xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 22 Ka 4500
controlfield tag 007 cr-bnu---uuuuu
008 s2010 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003450
Testing the plausibility of a series of causal minor cyberloafing models
h [electronic resource] /
by Kevin Askew.
[Tampa, Fla] :
b University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains X pages.
Thesis (M.A.)--University of South Florida, 2010.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
ABSTRACT: Cyberloafing is the nonsanctioned recreational use of the computers/internet during work hours. Although research is increasing, the processes related to cyberloafing are not well understood. In the current studies, I developed, tested, and evaluated a series of causal minor-cyberloafing models. In Study 1, I empirically compared four minor-cyberloafing taxonomies and selected two of these models as my working taxonomies for minor cyberloafing. In Study 2, I tested and evaluated eight causal minor-cyberloafing models using structural equation modeling techniques and various model-data fit indices. Results of Study 2 indicated that the models were not plausible, bringing into question the value of the proposed models. Despite the poor primary results, I did find a number of potentially important results in the subsequent exploratory analyses. First, I observed high correlations between minor cyberloafing and four of my exploratory variables. Second, I found that one's perception of the descriptive cyberloafing norms predicted minor cyberloafing above and beyond one's perception of the injunctive cyberloafing norms. Finally, I found that the predictors cyberloafing attitudes and perceived descriptive norms accounted for a substantial amount of variance in minor cyberloafing. I discuss the theoretical implications of the exploratory results and future directions for research in the discussion section.
Advisor: Michael Coovert, Ph.D.
Counterproductive work behavior
t USF Electronic Theses and Dissertations.
Testing the Plausibility of a Series of Causal Mino r Cyberloafing Models by Kevin Askew A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Psychology College of Arts and Sciences University of South Florida Co-Major Professor: Michael Coovert, Ph.D. Co-Major Professor: Joseph Vandello, Ph.D. Russell Johnson, Ph.D. Stephen Stark, Ph.D. Date of Approval: April 1, 2009 Keywords: cyberslacking, withdrawal, CWB, counterproductive w ork behavior, computers Copyright 2010, Kevin Askew
i Table of Contents List of Tables iii Abstract iv Chapter One Â– Introduction 1 Chapter Two Â– Counterproductive Work Behavior Literature Review 3 Conceptualizations of CWB 3 Types of CWB 5 Antecedents of CWB 7 Consequences of CWB 9 CWB Concluding Statement 10 Chapter Three Â– Cyberloafing Literature Review 12 Taxonomy of Cyberloafing 12 Cyberloafing Antecedents 16 Prevalence and Consequences 17 Cyberloafing Concluding Statement 19 Chapter Four Â– Goals of the MasterÂ’s Thesis 20 Chapter Five Â– Study 1 21 Method 21 Results 24 Discussion 27 Chapter Six Â– Study 2 28 Perceived Injunctive Norms 28 Job Boredom 29 Task Performance 30 The Ability to Hide Cyberloafing 31 Chapter Seven Â– The Ability to Hide Cyberloafing Pilot Study 33 Method 33 Results 35 Discussion 38
ii Chapter Eight Â– Study 2 Continued 39 Causal Minor-Cyberloafing Models 39 Method 39 Results 43 Exploratory Analyses 47 Discussion 49 Chapter Nine Â– General Discussion 51 Contributions to the Literature 51 Limitations 53 Future Directions 55 Summary and Conclusion 56 References 57
iii List of Tables Table 1 Model-Data Fit Statistics for the Minor Cyb erloafing Models 25 Table 2 Correlations Among the Ability to Hide Cybe rloafing Factors 37 Table 3 Model-Data Fit Statistics for ModelÂ’s Using LimÂ’s Taxonomy 43
v Testing the Plausibility of a Series of Causal Mino r Cyberloafing Models Kevin Askew ABSTRACT Cyberloafing is the nonsanctioned recreational use of the computers/internet during work hours. Although research is increasing, the proces ses related to cyberloafing are not well understood. In the current studies, I developed, t ested, and evaluated a series of causal minor-cyberloafing models. In Study 1, I empirical ly compared four minor-cyberloafing taxonomies and selected two of these models as my w orking taxonomies for minor cyberloafing. In Study 2, I tested and evaluated e ight causal minor-cyberloafing models using structural equation modeling techniques and v arious model-data fit indices. Results of Study 2 indicated that the models were not plaus ible, bringing into question the value of the proposed models. Despite the poor primary r esults, I did find a number of potentially important results in the subsequent exp loratory analyses. First, I observed high correlations between minor cyberloafing and fo ur of my exploratory variables. Second, I found that oneÂ’s perception of the descriptive cyberloafing norms predicted minor cyberloafing above and beyond oneÂ’s perceptio n of the injunctive cyberloafing norms. Finally, I found that the predictors cyberl oafing attitudes and perceived descriptive norms accounted for a substantial amoun t of variance in minor cyberloafing. I discuss the theoretical implications of the explo ratory results and future directions for research in the discussion section.
1 Chapter One Introduction Roughly 300 years ago, a major shift began to occur : wage earning began to overtake farming as the dominant way people made a living (Christian, 2008). As more people became wage earners, more people began to wo rk in organizations. Eventually, people began to systematically study work in organi zations. Today, this field is known as I/O psychology. I/O psychology consists of two subfields: (1) indus trial psychology, and (2) organizational psychology. The Â“I-sideÂ” of the fie ld focuses on topics such as recruitment, selection, and training; whereas the Â“ O-sideÂ” focuses on areas such as motivation, well-being, attitudes, and the social c ontext within the organization. In general, the I-side focuses on the management of hu man resources in organizations, and the O-side focuses on understanding and predicting behavior within organizational settings (Jex & Britt, 2008). In practice, however the distinction between the I-side and the O-side is not clear cut, and many research prog rams span both sides. In the last three decades, I/O psychology has deve loped a much better understanding of Â“good work behaviorÂ” and Â“bad work behaviorÂ”. Good work behavior, or organizational citizenship behavior, is employee behavior that contributes to the goals of the organization but is not a formal part of the job. Examples of organizational
2 citizenship behavior are helping a coworker with th e fax machine, speaking positively about the organization to friends, and volunteering to work on Saturday. Organizational citizenship behaviors have been linked to a number of important organizational variables, such as job satisfaction, organizational commitment and organizational justice (Dalal, 2005). This research, however, is about bad work behavior; or as many I/O psychologist call itÂ— counterproductive work behavior More precisely, this concerns a special kind of counterproductive work behavior: counterproductive work behavior involving a computer and/or the internet. Many I/O psychologists refer to this kind of counterproductive behavior as cyberloafing This paper will present two studies designed to ex tend the fieldÂ’s knowledge of cyberloafing. In order to do this, it is necessary to review the appropriate literatures, so that it is clear why the two presented studies were conceived and conducted. The two appropriate literatures in this situation are the c ounterproductive work behavior and cyberloafing literatures. The counterproductive wo rk behavior literature will be reviewed first, followed by the cyberloafing literature. Af ter the literature reviews, I present the two cyberloafing studies. The ultimate goal of thi s paper is to develop, test, and evaluate a series of causal minor-cyberloafing models.
3 Chapter 2 Counterproductive Work Behavior Literature Review Counterproductive work behavior (CWB) is behavior that harms, or has the potential to harm, the goals of an organization. E xamples of CWB are stealing pens from work, making unwarranted personal phone calls, and physically assaulting a coworker. CWB has been studied under a variety of terms; some of these terms refer to a broad collection of Â“bad work behaviorsÂ”, and therefore c an be considered more or less synonymous with the term CWB (e.g., organizational misbehavior); whereas other terms refer to a specific kinds of Â“bad work behaviorÂ”, a nd therefore can be consider specific types of CWB (e.g., violence, cyberloafing). This section of the paper reviews the CWB literatu re in preparation for the two cyberloafing studies. The CWB literature is extens ive, so a completely exhaustive review of the CWB is beyond the scope of this paper. None theless, this review will cover the major topics and findings in the CWB literature. Conceptualizations of CWB The conceptual definition for CWB given in this pa per is behavior that harms, or has the potential to harm, the goals of the organiz ation. This definition, while common, is not used by all CWB researchers. Other research ers have approached CWB from different perspectives, and have consequently defin ed CWB in slightly different ways.
4 Robinson and Bennett (2003) group the different con ceptualizations of CWB into one of the three different categories of approaches. The first approach identified by Robinson and Benn ett (2003) is to define CWB as deviant behavior in the workplace that results f rom a particular cause Some researchers have identified CWB as deviant behavior caused by aspects of the work environment that the organization is responsible fo r, and these researchers have consequently focused on organization-directed CWBs (OÂ’Leary-Kelly, Griffin, & Glew, 1996). Other researchers have focused on interpers onal-directed CWBs caused by the mistreatment (Bies & Tripp, 1998; Skarlicki & Folge r, 1997; Stuckless & Goranson, 1992). And still other researchers have focused on CWBs as a response to frustration (Spector, 1975; Spector, 1997). The second approach identified by Robinson and Ben nett (2003) is to define CWB as deviant behaviors in the workplace that are purposely harmful to the organization (Baron & Neuman, 1996; Giacalone & Gre enberg, 1997) or individuals within the organization (Ashforth, 1994; Perlow & L atham, 1993). Including intent as a necessary component of the definition of CWB makes it clear that accidents (e.g., a waitress accidently dropping and breaking a coffee cup) and poor task performance (e.g., not making enough widgets due to lack of widget-mak ing-ability) are not CWB. Finally, the third approach identified by Robinson and Bennett (2003) is to define CWB as deviant behavior in the workplace that violates organizational norms (Andersson & Pearson, 1999; Puffer, 1987; Vardi & W einer, 1996). Organizational norms vary from job-to-job, and the behaviors that employees consider counterproductive
5 likely vary as well. Including the breaking of org anizational norms as a necessary component of the definition of CWB can potentially take these differences across jobs into account, and as a result, be closer to what em ployees consider CWB. In sum, researchers do not agree on an exact conce ptualization of CWB. However, although there are slight differences in t he conceptualizations of CWB, the end result is that researchers are measuring much of th e same thing. A number of consistent themes appear in the conceptual definitions of CWB, and examination of the CWB items from various scales shows that the sets of behavior s measured by different researchers overlap greatly (Robinson & Bennett, 2003). Types of CWB CWB has been useful as a broad construct for vario us kinds of bad behavior at work. The construct of CWB ties together a lot of different behaviors and highlights the similarity between them. However, if different com ponents of CWB have different antecedents and consequences, a more fine-grained a nalysis is needed. Because this is sometimes the case, CWB researchers often distingui sh between different types of CWB. This subsection highlights the common types, or dis tinctions, of CWB used in the literature, and places cyberloafing within these fr ameworks. The most common distinction in the literature is b etween CWB directed towards the organization (CWB-O) and CWB directed towards i ndividuals (CWB-I). CWB-O is counterproductive work behavior that harms the organization such as stealing pens or taking an extended lunch break. CWB-I is counterpro ductive work behavior that harms organizational employees or customers such as spreading rumors or teasing an employee.
6 CWB-O is hypothesized to be more strongly related t o stressors stemming from the organization or job itself, such as job dissatisfac tion and situational constraints; whereas, CWB-I is hypothesized to be more strongly related t o stressors related to other individuals, such as interpersonal conflict (Hershc ovis et al., 2007). A recent metaanalysis by Hershcovis supports these hypotheses (H ershcovis et al., 2007). A second common taxonomy is Robinson and BennettÂ’s (1995) taxonomy of deviant work behavior. In addition to distinguishi ng CWBs based on their target (i.e., CWB-O, CWB-I), Robinson and Bennett distinguish CWB s based on severity of the behavior. Thus, Robinson and BennettÂ’s taxonomy po sits that CWBs differ along two dimensions: (1) the seriousness of the behavior; an d (2) the target (Robinson & Bennett, 1995). The two orthogonal dimensions divide devian t workplace behavior into four quadrants: personal aggression (serious-interpersonal), property deviance (seriousorganizational), political deviance (minor-interpersonal), and production deviance (minor-organizational). Depending on the severity of the behavior, cyberlo afing can be considered either production deviance (minor-organizational) or prope rty deviance (serious-organizational) [Blanchard & Henle, 2008; Blau et al., 2006]. Cybe rloafing can be considered production deviance when an employee engages in rel atively minor behaviors, such as sending a personal email at work. Cyberloafing can be considered property deviance when an employee engages in more serious behaviors, such as sharing proprietary company information at work. Parallel to findings in the counterproductive work behavior literature, researchers have found that mi nor cyberloafing behaviors are fairly
7 common, whereas serious cyberloafing behaviors are rare (Blanchard & Henle, 2008; Blau et al., 2004; Lim & Teo, 2005; Mastrangelo et al., 2006). A third taxonomy is offered by Spector and his col leagues (Spector et al., 2006). Spector and his colleagues had subject matter exper ts sort CWB items into different categories, and found evidence for five types of CW B: (1) abuse [e.g., harassing a coworker], (2) production deviance [e.g., purposely working inefficiently], (3) sabotage [e.g., destroying company property], (4) theft [stealing pens], and (5) withdrawal [e.g., taking an extended lunch break]. Spector et al. (2 006) demonstrated the utility of this taxonomy by showing that the different types of CWB differentially correlated with boredom, job satisfaction, and anger. Furthermore, Spector and his colleagues showed that these distinctions would be obscured if their CWB types were combined into CWBO and CWB-I. In sum, CWB can be broken down into a number of di fferent types, depending on how fine-grained an analysis one desires. The bett er recognized taxonomies are described above. The next two sections break away from the conceptualizations of CWB and discuss the empirical findings of the CWB liter ature, specifically the antecedents and consequences of CWB. Antecedents of CWB Given the prevalence and costs of CWB to organizat ions, it is not surprisingly that a lot of research has been done on the antecedents of CWB. This subsection will summarize the findings on the antecedents of CWB. To aid in the summarization, trends
8 identified by Robinson and Bennett (2003) in their seminal work, The Past, Present, and Future of Workplace Deviance Research, will be used. The first trend identified by Robinson and Bennett is to treat CWB as a reaction to experiences at work Researchers who take this perspective typically focus on CWB as an emotional response to either frustration, per ceived injustice, lack of control, or threats to oneÂ’s status. Spector and his colleague s, for example, have provided convincing evidence that CWB is often a result of a n emotional response to frustrating job stressors (Fox & Spector, 1999; Fox, Spector, & Miles, 2001; Spector, 1997). Other researchers have examined CWB as a response to perc eived injustices in the workplace, and have found that perceived injustice is related to a number of CWBs, including aggression (Folger & Baron, 1996; OÂ’Leary-Kelly, Gr iffin, & Glue, 1996; Skarlicki & Folger, 1997), theft (Greenberg, 1990; Greenberg, 1 993) interpersonal deviance (Burroughs, 2001) and sabotage (Ambrose, Seabright, & Schminke, 2001). Still other researchers have examined CWB as an emotional respo nse to shame (Tangney et al., 1996) or feelings of powerlessness (Ambrose, Seabri ght, Schminke, 2001; DiBattista, 1991; Perlow & Latham, 1993). A second trend identified by Robinson and Bennett is to treat CWB as a reflection of oneÂ’s personality In this view, CWB is the result of employees hav ing certain personality traits. Personality traits that have b een shown to predict CWB include: dispositional aggressiveness (Sablynski, Mitchell, James, & McIntytre, 2001), negative affect (Spector & OÂ’Connell, 1994), trait anger (De ffenbackher, 1992; Fox & Spector, 1999) and low conscientiousness (Lee, Ashton, & Shi n, 2001). The personality profile of
9 high-extraversion-low-agreeableness has also been f ound to predict CWB (Lee, Ashton, & Shin, 2001). The final trend identified by Robinson and Bennett (2003) is to treat CWBs as an adaptation to the social context Social norms, in particular, have been found to be important. For example, Robinson and OÂ’Leary-Kelly (1998) found that the extent to which oneÂ’s coworkers engages in antisocial behavio r was the best predictor of workplace antisocial behavior, and social norms hav e been found to strongly predict cyberloafing (Blanchard & Henle, 2008). Other rese archers have looked at CWBs as a learned behavior that is reinforced in certain envi ronments (OÂ’Leary-Kelly, Griffin, & Glew, 1996). In sum, a number of different antecedents for CWB have been identified. These antecedents can be grouped together based on whethe r they are situational-based, personality-based, or adaptation-based. The final subsection of the CWB literature review deals with the consequences of CWB. Consequences of CWB The consequences of CWB can be grouped into two br oad categories: consequences for the organization, and consequences for the employees. To date, most of the research in this area has focused on the con sequences of CWB for the organization, with most of that research focusing on the cost of specific CWB behaviors. Not surprisingly, the consistent finding is that CWB is expensive; the annual costs of CWB to the organization range from $4.2 billion for violen ce (Bensimon, 1997) to $200 billion for theft (Buss, 1993).
10 The literature on the consequences of CWB for the employees has primarily focused on the effects of abusive supervision. Not surprisingly, abusive supervision is associated with a number of negative outcomes for t he victim, including negative personal (cognitive, physical), interpersonal (aggr essive behaviors, interpersonal conflict), professional (job satisfaction, turnover ), and organizational functioning (productivity, commitment) [Keashly & Jagatic, 2003 ]. Researchers have described abusive supervision as having two effects: (a) Â“a s piraling effectÂ”, where the abused employee withdraws as a response to the supervisorÂ’ s abuse, which leads to decreased task performance, which elicits even greater abuse from the supervisor (Ashforth, 1994); and (b) Â“a spillover effectÂ”, where the negative ef fect of abusive supervision spills-over into the victimÂ’s home life, affecting the victimÂ’s friends and family as well (Ashforth, 1994). In sum, CWB has high direct and indirect costs to the organization. The long term consequences of CWB for the victims are mostly unknown. However, abusive supervision is known to have a number of detrimenta l effects on the victim. CWB Concluding Statement The previous section gave an overview of the CWB l iterature. The review discussed the different conceptualizations of CWB, CWB taxonomies, and the antecedents and consequences of CWB. The next sect ion reviews the cyberloafing literature. To aid in this review, the cyberloafin g literature will be broken down by the three common topics found in the cyberloafing liter ature: (1) the taxonomy of
11 cyberloafing, (2) the antecedents of cyberloafing, and (3) the prevalence and consequences of cyberloafing.
12 Chapter 3 Cyberloafing Literature Review Cyberloafing is the misuse of computers and/or the internet during work hours (Lim, 2002). In other words, cyberloafing is when one is suppose to be working, but really, he or she is engaged in another activity, s uch as: chatting on Instant Messenger, checking Facebook, or watching videos on Youtube. Some cyberloafing behaviors can be considered relatively harmless, especially if do ne in moderation (e.g., checking sports scores, writing personal emails). Other cyberloafi ng behaviors, however, are more of a problem because the behaviors are either time consu ming (e.g., planning a vacation online), place the organization at risk of litigati on (e.g., downloading copyrighted material), or are directly harmful to the goals of the organization (e.g., sharing proprietary company information). Taxonomy of Cyberloafing The primary focus of cyberloafing literature has be en identifying the taxonomy of cyberloafing. One of the first cyberloafing taxon omies was proposed by Lim in 2002. LimÂ’s taxonomy states that cyberloafing consists of two factors: (1) web-browsing, and (2) emailing. The web-browsing factor refers to re ading general news sites (e.g., CNN.com), shopping online (e.g., amazon.com) and an y other non-email activities that involve a web-browser. The email factor refers to checking and sending non-work
13 related emails. LimÂ’s taxonomy was supported by a confirmatory factor analysis in a later study (Lim & Teo, 2005). Lim defined cyberloafing as the misuse of the inter net during office hours (Lim, 2002). However, there are many types of behaviors that meet LimÂ’s conceptual definition of cyberloafing that are not captured by the two factors Lim proposed. For example, moonlighting (using the internet to gain a dditional income), posting messages, downloading non-work related information, using cha trooms, and playing games online all fit LimÂ’s conceptual definition of cyberloafing but are not covered by LimÂ’s cyberloafing factors or the items in LimÂ’s scale. To address this issue, two teams of researchers ind ependently created new scales with items covering more of the cyberloafing constr uct. Blau and his colleagues (Blau, Yang, & Ward-Cook, 2004) created a new measure by e xtending LimÂ’s cyberloafing scale to cover more of LimÂ’s conceptual definition of cyberloafing. Examples of some of the items added by Blau et al. are, Â“Chat with othe r people with instant messengerÂ”, and Â“Play online gamesÂ”. When the data were factor ana lyzed, LimÂ’s original items loaded onto a web-browsing factor and email factor, and Bl au et al.Â’s additional items loaded onto a third factor, which Blau and his colleagues called Â“Interactive CyberloafingÂ”. Blau et al. describe interactive cyberloafing as a type of cyberloafing that involves more dynamic responding, either with other humans (e.g., instant messenger) or with software (e.g., online games). Thus, BlauÂ’s research team p roposed that cyberloafing consisted of three factors: (1) web-browsing, (2) e-mailing, and (3) interactive. Blau et al.Â’s threefactor solution was replicated on a validation samp le.
14 Mahatanankoon and his colleagues (2004) were the se cond research team to address the criterion deficiency of LimÂ’s original scale. Instead of extending LimÂ’s original cyberloafing scale, like Blau and his coll eagues did, Mahatanankoon et al., created a new scale from scratch. To develop their scale, Mahatanankoon and his team had MBA students generate a list of various cyberlo afing behaviors. The list was examined by the researchers for clarity and redunda ncy, and, after pilot testing, eventually condensed into a final pool of 11 statem ents. After further testing, the data were factor-analyzed, and a three-factor solution e merged. Factor 1 consisted of items related to shopping and purchasing goods online (e. g., conducting personal on-line shopping); factor 2 consisted of items related to s eeking and viewing information on the internet (e.g., researching personal hobbies); and factor 3 consisted of items related to personal communication (e.g., using personal web-ba sed e-mail, such as hotmail, yahoo, etc.). Based on item content, Mahatanankoon et al. named these factors: (1) e-commerce, (2) information research, and (3) personal communic ation, respectively. Mahatanankoon et al.Â’s three-factor solution was later replicated on a validation sample. A fourth cyberloafing taxonomy was proposed by Blan chard and Henle (2008). Blanchard and Henle agreed with other researchers t hat cyberloafing is a multifaceted construct. However, Blanchard and Henle believed t he distinction between minor cyberloafing behaviors (e.g., viewing a CNN webpage ) and serious cyberloafing behaviors (e.g., viewing adult-oriented websites) w as important, and criticized past taxonomies for not making this distinction. Blanch ard and Henle argued that the distinction between minor and serious cyberloafing is critical because minor cyberloafing and serious cyberloafing are likely to have differe nt antecedents and relations with other
15 variables. Blanchard and Henle added additional it ems to LimÂ’s original scale, and factor analyzed the subsequent data: A two-factor solutio n consistent with Blanchard and HenleÂ’s theorizing emerged. Blanchard and Henle th erefore proposed that cyberloafing consists of two broad factors: (1) minor cyberloafi ng, and (2) serious cyberloafing. Finally, yet another cyberloafing taxonomy was prop osed by Mastrangelo and his research team (Mastrangelo et al., 2006). Mastrange lo and his colleagues, similar to MahatanankoonÂ’s research group, developed their own scale instead of extending LimÂ’s cyberloafing scale. Mastrangelo and colleaguesÂ’ cy berloafing scale asks participants to rate the frequency, on a 7-point scale (1= never did this or not in the past 6 months ; 7= almost constantly ), of an extensive list of 40 cyberloafing behavior s. Mastrangelo et al. (2006) conducted a factor analys is on the responses to their scale, and argued for a two-factor solution consist ing of the factors: (1) nonproductive computer use, and (2) counterproductive computer us e. Non productive computer use occurs when an employee uses the computer during wo rk hours for activities that are unproductive, but are not potentially destructive t o the organization (e.g., reading a news website). Counter productive computer use occurs when an employee eng ages in behavior that could conflict with the companyÂ’s goa ls (e.g., sending proprietary company information to a third party) [Mastrangelo, Everton & Jolton, 2006]. To summarize, a major focus of the nascent cyberloa fing literature has been identifying a taxonomy of cyberloafing. Numerous c yberloafing taxonomies have been proposed. Some taxonomies classify a broad range o f cyberloafing behaviors (e.g., Blanchard & HenleÂ’s taxonomy, Mastrangelo et al.Â’s taxonomy); other taxonomies
16 classify the more common, minor forms of cyberloafi ng (e.g., LimÂ’s taxonomy, Blau et al.Â’s taxonomy, Mahatanankoon et al.Â’s taxonomy). Cyberloafing Antecedents A second focus of the cyberloafing literature has been identifying antecedents of cyberloafing. The goal of this section is to brief ly review the known antecedents of cyberloafing. To help summarize, the antecedents w ill be grouped based on whether they are personality-based, situation-based, or based on some non-personality individual difference variable. Both higher-order and lower-order personality chara cteristics have been found to predict cyberloafing. Wyatt and Phillips (2005) ha ve implicated low agreeableness and high extraversion in cyberloafing, and other resear chers have observed significant correlations between conscientiousness and cyberloa fing (Everton, Mastrangelo, & Jolton, 2005). Lower order personality characteris tics, such as impulsivity (Davis, Flett, & Besser, 2002; Everton et al., 2005), sensation se eking (Everton et al., 2005), external locus of control (Blanchard & Henle, 2008), and tra it procrastination (Davis, Flett, & Besser, 2002) have also been implicated in cyberloa fing. A number of non-personality individual differences variables have also been found to predict cyberloafing. Not surprisingly, i ndividual difference variables that predict general computer-use often predict cyberloa fing as well. For example, age (De Lara, 2007; Everton, Mastrangelo, & Jolton, 2005; G arrett & Danziger, 2008) time spent on the internet at home (Blanchard & Henle, 2008), Â“internet skillÂ” (Blanchard & Henle,
17 2008), and gender (Mastrangelo, Everton, and Jolton 2006; Everton et al., 2005; Mastrangelo et al., 2006) have all been implicated in cyberloafing. Finally, a few situational variables have been foun d to predict cyberloafing. Social norms have been one of the strongest predict ors of cyberloafing (Blanchard & Henle, 2008). Other variables, such as employee st atus (Garrett & Danziger, 2008), job autonomy (Garrett & Danziger, 2008), job type (Garr ett & Danziger, 2008), and oneÂ’s connection speed at work compared to at home (Mastr angelo, et al., 2006) have also been implicated in cyberloafing. In sum, a number of different antecedents for cyber loafing have been identified. These antecedents can be grouped together depending on whether they are personalitybased, situation-based, or non-personality individu al-difference based. Generally, variables that predict computer-use also predict cy berloafing. Prevalence and Consequences The third major focus of the cyberloafing literatur e has been estimating the prevalence of cyberloafing and determining the cons equences of cyberloafing. Numerous estimates of the prevalence of cyberloafin g have been made, and although the estimates vary substantially, they all converge on the idea that cyberloafing is widespread. In a study by Vault.com, an online ana lyst firm, 37% of employees admitted to surfing constantly at work, and an additional 32 % of employees admitted to surfing the internet a few times a day. Greenfield and Davis ( 2002) estimate the average employee spends three hours per week cyberloafing, whereas M ills, Hu, Beldona, and Clay (2001) estimate the average employee spends two and a half hours per day cyberloafing.
18 Surfwatch software paints an even grimmer picture, estimating that almost one third of American workersÂ’ time on the internet is spent Â“ch eating the boss out of real workÂ” (Naughton, Raymond, & Shulman, 1999). Indeed, Mala chowski (2005) found that cyberloafing is now the most common way employees w aste time at work. The rise in cyberloafing has not gone unnoticed by organizations. Findings from a survey by Telemate.Net indicated that 83% of surv eyed companies were concerned with employees misusing the internet at work, and o ver 70% of companies indicated that cyberloafing results in real costs to their compani es (Business Wire, 2002). Estimates for the cost of cyberloafing vary substantially, but th ose for United States businesses as a whole are usually in the billions of dollars per ye ar (e.g., Foster, 2001; Naughton et al., 1999). In addition to productivity loss, cyberloafing can cause the organization legal problems in cases where employees download copyrigh ted material and view or send offensive electronic material (Lichtash, 2004; Mill s, et al., 2001; Panko & Beh, 2002; Scheuermann & Langford, 1997). Furthermore, bandwid th intensive cyberloafing can bog down computer resources and degrade system perf ormance (Sipior & Ward, 2002). Many organizations have responded to cyberloafing b y implementing internet monitoring systems (American Management Association, 2001). H owever, studies looking at the effectiveness of internet monitoring systems to red uce cyberloafing have found mixed results (Galletta & Polak, 2003; Lee, Lee, & Kim, 2 004).
19 In sum, estimates of the prevalence and cost of cyb erloafing differ substantially, but they all converge on the idea that cyberloafing is widespread and expensive to the organization. Cyberloafing Concluding Statement Although research on cyberloafing is increasing, c yberloafing is still not well understood. One glaring short-coming of the cyberl oafing literature is the dearth of empirical studies testing causal models of cyberloa fing. To make progress, cyberloafing researchers need to move beyond descriptive studies Â—which have focused on the taxonomy, antecedents, and prevalence of cyberloafi ngÂ—and start empirically testing causal models.
20 Chapter 4 Goals of the MasterÂ’s Thesis The goal of the present studies was to develop, te st, and evaluate a series of causal minor -cyberloafing models. The review of the cyberloafi ng literature showed that many studies have examined the taxonomy, antecedent s, and prevalence of cyberloafing, but that few studies have empirically tested causal cyberloafing models. I conducted the present studies to begin to fill this gap in the cy berloafing literature. The purpose of Study 1 was to select a working tax onomy of minor cyberloafing for use in the models. This was accomplished by de riving factor-models based on four taxonomies, comparing the model-data fit of the fou r factor-models, and selecting the taxonomy with the best fitting factor-model. Study 1 was necessary because multiple cyberloafing taxonomies are used in the cyberloafin g literature, and there was previously no empirical or theoretical reason to favor one of the taxonomies over the others. In Study 2, a series of causal minor-cyberloafing models was tested and evaluated. This was accomplished using structural equation modeling (SEM) and various model-data fit indices. Study 2Â’s data was cross-s ectional, so it is not possible to determine if any of the causal models are correct ( i.e., causation cannot be determined). However, SEM does allow one to determine if the cau sal models are plausible
21 Chapter 5 Study 1 The goal of Study 1 was to select a working minorcyberloafing taxonomy. The four taxonomies examined in Study 1 are: (a) LimÂ’s  taxonomy, (b) Blau et al.Â’s  taxonomy, (c) Mahatanankoon et al.Â’s  taxonomy, (d) and a general 1-factor taxonomy. Blanchard and HenleÂ’s (2008) and Mastran gelo et al.Â’s (2006) taxonomies are not investigated in this study because these taxono mies classify extreme behaviors (e.g., using work computers to traffic illicit drugs)Â—in a ddition to more common behaviorsÂ— and are therefore not taxonomies of minor cyberloafing. Method Participants. Participants were university students gathered from SONA, an electronic system designed to manage and schedule s tudies. Participants were prescreened based on their answers to two questions: ( 1) Â“Do you have a job that involves working with a computer?Â” and (2) Â“Do the computer( s) you use at work have internet access?Â” Four-hundred one men and women answered Â“ yesÂ” to both questions and were therefore eligible for this study. Materials. Participants completed three minor cyberloafing sca les and two exploratory measures. The three minor cyberloafing scales were, (a) Blanchard and HenleÂ’s  minor cyberloafing scale, (b) Mahata nankoon et al.Â’s  cyberloafing
22 scale, and (c) Mastrangelo et al.Â’s  nonprodu ctive cyberloafing scale. These scales were chosen based on their frequency in the literat ure, and their use of separate validation studies. Although there is much overlap in item co ntent across the different cyberloafing scales, each scale also measures unique behaviors ( e.g., Mastrangelo et al.Â’s scale is the only scale to measure cyberloafing behaviors relate d to building websites; Mahatanankoon et al.Â’s scale is the only scale to m easure the sending of e-cards). An analysis containing a comprehensive set of cyberloa fing behaviors is desirable so the entire minor cyberloafing domain can be represented Blanchard and HenleÂ’s scale. Blanchard and HenleÂ’s (2008) minor cyberloafing scale consists of 9 items. The lead-in question is Â“How often do you engage in each activity during work hours?Â” Participants rate the frequency of the behaviors on a fourpoint scale, from hardly ever (once every few months or less) to frequently (at least once a day) An example item is Â“Checked online personalsÂ”. Coefficient alpha was .85 in Study 1. Mahatanankoon et al.Â’s scale. Mahatanankoon et al.Â’s (2004) cyberloafing scale consists of 11 items. The lead-in question is, Â“Ho w often do you perform these activities at work?Â” Participants rate the frequency of the behaviors on a five-point scale, from never to always An example item is Â“Researching any products or services related to personal interests.Â” In Study 1, coefficient alpha was .77, .85, and .60 for the ecommerce, information research, and communication s ubscales, respectively. Mastrangelo et al.Â’s scale. MastrangeloÂ’s (2006) nonproductive cyberloafing scale consists of 15 items. The lead-in question i s, Â“Have you done these at work?Â”
23 Participants rate the frequency of the behaviors on a seven-point scale, from not in the past 6 months to almost constantly An example item is Â“Used the Internet while at w ork to visit sweepstakes sites that award prizes (iwon. com, etc.).Â” Coefficient alpha was .83 in Study 1. Exploratory scales. Two exploratory scales were included: Blanchard and HenleÂ’s (2008) serious cyberloafing scale and Mastr angelo et al.Â’s (2006) counterproductive cyberloafing scale. These scales were included to examine the relations among the serious cyberloafing scale, the counterproductive cyberloafing scale, and the minor cyberloafing scales. Procedure. Participants completed the cyberloafing scales as p art of mass testing in SONA. Participants had to complete all the scal es in mass-testing before they were allowed to sign up for a study. Since the cyberloa fing scales were similar in content, participant boredom and response tendencies were a concern. To partially mitigate this concern, cyberloafing scales from different researc h teams were separated by at least two non-cyberloafing scales. Ideally, the presentation order of the cyberloafing scales should be controlled for. However, in mass-testing it is not possible to alter the order of the scales for different participants. I assumed parti cipants would be most attentive at the beginning of mass-testing and least attentive at th e end of mass-testing, so I arranged the cyberloafing scales from longest to shortest. Thus all participants completed the cyberloafing scales in the following order: (1) Mas trangelo et al.Â’s scales, (2) Blanchard and HenleÂ’s scales, (3) Mahatanankoon et al.Â’s scal e, with at least two non-cyberloafing scales between each set of cyberloafing items.
24 Analysis. Four different confirmatory factor models for minor cyberloafing were tested and compared for fit using LISREL. The fact or models tested were: (a) a 1-factor general minor cyberloafing model, (b) a 2-factor mo del based on LimÂ’s  taxonomy, (c) a 3-factor model based on Mahatanankoon et al.Â’ s  taxonomy, and (d) a 3-factor model based on Blau et al.Â’s  taxonomy. The final factor structure for each taxonomy was arrived at with the same procedure: (1 ) Factor loadings were hypothesized based on previous factor loadings and item content; (2) a model based on the hypothesized structure was run; (3) non-significant paths were eliminated one-by-one based on theoretical considerations first and t-val ues second. Once all factor loadings were significant, the model was considered finished and ready to be compared against the other factor models. Models were compared using in cremental fit statistics (i.e., TLI, NFI, CFI, GFI), discrepancy-based fit statistics (i .e., RMSEA, SRMR), and the EVCI statistic. Results Item means were lower in the current sample than s amples reported in the literature (e.g., Blanchard & Henle, 2008; Mastrang elo et al., 2006). Perhaps the students in the current sample had jobs with lower autonomy and lower statusÂ—conditions that have been found to result in less cyberloafing (Gar rett & Danziger, 2008)Â—than previous samples. Generally, students in the current sample indicated that they rarely engage in most of the cyberloafing behaviors. Model-data fit. Table 1 shows a comparison of the model-data fit of the different factor-models. Two items had nonsignificant factor loadings across all analyses: Â“Played
25 computer games against your computer while at workÂ” and Â“Downloaded computer programs/applications (NOT job related)Â”. Consiste nt with protocol, final fit statistics were calculated without these items. Table 1 Model-Data Fit Statistics for the Minor Cyberloafin g Factor-Models Model 2 exact df p exact RMSEA ECVI TLI CFI GFI General 2,777.44 495 <.01 .12 9.01 .88 .89 .66 Lim 2,471.46 494 <.01 .11 7.92 .89 .90 .69 Mahat 2,421.92 492 <.01 .11 7.90 .90 .90 .69 Blau 3,535.61 492 <.01 .12 10.13 .85 .86 .64 Note GFI = Goodness of Fit Index The General 1-Factor Model. The General 1-Factor Model posits that there is one overall minor cyberloafing factor. The General 1-Factor model did not fit the data well, 2(495) = 2777.44, p < .05. Incremental fit indices were below the reco mmended .90 cut-off value (TLI = .88, NFI = .86, CFI = .89, GFI = .66) and the RMSEA was higher than the recommended .08 cut-off value (RMSE A =.12). An adequate fit value was observed for SRMR statistic (SRMR = .08), but o verall the results suggest that the General 1-Factor Model is not an appropriate model for minor cyberloafing. LimÂ’s 2-Factor Model. LimÂ’s 2-Factor Model posits that there are two cyberloafing factorsÂ—email and web-browsing (Lim, 2 002). LimÂ’s 2-Factor Model had comparable fit to the General 1-Factor Model based on the incremental (TLI = .89, NFI = .88, CFI = .90, GFI = .69) and discrepancy-based (R MSEA = .11, SRMR = .09) fit
26 indices. However, LimÂ’s 2-Factor Model had better fit than the General 1-Factor Model based on the EVCI index (EVCILim = 7.92, EVCIGeneral = 9.01). Mahatanankoon et al.Â’s 3-Factor Model. Mahatanankoon et al.Â’s 3-Factor Model posits that there are three cyberloafing fact orsÂ—e-commerce, information research, and communication (Mahatanankoon et al, 2 004). Mahatanankoon et al.Â’s 3Factor Model showed comparable fit to LimÂ’s 2-Facto r Model based on the incremental fit indices (TLI = .90, NFI = .88, CFI = .90, GFI = .69), the discrepancy-based fit indices (RMSEA = .11, SRMR = .11), and the EVCI (7.90). Blau et al.Â’s 3-Factor Model. Blau et al.Â’s 3-Factor Model posits that there are three cyberloafing factorsÂ—email, web-browsing, and interactive cyberloafing (Blau et al., 2004). Of the four factor-models, Blau et al. Â’s 3-Factor Model had the worst fit. The incremental fit indices (TLI = .85, NFI = .83, CFI = .86, GFI = .64) and the RMSEA (RMSEA = .12) were comparable to the other factor-m odels, but the EVCI and SRMR indices were considerably worse (SRMR = .22, EVCI = 10.13). Selecting a factor-model. Model-data fit was generally poor across the four tested models, suggesting that more work is needed on how to categorize cyberloafing behaviors. However, LimÂ’s 2-Factor Model and Mahat anankoon et al.Â’s 3-Factor Model showed considerably better fit than the other two m odels on the ECVI indexÂ—an index that is increasingly becoming favored by researcher s for comparing non-nested models (Brown & Cudeck, 1993). For this reason, I chose t o use LimÂ’s 2-Factor Model and Mahatanankoon et al.Â’s 3-Factor Model as my working minor cyberloafing models for Study 2.
27 Selecting items. Next, I had to select items to represent the differ ent minor cyberloafing factors. The selected items were to s erve as indicators for their respective factors in the structural equation models in Study 2. For each factor, I chose the three highest loading items to represent that factor. Th is resulted in the same items being chosen for the email and communication factors, and the same items being chosen for the web-browsing and information research factors. In other words, the only difference between my two working taxonomies is the presence o r absence of the e-commerce factor. One last step was needed to prepare the items for S tudy 2. Since the items came from different scales, they had different lead-in s tatements, and often different tenses. To make the scales easier to read, items were changed to a common tense. For example, the item Â“Conducting on-line shoppingÂ” was changed to Â“ Conduct on-line shoppingÂ” and the item Â“Checked non-work related emailÂ” was changed t o Â“Check non-work related emailÂ”. Discussion The purpose of Study 1 was to choose a working tax onomy of minor cyberloafing for Study 2. This was accomplished by comparing th e model-data fit of four factormodels based on four minor cyberloafing taxonomies. Comparable fit was found for LimÂ’s 2-Factor Model and Mahatanankoon et al.Â’s 3-F actor Model. Rather than arbitrarily choosing one of models, the decision wa s made to use both models in Study 2. Although the fit of the two models was not ideal, S tudy 1 allowed me to do two things necessary for Study 2: (a) rule out two minor cyber loafing models and focus on the selected working models, and (b) develop the subsca les needed for Study 2.
28 Chapter 6 Study 2 The goal of Study 2 was to test and evaluate a ser ies of causal minor-cyberloafing models. In order to do this, I had to finish propo sing the models. The minor cyberloafing factors to be included in the causal models had bee n selected in Study 1. The next step was to identify all the other variables to be inclu ded in the models (i.e., the variables proximal to minor cyberloafing). The next few subs ections describe these variables, and why they were hypothesized to be important to minor cyberloafing. Perceived Injunctive Norms Social norms are behavioral expectations of what is and is not acceptable behavior within a group or society. There are two types of social norms: injunctive norms norms and descriptive norms. Injunctive norms are what p eople say others should do, and descriptive norms are what people actually do. The two types of norms are not always in agreement: For example, a group of smokers may say that you shouldnÂ’t smoke (injunctive norm), even though everyone in the grou p does smoke (descriptive norm). Although there was some initial controversy over th e role of norms in predicting behavior (Schultz et al., 2007), since then researc h has clearly established that norms are important in guiding peopleÂ’s actions (Aarts & Dijk sterhuis, 2003; Cialdini, Kallgren, & Reno, 1991; Darley & Latane, 1970; Kerr, 1995; Terr y & Hogg, 2001). Indeed,
29 perceived injunctive norms have been the best predi ctor of cyberloafing found to date (Blanchard & Henle, 2008). The high correlation between cyberloafing and perce ived injunctive norms found in previous research ( r = .43; Blanchard & Henle, 2008) justifies the incl usion of perceived injunctive norms in the models. At the i ndividual level, norms are usually considered to be antecedents to behavior; thus, per ceived injunctive norms are included in the models as an antecedent to cyberloafing. It is expected that the results of this study will replicate results found by Blanchard and Henle (2008). Job Boredom Job boredom is the individualÂ’s subjective appraisa l of how dull or exciting his or her job is (Bruursema, 2007). Individuals who find their job boring, often experience state boredomÂ—a dissatisfying, low-arousal state, o ften attributed to lack of stimulation from the environmentÂ—at work (Farmer & Sundberg, 19 86; OÂ’Hanlon, 1981). There are two reasons to expect job boredom to stro ngly correlate with minor cyberloafing. First, cyberloafing can be considere d a type of withdrawal CWBÂ— behavior that restricts the amount of time one work s to less than what is expected (Spector et al., 2006)Â—and withdrawal behavior has been found to strongly correlate with job boredom ( r = .52; Bruursema, 2007). If cyberloafing is a typ e of withdrawal behavior, it should have the same relations with ot her variables as other withdrawal behaviors. Therefore, the high correlation found b etween job boredom and withdrawal CWB should also exist between job boredom and minor cyberloafing.
30 The second reason to expect job boredom to strongly correlate with minor cyberloafing is theoretical. As stated earlier, jo b boredom refers to oneÂ’s subjective appraisal of how dull or exciting his or her job is (Bruursema, 2007) and individuals who find their job boring, often experience state bored omÂ—a dissatisfying, low-arousal stateÂ—at work (Farmer & Sundberg, 1986; OÂ’Hanlon, 1 981). Since boredom is a dissatisfying state, when one experiences boredom he or she is m otivated to reduce his or her feelings of boredom. In the work context, wher e the number of engaging activities is often limited, cyberloafing can be an effective and discreet way to reduce boredom. Because job boredom is hypothesized in the CWB lite rature to be an antecedent to withdrawal behavior, and because job boredom is hyp othesized to motivate increases in minor cyberloafing, job boredom will be included in the models as an antecedent to minor cyberloafing. Task Performance Task performance is employeesÂ’ performance on the c ore parts of their job. Task performance for a salesperson may refer to how many sales he or she made in a given time span; the task performance for a McDonaldÂ’s em ployee might be how quickly he or she makes cheeseburgers. The defining feature of t ask performance is that it refers to core aspects of the employeeÂ’s job. Although the conceptual definition of cyberloafing implies that cyberloafing is harmful to task performance, a number of researcher s have suggested that cyberloafing is sometimes beneficial to task performance. These re searchers argue that cyberloafing can provide a much needed break, which can lead to impr oved task performance once the
31 employee resumes work (Anandarajan, Devine, & Simme rs, 2004; Anandarajan & Simmers, 2003; Belanger & Van Slyke, 2002; Block, 2 001; Greenfield & Davis, 2002; Oravec, 2002; Stanton, 2002). This idea, that short cyberloafing breaks can boos t task performance, is plausible. However, the focus of this study is on the overall relations among cyberloafing and its antecedents and consequences, as these relations oc cur in Â“the wildÂ”. In certain circumstances cyberloafing may be beneficial, but w hat is the overall relation between cyberloafing and task performance in organizations today? The studies on the prevalence and effects of cyberloafing, mentioned earlier, imp ly that what is occurring in organizations is not that employees are taking shor t breaks, but rather employees are spending considerable amounts of time cyberloafing. Furthermore, even if cyberloafing increases task p erformance once the employee resumes work, in order for the relation between cyb erloafing and task performance to be positive, the performance gain would have to be big enough to compensate for the time lost cyberloafing. Given the high base rates of cy berloafing discussed in the introduction, it seems unlikely that the benefits of cyberloafing will compensate for the productivity that could have occurred. Minor cyberloafing is th erefore hypothesized to negatively relate to task performance. The Ability to Hide Cyberloafing The ability to hide cyberloafing refers to a worker Â’s ability to hide his or her computer activity from his or her coworkers and sup ervisors based on variables in his or her work environment. Variables that are likely to be important to an employeeÂ’s ability
32 to hide cyberloafing include: (a) the visibility of the computer screen to coworkers and supervisors [e.g., computer screen facing the hallw ay vs. computer screen facing the wall], (b) the location of the employeeÂ’s computer [e.g., in an isolated corner vs. next to a busy hallway], (c) the employeeÂ’s ability to detect someone approaching [i.e., can the employee see his or her supervisor approaching?], a nd (d) whether or not the employeeÂ’s computer activity is recorded [e.g., whether or not the IP addresses they visit are logged]. The ability to hide cyberloafing is likely to be an important predictor of cyberloafing because it presumably affects the chan ces of being reprimanded for cyberloafing. Simply put, the ability to hide cybe rloafing lowers the risk of cyberloafing, which raises the expected value of cyberloafing. T hus, all other things being equal, an employee with a high ability to hide cyberloafing i s more likely to cyberloaf than an employee with a low ability to cyberloafing. Despite its potential importance, the ability to hi de cyberloafing has not been examined by cyberloafing researchers. This create s two problems for the current investigation. The first problem is conceptual: it is necessary to know which ability to hide cyberloafing factors are important to minor cy berloafing so those factors can be included in the causal-models. The second problem is practical: scales are needed to measure the relevant ability to hide cyberloafing f actors. These problems need to be addressed in a pilot study before the causal models can be finalized.
33 Chapter 7 The Ability to Hide Cyberloafing Pilot Study A pilot study was conducted to (1) examine the fact or structure of an initial ability to hide cyberloafing scale, (2) determine which factors are likely to be important antecedents to cyberloafing, and (3) finalize scales to measure th ese factors. Method Participants. Participants were 63 employees from various compani es. The sample was mostly male (71.2%), with a mean age 41. 86 years old (SD = 10.34). Many participants in the sample had high incomes: Over h alf of participants who completed the annual-household-income item indicated that they ha ve household incomes exceeding $100,000 a year. Participants also indicated that they worked many hours a week (M = 50.98, SD = 10.61). Materials. The ability to hide cyberloafing was measured with 17 items created for this study. Items covered various reasons why employees might have the ability to hide cyberloafing, including: (a) the visibility of the computer screen to coworkers and supervisors, (b) the location of employeeÂ’s compute r, (c) the employeeÂ’s ability to detect someone approaching, and (d) whether or not the emp loyeeÂ’s computer activity is recorded. In addition, global ability to hide cybe rloafing items were also included.
34 Cyberloafing was measured with Blanchard and Henle Â’s (2008) cyberloafing scale. Ideally, it would have been best to use the minor cyberloafing scales developed in Study 1, since those are the scales I included in S tudy 2. However, at the time of data collection, the results from the first study were i ncomplete. Since the purpose of including a cyberloafing scale in the pilot study i s to provide criteria from which to evaluate the predictive validity of the ability to hide cyberloafing factors, Blanchard and HenleÂ’s scale was deemed sufficient for this purpos e. Procedure. I approached travelers individually at a departure gate of a large international airport. Travelers were only approac hed if they were not engaged in another activity, such as reading or talking on a cell phon e. This selection strategy appeared to work well as it limited the number of potential par ticipants in a given area to a manageable fewÂ—possibly limiting selection bias on my part. Once a potential participant was selected, I appro ached him and asked if he would mind filling out a one-page questionnaire. If the participant agreed, I asked him Â“Do you have a job that involves working with a computer wi th internet access?Â” If he said Â“YesÂ”, I handed him the informed consent form and t he questionnaire. If he said Â“NoÂ”, I thanked him, but kindly explained that he was not e ligible to participate. Occasionally, a traveler would see me distributing the questionnair e, and ask to participate in the study. When a participant was filling out the questionnai re, I waited nearby. The survey took most participants about 10 minutes to complete When a participant completed the survey, he would hand the survey to me, and I would thank him for participating.
35 Results Items demonstrated sufficient variability: Most ite ms had mean responses around 4 ( neither agree, nor disagree ) and a standard deviation around 2. These results suggest that participants in different work situations do d iffer in their ability to hide cyberloafing. Factor structure. I first ran a principal axis exploratory factor ana lysis with no rotation to determine the number of factors to extr act. Three Â“elbowsÂ” were present in the scree plot, the locations of the elbows suggest ing a 1-, 2-, or 5-factor solution. I then ran three exploratory factor analyses, extracting o ne, two, and five factors. In order to select among the solutions, I needed to choose a cr iterion to compare them. I chose interpretability as my primary criterion sinceÂ—in o rder to appropriately place the factors within the causal-modelsÂ—it is necessary that the f actors be interpretable. Both the 1and 5-factor solutions were easy to interpret. Bec ause the purpose of the exploratory factor analysis was to identify factors which could possibly relate to minor cyberloafing, I decided thatÂ—all things being equalÂ—it was better t o have more factors than fewer. Thus, I decided the 5-factor solution was the best representation of the data for my purposes. Five-factor solution. The five-factor solution yielded five easily interp retable factors. Items that loaded highly on Factor 1 were items related to oneÂ’s global assessment of oneÂ’s ability to hide cyberloafing. The items that loaded highly on Factor 1 were, Â“I COULD hide my computer activity if I wan ted toÂ” ( = 1.03), Â“I COULD pretend to be working on my computer without anybod y knowingÂ” ( = .93), Â“Other employees donÂ’t know what I do on my computerÂ” ( = .63), Â“I COULD hide what I do
36 on my work computer from other employeesÂ” ( = .60), and Â“I COULD watch a 30minute video on my computer without anybody knowing Â” ( = .53). I named Factor 1 Â“ Perceived Ability to Hide CyberloafingÂ”. Items that loaded highly on Factor 2 were items re lated to how easily other employees could see oneÂ’s computer screen. The ite ms that loaded highly on Factor 2 were, Â“It is easy for people to see my computer scr een without me knowingÂ” ( = .88), Â“My computer screen is highly visible to other empl oyeesÂ” ( = .85), Â“There are a lot of people around me when I am workingÂ” ( =.66), and Â“Many people walk by my cubicle/office during the dayÂ” ( = .65). I named Factor 2 Â“Visibility of the Compu ter ScreenÂ”. Items that loaded highly on Factor 3 were items re lated to oneÂ’s ability to detect people approaching his or her work station. The it ems that loaded highly on Factor 3 were, Â“I can see people approaching my work station Â” ( = .83), Â“I can hear people approaching my work stationÂ” ( = .73), and Â“It is impossible for people to sneak up on me at workÂ” ( = .72). I named Factor 3 Â“Ability to detect peopl e approachingÂ”. Factors 4 and 5 were represented with two items and one item respectively, which is below the recommended minimum of three items for each factor. However, since the purpose of the exploratory factor analysis was to c reate factors which could possibly relate to minor cyberloafing, I decided to retain t he factors. Two items loaded highly on Factor 4, and both were related to the amount of monitoring from the organization. The items that l oaded highly on Factor 4 were, Â“My company keeps records on my computer activityÂ” ( = .93) and Â“My company monitors
37 my computer activityÂ” ( = .86). In order to maintain the same positive di rectional hypotheses as the other ability to hide cyberloafin g factors, I named Factor 4 Â“ Lack of Company MonitoringÂ”. Only one item loaded highly o n Factor 5: Â“I have an assigned computer at workÂ” ( = .78). I named Factor 5 Â“Assigned computerÂ”. Correlations with minor cyberloafing. The second goal of the pilot study was to determine which of the ability to hide cyberloaf ing factors are likely to be antecedents to minor cyberloafing. To accomplish this goal, I examined the relation between each factor and minor cyberloafing. Composite scores fo r each participant on a given factor were created by taking their mean response of all i tems whose (a) loadings were highest on the given factor, and (b) whose factor loadings were greater than .30. Composite scores for each factor were then correlated with mi nor cyberloafing. The correlations among composite scores and minor cyberloafing can be seen in Table 2. I decided a priori to use a 1-tailed sign ificance test since I had lower-thanexpected power (due to a lower-than-expected sample size). One-tailed significance tests were appropriate since I had clear directional hypo theses (Hayes, 1994). Table 2 Correlations Among the Ability to Hide Cyberloafing Factors Variable 1 2 3 4 5 1. Minor cyberloafing .82 2. Perceived AtHC .85 -.04 3. Visibility of the computer screen .82 -.12 .52 4. Ability to detect people .78 .08 .37 .38 5. Lack of company monitoring .93 .23 .36 .26 .02 6. Assigned computer -.10 .00 -.04 -.16 .04 Note. Correlations equal to or higher than .23 are sign ificant at the .05 level.
38 Out of the five ability to hide cyberloafing factor s, only the lack of company monitoring factor significantly correlated with minor cyberloa fing, r (63) = .23, p < .05. Selecting items. The last goal of the pilot study was to create a sh ort ability to hide cyberloafing scale. Since only two items load ed highly on the lack of company monitoring factor, eliminating items was unnecessar y. However, it is recommended that each latent construct in a structural equation mode l be represented by at least three indicators (Bollen, 1989). In order to meet this r ecommendation, a third item that closely resembled the first two items was created: Â“My comp any keeps logs of the websites I visitÂ”. Thus, the final lack of company monitorin g scale consisted of the following items: Â“My company monitors my computer activityÂ”, Â“My company keeps records of my computer activityÂ”, and Â“My company keeps logs o f the websites I visitÂ”. Coefficient alpha of the 2-item scale was .93 in the pilot stud y. Discussion The construct of the ability to hide cyberloafing was examined in this study. A number of factor analyses were conducted, and based the interpretability of the solutions, the five factor solution was chosen. Composite sco res for each participant for each factor were created, and these factor scores were correlat ed with minor cyberloafing. The lack of company monitoring factor was found to correlate significantly with minor cyberloafing, suggesting that lack of company monit oring might be an antecedent to minor cyberloafing. The results also suggest that if the ability to hide cyberloafing is an antecedent to cyberloafing, the relation is driven primarily by the lack of company monitoring factor.
39 Chapter 8 Study 2 Continued Causal Minor-Cyberloafing Models Given the relations hypothesized in Study 2Â’s intr oduction and the results from the Ability to Hide Cyberloafing Pilot Study, I now have two causal minor-cyberloafing models. The models posit that company monitoring, perceived injunctive norms, and job boredom affect the two/three minor cyberloafing fac tors, and that the two/three minor cyberloafing cyberloafing factors affect self-rated task performance. The last step is to test and evaluate these proposed models, as well as a number of plausible alternative models. In order to distinguish between models using diffe rent taxonomies, an Â“MÂ” suffix will be added to the name of models using Mahatanan koon et al.Â’s taxonomy, and an Â“LÂ” suffix will be added to the names of models using L imÂ’s taxonomy. For example, I will call the two above-mentioned models Â“Model 1MÂ” and Â“Model 1LÂ”. Method Power analysis. To conduct the power analysis, I used a table from a seminal SEM power-analysis article (MacCallum, Browne, and Sugawara 1996). I wanted at least 80% power to reject the null hypothesis of not -close fit. To ensure adequate power for each test, I based my power analysis off the ca usal model with the fewest degrees of
40 freedom (Model 2L, df = 177). Results indicated that I needed a sample size of 178 participants. However, so that I could drop proble matic participants (e.g., participants who do not work with a computer) and still maintain the designated level of power, I decided a minimum sample size of 200 participants w as needed. Participants and procedure. Participants were 220 male and female employees from downtown Tampa. Potential participants were a pproached by myself or one of my research assistants and asked to complete a short o ne-page survey. Participants were asked the following qualifying question: Â“Do you ha ve a job that involves working with a computer with internet access?Â” Participants who a nswered affirmatively were handed the survey, while I or a researcher assistant waite d nearby. Participants were offered a bottle of water for their participation, although t he large majority of participants declined the bottled water. Most participants took about 15 minutes to complete the survey. Materials. A one-page, front-and-back survey was created for S tudy 2. The survey consisted of 12 scales (some exploratory), f ive demographic and exploratory items, and one item to check the integrity of the d ata. Perceived injunctive norms. Perceived injunctive norms towards cyberloafing were measured with a 4-item cyberloafing norms scal e developed by Blanchard and Henle (2008). Participants were asked to rate thei r beliefs that their coworkers would approve of them using the internet for personal use on a 5-point scale (1 = strongly disapprove 5 = strongly approve ). An example item is, Â“My coworkers would approve of me using the Internet for non-work related purpo sesÂ”. Coefficient alpha was .89 in Study 2.
41 Job boredom. Job boredom was measured using four items from Le eÂ’s (1986) Job Boredom Scale. Participants were asked to resp ond to questions about how dull or exciting their job is on a 7-point Likert scale (1 = never 7 = always ). An example item is, Â“Do you get bored with your work?Â” Coefficient alpha was .80 in Study 2. Company monitoring. Company monitoring was measured using the 3-item scale developed in the pilot study. Participants r esponded to each item using a 7-point Likert scale (1 = strongly disagree 7 = strongly agree ). An example item is Â“My company keeps records of my computer activityÂ”. Co efficient alpha was .94 in Study 2. Minor cyberloafing. Minor cyberloafing was measured using the cyberloaf ing scales developed in Study 1. Each of the Â“fiveÂ” sc ales contained three items. Participants rated how much they engage in each act ivity on a 7-point Likert scale (1 = never did this 7 = almost constantly ). Coefficient alphas for the minor cyberloafing scales ranged from .77 to .86 in Study 2. Task performance. Self-rated task performance was measured using Wi lliam and AndersonÂ’s (1991) 7-item in-role behavior subscale. Participants rated their performance at work compared to their coworkers with the same j ob on a 5-point Likert scale (1 = a lot less than others 5 = a lot more than others ). An example item is, Â“I adequately complete assigned dutiesÂ”. Task performance items 5-7 (mean ritem-total = .32) had considerably lower corrected item-total correlation s than task performance items 1-4 (mean ritem-total = .80). Since I planned to use SEMÂ—and since SEM op erates at the factor levelÂ—I dropped task performance items 5-7 to creat e a more homogenous factor. Coefficient alpha for a scale consisting of items 1 -4 was .96 in Study 2.
42 Demographics and exploratory items. A number of additional items were included for exploratory and control purposes. Som e of these additional items measured demographic information (e.g., age, gender, job cat egory, hours worked per week), others measured cyberloafing (e.g., social networking site s, percentage of the work day spent cyberloafing), and still others measured potential antecedents and moderators of cyberloafing (e.g., cyberloafing intentions, cyberl oafing attitudes, computer knowledge, perceived ability to hide cyberloafing, descriptive norms). Analysis. The data were initially screened using SPSS. Three participants were dropped because they indicated that they did not wo rk with a computer with internet access. For each item, the mean, standard deviatio n, and corrected item-total correlations were calculated. The plausibility of each of the causal minor-cyber loafing models was tested using structural equation modeling (SEM) as implemented b y the program LISREL. In order to test the proposed models in LISREL, a number of ste ps were taken. First, the covariance matrix of the observed variables (i.e., the items) was calculated using SPSS. Then the covariance matrix, along with a template from an SE M course, was used to create the input file for Model 1M. Input files for ModelÂ’s 2 M, 3M, and 4M were subsequently created by modifying the input file for Model 1M. This three step process was repeated to generate input files for Models 1L, 2L, 3L, and 4L. After creating the input files, I tested the propo sed models using LISREL. After running each model, I examined the output to make s ure LISREL converged on a solution. After that, I examine the significance o f the path loadings between the latent
43 constructs and the observed variables, and the path loadings between latent variables and other latent variables. Next, model-data fit was e xamined using incremental fit indices (i.e., TLI, NFI, CFI, GFI), discrepancy-based fit i ndices (i.e., RMSEA, SRMR), and the test of not-close fit. The fit indices and statistical tests allowe d me to examine whether the proposed models were plausible representations of the data. Finally, I used the fit indices, including the EVCI, to compare the fit of the different models. Results Results of the SEM analyses are discussed below. A summary of the SEM analyses are shown in Table 3. Table 3 Model-Data Fit Statistics for Models Using Lim's (2 002) Taxonomy Model 2 exact df p exact RMSEA ECVI TLI CFI GFI Null Model 3,916.33 210 Model 1L 523.56 178 <.01 .09 2.97 .89 .91 .81 Model 2L 521.50 177 <.01 .09 2.97 .89 .91 .81 Model 3L 531.68 181 <.01 .09 2.98 .89 .91 .81 Model 4L 529.51 180 <.01 .09 2.98 .89 .91 .81 Note. GFI = Goodness of Fit index. Model 1M. Model 1M posits that company monitoring, perceived injunctive norms, and job boredom affect MahatanankoonÂ’s three types of cyberloafingÂ—ecommerce, information research, and communicationÂ—a nd that the three types of cyberloafing affect self-rated task performance. M odel 1M showed a moderate improvement over the null model as shown by the inc remental fit indices (TLI = .89, NFI = .87, CFI = .91, GFI = .78). However, only one of the incremental fit values was higher
44 than the recommended .90 cut-value. The RMSEA and SRMR for Model 1M were .10 and .12 respectively, suggesting poor model-data fi t. Overall, model-data fit was somewhat poor for Model 1M. Model 2M. Model 2M posits that company monitoring, injunctive norms, and job performance affect Mahatankoon et al.Â’s three cyber loafing factors, and that the three cyberloafing factors and job boredom directly affect task performance. Mod el 2M had similar incremental fit values to Model 1M (TLI = 89, NFI = .87, CFI = .91, GFI = .78). Again, two of these values were somewhat below the recommended .90 cut-value, and one of these values (i.e., the GFI value) was well below the recommended .90 cut-value. The RMSEA and SRMR for Model 2M were .10 and .12, s uggesting that there was discrepancy between the observed data and what we w ould expect to observe based on the model. The EVCI for Model 2M was the same as M odel 1M (EVCI = 3.97). Overall, Model 2M had comparable fit to Model 1M. Model 3M. Model 3M posits that company monitoring affects i njunctive norms, which influences the three types of cyberloafing, w hich affects task performance. Additionally, Model 3M posits that job boredom also influences the three types of cyberloafing. Model 3M had similar incremental fit values (TLI = .89, NFI = .87, CFI = .89, GFI = .76), and the discrepancy-based fit valu es (RMSEA =.09, SRMR =.12) to the first two models. The EVCI for Model 3M was 4.00, which is slightly higher (i.e., worse) than Models 1M and 2MÂ’s EVCI value of 3.97. Model 4M. Model 4M posits that company monitoring affects inj unctive norms, which influences the three types of cyberloafing, w hich affects task performance.
45 Additionally, Model 4M posits that job boredom dire ctly affects the three types of cyberloafing and task performance. Model 4M had similar incrementa l fit values (TLI = .89, NFI = .87, CFI = .89, GFI = .76), and discrepa ncy-based fit values (RMSEA =.09, SRMR =.12) to the first three models. The EVCI for Model 4M was 4.00, which is slightly higher than Models 1MÂ’s and 2MÂ’s EVCI valu e and the same as Model 3MÂ’s EVCI value. Model 1L. Model 1L posits that company monitoring, injunct ive norms, and job boredom affect LimÂ’s two cyberloafing factorsÂ—email and web-browsingÂ—and that the two cyberloafing factors affect self-rated task per formance. Model 1L had similar values to Models 1-4M for the incremental fit indices (TLI = .89, NFI = .87, CFI = .91, GFI = .81) and the discrepancy based fit indices (RMSEA = .09, SRMR =.10). However, Model 1LÂ’s EVCI (2.97) was considerably lower than the EV CIs for Models 1-4M (3.97-4.00). The EVCI statistic suggests that Model 1L is a more parsimonious model than Models 14M. Model 2L. Model 2L posits that company monitoring, injunctive norms, and job boredom affect LimÂ’s two cyberloafing factors, and that the two cyberloafing factors and job boredom directly affect task performance. Mode l 2L had identical values to Model 1L on all incremental fit indices (TLI = .89, NFI = .87, CFI = .91, GFI = .81), discrepancy-based fit indices (RMSEA= .09, SRMR= .1 0), and the EVCI (2.97). Model 3L. Model 3L posits that company monitoring affects inj unctive norms, which influence the two types of cyberloafing, whic h affects task performance. Additionally, Model 3L posits that job boredom infl uences the two types of cyberloafing.
46 Model 3L showed comparable fit to Models 1L and 2L based on incremental fit indices (TLI = .89, NFI = .86, CFI = .91, GFI = .81) discre pancy-based fit indices (RMSEA = .09, SRMR = .11), and the EVCI (3.98). Model 4L. Model 4L posits that company monitoring affects i njunctive norms, which influences the two types of cyberloafing, whi ch affects job performance. Additionally, Model 4L posits that job boredom dire ctly affects the two types of cyberloafing and task performance. Model 4L showed comparable fit to Models 1-3L based on the incremental fit indices (TLI = .89, NF I = .86, CFI = .91, GFI = .81), discrepancy-based fit indices (RMSEA = .09, SRMR = .11), and the EVCI (3.98). Summary of the Results. Overall, model-data fit was poor for all eight test ed models: Fit indices values were mostly outside the recommended cut-off values, and for no model was I able to reject the null hypothesis o f not-close fit. The poor model-data fit is likely due to the variables job boredom and job performance, whichÂ—judging by the significance of the path loadingsÂ—did not relate to the other latent variables as hypothesized. Despite the less-than-ideal fit, the models showed improved fit over the basic measurement model. Model-data fit was highly similar across the eight models, especially when the same minor cyberloafing taxonomy was used. If I ha d to choose one of the eight proposed models, I would choose Model 1L because it had the lowest EVCI, the lowest SRMR, and the highest GFI of the eight models. How ever, because Model 1L also had poor model-data fit, it is unlikely to be an accura te representation of how cyberloafing relates to the other studied variables.
47 Exploratory Analyses In addition to the primary analyses, I conducted a number of exploratory analyses. Within each set of exploratory analyses, the analys es were conducted using multiple cyberloafing variables as the dependent variable (e .g., all minor CL items, web-browsing, email). Two patterns were present across all analy ses: (a) The magnitude of the relations was greater when all minor CL items was used as the dependent variable, and (b) the magnitude of the relations were less when e-commerce was used as the dependent variable. Besides these two exceptions, results we re largely consistent in pattern and magnitude regardless of the cyberloafing variable u sed as the dependent variable. My first set of exploratory analyses examined the b ivariate correlations among the exploratory variables and various minor cyberloafin g variables. I found that the exploratory variables (a) perceived ability to hide cyberloafing, (b) perceived descriptive norms, (c) cyberloafing attitudes, and (d) cyberloa fing intentions were strongly correlated with all examined cyberloafing variables. For exam ple, the correlations between webbrowsing and perceived ability to hide cyberloafing, descri ptive norms, cyberloafing attitudes, and cyberloafing intentions were .36, .5 7, .58, and .57, respectively. To explore the combined predictive power of these n ew variables, I ran a number of regression models. The first question I had was Â“Do descriptive norms predict cyberloafing incremental to injunctive norms?Â” To answer this question, I conducted a hierarchical regression with injunctive norms in th e first step, and descriptive norms added in the second step. I used all the cyberloaf ing variables as criteria (e.g., web-
48 browsing, email), but the results were so consisten t in pattern and magnitude that only the results using web-browsing as the criterion will be reported. Adding perceived descriptive norms to the model res ulted in a significant change in R2 web, F (1, 199) = 44.12, p < .01, suggesting that descriptive norms did accou nt for variance in web-browsing unaccounted for by injunct ive norms. In fact, descriptive norms accounted for a substantial amount of varianc e unaccounted for by injunctive norms: Adding descriptive norms to the model increa sed the adjusted R2 by .14 unitsÂ— both variables together accounting for a surprising 35% of the variance in web-browsing. Furthermore, examination of the betas revealed that most of the variance was being accounted for by descriptive norms (desc = .45, inj = .22). My next question was, Â“What is the most amount of variance I can account for in minor cyberloafing while still keeping a relatively simple model?Â” To answer this question, I examined various combinations of the va riables that were found to significantly predict cyberloafing. Ultimately, th e model I came to favor was a linear regression model with perceived descriptive norms a nd cyberloafing attitudes as predictors. These two variables accounted for 45% of the variance in web-browsingÂ— even more than injunctive norms and descriptive nor ms. Examination of the betas revealed that each variable contributed about equal ly to the variance accounted for in web-browsing (desc = .38, att = .41). Discussion
49 The purpose of Study 2 was to test and evaluate a series of causal minorcyberloafing models. Contrary to expectations, the models did not fit the data well: No model had acceptable fit statistics, and for no mod el was I able to reject the null hypothesis of not-close fit. Model 1L had somewhat better fit than the other models, but not by much. The overall lack of model-data fit wa s likely due to the variables job boredom and self-rated task performance, which did not relate to the other latent variables as hypothesized. The results of the primary analyses were underwhel ming. However, three noteworthy findings came out of the exploratory ana lyses. First, four variables, previously untested in relation to cyberloafing, we re found to strongly predict minor cyberloafing. Second, descriptive norms were found to predict incremental to injunctive norms. And third, a parsimonious model consisting of the variables descriptive norms and cyberloafing attitudes was found to account for a substantial amount of the variance in minor cyberloafing. The findings from the exploratory analyses are emp irically interesting, but the findings have potential theoretical importance as w ell. First, recall that CWB researchers take different perspectives on the nature of CWB (R obinson & Bennett, 2003). Some researchers view CWB as an emotional reaction to ex periences at work, other researchers view CWB as reflection of oneÂ’s personality, while still others view CWB as an adaptation to the social context. CWB researchers who view CWB as an adaptation to the social context, typically draw off theories suc h Social Information Processing Theory (Salancik & Pfeffer, 1978) and Social Learning Theo ry (Bandura, 1977), which state that much of what we learn about the appropriateness of behaviors comes from other people
50 in the environment. The strong relation between mi nor cyberloafing and social norms found in the second set of exploratory analysesÂ—in combination with the fact that personality variables and emotional variables have only weakly correlated with minor cyberloafing in past studiesÂ—suggests that minor cy berloafing is perhaps best viewed from the adaptation to the social context perspecti ve. A second potential theoretical contribution can be extrapolated from the finding that a large amount of variance in minor cyberloafi ng was accounted for by the variables descriptive norms and cyberloafing attitudes. Thes e results suggest that the Theory of Reasoned ActionÂ—which posits that perceived social norms and attitudes influence intention to behave, and that intention to behavior influences behavior (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975)Â—may be an appropriate model for minor cyberloafing. However, more research is needed before any firm co nclusions can be made.
51 Chapter 9 General Discussion The goal of the presented investigation was to dev elop, test, and evaluate a series of causal minor-cyberloafing models. Obtaining thi s goal required a few intermediate steps, including: (1) selecting a working minor cyb erloafing taxonomy, (2) identifying important proximal variables to minor cyberloafing, and (3) hypothesizing a series of causal-minor cyberloafing models. These intermedia te steps were completed in Study 1, the pilot study, and the introduction to Study 2. In Study 2, I tested the model-data fit for each o f the eight models. Model-data fit was consistently poor: Most fit statistics were out side the recommended values and for no model was I able to reject the null hypothesis of n ot-close fit. The poor model-data fit suggests that the models are not reasonable represe ntations of minor cyberloafing and its relations with the other examined variables. Contributions to the Literature Despite the poor primary findings, the current inve stigation makes at least three contributions to the cyberloafing literature. The first contribution is that a number of strong, previously-untested relations were found am ong minor cyberloafing and some of the exploratory variables. In fact, the correlatio ns between minor cyberloafing and descriptive norms, cyberloafing attitudes, and cybe rloafing intentions are the highest
52 correlations with cyberloafing I am aware of. And the correlation between perceivedability-to-hide-cyberloafing and minor cyberloafing is almost as high. Thus, one contribution of the present studies is the identifi cation of four previously untested, but potentially important, cyberloafing antecedents. A second contribution of the present study is the f inding of the incremental power of measuring perceived injunctive and descriptive n orms. Injunctive norms were previously the best known predictor of minor cyberl oafing (Blanchard & Henle, 2008) and descriptive norms have been suggested (using di fferent terminology) as a possible predictor of minor cyberloafing. Study 2 showed th at descriptive norms are not only important, but that they predict incremental to inj unctive normsÂ—accounting for approximately 35% of the variance in minor cyberloa fing. But perhaps the most important finding in the pres ent investigation was the finding that descriptive norms and cyberloafing att itudes account for a surprising amount of the variance in minor cyberloafing (45% of the v ariance in Study 2). Why do I consider this the most important finding of the pre sent investigation? Because when combined with the fact that intentions were highly correlated with minor cyberloafing, the findings suggest that the well-established Theo ry of Reasoned Action (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975) is a useful theory for understanding cyberloafing. Therefore, although my goal to provi de a plausible causal minorcyberloafing model was not obtained in the primary analyses, a plausible model emerged accidently during the exploratory analyses. More r esearch is needed, but the strong relations found are certainly promising.
53 Limitations A number of limitations need to be acknowledged. F irst, self-rated task performance may have been a poor proxy for actual task performance. It is impossible to tell with the current data whether there is no rela tion between actual task performance and minor cyberloafing, or if no relation was found because actual task performance was not adequately measured. Future research should us e supervisor-rated task performance instead of self-rated task performance to better ex amine the relation between minor cyberloafing and task performance. A second limitation of the current investigation is that it is impossible to tell the importance of actual norms towards cyberloafing in determining cyberloafing. Are actual norms the real drivers of cyberloafing, and perceiv ed norms the mere mediators? Or are perceived norms influenced by other factors? Futur e research should examine the extent to which subjective norms and actual norms agree. To examine this, one could gather a sample of work groups, have each member in those wo rk groups fill out the injunctive and descriptive norms scales, and then look at the intraclass correlations. A high intraclass correlation would suggest that people ar e accurate in perceiving the cyberloafing norms, and would be consistent with th e hypothesis that objective norms are the real drivers of cyberloafing. A low intraclass correlation would suggest that there are no objective cyberloafing norms, and would be consi stent with the hypothesis that perceived norms are substantially influenced by oth er factors. A third limitation is that the perceived descripti ve norms-minor cyberloafing relation is slightly overstated by the Pearson corr elation. The perceived descriptive
54 norms-cyberloafing correlations in Study 2 are base d on participants who responded to at least one of the descriptive norms items. Fifteen of the 217 participants indicated that they did not know the descriptive norms of their co workers by writing something like Â“DonÂ’t KnowÂ” and leaving the descriptive norms item s blank. Therefore, the descriptive norms correlations should be interpreted as the cor relations among people who are aware of their coworkersÂ’ computer behavior. A fourth limitation is that job boredom was not di stinguished from excessive freetime. A person may think his or her job is boringÂ— not because he or she has nothing to doÂ—but because he or she finds the work itself bori ng. Perhaps excessive free-timeÂ—not job boredomÂ—is the critical variable. Future resea rch should tease apart the effect of free-time from the effect of job boredom in relatio n to cyberloafing. A fifth limitation is that a model based on the Th eory of Reasoned Action was not directly tested using SEM. Testing the model with current data is inappropriate since I measured past cyberloafing and intentions to cyberloaf in the future If the Theory of Reasoned Action was tested using the current data, the model would posit that intention to cyberloaf in the future causes past minor cyberloafing! Future research should test t he Theory of Reasoned Action model using appropriate d ata from a longitudinal study. A sixth limitation is that serious cyberloafing wa s excluded from the present investigation. Serious cyberloafing is likely to h ave different antecedents than minor cyberloafing (the person watching cat videos is pro bably different from the person watching pornography) and so separate models, at le ast for now, are appropriate. Whereas minor cyberloafing seems to be strongly inf luenced by social norms, there is at
55 least preliminary evidence that social norms are le ss important for serious cyberloafing (Blanchard & Henle, 2008). Perhaps serious cyberlo afing is driven more by individual personality characteristics, such as impulsivity an d machiavellism. Future research should focus on developing separate causal-models f or serious cyberloafing. Finally, most of the present studyÂ’s contributions are based on post hoc analyses. Future studies are needed to make sure the findings are robust. Future Directions Although LimÂ’s (2002) definition of cyberloafing (i .e., the misuse of the computer at work) has been fruitful, changes in technology a nd the way technologies are used suggest that LimÂ’s definition may be deficient, or at least need some clarification. For example, is cyberloafing limited to personal use of work computers, or is the use of personal devices at work (i.e., browsing the web on your phone) also cyberloafing? Is cyberloafing qualitatively different from general l oafing, or is it a different manifestation of general loafing? And is cyberloafing conceptual similar across different job types (Â“nine-to-five jobsÂ” vs. jobs where one often works from home)? Furthermore, the construct of cyberloafing is beginning to get fuzzi er as the boundary between being online and offline is blurred (e.g., an increasing portion of the population carries cellphones which are constantly connected to the intern et). One possible solution to these issues is to includ e harm-to-the-organization as a necessary component for cyberloafing. But this rai ses other issues: If a computer-related behavior reduces productivity, but makes it less li kely that the employee engages in larger CWBs (e.g., stealing), should that be consid ered cyberloafing? One could address
56 this discrepancy by specifying that cyberloafing is computer-related behavior that is harmful to the organization in the long-term however, this definition could be problematic if the same behavior has similar shortterm consequences but different longterm consequences in different organizations. A final conceptual issue that needs to be addresse d is how to model minor cyberloafing. The current practice is to model min or cyberloafing using effects indicators, but this sometimes results in seemingly good items being dropped [e.g., in the current study, the items Â“Played computer games aga inst your computer while at workÂ” and Â“Downloaded computer programs/applications (NOT job related)Â”]. Perhaps causal indicator models would be a more appropriate way to model minor cyberloafing. Once these conceptual issues are worked out, cyber loafing researchers can begin to work on other important problems. For example, how important is oneÂ’s perception of equity? If you work from home, are you more likely to find it justifiable to engage in personal activities at work? How do personality va riables interact with the situational variables to cause cyberloafing? Although in past studies personality variables have only weakly correlated with minor cyberloafing, perhaps certain lower personality variables (e.g., industriousness) are important to minor cybe rloafing. Summary and Conclusion In short, the proposed causal models were a bust. Descriptive norms predicted minor cyberloafing above and beyond injunctive norm s. And exploratory analyses strongly suggest that the Theory of Reasoned Action is an appropriate model for minor cyberloafing, but more research is needed.
57 References Aarts, H., & Dijksterhuis, A. (2003). The silence o f the library: Environment, situational norm, and social behavior. Journal of Personality and Social Psychology 84 18Â– 28. Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behav ior Englewood Cliffs, NJ: Prentice-Hall. American Management Association. (2001). Work-place testing: monitoring and surveillance. Available at: www.amanet.org/research /summ.htm. Accessed September 13, 2006. Anandarajan, M., & Simmers, C. A. (2003). Constructive and Dysfunctional Personal Web Usage in the Workplace: Mapping Employee Attitu des, Personal Web Usage in the Workplace London, England: Information Sciences Publishing. Anandarajan, M., Devine, P., & Simmers, C. (2004). Personal web usage in the workplace: A guide to eective human resource management Hershey, PA: Information Science Publishing. Andersson, L. M., & Pearson, C. M. (1999). Tit for tat? The spiraling effect of incivility in the workplace. Academy of Management Review 24(3) 452-471. Ashforth, B. (1994). Petty tyranny in organizations Human Relations 47 755-777. Bandura, A. (1977). Towards a unifying theory of b ehavioral change. Psychological Review 84 191-215.
58 Baron, R., & Neuman, J. (1996). Workplace violence and workplace aggression: Evidence on their relative frequency and potential causes. Aggressive Behaviour 22 161-173. Belanger, F., & Van Slyke, C. (2002). Abuse or lear ning? Communications of the ACM 45 64Â–65. Bensimon, H. (1997). What to do about anger in the workplace. Training and Development 28-32. Bies, R. J., & Tripp, T. M. (1998). Revenge in org anizations: The good, the bad, and the ugly. In R. W. Griffin, A. O'Leary-Kelly, & J. M.C ollins (Eds.), Dysfunctional behavior in organizations: Violent and deviant beha vior (pp. 221-239). Stamford, CT: JAI Press. Blanchard, A. L., & Henle, C. A. (2008). Correlates of different forms of cyberloafing: The role of norms and external locus of control. Computers in Human Behavior, 24 (3), 1067-1084. Blau, G., Yang, Y., & Ward-Cook, K. (2004). Testing a Measure of Cyberloafing. Journal of Allied Health, 35 (1), 9-17. Block, W. (2001). Cyberslacking, business ethics an d managerial economics. Journal of Business Ethics, 33 225Â–231. Bock, G., Kuan, H., Liu, P., & Sun, H. (2007). The role of task characteristics and organization culture in non-work related computing (NWRC). Lecture Notes in Computer Science, 4550 681. Bollen, K. A. (1989). Structural equations with latent variables Hoboken, NJ: John Wiley & Sons, Inc.
59 Browne, M. W., & Cudeck, R. (1993). Alternative way s of assessing model fit. Testing structural equation models, 154 136Â–162. Bruursema, K. (2007). How individual values and trait boredom interface w ith job characteristics and job boredom and their effects o n counterproductive work behavior Doctoral dissertation, University of South Flori da, Tampa. Burroughs, S. M. (2001). The role of dispositional aggressiveness and organi zational injustice on deviant workplace behavior Unpublished doctoral dissertation, University of Tennessee, Knoxville. Buss, D. (1993). Ways to curtail employee theft. NationÂ’s Business 36-38. Business Wire. (2000). A landmark survey by telemat e.net software shows that 83% of companies are concerned with the problem of interne t abuse. July 31. Christian, D. (2008). Big history [avi]. (Availab le from The Teaching Company www.teach12.com). Cialdini, R. B., Kallgren, C. A., & Reno, R. R. (19 91). A focus theory of normative conduct. Advances in Experimental Social Psychology 24, 201Â–234. Dalal, R. S. (2005). A meta-analysis of the relatio nship between organizational citizenship behavior and counterproductive work beh avior. Journal of Applied Psychology, 90 (6), 1241-1255. Darley, J.M., & Latane, B. (1970). Norms and normat ive behavior: Field studies of social interdependence. In J. Macaulay & L. Berkowitz (Eds .), Altruism and helping behavior (pp. 83Â–102). New York, NY: Academic Press.
60 Davis, R. A., Flett, G. L., & Besser, A. (2002). Va lidation of a new scale for measuring problematic internet use: Implications for pre-empl oyment screening. CyberPsychology & Behavior, 5 (4), 331-345. Deffenbacher, J.L. (1992). Trait anger: theory, fin dings, and implications. In C.D. Spielberger & J. N. Butcher (Eds.), Advances in personality assessment, Vol. 9 (pp. 177-201). Hillsdale, NJ: Lawrence Erlbaum Asso ciates. de Lara, P. Z. M. (2007). Relationship between orga nizational justice and cyberloafing in the workplace: Has 'anomia' a say in the matter? CyberPsychology & Behavior, 10 (3), 464-470. DiBattista, R. A. (1991). Creating new approaches to recognize and deter sabotage. Public Personnel Management, 20 347-352. Everton, W. J., Mastrangelo, P. M., & Jolton, J. A. (2005). Personality correlates of employees' personal use of work computers. CyberPsychology & Behavior, 8 (2), 143-153. Farmer, R., & Sundberg, N. D. (1986). Boredom prone ness: The development and correlates of a new scale. Journal of Personality Assessment, 50 (1), 4 17. Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behavior: An introd uction to theory and research Reading, MA: Addison-Wesley. Folger, R., & Baron, R. A. (1996). Violence and hos tility at work: A model of reactions to perceived injustice. In G. R. VandenBos & E. Q. Bulatao (Eds.), Violence on the job: Identifying risks and developing solutions (pp. 51-85). Washington, DC: American Psychological Association.
61 Foster, M. (2001). Be alert to the signs of employe e Internet addiction. National Public Accountant 46 39Â–40. Fox, S., & Spector, P. E. (1999). A model of work f rustration-aggression. Journal of Organizational Behavior, 20 915-931. Fox, S., Spector, P. E., & Miles, D. (2001). Count erproductive work behavior (CWB) in response to job stressors and organizational justic e: Some mediator and moderator tests for autonomy and emotions. Journal of Vocational Behavior, 59 1-19. Galletta, D. F. & Polak, P. (2003). An Empirical I nvestigation of Antecedents of Internet Abuse in the Workplace, Proceedings of the 2nd Annu al Workshop on HCI Research in MIS (Seattle, WA), pp. 12Â–13. Garrett, R. K., & Danziger, J. N. (2008). On cybers lacking: Workplace status and personal internet use at work. CyberPsychology & Behavior, 11 (3), 287-292. Giacalone, R.A., & Greenburg, J. (1997). Antisocial behavior in organizations Thousand Oaks, CA: Sage. Greenberg, J. (1990). Employee theft as a reaction to underpayment inequity: The hidden cost of pay cuts. Journal of Applied Psychology, 75(5) 561-568. Greenberg, J. (1993). Stealing in the name of justi ce: Informational and interpersonal moderators of theft reactions to underpayment inequ ity. Organizational Behavior and Human Decision Processes, 54 81-103. Greeneld, D. N., & Davis, R. A. (2002). Lost in cy berspace: The web @ work. CyberPsychology and Behavior 5, 347Â–353. Hayes, W. L. (1994). Statistics Florence, KY: Wadsworth Publishing Company.
62 Hershcovis, M. S., Turner, N., Barling, J., Arnold, K. A., Dupr, K. E., Inness, M., et al. (2007). Predicting workplace aggression: A Meta-ana lysis. Journal of Applied Psychology, 92 (1), 228-238. Jex, S. M. & Britt, T. W. (2008). Organizational Psychology: A Scientist-Practitioner Approach Hoboken, NJ: John Wiley & Sons, Inc. Kass, S. J., Vodanovich, S. J., & Callender, A. (20 01). State-trait boredom: Relationship to absenteeism, tenure, and job satisfaction. Journal of Business and Psychology 16, 317-327. Keashly, L. & Jagatic, K. (2003). By any other nam e: American perspectives on workplace bullying. In S. Einarsen, H. Hoel, D. Za pf, & C. Cooper (Eds.), Bullying and emotional abuse at work: Internationa l perspectives on research and practice London, UK: Taylor Francis. Kerr, N. L. (1995). Norms in social dilemmas. In D. Schroeder (Ed.), Social dilemmas: Perspectives on individuals and groups (pp. 31Â–48). Westport, CT: Praeger. Lavoie, J. A., & Pychyl, T. A. (2001). Cyberslackin g and the procrastination superhighway: A web-based survey of online procrast ination, attitudes, and emotion. Social Science Computer Review, 19 (4), 431. Lee K., Ashton M. C., Shin K., (2001). Personality correlates of workplace antisocial behavior Paper presented at the meeting of the Academy of Management, Washington, D.C. Lee, T.W. (1986). Toward the development and valid ation of a measure of job boredom. Manhattan College Journal of Business 15 22-28.
63 Lee, Z., Lee, Y., & Kim Y. (2004). Personal Web Pa ge Usage in Organizations, in M. Anandarajan and C. Simmers (eds.), Personal Web Usage in the Workplace: A guide to Effective Human Resources Management (pp.28-46). (Hershey, PA; Information Sciences Publishing.) Lim, V. K. G. (2002). The IT way of loafing on the job: Cyberloafing, neutralizing and organizational justice. Journal of Organizational Behavior, 23 (5), 675-694. Lim, V. K. G., Teo, T. S. H., & Loo, G. L. (2002). How do I loaf here? Let me count the ways. Communications of the ACM, 45 (1), 66-70. Lim, V. K. G., & Teo, T. S. H. (2005). Prevalence, perceived seriousness, justification and regulation of cyberloafing in Singapore: An exp loratory study. Information & Management, 42 (8), 1081-1093. Lichtash, A. E. (2004). Inappropriate use of e-mail and the Internet in the workplace: The arbitration picture. Dispute Resolution Journal, 59 26Â–36. MacCallum, R. C., Browne, M. W., & Sugawara, H. M. (1996). Power analysis and determination of sample size for covariance structu re modeling. Psychological Methods, 1 (2), 130-149. Mahatanankoon, P., Anandarajan, M., & Igbaria, M. ( 2004). Development of a measure of personal web usage in the workplace. CyberPsychology & Behavior 7(1), 93104. Malachowski, D. (2005). Wasted time at work costing companies billions. Retrieved December 15, 2005, from http://www.salary.com/careers/layoutscripts/crel_di splay.asp?tab=cre&cat=nocat &ser=Ser374&part=Par555.
64 Mastrangelo, P. M., Everton, W., & Jolton, J. A. (2 006). Personal use of work computers: Distraction versus destruction. CyberPsychology & Behavior, 9 (6), 730-741. Mills, J. E., Hu, B., Beldona, S., & Clay, J. (2001 ). Cyberslacking! A liability issue for wired workplaces. Cornell Hotel and Restaurant Administration Quarter ly 42 34Â–47. Naughton K., Raymond J., & Shulman K. (1999, Novemb er). Cyberslacking. Newsweek 134 (22), 62Â–65. O'Hanlon, J. F. (1981). Boredom: Practical conseque nces and a theory. Acta Psychologica, 49 (1), 53-82. OÂ’Leary-Kelly, A., Griffin, R., & Glew, D. (1996). Organization-motivated aggression: A research framework. Academy of Management Review 21 225-253. Oravec, J. A. (2002). Constructive approaches to In ternet recreation in the workplace. Communications of the ACM, 45 60Â–63. Panko, R. R., & Beh, H. G. (2002). Monitoring for p ornography and sexual harassment. Communications of the ACM, 45 84Â–87. Perlow, R., & Latham, L. L. (1993). The relationsh ip of client abuse and locus of control and gender: A longitudinal study in mental retarda tion facilities. Journal of Applied Personality, 78(5) 831-834. Puffer, S. M. (1987). Prosocial behavior, non-comp liant behavior, and work performance among commission salespeople. Journal of Applied Psychology, 72(4) 615-621. Robinson, S. L., & Bennett, R. J. (1995). A typolog y of deviant workplace behaviors: A multidimensional scaling study. Academy of Management Journal, 38 555-555.
65 Robinson, S. L., & Bennett, R. J. (2003). The past present, and future of workplace deviance Research. In J. Greenberg, Organizational Behavior: The State of the Science (pp. 235-268). Mahwah, NJ: Lawrence Erlbaum Assoc iates. Robinson, S. L., & OÂ’Leary-Kelly, A. M. (1998). Mon key see, monkey do: The influence of work groups on the antisocial behavior of employ ees. Academy of Management Journal, 41 658-672. Sablynski, C. J., Mitchell, T.C., James, L.R., & Mc intyre, M.D. (2001, August). Identifying aggressive individuals via conditional reasoning: An experimental study Paper presented to the meeting of the Academy of Management, Washington, DC. Salancik, G. R. & Pfeffer, J. (1978). A social inf ormation processing approach to job attitudes and task design. Administrative Science Quarterly 23 224-253. Scheuermann, L. S., & Langford, H. P. (1997). Perce ptions of Internet abuse, liability, and fair use. Perceptual and Motor Skills 85 847Â–850. Schultz P. W., Nolan, J. M., Cialdini, R. B., Golds tein, N. J., & Griskevicius, V. (2007). The constructive, destructive, and reconstructive p ower of social norms. Psychological Science, 18 (5), 429-434. Sipior, J. C., & Ward, B. T. (2002). A strategic re sponse to the broad spectrum of Internet abuse. Information Systems Management, 19 71Â–79. Skarlicki, D. P., & Folger, R. (1997). Retaliation in the workplace: The role of distributive, procedural, and interactional justice Journal of Applied Psychology, 82(3) 434-443.
66 Spector, P.E. (1975). Relationships of organization al frustration with reported behavioral reactions of employees. Journal of Applied Psychology, 60 635-637. Spector, P.E. (1997). The role of frustration in a ntisocial behaviour at work. In : Giacalone, R.A. & Greenburg J. (Eds.), Antisocial Behavior in Organizations (pp. 1-17). Thousand Oaks, CA: Sage. Spector, P. E., Fox, S., Penney, L. M., Bruursema, K., Goh, A., & Kessler, S. (2006). The dimensionality of counterproductivity: Are all coun terproductive behaviors created equal? Journal of Vocational Behavior, 68 (3), 446-460. Spector, P. E., OÂ’Connell, B. J. (1994). The contri bution of personality traits, negative affectivity, locus of control and Type A to the sub sequent reports of job stressors and job strains. Journal of Occupational and Organizational Psycholo gy, 67 1-11. Stanton, J. M. (2002). Company prole of the freque nt Internet user. Communications of the ACM, 45 55Â–59. Stuckless, N., & Goranson, R. (1992). The vengeance scale: development of a measure of attitudes toward revenge. Journal of Social Behavior and Personality, 7(1) 25-42. Tangney, J. P., Wagner, P. E., Hill Barlow, D., Mar schall, D. E. & Gramzow, R. (1996). Relation of shame and guilt to constructive versus destructive responses to anger across the lifespan. Journal of Personality and Social Psychology, 70 797-809. Terry, D. J., & Hogg, M. A. (2001). Attitudes, beha vior, and social context: The role of norms and group membership in social inuence proce sses. In J.P. Forgas & K.D. Williams (Eds.), Social inuence: Direct and indirect processes (pp. 253Â–270). Philadelphia, PA: Psychology Press.
67 Vardi, Y., & Weiner, Y. (1996). Misbehavior in orga nizations: A motivational framework. Organizational Science, 7(2) 151-165. Young, K. S., & Case, C. J. (2004). Internet abuse in the workplace: New trends in risk management. CyberPsychology & Behavior, 7 (1), 105-111. Williams, L. J., & Anderson, S. E. (1991). Job sati sfaction and organizational commitment as predictors of organizational citizens hip and in-role behaviors. Journal of Management, 17 (3), 601-610. Wyatt, K., & Phillips, J. G. (2005). Personality as a predictor of workplace Internet use. Proceedings of OZCHI 2005. ACM ISBN : 1-59593.