Examining the Applicability of Lifestyle-Routine Activities Theory for Cybercrime Victimization

Abstract A great deal of criminological research has attempted to understand and identify the causes of victimization using the lifestyle-routine activities theory. Recent researchers have argued that the lifestyle-routine activities theory may be able to explain the increasingly significant phenomenon of computer and cybercrime. This claim has been contested by Yar (Citation2005), however, who argues that routine activities theory is limited in explaining cybercrime. Few empirical tests exist to address this important issue. Thus, this study attempts to explore this gap in the research literature by examining a specific form of cybercrime, on-line harassment. Using a sample of students at a southeastern university, this analysis found some support for elements of lifestyle-routine activities theory. Individual and peer involvement in computer crime and deviance also significantly increased the risk of victimization. The implications of these findings for theorists and researchers are also explored. Acknowledgments The authors thank Janet Lauritsen, Michael G. Turner, Robert Brame, and the anonymous reviewers for their comments and assistance on previous drafts. Also, this work was supported, in part, by funds provided by the University of North Carolina at Charlotte. Notes 1There are several venues for real time communication on-line, including chatrooms, instant messaging, and Internet Relay Chat (IRC), although there are distinct differences across these outlets (Taylor et al. Citation2006). For example, chatrooms are often Web-based and allow groups of people to communicate simultaneously. All messages, however, can be seen by any person logged into the chat room. Alternatively, instant messaging services run through stand alone software programs like AIM or Yahoo and send messages directly to a single individual rather than groups. Finally, IRC operates on a separate portion of the Internet and special software is needed to access the system. Individuals must then log in to category-based channels, such as sports or movies. IRC provides group-based chat functions much like chatrooms, although private messages can also be created in the same fashion as instant messaging services (see Taylor et al. Citation2006). 2Of the 578 cases analyzed, 13.7% reported being victimized 1–2 times, 4% reported 3–5 times, .5% (n = 3) reported being victimized 6–9 times, and .7% (n = 4) reported 10 or more victimizations. In our initial analyses, we collapsed the last three categories into one category (three or more) because of the small n in these categories and created a three-point ordinal measure. We then ran ordered logistic regression models. Because of a large number of independent variables that are not dichotomous and only 30 respondents reporting three or more victimizations, 66.7% of the cells had zero frequencies. In order to address this empty cell issue, but not wanting to lose variation in our independent variables, we collapsed the dependent variable into two categories (0 = non-victimization; 1 = victimization) and ran logistic regression models. The logistic regression models (as presented in the tables) are substantively similar to the ordered logistic regression models. Additionally, we would not have been able to partition the ordered logistic regression models by sex because of the small number of males who have been victimized and the large number of independent variables. Tobit regressions were not performed because our dependent variable consisted of a single ordinal item that can be examined via ordered logistic regression. Once the dependent variable was collapsed into a dichotomous measure, and 18.9% of the sample fell into one category, special forms of regression such as tobit or rare-case logistic regression were not necessary because logistic regression can adequately handle this distribution. 3A fourth category was provided for this question, stating "I am afraid of computers and don't use them unless I absolutely have to." Only one student in the data set and no students in the 578 cases analyzed indicated they fell into this category, indicating that this sample is relatively computer literate. 4For the measure assessing how many hours per week the respondents spend on a computer for work or school, the respondents reported that 26.6% spent less than 5 hours, 32.7% spent 5–10 hours, 17.5% were on the computer 11–15 hours, 8.5% reported being on the computer 16–20 hours, and 14.7% reported 21 hours or more. Regarding the number of hours on a computer outside of work or school, 24.6% reported being on a computer less than 5 hours, 33.2% reported 5–10 hours, 16.4% spent 11–15 hours, 10.2% reported 16–20 hours, and 15.6% reported 21 or more hours. p ≤ .05*, p ≤ .01**, p ≤ .001***. Model 1: Chi-square of 68.136*** and − 2LL of 491.552; Model 2: Chi-square of 72.293 and − 2LL of 487.395. p ≤ .05*, p ≤ .01**, p ≤ .001***. Male sample: Chi-square = 46.355*** and − 2LL of 147.300; Female sample: Chi-square of 48.733*** and − 2LL of 308.993. 5The original survey question also included an option of "all of them." Because of the small number of respondents who reported all of their friends pirated software or committed any of the "hacker-like" behaviors, "all of them" were combined with "more than half." All models were also run with a scale created from the non-recategorized questions; these models were substantively similar to the results in Tables 2 and 3. 6These measures do not create a reliable scale (alpha = .5053). We ran additional analyses for the full sample and for each of the subsamples with the scale included instead of the seven separate measures. The scale was not significantly related to on-line harassment. Therefore we included the seven separate measures instead of incorporating an unreliable scale. 7Respondents were asked to identify themselves as white, African American, Hispanic, Asian, or another racial group. Asians, Hispanics, and other race categories only comprise 5.2%, 2.8%, and 3.1% of the respective sample. The original models analyzed included separate dummy measures for each group; however, no race measure was significant. Thus, these race groups were combined into one category to simplify the model. This combination of groups did not affect any results. 8Multicollinearity was not an issue for the analysis using the full sample (n = 578 cases) or for the analyses partitioned by sex. No VIF was over 10 and no tolerance level was below .2. The two measures with the lowest tolerance levels and highest VIFs are the computer deviance scale (tolerance of .513 and VIF of 1.951) and the friends' computer deviance scale (tolerance of .534 and VIF of 1.874). The issue was slightly worse for the male sample analysis (computer deviance scale had tolerance of .497 and VIF of 2.012 and friends' computer deviance scale had tolerance of .512 and VIF of 1.954), but these scores do not raise concerns. 9To further examine whether skill level is related to on-line harassment, we also ran models using dummy measures assessing whether they have used and can operate Windows 95/98, Windows NT/2000, Windows XP, Macintosh, UNIX, and Linux (adapted from Rogers Citation2001). We ran zero-order models and full models (including and excluding the skill level measure) for the full sample and the male and female subsamples. In no model did the ability to run a certain program affect on-line harassment, nor did the inclusion of these measures affect any of the results presented in the analyses section. Additionally, the six dummy measures do not create a reliable scale (alpha = .5424) and cannot be used as a measure of computer skill. 10To further examine this issue, we ran all full models without the specific computer activity measures (shopping, video game, e-mail, chatrooms, and programming) to further assess whether a relationship between general hours on the computer and cyberspace harassment victimization exists. We also ran models including only one of the general hours on the computer measures. We believed that this would provide the "general hours on the computer" measure the best opportunity to be statistically significant. In no model was a significant relationship between general hours on the computer (for work and school and outside of work and school) and chatroom victimization found. 11Pirating software and media are not included together because they create an unreliable scale (a = .5681). Furthermore, pirating software and media do not predict victimization similarly. The three hacking items do create a reliable scale (alpha = .8585). 12The pseudo R-square value for the male model (.316) is greater than the female (.207) and full models (.179). It must be noted that the pseudo R-squares presented are Nagelkerke R squares. If Cox and Snell R squares were presented, the pseudo R-squares for the two models would be more comparable (male = .172; female = .136). Although the z-tests only showed a significant difference between skill level, several measures ownership, computer speed (dial-up and T-1), and race (African American and other) were larger in one model over the other, but were not found to be significant across models because of large standard errors. To examine whether the inclusion of these measures explains the disparity, we ran additional models excluding them. These analyses illustrated that much of the difference can be explained by the three types of variables listed earlier. When computer speed is excluded, the male model has a Nagelkerke R-square of .290 and a Cox and Snell R-square of .159. The female model's Nagelkerke R-square is .206 with a Cox and Snell of .136. When computer speed and computer ownership are excluded, the male model's Nagelkerke R-square drops to .250 and the Cox and Snell is .137. The female model's Nagelkerke and Cox and Snell remain stable at .205 and .135, respectively. Finally, when all three categories of variables are excluded, the Nagelkerke for the male model continues to drop to .233 with a Cox and Snell of .127. The female model's Cox and Snell is actually larger than that of the males (.130) and the Nagelkerke only drops to .198. As a result, much of the disparity in pseudo R-squares can be explained by the decision to provide Nagelkerke R-squares and that the measures of computer speed, ownership, and race helped explain a small number of male victimizations and not female victimizations, though they are not significant across models. p ≤ .05*, p ≤ .01**, p ≤ .001***. Chi-square = 141.726; − 2LL = 585.715. Additional informationNotes on contributorsThomas J. Holt Thomas J. Holt is an Assistant Professor of Criminal Justice at the University of North Carolina at Charlotte. He has a doctorate in criminology and criminal justice from the University of Missouri—Saint Louis. His research focuses on computer crime, cybercrime, and the role that technology and the Internet play in facilitating all manner of crime and deviance. Adam M. Bossler Adam M. Bossler is an Assistant Professor of Criminal Justice at Armstrong Atlantic State University. He has a doctorate in criminology and criminal justice from the University of Missouri—St. Louis. His research interests include testing criminological theories that have received little empirical testing such as control balance theory, evaluating the application of traditional criminological theories to cybercrime offending and victimization, and evaluating policies and programs aimed at reducing youth violence.