Abstract
Cyber fraud victimization among adolescents remains mechanistically underexplored despite growing global concern. Using self-reported data from 929 Chinese high school students, this study employs structural equation modeling informed by the general theory of crime, lifestyle–routine activity theory, and online risk-taking. Treating victimization as sequential—fraud targeting followed by financial loss—findings reveal that a deviant online lifestyle is the strongest predictor across both stages. After accounting for online risk-taking, the direct effects of low self-control on cyber fraud victimization become non-significant. Paradoxically, non-deviant online routines increase targeting exposure while reducing financial loss likelihood. These findings highlight the importance of risk-oriented mechanisms in explaining adolescent cyber fraud victimization.
Introduction
Cyber fraud offenses have exhibited significant growth in recent years, accounting for 30% to 60% of total crime incidents in many countries worldwide, including various regions of China (Government Information Bureau, 2025; Hong Kong Police Force, 2025; Sina News, 2024). At its core, cyber fraud offenses seek to exploit individuals’ vulnerabilities through deception, manipulation, and coercion to obtain money or personal information (F. Wang & Zhou, 2023; J. Xu et al., 2024; Zhou et al., 2023). In this context, individuals’ levels of self-guardianship play a critical role in mitigating their experiences with such victimization (Button & Cross, 2017; Chang et al., 2023; Vakhitova et al., 2023). Although prior research has emphasized the importance of criminological theory in explaining cyber fraud victimization, few studies have systematically examined the mechanisms linking different theoretical components in both predicting and interpreting cyber fraud victimization.
To date, limited scholarship has investigated the interrelated pathways between various criminological theories and cyber fraud victimization. While existing studies have frequently assessed the explanatory capabilities of mainstream criminological theories—such as the general theory of crime (M. R. Gottfredson & Hirschi, 1990), routine activity theory (L. E. Cohen & Felson, 1979), and lifestyle-exposure theory (Hindelang et al., 1978)—most have examined these theoretical components in parallel, focusing on their independent predictive power without specifying the underlying relationships among them (Chen & Xia, 2025; Lin, Wu, et al., 2025; Mesch & Dodel, 2018). Examining these theoretical interrelations allows researchers to identify how individual propensities (e.g., low self-control) may shape risky or deviant online lifestyles, which in turn influences exposure to motivated offenders. Failure to address these underlying relationships not only limits a holistic theoretical understanding of cyber fraud victimization but also constrains the development of prevention strategies suited to the evolving nature of cyber fraud behaviors.
Given this gap in the extant literature, the current study examined a large sample of Chinese high school students (N = 929) from Shandong Province, China, to explore the theoretical predictors of cyber fraud victimization and assess the theoretical interrelations among these underlying components. Specifically, a structural equation model (SEM) was conducted drawing on the general theory of crime (M. R. Gottfredson & Hirschi, 1990), routine activity theory (L. E. Cohen & Felson, 1979), and lifestyle-exposure theory (Hindelang et al., 1978) to examine two stages of cyber fraud victimization: fraud targeting and fraud resulting in financial loss (hereafter “financial loss”). In addition to these theoretical perspectives, the concept of online risk-taking was incorporated to provide a more comprehensive assessment of the underlying victimogenic mechanisms (Weber, 1998; Weber et al., 2002; Zuckerman, 1994). By testing the mediating relationships among these theoretical constructs, the current study offers a context-specific examination of cyber fraud victimization among Chinese adolescents in a rapidly digitizing environment.
To that end, this study presents four novel research findings: (1) the diminished interpretive value of non-deviant online routine activities, (2) the robust predictive power of online deviant lifestyles, (3) the salient mediating role of online risk-taking, and (4) the non-significant effect of low self-control. Taken together, these observations highlight the importance of moving beyond conventional victimological frameworks and toward a risk-focused conceptualization of cyber fraud victimization, particularly within adolescent online contexts (Engström, 2021; Kulig et al., 2017; Pratt & Turanovic, 2016; S. Y. K. Wang et al., 2025).
Two Stages of Cyber Fraud Victimization: Fraud Targeting and Financial Loss
There has been considerable debate within the academic literature about whether direct financial loss or the loss of valuable goods is required to constitute cyber fraud victimization (Chen & Xia, 2025; Fan & Yu, 2021; Holtfreter et al., 2008). Proponents of this standard are often influenced by legal statutes that consider the incurrence of tangible damage as a prerequisite to fraud victimization (e.g., the Computer Fraud and Abuse Act in the USA). While it is understandable that legal practice requires a degree of tangibility and precision to avoid the potential abuse of the “Rule of Lenity” (Price, 2004) by offenders, this definition does not account for any intangible damage equally consequential to those targeted for fraud, such as psychological, relational, and reputational harms (Button & Cross, 2017; Maher & Hayes, 2024; Nevin et al., 2025). Given the complexity of online victimization and its widespread consequences, the current study embraces a broader definitional framework by considering fraud targeting as a distinct stage and form of cyber fraud victimization.
A growing number of empirical studies have also attempted to differentiate fraud targeting from financial loss. Aligning with antecedent studies, fraud targeting often captures individuals’ exposure to fraudulent attempts (e.g., receiving fraudulent messages or deceptive offers), whereas financial loss denotes the outcome when such attempts succeed in eliciting monetary or personal information transfer (Chen & Xia, 2025; Fan & Yu, 2021; Holtfreter et al., 2008; Lin, Wu, et al., 2025; Mesch & Dodel, 2018; Pratt et al., 2010). While both share common attributes—such as exposure, harm, and manipulation of trust—they diverge in underlying mechanisms: fraud targeting is more closely linked to situational visibility and routine activity patterns, whereas fraud resulting in loss involves psychological and emotional processes, such as impulsivity, persuasion, and compliance (Holtfreter et al., 2008; Pratt et al., 2010). This dichotomy acknowledges that being a target of cyber fraud is a form of harm regardless of the presence of economic damage.
Specifically, although cyber fraud research often treats financial loss as the defining outcome of victimization, the initial act of fraud targeting constitutes a meaningful stage in the victimization process. From a victimological perspective, crime is not limited to completed offenses but includes attempts and exposures that produce psychological, behavioral, and cognitive harms (Button, Lewis, & Tapley, 2014; Cross, 2015; J. Wang et al., 2024). In this way, individuals who are targeted by fraudulent schemes may experience stress, uncertainty, and reduced trust in online environments, even when no financial loss occurs (Buchanan & Whitty, 2014; Cross et al., 2016; J. Wang et al., 2024). Moreover, fraud targeting is rarely a randomized offense, as perpetrators frequently select individuals based on perceived vulnerabilities, resulting in patterned exposure across demographic and behavioral groups (Button, Nicholls, et al., 2014; Deevy et al., 2012). As such, fraud targeting represents the first stage of a broader victimization trajectory and a necessary precursor to financial loss. Examining this stage of the cyber fraud process as a distinct form of victimization provides important insights into risk exposure, offender decision-making, and opportunities for early intervention.
Nevertheless, the theoretical mechanisms shaping different stages of cyber fraud victimization remain largely underexplored. Without distinguishing fraud targeting from financial loss, existing research risks a biased theoretical understanding of cyber fraud victimization by conflating risk exposure with harm realization. This lack of stage-specific theorization also undermines the effectiveness of prevention strategies and victim support, as individuals who experience financial loss may exhibit qualitatively different vulnerabilities (e.g., heightened risk-taking) relative to those who are solely targeted. Given these concerns, the current study differentiates fraud targeting from financial loss and examines how each stage can be explained by relevant criminological theories, including the general theory of crime, lifestyle–routine activity theory, and the concept of online risk-taking.
Self-Control and Cyber Fraud Victimization
The general theory of crime, also commonly referred to as self-control theory, posits that individuals with low levels of self-control are more likely to engage in both criminal and deviant (analogous) behaviors (M. R. Gottfredson & Hirschi, 1990). Specifically, M. R. Gottfredson and Hirschi (1990) specified six dimensions of low self-control (i.e., impulsivity, preference for physical tasks, risk-seeking, simple-task oriented, self-centered, and short-tempered) that they argued were developed due to inadequate child-rearing behaviors. More recently, M. Gottfredson and Hirschi (2019) proposed a more nuanced interpretation of self-control, referring to “the tendency to forego immediate or near-term pleasures that also have negative consequences for the actor and the tendency to act in favor of longer-term interests” (p. 6). To that end, victimization has been identified as one of the “negative consequences” caused by risky, short-sighted, and gratification-driven behaviors, with compromised capability to prevent individuals from being targeted and victimized by predatory offenders (see Berg & Schreck, 2022; L. E. Cohen & Felson, 1979; M. Gottfredson & Hirschi, 2019; Schreck, 1999).
Antecedent literature has examined the links between low self-control and both forms of cyber fraud victimization. In terms of cyber fraud targeting, earlier studies found mixed results regarding the impact of low self-control. For one, using a sample of 922 adult respondents residing in Florida, Holtfreter et al. (2008) found a non-significant association between low self-control and consumer fraud victimization. More recently, Mesch and Dodel (2018) found that low self-control exerted a significant effect on cyber fraud targeting, even after controlling for online routine activities (e.g., online shopping). Compared to cyber fraud targeting, research on fraud resulting in financial loss reveals more consistent observations. Studies have found that the direct effect of low self-control on fraud loss remains significant even after accounting for confounding factors, including demographic traits, online routine activities, and online deviant behaviors (e.g., Chen & Xia, 2025; Holtfreter et al., 2008; Mesch & Dodel, 2018; Reyns & Randa, 2020; Zhang & Ye, 2022).
Although it is reasonable to maintain the assumption of a non-significant relationship between low self-control and cyber fraud targeting, the increasing digitalization of society may introduce new parameters that contribute to covert and unexplained residual variance behind this direct effect. For instance, the decentralized nature of various data and digitization of sensitive information has facilitated a range of cyber fraud opportunities through hacking offenses, data breaches, and data trafficking (Cross & Holt, 2023; Lee, 2023; Smith, 2010), where low self-control—as a personality vulnerability—interacts with negligence, agreeableness, and cognitive biases to increase the odds of being both targeted and incurring financial loss by motivated offenders (Schreck, 1999; Shi & Li, 2023). Thus, it is reasonable to expect significant associations between low self-control and both forms of cyber fraud victimization, leading to the first research hypothesis (
Lifestyle-Routine Activities as a Mediator: Applications and Limitations
Studies have also examined tenable mediators between low self-control and cybercrime victimization generally, such as lifestyle-routine activities (Herrero et al., 2021; Partin et al., 2022). Lifestyle-routine activity theory is an integrated framework combining lifestyle-exposure theory (Hindelang et al., 1978) and routine activity theory (L. E. Cohen & Felson, 1979). The framework posits that individuals who adopt certain lifestyles or routine activities may amplify their exposure to criminogenic circumstances (i.e., enhance their target suitability, increase their proximity to motivated offenders) and ultimately increase their likelihood of being victimized (e.g., Garofalo, 1987; Miethe & Meier, 1990). This framework has been extensively applied to understand a wide range of cybercrime victimization behaviors (e.g., Gainsbury et al., 2019; Hsieh et al., 2021; Holt et al., 2020; Holtfreter et al., 2008; Mesch & Dodel, 2018; Pratt et al., 2010).
In the context of cyber fraud, online lifestyle-routine activities may play a salient mediating role in facilitating victimization. Studies have found that individuals with lower levels of self-control tend to engage in deviant online activities—such as trading personally identifiable information for profit and engaging in doxxing activities—which in turn enhances the risk of cyber fraud victimization (Herrero et al., 2021; Lin, Wu, et al., 2025; Partin et al., 2022; Xin et al., 2026). Lifestyle-routine activity theory provides an appropriate framework to understand the specific rationales for the effects of lifestyle-routine activities on victimization—that is, victims’ exposure, target suitability, and proximity to motivated offenders can be impacted by their frequent exposure in risky situations induced by low self-control, making them more vulnerable targets to cyber fraud victimization (Herrero et al., 2021; Miethe & Meier, 1990). Nevertheless, extant literature has not examined the mediating role of online lifestyle-routine activities by differentiating cyber fraud targeting from fraud resulting in financial loss. As a result, it is unclear whether the mediating effect on financial loss remains significant after accounting for cyber fraud targeting, necessitating a more nuanced investigation into this inquiry.
It is important to note that the general application of lifestyle-routine activities has been challenged by various scholars (e.g., Engström, 2021; Pratt & Turanovic, 2016; Tittle, 1995). First, researchers have critiqued the broader definitions of lifestyle and routine activities, arguing that the theoretical focus should be leveraged onto the “conception of risk” rather than non-risky counterparts (Pratt & Turanovic, 2016, p. 346). Second, based on a systematic literature review of 101 studies, Engström (2021) identified inconsistencies in how lifestyle/routine activity has been operationalized, often measuring occasionally performed behaviors rather than habitually engaged activities. These concerns emphasize the need for theoretical and methodological specifications in applying these core concepts to victimization behaviors. Theoretically, deviant lifestyle-routine activities should be differentiated from non-deviant activities (Pratt & Turanovic, 2016; Tittle, 1995). Methodologically, the operationalization of lifestyle/routine activities should be framed in terms of frequency, such as “how many times a week do you use the dark web” (Engström, 2021). These implications also inform the current study to evaluate the roles of both deviant and non-deviant lifestyle-routine activities in cyber fraud victimization, informing the second and third research hypotheses (H2 & H3):
Online Risk-Taking: Bridging Self-Control and Online Deviant Lifestyles
While extant literature has examined the direct and indirect effects outlined above, studies exploring the underlying mechanisms linking low self-control and deviant lifestyles remain scarce. This gap is theoretically important, as it leaves unclear the processes through which stable individual traits translate into patterned, high-risk online behaviors. To address this issue within the cyber fraud literature, the present study incorporated the concept of risk-taking as a key intervening mechanism between low self-control and online deviant lifestyles.
Risk-taking has been frequently examined in the field of psychology to denote an individual’s inclination to engage in behaviors that involve the potential for negative consequences in pursuit of uncertain outcomes (Zuckerman, 1994). Different from self-control, which is prescribed in the criminological literature as a life-persistent trait (Burt, 2020; M. R. Gottfredson & Hirschi, 1990, p. 90), risk-taking is defined as a domain-specific choice inclination that may vary across situations (Weber et al., 2002; Zuckerman, 1994). According to Weber et al. (2002), risk-taking is composed of two correlated yet conceptually distinct components: risk perception and risk attitude. Risk perception refers to an individual’s cognitive evaluation of the magnitude or likelihood of potential harm, which is shaped by factors such as domain-specific familiarity, perceived controllability, and prior experiences (Weber, 1998). Risk attitude, by contrast, reflects an individual’s dispositional orientation or preference toward engaging in behaviors despite their perceived riskiness, and is often conceptualized as shaping the weight individuals assign to perceived risks when making decisions (Weber et al., 2002). Importantly, these components are conceptually separable. For example, individuals may accurately perceive a behavior as risky (i.e., risk perception) yet differ in their willingness to engage in it, depending on their underlying risk attitudes.
Although self-control and risk-taking are two separate concepts, they are related in various ways, especially within online contexts. For one, the parameters of low self-control may affect individuals’ perceptions of risk in online settings. For example, characteristics of low self-control (e.g., impulsivity and simple-task orientation) are correlated with limited information collection (Hofmann et al., 2009) since executive functioning requires cognitive investment, which is energy-consuming (McCarthy, 2002, p. 420). To that end, individuals who process limited information are less likely to be comprehensively aware of negative consequences, raising the likelihood of making risky decisions (Slovic et al., 2016). This association has been exacerbated in online environments where emerging technologies not only generate an overwhelming volume of information but also enable digital impersonations that are increasingly challenging to distinguish (Chang et al., 2023; Cross & Layt, 2022; Zhou, Liu, et al., 2024).
More importantly, the concept of risk-taking provides theoretical justification for mediating the progressive trajectory from “low self-control as a trait” to “deviant behaviors as a lifestyle.” Conceptually, risk-taking operates as a proximal decision-making orientation that may precede and facilitate the development of patterned deviant routines, rather than constituting those routines themselves. 1 According to self-determination theory (Deci & Ryan, 1985), individuals’ behavioral patterns are developed and maintained by intrinsic and extrinsic motivations. Intrinsic motivation refers to behavioral inclination that originates from engagement in an activity itself without the need for external value indoctrination, as exemplified by hedonistic enjoyment (Deci & Ryan, 1985; Ryan & Deci, 2000). In contrast, extrinsic motivation refers to behavioral inclination driven by external reinforcements that are separable from the activity, including peer influence and social norms (Deci & Ryan, 1985). For individuals with lower levels of self-control, they may indicate a greater level of deviant intrinsic motivations, such as the propensity for hedonistic hacking and digital vandalism (Seigfried-Spellar & Treadway, 2014). Relatedly, as indicated by situational action theory (SAT), individuals with low self-control tend to experience deviant extrinsic reinforcement (e.g., deviant peer norms and subculture), which increases their probability of exhibiting deviant lifestyles (Wikström, 2004, 2006). The reinforced deviant intrinsic and extrinsic motivations collectively reshape individuals’ decision-making calculus, forging the lifestyle accustomed to engaging in deviant behaviors, which is not exclusive to the online context.
The current study provides a comprehensive framework to evaluate the developmental connection between low self-control and deviant lifestyle-routine activities in the context of cyber fraud victimization. It is reasonable to assume the direct effects of online risk-taking on both forms of cyber fraud victimization, especially given the increased propensity of exposure to risky online environments and situations. Based on the aforementioned theoretical associations, we propose the fourth and fifth research hypotheses (
All hypotheses included in the structural equation model are illustrated in Figure 1 (see below).
The hypothetical model.
Data and Methods
Data Collection and Procedures
Data were collected between October 2019 and January 2020 in a city located within Shandong Province, China. An online questionnaire was developed using Sojump (equivalent to SurveyMonkey), which is a widely used professional survey platform in China that has been adopted in numerous studies (Lin, Zhou, et al., 2025; Zhou, Tiwari, et al., 2024). Using a convenience sampling method, the survey was distributed via a Sojump link to students in Years 1 to 3 (equivalent to grades 10–12 in the U.S.) at three local high schools in Shandong Province, which were selected through a co-author’s connections. Participants filled in the questionnaire via the link without registering or providing any personal identifiers. No monetary incentives were provided for participation in the online survey. Researchers manually checked completion times and response patterns (e.g., identical responses) to ensure that the data were valid and reflected genuine human participation.
The data collection procedure was guided and supervised by the research team, following ethical protocols approved by a co-author’s affiliated institution. Specifically, two researchers (including one research assistant) were assigned to each classroom to introduce the study and oversee the research process (32 classrooms were approached in total). Students were instructed to use their personal electronic devices (e.g., smartphones, computers, and tablets) to complete the online questionnaire, though additional electronic devices (e.g., mobile phones) were made available by the research team for survey completion when a respondent did not have their own device or when technical issues with one’s personal device emerged. Each class was allocated 45 min to ensure adequate time for survey completion. Completed questionnaires were automatically uploaded to a password-protected Sojump account with multi-factor authentication and subsequently transferred to a secure cloud storage system.
Participation in the study was voluntary, and all survey responses were anonymized to ensure no personally identifiable information could be traced back to the respondents. Informed consent was obtained from all participants before participating in the study, and parental approval was secured for those under the age of 18. While a total of 965 students aged 14 to 20 years (M = 16.5, SD = 1.7) completed the online questionnaire (with a response rate of 72.8%), the final analytic sample comprised 929 respondents after removing ineligible cases (e.g., participants who reported never using the Internet) and those with missing values on key variables.
Measures
Dependent Variables
Two forms of cyber fraud victimization representing different stages of the victimization process were assessed as dependent variables in the current study. First, fraud targeting was measured by asking respondents to indicate their level of involvement (0 = never; 3 = several times a week) with the following item: “In the last three months, how often have you received the message of online and/or telecommunication fraud?” Financial loss was similarly measured by asking respondents to indicate their level of involvement (0 = never; 3 = several times a week) with the following item: “In the last three months, how often have you experienced financial loss in online and/or telecommunication fraud?” Given the skewed distribution of these measures, both variables were dichotomized, with “0” indicating no experiences with being targeted or experiencing financial loss due to online or telecommunication fraud and “1” indicating having any experiences with being targeted by cyber fraud or experiencing financial loss derived from the cyber fraud incident(s). Preliminary distributional diagnostics indicated substantial positive skewness and sparse observations in the higher-frequency categories, particularly for financial loss. Retaining the ordinal structure produced unstable estimates and limited variability in the upper categories within the SEM framework. Accordingly, dichotomization was employed to enhance model stability and to distinguish between the presence versus absence of each form of victimization. To that end, 60% (n = 557) of respondents self-reported being targeted by cyber fraud within the past three months, while 11% of the respondents (n = 102) reported experiences with financial loss associated with their cyber fraud incident(s) within the same timeframe.
Independent Variables
Several theoretical and conceptually relevant variables were measured to assess their relationship with both fraud targeting and financial loss derived from cyber fraud incidents. First, low self-control was measured using Grasmick et al.’s (1993) Low Self-Control Scale. Aligning with M. R. Gottfredson and Hirschi’s (1990) conceptual framework, Grasmick and colleagues utilized 24 items to measure the six dimensions of low self-control, with response values ranging from 1 (strongly disagree) to 4 (strongly agree). Examples of items that appeared in the scale include “sometimes, I will take a risk for fun,” “I lose my temper pretty easily,” and “excitement and adventure are more important to me than security.” Confirmatory factor analysis demonstrated a good model fit in this study, with a Comparative Fit Index (CFI) of 0.979, Tucker–Lewis Index (TLI) of 0.972, and root mean square error of approximation (RMSEA) of 0.040 (see Section 2.4 for details of index interpretations). Relatedly, the Cronbach’s alpha for the measure was .863, indicating strong reliability.
Online deviant lifestyle was measured by self-reported engagement in 12 online rule-violating activities. Aligning with its operationalization in previous literature (e.g., Choi & Lee, 2017; Engström, 2021; Lin, Zhou, et al., 2025), the current study asked individuals to indicate their weekly engagement in several online deviant behaviors, including cyberbullying, unauthorized access, digital piracy, online stalking, identity theft, doxxing, dark web access, and interacting within online illicit markets. The scale employed binary responses (0 = no, 1 = yes), with a higher overall score indicating a greater engagement in online deviant behaviors. The Cronbach’s alpha of this index scale was .858, indicating strong internal consistency.
It is important to note that although these behaviors vary in severity and specific motivations, they were aggregated to capture a broader lifestyle orientation characterized by sustained engagement in rule-violating online behavior. From a lifestyle-routine activities theory perspective, these behaviors share overlapping opportunity structures in that they increase (a) time spent in unregulated or loosely supervised digital spaces, (b) interaction with deviant peers or subcultures, (c) familiarity with anonymity-enhancing tools, and (d) greater exposure to motivated offenders operating within online spaces. Therefore, this online deviant lifestyle index reflects patterned involvement in deviant online contexts rather than equivalence in the seriousness of individual behaviors.
Online risk-taking was measured using a scale developed by Shadel et al. (2014). The scale contained seven items measuring an individual’s likelihood to participate in various risky online behaviors, including how likely a respondent would: (1) “tell your personal information to unknown persons online”; (2) “make profits by selling personal or others’ information online”; (3) “apply for an unknown online part-time job”; (4) “click the URLs from unknown SMS and/or emails”; (5) “trade with others via unknown online platforms”; (6) “scan QR codes from unknown sources”; and (7) “download documents from unknow online websites.” Response options were provided using a five-point Likert scale ranging from “extremely unlikely” (1) to “extremely likely” (5). An exploratory factor analysis demonstrated that the seven items significantly loaded onto a single factor, with all standardized factor loadings above 0.7, underscoring good convergent validity. A confirmatory factor analysis further demonstrated good model fit on CFI (0.991) and TLI (0.987) scores, and a nearly good model fit on RMSEA (0.065), providing evidence for structural validity. The Cronbach’s alpha for this measure was .858, indicating strong reliability.
Covariates
Frequency, duration, and age of first Internet use were measured to assess respondents’ overall online exposure and digital familiarity 2 (Leukfeldt & Yar, 2016; Pratt et al., 2010). These indicators capture the intensity and developmental timing of Internet use rather than the specific content of online activities, and are commonly included as controls to account for differences in digital familiarity and baseline exposure to online environments. In the present study, Internet use frequency was measured by asking respondents to indicate “How often do you use the Internet?” Response values ranged from “1” (about once a month) to “5” (every day). Due to a skewed distribution, the variable was recoded as a binary measure (0 = not daily, 1 = daily). Similarly, duration of Internet use was measured by asking respondents to indicate, “How much time do you usually spend online every day?” Responses ranged from “1” (less than 3 hr) to “6” (more than 15 hr). Lastly, first time Internet use was captured by asking respondents to indicate “When was the first time you used the Internet?” A 7-point response scale ranging from “1” (kindergarten) to “7” (later than high school) was provided.
Relatedly, a measure assessing non-deviant online routine activities was included to capture respondents’ engagement in non-delinquent online behaviors. This variable reflected individuals’ engagement in mundane online activities, such as using the Internet for secure entertainment, online learning, and secured commercial purposes. Conceptually, this measure aligns more closely with the lifestyle-routine activity theory perspective, which emphasizes the nature and context of everyday activities rather than overall exposure levels. 3 Specifically, nine items with binary responses (0 = no, 1 = yes) were summed to capture participants’ non-deviant routine online behaviors. Examples of items included: “Do you use social media platforms (e.g., WeChat) daily?”, “Do you use online payment applications (e.g., Alipay) daily?” and “Do you play online games daily?” A composite score was calculated by summing the number of “yes” responses, with higher scores indicating greater engagement in non-deviant online routine activities. The Cronbach’s alpha of this index scale was .776, indicating acceptable internal consistency.
To measure respondents’ level of social guardianship, the present study assessed the extent to which parents regulated their children’s online activities. Specifically, parental guardianship was measured by asking respondents to indicate whether their parents imposed any of the following rules or regulations on their online behaviors: (1) “Have your parents set a rule limiting your total online time”; (2) “Have your parents set a rule restricting the latest time you are allowed to use the Internet”; (3) “Have your parents restricted the locations where you can access the Internet”; (4) “Have your parents set rules about the types of websites you are allowed to visit”; and (5) “Have your parents imposed any other Internet-related rules?” Binary response options were provided for each item (0 = no, 1 = yes). Responses to all five items were summed to create an additive scale, with higher scores reflecting greater levels of parental guardianship. The Cronbach’s alpha of this index scale was .853, indicating acceptable internal consistency.
A measure of digital guardianship was also assessed by asking respondents how many anti-virus applications they had installed on their mobile phones. This concept aligns with the notion of target hardening under the framework of L. E. Cohen and Felson’s (1979) routine activity theory (see also Choi, 2008; Reyns et al., 2011). Specifically, respondents were asked about their use of seven types of anti-virus tools, including (1) antivirus programs, (2) email filters, (3) password managers, (4) firewalls, (5) intrusion detection systems, (6) system safety software, and (7) other similar applications. Each item was measured using a binary scale (0 = no, 1 = yes). The total score was calculated by summing the number of installed applications, with higher scores indicating a higher adoption of protective measures. The Cronbach’s alpha of this index scale was .736, indicating acceptable internal consistency.
Lastly, the study included several demographic variables as control measures, including gender, age, household type, academic performance, family income, parents’ education levels, and parental relationship. Gender was coded as a dichotomous variable (0 = male, 1 = female) while age was calculated by asking respondents to indicate their date of birth. Household type was classified as either agricultural (0) or urban (1), while academic performance was assessed on a 5-point scale, with “1” representing the bottom 20% and “5” representing the top 20%. Similarly, family income was measured on a scale ranging from “1” (below 2,500 RMB) to “5” (above 10,000 RMB). Lastly, parental relationship was measured as a binary variable (0 = unstable relationship, 1 = stable relationship).
Analytic Strategy
The analysis consisted of multiple steps. First, descriptive statistics for the study variables were generated, including the count of individuals who self-reported experiencing cyber fraud targeting and cyber fraud financial loss (see Table A1 in the Appendix 1). Next, a series of structural equation models (SEM) were conducted to understand the direct and indirect relationships between low self-control, online deviant lifestyles, and online risk taking on both cyber fraud targeting and financial loss, net of other relevant factors. Unlike standard multinomial regression models, SEM accounts for theoretical measurement errors through confirmatory factor analysis (CFA) and allows for an investigation of the structural relationships among theoretically related variables through path modeling (Kline, 2005). Multivariate normality is one of the fundamental assumptions of SEM (Gao et al., 2008). However, some variables examined in this study had skewed distributions, thus violating this assumption. To address this issue, the Weighted Least Squares Mean and Variance (WLSMV) adjusted estimator was used. WLSMV provides robust standard errors and a mean- and variance-adjusted chi-square statistic and does not require the assumption of normally distributed observed variables. It is an exclusive estimator of Mplus, which has been recommended for analyzing categorical data in SEM (see Muthén & Muthén, 2017). All SEM analyses were conducted in Mplus version 8.
The present study followed a two-step modeling approach. First, the fit of the measurement model was assessed, followed by an examination of the hypothesized structural model shown in Figure 1 (Kline, 2005). Model fit for both the measurement and structural models was evaluated using three commonly applied fit indices: the Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), and Root Mean Square Error of Approximation (RMSEA). CFI and TLI values greater than 0.95 indicate a good fit, while values above 0.90 are considered reasonable or acceptable. RMSEA values below 0.06 suggest a good fit, and values below 0.08 indicate a reasonable fit. The SEM revealed both good and reasonable fit: CFI = 0.94, TLI = 0.94, RMSEA = 0.04, χ2 = 1,645.30, df = 743, p < .001. The effect sizes of the standardized SEM results were interpreted using J. Cohen’s (1988) benchmarks as heuristic guidelines (β < .10 ≈ insubstantial, .10 ≤ β < .30 ≈ small, .30 ≤ β < .50 ≈ medium, and β ≥ .50 ≈ large).
To eliminate the multicollinearity concern, variance inflation factors (VIF) and Pearson correlation analysis were conducted via SPSS 24. The results revealed that the highest VIF score was 1.64, with financial loss as the dependent variable. This suggests the multicollinearity concern is very low in this study (i.e., when the VIF is less than 3, there is no significant concern; see more details in Kim, 2019). The Pearson correlation matrix was also provided in Appendix 2, which offers nuanced details of the collinearity between the study variables.
Lastly, the bootstrapping technique was used to estimate mediation effects. Statisticians regard bootstrapping as one of the most robust techniques for testing mediation (Hayes, 2009). Bootstrapping resamples the data repeatedly to estimate mediating effects without assuming a normal distribution. This study employed bias-corrected bootstrapping with 5,000 resamples.
Results
SEM Results: The Direct Effects
Table 2 presents the standardized direct effects of all independent variables and covariates on the endogenous variables (mediators and dependent variables). The results indicate that low self-control has a significant positive effect on online risk-taking (β = .49, p < .001, medium-to-large effect). Both low self-control (β = .10, p < .01, small effect) and online risk taking (β = .33, p < .001, medium effect) have significant positive effects on online deviant lifestyle, supporting H4. In terms of fraud targeting, only online deviant lifestyle (β = .47, p < .001, medium-to-large effect) and non-deviant online routine activities (β = .17, p < .001, small effect) show significant positive effects, supporting
Standardized Direct Effects in the SEM (N = 929).
Note. IVs = independent variables. Statistically significant coefficients are bolded.
p < .05. **p < .01. ***p < .001.
Regarding financial loss, while low self-control shows non-significant effect (p > .05) and non-deviant online activities show significant negative effect (β = −.15, p < .001, small effect), online risk-taking (β = .28, p < .001, small-to-medium effect), online deviant lifestyle (β = .34, p < .001, medium effect), and fraud targeting (β = .29, p < .001, small-to-medium effect) all exert significant positive effects, supporting
Moreover, the covariates also demonstrated significant effects on the endogenous variables. For online risk-taking, the significant covariates include gender (β = −.24, p < .01, small effect), academic performance β = −.08, p < .05, insubstantial effect), Internet use frequency (β = −.11, p < .01, small effect), Internet use duration (β = .09, p < .05, insubstantial effect), and social guardianship (β = −.017, p < .001, small effect). Concerning online deviant lifestyle, a slightly different set of covariates emerged as significant predictors: gender (β = −.17, p < .001, small effect), academic performance β = −.06, p < .05, insubstantial effect), Internet use frequency (β = −.18, p < .010, small effect), Internet use duration (β = .10, p < .01, small effect), and first time of Internet use (β = −.07, p < .01, insubstantial effect). Regarding fraud targeting, the only two significant covariates were digital guardianship (β = .12, p < .01, small effect) and online routine activities (β = .17, p < .001, small effect). In terms of financial loss, family income (β = −.09, p < .01, insubstantial effect), Internet use frequency (β = −.12, p < .01, small effect), and online routine activities (β = −.15, p < .01, small effect) were the significant covariates to the dependent variable. Figure 2 (see below) illustrates all the significant pathways identified in the SEM.
The standardized SEM results.
Bootstrapping Test: Indirect and Total Effects
A bias-corrected (BC) bootstrapping test with 5,000 resamples was employed to assess the significance of both indirect and total effects. If the 95% confidence interval (CI) does not span zero, the effect is considered significant at the .05 level. Table A2 (in Appendix 1) presents the results of the BC bootstrapping test for all mediation pathways.
For the indirect effects on fraud targeting, “Online risk-taking → Online deviant lifestyle → Fraud targeting” demonstrated the largest but overall small effect size (β = .16, 95% CI [0.12, 0.20]). In terms of indirect effects on financial loss, three paths showed small effect seizes on financial loss, including (1) “Low self-control → Online risk-taking → Financial loss” (β = .14, 95% CI [0.08, 0.21]), (2) “Online deviant lifestyle → Fraud targeting →Financial loss” (β = .14, 95% CI [0.07, 0.20]), and (3) “Online risk-taking → Online deviant lifestyle → Financial loss” (β = .10, 95% CI [0.07, 0.14]).
Total indirect and total effects were also analyzed to demonstrate the varying cumulative impacts of each independent variable on the dependent variables, as presented in Table 2. Regarding the total effect on fraud targeting, online deviant lifestyle exhibited the strongest and medium-to-large effect (β = .47, 95% CI [0.37, 0.58]. Low self-control (β = .17, 95% CI [0.08, 0.25]) and online risk-taking (β = .14, 95% CI [0.02, 0.26]) ranked second and third, which all showed small effect sizes. Similarly, for financial loss, online deviant lifestyle again emerged as the strongest predictor among all independent variables (β = .47, 95% CI [0.40, 0.53]), showing a medium-to-large effect. Online risk-taking followed closely, with a slightly lower medium effect (β = .43, 95% CI [0.30, 0.55]). Low self-control demonstrated the weakest, though still medium, total effect on financial loss (β = .31, 95% CI [0.20, 0.41]).
Total Effects with Bootstrapping Test (5,000 resamples).
Discussion and Conclusion
Cyber fraud victimization, which continues to emerge as a widespread cybercrime concern worldwide, requires a comprehensive investigation into its theoretical underpinnings and evidence-based crime prevention strategies. Analyzing self-reported online survey data from 929 high school students in Shandong Province, China, this study compared the predictive effects across several criminological frameworks on two distinct, yet often conflated, forms of cyber fraud victimization—namely, cyber fraud targeting and cyber fraud resulting in financial loss. Based on the theoretical associations between low self-control, online deviant lifestyles, online routine activities, and online risk-taking, results from the SEM and postestimation mediation analyses revealed several key findings: (1) non-deviant online routine activities are risk factors for fraud targeting but protective factors for financial loss; (2) online deviant lifestyle remains the strongest predictor for both forms of cyber fraud victimization; (3) online risk-taking mediates more than 60% of the relationship between low self-control and online deviant lifestyle; and (4) low self-control has non-significant effects on fraud targeting and financial loss. The following sections will unfold the research findings and limitations by engaging with broader scholarly dialogs.
First, consistent with our expectation, cyber fraud targeting and financial loss are related yet distinct victimization behaviors. While the two phenomena share various risk factors (e.g., online deviant lifestyle), there are several divergent pathways through which different domains of victimization factors exert their influence. Consistent with previous research on fraud targeting (Chen & Xia, 2025; Holtfreter et al., 2008; Lin, Wu, et al., 2025; Mesch & Dodel, 2018; Pratt et al., 2010), non-deviant online routine activities (e.g., entertainment, learning, and commercial use) were found to significantly increase fraud targeting in the current study (
This is a novel finding that demonstrates the protective role of non-deviant online routine activities in cyber fraud victimization resulting in financial loss, since previous studies found either non-significant or positive associations (Chen & Xia, 2025; Holtfreter et al., 2008; Lin, Wu, et al., 2025; Mesch & Dodel, 2018). An explanation of this discrepancy is that the current study deliberately split the non-deviant online routine activities from the online deviant lifestyle construct by measuring the non-risky online exposure (e.g., entertainment, learning, and commercial use). The differentiated measurements also suggest the importance of (re)evaluating the concept of risk under the lifestyle-routine activity theory framework through theoretical and methodological innovation (Engström, 2021; Pratt & Turanovic, 2016). Accordingly, future studies would benefit from cautiously conceptualizing ordinary engagement as deviant routine activities but employing a risk-specific measurement that purposely captures risk inclination (e.g., risk-taking) and involvement (e.g., deviant lifestyle).
Among all the covariates, online deviant lifestyle (i.e., an indicator of lifestyle-routine activity theory) emerged as the most salient predictor for both forms of cyber fraud victimization. To a certain degree, this finding corroborates recent studies that found significant victim-offender overlap across various types of cybercrime, including cyber violence, cyber trespass, and cyber theft (Burden, 2023; Lin, Zhou, et al., 2025; Weulen Kranenbarg et al., 2019; B. Xu, 2025). The current study’s operationalization of the theoretical constructs and modeling of the etiological pathways followed a faithful interpretation of Hindelang et al.’s (1978) original thesis: an individual’s expected social roles and position influence their lifestyle—consisting of vocational and leisure activities—which in turn, shapes their risk of engaging in deviance and experiencing victimization. By extension, the current findings show that online deviant lifestyles not only elevate the risk of being exposed to motivated offenders (
While the concept of online risk-taking is introduced in understanding cybercrime victimization, it offers a “trait-choice-lifestyle” model that coherently links the general theory of crime with the lifestyle-routine activity theory framework. Different from self-control that has been widely accepted as a relatively persistent trait (M. R. Gottfredson & Hirschi, 1990), risk-taking is defined as a domain-specific choice inclination that varies across contexts and situations (Weber, 1998; Weber et al., 2022). For example, individuals with a greater level of financial risk-taking may not demonstrate a similar extent of inclination in physical activities (Weber et al., 2002). Given its domain-sensitive feature, the incorporation of online risk-taking provides a niche theoretical hinge that can coherently explain the developmental trajectory between low self-control and online deviant lifestyle (Kulig et al., 2017; Lin, Zhou, et al., 2025). In this regard, the current findings revealed that online risk-taking accounted for more than 60% of the total effect in the relationship between low self-control and online deviant lifestyle (
Interestingly, low self-control had a non-significant effect on both types of cyber fraud victimization (
Notably, the current findings should be interpreted within the developmental context of adolescence. Compared to adults, adolescents tend to exhibit heightened levels of impulsivity (M. R. Gottfredson & Hirschi, 1990) and skewed risk perception (Duell et al., 2018), while simultaneously being subject to parental and institutional guardianship (Vakhitova et al., 2023). These age-related conditions may amplify the roles of online risk-taking and online deviant lifestyle while attenuating the direct effect of low self-control. As such, the observed pathways may differ in magnitude or configuration among adult populations who tend to possess decreased levels of risk inclination and increased levels of self-disciplinary skills (Duell et al., 2018).
Limitations and Implications for Future Studies
Despite its valuable theoretical and empirical contributions, several limitations should be acknowledged when interpreting the findings. First, the measurement of cyber fraud victimization was intentionally designed to capture a generalized conceptualization of both fraud targeting and fraud resulting in financial loss rather than specific cyber fraud subtypes (e.g., government impersonation fraud, romance fraud, general phishing fraud). As a result, this broad operationalization did not differentiate between distinct forms of fraud, such as spear phishing, advance fee fraud, or government impersonation schemes. Future research would benefit from assessing distinct types of cyber fraud victimization to determine whether these findings remain across different cyber fraud contexts.
Another limitation of the current study concerns the representativeness and generalizability of the data, which were collected from a sample of Chinese high school students in Shandong Province, China. While this sample offers valuable insights into Chinese adolescents’ experiences of cyber fraud victimization, its limited representativeness restricts the generalizability of the findings to both a non-Chinese context and a non-Shandong Province context. Future research would benefit from replicating the analysis using samples drawn from broader age groups and across different socio-cultural contexts. Relatedly, the data used in this study was collected before the prevalence of generative AI and other emerging technologies. It remains uncertain to what extent the research findings on cyber fraud victimization can be affected by the malleable modalities of technology and digitalization (Zhou et al., 2026), as have been indicated by the recent studies (Akartuna & Manning, 2026; Sun et al., 2026). Future studies would benefit from testing the temporal generalizability of the current findings to current contexts.
It is important to note that the current study’s theoretical focus was also restricted to only three criminological perspectives. While these frameworks provide a coherent explanation of individual and situational correlates of cyber fraud victimization, they do not represent the full spectrum of criminological explanations. The absence of other relevant perspectives (e.g., social learning theory, general strain theory, and techno-social process theories) thus constrains theoretical coverage of the current analysis. Similarly, the present study was also limited in its use of a cross-sectional research design and reliance on self-reported survey measures. The study’s cross-sectional design restricted its ability to draw causal inferences between the examined variables, including the association between fraud targeting and financial loss. Although the analytical model specified a directional pathway from fraud targeting to financial loss, this ordering was theoretically motivated rather than empirically established. Since all measures were collected at a single time point, the temporal sequencing between these stages of victimization cannot be definitively determined. Recall bias, telescoping effects, and heightened awareness of targeting following a loss may similarly affect the validity of causal inferences in the current study. For example, individuals who experienced financial loss may retrospectively report higher levels of fraud targeting, or may become more attentive to fraudulent contacts after the loss occurs, thus inflating the observed association. Future research employing longitudinal or experimental research designs would clarify these temporal dynamics and validate the proposed theoretical model. Panel data or event-based designs would be particularly useful in disentangling the sequencing between fraud targeting/exposure and subsequent financial victimization.
The study’s reliance on self-reported measures is another limitation of the current study that must be noted. While self-report surveys remain a standard methodological approach in criminology and social science research generally, they may be subject to social desirability bias, particularly when assessing deviant online behaviors among adolescents. Future studies would benefit from incorporating a multi-source data collection method (e.g., corroborating survey items with digital activity logs, parental/peer reports of online behavior) to strengthen measurement validity and minimize common biases found in self-reported social science research. Relatedly, although the research team provided terminological explanations of key survey items before data collection, it should be noted that some respondents may have had a limited or biased understanding of some key terms used in the questionnaire, such as “online or telecommunications fraud.” Opting for behavioral descriptions of key survey items rather than employing terminological descriptions could have reduced bias, misunderstanding, or respondents independently interpreting survey constructs.
The dichotomization of fraud targeting and financial loss also raises concerns about loss of conceptual and empirical richness. Collapsing frequency-based responses into binary variables obscures gradations in cyber fraud exposure and harm, especially in the analysis of SEM. Although this approach improved model stability given the skewed distributions and low base rate of repeated financial loss, it reduced variability in the intensity and recurrence of victimization experiences. As a result, the present findings relate more directly to the occurrence of victimization rather than differences in severity or chronicity. When it permits, future studies should consider retaining frequency-based or count-based measurements, exploring how the theories explain the variation of fraud targeting and financial loss. Studies with larger samples or higher base-rate victimization may benefit from ordinal SEM estimators or count-based modeling strategies that preserve gradations in exposure and harm.
Moreover, the current study measured online deviant lifestyle through a single index, which could have obscured meaningful differences in how specific forms of deviance relate to victimization. For instance, the aggregated measure did not distinguish between lower-severity behaviors (e.g., digital piracy) and more serious or technologically sophisticated acts (e.g., identity theft or dark web engagement). Although the index was theoretically justified as capturing involvement in deviant online contexts, it may mask heterogeneous effects across specific behaviors. Future studies should consider operationalizing different types of online deviant behaviors separately to assess their likely heterogeneous effects. Disaggregated analyses may help determine whether particular categories of deviance disproportionately drive exposure to cyber fraud victimization.
A related limitation concerns the measurement of non-deviant online routine activities. The use of binary “daily use” indicators may collapse meaningful distinctions among intensity of use, digital skill, and online literacy. As a result, the observed protective effect of non-deviant routine activities on financial loss may partially reflect higher levels of digital familiarity or competence rather than routine activity structures alone. Although the study included Internet-use frequency, duration, and age of first use as covariates to mitigate this concern, these proxies may not fully capture differences in digital literacy or security awareness. Future research should employ more nuanced measures that distinguish routine activity patterns from digital competence, allowing for clearer theoretical tests of lifestyle-routine activity mechanisms in cyber fraud victimization contexts.
Policy Implications
Despite its numerous limitations, the current findings yielded several evidence-based policy implications. First, cyber prevention efforts should focus on intervening in individuals’ risk-taking propensity before habitual behaviors (e.g., online deviant lifestyle) are shaped. In practice, educational programs may prioritize building adolescents’ capacity for online risk assessment in unknown but risky digital scenarios. For example, scenario-based training that simulates scamming attributes would be an effective way because it directly targets the cognitive–behavioral mechanism (e.g., risk-taking), which is found as the key bridge between personality traits and risky online lifestyles.
Another implication challenges the common policy emphasis on restricting adolescents’ Internet exposure. While non-deviant online routine activities increased fraud targeting victimization, it reduced the propensity of experiencing cyber fraud victimization, resulting in financial loss. This divergence between the two stages of victimization is critical. It indicates that routine online engagement enhances digital familiarity and defensive capability, enabling adolescents to recognize and resist fraudulent activities even when they are frequently exposed. Therefore, policies centered around generalized Internet restriction may be counterproductive. Instead, guided and safe participation in ordinary online activities should be encouraged as a form of experiential digital literacy that reduces harm at the financial loss stage.
Third, the model revealed an important contrast between social guardianship and digital guardianship. Parental guardianship significantly reduced online risk-taking, whereas digital guardianship (e.g., reliance on anti-malware tools) was positively associated with fraud targeting, 4 thus challenging the notion that technology alone is sufficient to improve cybersecurity (Choi, 2008; Reyns et al., 2011). Human oversight and parental guidance appear more effective because human beings can actively and dynamically intervene during social engineering scenarios at different stages, which cannot be accurately identified by digital technologies because many fraudulent scripts do not abuse technological vulnerabilities (Zhou et al., 2023). Therefore, cyber fraud prevention practices would benefit from prioritizing social guardianship over excessive reliance on technological protections.
In conclusion, the present study provides an integrative theoretical model that explains how individual traits, online choices, and lifestyle routines converge to impact adolescents’ vulnerability to cyber fraud victimization. The findings call for both adaptive theoretical frameworks and evidence-based prevention strategies that move beyond simplistic deterrence or technological control measures. Future research would benefit from refining these models through longitudinal, cross-cultural, and fraud-specific analyses, thus contributing to a more comprehensive and dynamic understanding of cyber fraud behaviors in the digital age.
Footnotes
Appendix 1
Correlation Analysis.
| Variables | Gender | Age | Household | Academic performance | Family income | Parental Relatioship | Frequency | Duration | First time | Social guardianship | Digital guardianship | Low self-control | Online risk-taking | Online deviant lifestyle | Non-deviant online routine | Fraud targeting | Financial loss |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Gender | 1 | ||||||||||||||||
| Age | .125*** | 1 | |||||||||||||||
| Household | −.038 | .038 | 1 | ||||||||||||||
| Academic performance | −.182*** | −.196*** | −.010 | 1 | |||||||||||||
| Family income | −.246*** | −.041 | .162*** | .129*** | 1 | ||||||||||||
| Parental relationship | −.031 | −.005 | −.095*** | −.017 | −.008 | 1 | |||||||||||
| Frequency | .167*** | −.006 | .017 | .070* | −.021 | −.001 | 1 | ||||||||||
| Duration | .001 | −.010 | .066* | .186*** | .085** | −.069* | .217*** | 1 | |||||||||
| First time | .185*** | .134*** | −.090** | −.097** | −.077* | −.030 | −.195*** | −.115*** | 1 | ||||||||
| Social guardianship | .045 | −.131*** | −.013 | −.057 | −.078* | .014 | .039 | −.109** | .036 | 1 | |||||||
| Digital guardianship | −.084* | −.025 | .001 | −.013 | −.043 | .081* | −.009 | −.067* | −.069* | .223*** | 1 | ||||||
| Low self-control | −.159*** | −.067* | .008 | .093** | .089** | −.072* | .012 | .166*** | −.185*** | −.206*** | −.089** | 1 | |||||
| Online risk-taking | −.256*** | −.037 | .064 | −.003 | .060 | −.069* | −.135*** | .090** | −.087** | −.179*** | −.069* | .447*** | 1 | ||||
| Online deviant lifestyle | −.320*** | −.099** | .062 | .006 | .096** | −.070* | −.244*** | .089** | −.101** | −.104** | −.030 | .301*** | .514*** | 1 | |||
| Non-deviant online routine | .173*** | .095** | .034 | −.097** | .003 | .025 | .308** | .069* | −.133*** | .039 | .141*** | .039 | −.089** | −.187*** | 1 | ||
| Fraud targeting | −.109** | −.014 | −.001 | .001 | .042 | −.028 | .008 | .048 | −.075* | −.036 | .100** | .151*** | .159*** | .272*** | .127*** | 1 | |
| Financial loss | −.212*** | −.027 | .024 | −.021 | .020 | −.052 | −.228*** | .037 | −.074* | −.096** | −.047 | .217*** | .393*** | .757*** | −.140*** | .219*** | 1 |
Correlation is significant at the .001 level (two-tailed).
Correlation is significant at the .01 level (two-tailed).
Correlation is significant at the .05 level (two-tailed).
Ethical Considerations
Ethical approval was obtained from Jining Industrial Technician College (XSC201911).
Consent to Participate
All participants and participants’ parents provided written or verbal informed consent.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author* upon reasonable request. The data is not publicly available due to ethical considerations.