed Effects, Coefficient Standard Error p-value Neighborhood Mean -3.478 .092 .OW Risk of Multiple Victimization & -.268 , .058 .ooo - Sex .303 .I64 .064 Race -.382 .277 .168 Dangerous .351 .lo1 .oo 1 Activities Home Unoccupied Family Income Expensive Goods Carry Valuables Safety Precautions Live Alone 35 .060 .042 .155 -.095 .074 .198 .073 .062 .24 1 .045 .03 1 .149 .091 .059 .122 .I21 .195 .535 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice. NOTES Ideally, this analysis would have used a hierarchical bivariate probit to capture both the effects of neighborhood 0 1 on violent and property victimization and the relationship between the two types of victimization. However, &re is no current statistical program than can estimate such a model. Instead, we did these analyses in two separate steps. First, the hierarchical logistic regression indicated that neighborhood factors did not significantly affect the risk of multiple victimization. Second, a bivariate probit analysis indicated that there was a small but significant correlation (. 19) between violent and property victimizations. Together, these two analyses suggest that multiple victimization is a meaningful construct but that it does not vary by neighborhood. Because these analyses comparing one-time victims to multiple victims included only 2 1 neighborhoods in which there were sufficient numbers of both types of victims, there is clearly little statistical power to detect contextual effects. ..- . .. . . . .-.-..= ., ~ . ,. . - ,... -.- .. a.,*: ;.. --- 36 position or policies of the U.S. Department of Justice. expressed are those of the author(s) and do not necessarily reflect the official This report has not been published by the Department. Opinions or points of view This document is a research report submitted to the U.S. Department of Justice.