Anxiety: According to self-efficacy theory, information obtained through one’s physiological states represents another source of self-efficacy information, although a relatively weak one. Since the literature suggests that information technology plays a key role in enabling remote work, the employee’s physiological state towards computers (i.e., computer anxiety) was included in the current research model. While computer anxiety does not represent anxiety towards all forms of IT, it does appear to capture a large component of general IT anxiety. Self-efficacy theory predicts that high computer anxiety should lead to lower self-efficacy:
Hypothesis 4: The higher the employee’s computer anxiety level, the lower the employee's remote work self-efficacy.
Environmental Factors: The last source of self-efficacy information in the current research model deals with environmental factors. As suggested by Gist and Mitchell (1992), the environment can contribute a variety of cues that influence self-efficacy assessments. Two environmental factors were included in the model examined here: physical conditions and level of connectivity. Physical working conditions was suggested by Gist and Mitchell (1992) as a possible environmental factor which could affect self-efficacy judgments, and it was found to be an important issue for remote workers (Staples, 1996). Gist and Mitchell (1992) suggested that a good physical working environment will positively impact an individual’s performance beliefs. Distractions, noise, and interruptions will all likely hurt performance, as will inadequate physical conditions (e.g., inadequate furniture and space). Thus:
Hypothesis 5: Positive physical working conditions will be associated with higher levels of employee remote work self-efficacy.
The second environmental factor included in the model was level of connectivity. The level of connectivity that the remote employee has to his/her manager and co-workers, defined here by the types of information technology available for use, is likely to affect both the ease of and amount of communication undertaken. This is consistent with the literature suggesting that virtual organizations are enabled by IT (e.g., Freedman, 1993; Handy, 1995; Illingworth, 1994; Lucas, 1996; Lucas & Baroudi, 1994; Mowshowitz, 1994). Consequently, level of connectivity is likely to be an important cue to individuals in determining their beliefs that they can carry out certain tasks:
Hypothesis 6: The greater the level of connectivity provided to an employee, the higher the employee's remote work self-efficacy.
Self-Efficacy The ability to use IT represents an important component in an employee's ability to perform effectively in a remote management environment. Therefore, high levels of IT self-efficacy should also enhance the remotely-managed employee's remote work self-efficacy, and their ability to work effectively in a remote management setting:
Hypothesis 7: Higher levels of information-technology self-efficacy will be associated with higher levels of remote work self-efficacy.
Outcomes of Self-Efficacy
Five outcome variables were included in the research model, as shown on the right-hand side of Figure 2. The first outcome, performance, is an integral part of self-efficacy theory. The other four outcomes were suggested by previous research and the results of 19 focus group sessions (Staples, 1996, 1997). The focus groups were carried out with 104 remote managers and remotely-managed employees from five North American organizations to identify the key issues of remote management and remote work and to identify potential best practices. The findings of this research suggested that the other four outcome variables in Figure 2, namely job satisfaction, coping ability, organizational commitment, and job stress were all important in a remote management context.
Performance: Previous studies have found that self-efficacy is linked closely to task performance (e.g., Bandura, 1978; Gist & Mitchell, 1992; Locke, 1991). In the present context, this body of research suggests that employees who have a high level of remote work self-efficacy are likely to believe that they are more effective at performing tasks that enable remote work, and thus believe that they are more effective remote workers overall:
Hypothesis 8: High levels of employee self-efficacy on remote work-enabling tasks will be related to employees' positive perceptions of their performance.
Job Satisfaction: In his summary of previous research, Locke (1976) suggested that information on the task activities, achievement, rewards, working conditions, and management practices can all have an impact on individuals' perceptions of job satisfaction. These constructs are all viewed as antecedents to job satisfaction in our research model. Furthermore, participants in the focus group research by Staples (1996) suggested that perceptions of job satisfaction in a virtual environment would vary depending upon the support and activities of management and on the remote individual's competence in working remotely. Taken together, these results suggest that positive judgments about one's ability to perform tasks (i.e., self-efficacy) should have a positive impact on the satisfaction associated with one's doing those tasks (i.e., job satisfaction). Although this prediction is not based on existing self-efficacy theory, it represents a natural extension of the theory:
Hypothesis 9: High levels of employee remote work self-efficacy will lead to higher levels of remote job satisfaction.
Ability to Cope: Previous research has found significant relationships between self-efficacy and ability to cope (Bandura, 1982; Saks, 1995). Specifically, self-efficacy has been found to be positively associated with an individual's ability to cope with a variety of different situations. Thus:
Hypothesis 10: High levels of employee remote work self-efficacy are positively related to the employee’s ability to cope.
Organizational Commitment: Staples (1996) found that feelings of isolation could lower an individual's organizational commitment in virtual work settings. However, he also found that effective management practices and communication often reduce such feelings of isolation while increasing organizational commitment. It therefore appears reasonable to hypothesize the following:
Hypothesis 11: High levels of remote work self-efficacy in an employee will be positively related to his or her level of organizational commitment.
Job Stress: Previous research has demonstrated a significant relationship between self-efficacy and stress (e.g., Bandura, 1982; Saks, 1995). Specifically, increases in self-efficacy have been found to be negatively associated with stress:
Hypothesis 12: High levels of employee remote work self-efficacy will be related to lower levels of employee job stress.
Method
A quantitative research design was chosen to examine the proposed relationships among the various constructs in the research model. A questionnaire was used in order to facilitate collection of information from a large and geographically dispersed sample. This section describes the sampling method, construct measures, and analysis methods employed.
Sample
A questionnaire was sent to 1,343 individuals working in 18 North American organizations, who (1) employed individuals who worked remotely from their managers, and (2) were interested in participating in a study of remote management. Completed questionnaires from 631 respondents were returned, for an overall response rate of 47%. Use of the procedure suggested by Armstrong and Overton (1977) indicated no significant differences between respondents and non-respondents on a variety of demographic variables included in the questionnaire. Thus, non-response bias did not appear to be a major problem.
In the current study, workers were defined as remote or non-remote in terms of their physical proximity to their managers. If employees worked in a different building than their managers (which could be across the city, the state, the country, or even the globe), the employees were considered to be remote workers, since they were working remotely from their managers. In virtual organizations, many employees regularly conduct work in locations that are remote from managers and co-workers.
A total of 376 of the returned questionnaires were from remotely-managed employees, representing the actual sample of interest for the study reported here.3 Forty-seven percent of these respondents worked in private sector high technology firms, 22% worked in private sector financial service firms, and the remaining 31% worked in the public sector. Although all employees included in this sample worked remotely, only seventeen per cent actually worked at home, with the vast majority of the seventeen percent indicating that it was easy for them to do so. The median distance between the respondents’ office and their managers’ office was 483 kilometres.
Construct Measurement
The questionnaire completed by the respondents contained multiple measurement items relating to each of the constructs in the research model. Wherever possible, appropriate scales that had demonstrated good psychometric properties in previous studies were employed. However, for the remaining constructs, sets of items were generated based on reviews of previous relevant literature and expert opinion.
In order to achieve acceptable levels of measurement reliability and validity, both a pre-test and a pilot study were carried out, following the guidelines suggested by Dillman (1978). Questionnaire pre-testing was first completed using faculty, graduate student, and practitioner input. This information was used to refine the original survey instrument. A preliminary pilot study questionnaire was then administered to remote employees in one insurance firm, resulting in 64 responses. The resulting data were analyzed and used to further modify the questionnaire items for the full study.
Appendix A contains a list of the items used to measure the constructs. The number of items used to measure each construct and the resulting internal consistencies of the constructs are provided in Table 1.
Five of the constructs were measured using scales taken from the literature. A short form of the Computer Anxiety Rating Scale (Heinssen, Glass & Knight, 1987), developed by Compeau (1992), was used to assess computer anxiety of respondents. Four items from House, Schuler, and Levanoni's (1983) role ambiguity/coping ability scale were used to measure ability to cope. This short form of the House et al. (1983) scale had previously been used successfully by Saks (1995). Four items from the short version of the Mowday, Steers and Porter (1979) Organizational Commitment Questionnaire were used to measure organizational commitment. A five item scale developed by Rizzo, House and Lirtzman (1970) was used to measure job stress.
Finally, job satisfaction was measured initially using a 15 item scale developed by Warr, Cook and Wall (1979). Although the reliability of this scale had been found to be adequate in the past (e.g., British Telecom, 1984), the results of the pilot test for the current study found the scale to be multi-dimensional in the virtual organization context. One subset of items broke into a dimension dealing with satisfaction with management, while the other dimension appeared to deal with issues about other aspects of the job (i.e., physical work conditions, rate of pay, hours of work, variety in the job, and job security). After further review, five items were used to measure the satisfaction with management construct and five items were used to measure the construct dealing with other job satisfaction factors.
The remaining eight constructs in the research model were measured with scales developed for this study. The modeling best practices by the manager construct was measured with 14 items. These items were drawn from Staples (1996), in which key remote management best practices such as using IT effectively to facilitate communication were identified. The best practice items were further validated through a series of structured interviews with remotely-managed employees.
Three items were developed to measure the connectivity construct. One item addressed remote-access capability and was created by summing items which asked respondents about their ability to use their e-mail, groupware, and telephone / voicemail systems remotely. The second item was a sum of the responses to questions dealing with respondents' use of various IT tools (i.e., laptops, desktop PC's, modems, fax, cellular phones, and pagers). The third item assessed respondents' access to voice mail, e-mail, groupware, and videoconferencing systems, and was created by summing responses to four questions which determined whether or not they had access to each of the specific technologies/systems at their place of work.
The items for the IT Self-Efficacy construct assessed employees' self-efficacy judgments of their ability to perform specific IT tasks. The measurement of this construct was made up of four items each representing one IT subscale. These subscales were comprised of several questions dealing with specific tasks possible with each type of IT (i.e., telephone/voice mail, e-mail, groupware and videoconferencing). Respondents were first asked in each case if they could do a specific activity using the relevant IT to help perform their work or communicate with their manager. If they answered yes, they were further instructed to rate, on a 1 to 9 scale, their confidence in the judgment that they could do that task. If they answered no, the answer was coded as zero; otherwise it was coded as the value of the judgment (i.e., 1 to 9). This method of measuring self-efficacy was found to have high validity in a study of different methods of measuring self-efficacy by Lee and Bobko (1994).
Remote work self-efficacy relates to the employee’s belief or judgments that they can carry out tasks that are required to work effectively in a remote environment. The focus groups, mentioned earlier, consistently identified a number of tasks as being important across a wide range of job categories (Staples, 1996). These were relatively generic tasks that were independent of the employee’s specific job function. Sixteen items in the questionnaire were used to assess the respondents' self-efficacy of performing these tasks. An overall self-efficacy score was computed by summing these 16 items. This is standard with self-efficacy measures (Lee & Bobko, 1994).
The last construct relates to performance. The general self-efficacy model shown in Figure 1 indicates that previous performance accomplishments act as inputs to an individual's self-efficacy assessments, which in turn affect subsequent behavior and performance. We did not attempt to measure the level of performance of the various tasks that comprised the remote work self-efficacy scale. This was because we felt it would be difficult to get accurate self-assessments on actual levels of performance on the wide range of tasks faced by remote employees. Instead, we measured the respondents' beliefs about the effectiveness of working remotely in general as well as their own overall perceived productivity. In this way, we relied on previous research that has shown strong links between self-efficacy judgments and the performance of tasks. Although the self-reported measures of performance used here might be viewed as potential inputs to self-efficacy, we believe they are more correctly modeled as outcomes of the self-efficacy assessment process. Specifically, the items used to measure the overall productivity construct determined the individual's general productivity, rather than remote work-specific productivity. Thus, overall productivity cannot really be considered as an appropriate input to remote work self-efficacy. The items used to measure remote work effectiveness deal with the individual's general perception of remote work effectiveness, but do not assess the individual's own ability to perform remote work.
Analysis
A structural equation modeling technique called Partial Least Squares (PLS) was chosen for analyzing the research model (Wold, 1985). PLS is a technique that uses a combination of principal components analysis, path analysis, and regression to simultaneously evaluate theory and data (Pedhazur, 1982; Wold, 1985). The path coefficients in a PLS structural model are standardized regression coefficients, while the loadings can be interpreted as factor loadings. A detailed discussion of the implementation of PLS in an information systems context is provided by Barclay, Higgins and Thompson (1995), who also compare PLS and LISREL. PLS is ideally suited to the early stages of theory development and testing -- as is the case here -- and has been used by a growing number of researchers from a variety of disciplines (e.g., Birkinshaw, Morrison & Hulland 1995; Green, Barclay & Ryans 1995; Higgins, Duxbury and Irving 1992).
The explanatory power of the model is tested by examining the size, sign, and statistical significance of the path coefficients between constructs in the model (Davies, 1994). The statistics for the paths are generated using a jackknifing technique (Fornell & Barclay, 1983). The predictive capacity of a PLS model can also be evaluated by examining the variance explained (i.e., R2) in the dependent (or endogenous) constructs. The objective of a PLS analysis is to explain variance in the endogenous constructs, rather than to replicate the observed covariance matrix, as is the case with covariance structure techniques (such as LISREL). One consequence of using a variance-minimization objective is the absence of overall fit statistics for PLS models (Hulland, 1998).
Results
Given both the exploratory nature of the proposed research model and the relatively large sample size, a decision was made to randomly split the data into two subsets. The first of these subsets (n = 184) was used to test the initial research model, as shown in Figure 2. Based on what was learned from the estimation of this model, we identified a refined model. We then used the second data subset (n = 185) to validate this reduced model. The results of the initial model analysis are presented first, starting with an assessment of the measurement model (i.e., reliability and validity of the measures), and followed by a formal test of the hypotheses. The results for the refined model are then presented.
Measurement Model Assessment
Table 1 reports internal consistency values for each of the constructs in the research model (using both a measure proposed by Fornell and Larcker (1981) and Cronbach's alpha), and average variance extracted (a measure used to assess discriminant validity). Table 2 presents the inter-correlations of constructs. The diagonal element of Table 2 is the square root of the average variance extracted. This table can be used to assess the discriminant validity of the constructs.
Table 1: Internal Consistency of the Constructs
Table 2: Discriminant Validity Analysis
The bold diagonal elements are the square root of the variance shared between the constructs and their measures (i.e., the average variance extracted). Off diagonal elements are the correlations between constructs. For discriminant validity, the diagonal elements should be larger than any other corresponding row or column entry.
With few exceptions, the constructs all had acceptable reliability and validity. Two constructs (e.g., IT self-efficacy, and satisfaction with other job factors) had somewhat lower Cronbach's alphas; however, each of them had acceptable internal consistency values as assessed using the Fornell and Larcker (1981) measure. The latter is calculated independently of the number of items employed for a construct, whereas alpha is not, and it thus provides a more robust assessment of internal consistency. Furthermore, the Fornell and Larcker approach (1981) uses the observed loadings, and therefore more accurately reflects the relative importance of each of the underlying measures. Thus, use of the Fornell and Larcker (1981) internal consistency values is preferred, and it was concluded that all of the constructs used here had acceptable internal consistency.
Remote work self-efficacy was a single item construct, so its internal consistency and validity could not be directly assessed within the PLS model. However, the Cronbach's alpha for the 16 items which were used to construct the single score was 0.84, indicating strong internal consistency. Also, the item did not load highly on any other construct and the correlations with other constructs were generally low (Table 2), implying adequate discriminant validity.
An examination of Table 2 shows that the discriminant validity was weak between IT experience and training, connectivity, and IT self-efficacy. This can be seen by examining the correlations among the three constructs and the square root of the average variance extracted. In all cases, these correlations exceeded the average variance extracted indicating weak discriminant validity. The implications of this weak discriminant validity will be dealt with more fully later, but its presence implies that all three constructs are closely inter-related.
Three measurement items were retained despite having individual loadings of less than 0.60. The first of these was an item dealing with telephone self-efficacy (part of the set of measures for the IT self-efficacy construct). Although the reliability of this item was rather weak (0.53), it was retained in the model since this is a central construct and the four items together displayed adequate internal consistency (0.78). The two other problematic measures were both related to the Satisfaction with Other Job Factors construct (i.e., two items had loadings of 0.50 and 0.57, respectively). However, all five items were retained in order to reflect the full range of job factors typically used to assess job satisfaction (e.g., Locke, 1976).
Assessment of the Structural Model
Given our acceptance of an adequate measurement model in the previous section, it is appropriate to now turn to an examination of the structural model. This was done in two steps. The predictive power of the model was assessed first, followed by an analysis of the hypothesized relationships among the constructs.
The Predictive Power of the Model
The predictive power of the model is summarized in Table 3. The model explained 35.5% of the variance in remote work self-efficacy and 82.4% of the variance in IT self-efficacy. This latter result is clearly an artifact of poor discriminant validity and will be dealt with later. About 9.7% of the variance was explained in remote work effectiveness construct, while 13.8% of the variance in overall productivity was explained. The two constructs that dealt with job satisfaction, satisfaction with management and satisfaction with other job factors, had R2 values of 15.7% and 14.6%, respectively. The variance explained in the ability to cope construct was 11.4%. Finally, the model explained 13.6% of the variance in organizational commitment and 12.8% of the variance in job stress. Overall, the amount of variance explained by the initial model appeared reasonable. For all of the outcome constructs, self-efficacy would be one of many things affecting the respondents' attitudes and behaviors, resulting in the relatively modest R2 values.
Table 3: The Predictive Power of the Model
Hypothesis Testing
Figure 3 indicates both the variance explained for the individual endogenous constructs and the estimated path coefficients, while Table 4 contains a summary of the hypotheses, the path coefficients obtained from the PLS analysis of the initial model, and the t-values (and associated significance levels) for each path. For the antecedent constructs in the model, the following paths were statistically significant: remote work experience and training to remote work self-efficacy (hypothesis 1); IT experience and training to IT self-efficacy (hypothesis 2); modeling by the manager to both remote work self-efficacy and IT self-efficacy (hypothesis 3); computer anxiety to remote work self-efficacy and IT self-efficacy (hypothesis 4); connectivity to IT self-efficacy (hypothesis 6); and IT self-efficacy to remote work self-efficacy (hypothesis 7). Three paths were not significant: physical conditions to IT self-efficacy (hypothesis 5), physical conditions to remote work self-efficacy (hypothesis 5), and connectivity to remote work self-efficacy (hypothesis 6).
On the right-hand side of the model, the following paths were significant: remote work self-efficacy to both remote work effectiveness and overall productivity (hypothesis 8); remote work self-efficacy to both satisfaction with management and satisfaction with other job factors (hypothesis 9); remote work self-efficacy to ability to cope (hypothesis 10); remote work self-efficacy to organizational commitment (hypothesis 11); and remote work self-efficacy to job stress (hypothesis 12). All of these paths were in the directions hypothesized.
Figure 3: The Research Model with Path Coefficients and R2 Values
* p < .05; ** p < .01; ***p < .001 (2 tailed test)
Table 4: Summary of Path Coefficients and Significance Levels
* p < .05; ** p < .01; ***p < .001 (2 tailed test)
Table 4 continued: Summary of Path Coefficients and Significance Levels
* p < .05; ** p < .01; ***p < .001 (2 tailed test)
To summarize, fifteen of the eighteen estimated path coefficients in the initial model were statistically significant in the predicted direction, providing strong overall support for the proposed model. However, hypothesis 5 (dealing with physical conditions) was not supported at all, while hypothesis 6 was only partially supported.
Refined Model Results
Although the IT experience and training, connectivity, and IT self-efficacy constructs are conceptually distinct, the initial research model results indicated that they were highly inter-related at the empirical level. Further examination of the three constructs' measures suggested that all three could be viewed as indications of the respondents' capabilities with IT. Thus, for the refined model we decided to combine these three IT-related constructs into a single, over-arching construct (IT Capabilities). The measures used for this new construct were the same as those used in the prior analysis (i.e., 13 items; internal consistency = 0.92; Cronbach's alpha = 0.81). In addition, the Physical Conditions construct was dropped from the refined model entirely, since the results from the initial model indicated no support for the link between this construct and self-efficacy. Figure 4 shows the simpler, refined model resulting from these changes.
Figure 4: The Revised Research Model with Path Coefficients and R2 Values
* p < .05; ** p < .01; ***p < .001 (2 tailed test)
The refined model was analyzed using the second data subset (n = 185), again using PLS. The explanatory power of the model was reduced only slightly (i.e., the remote work self-efficacy R2 = 0.31 versus 0.36 previously), suggesting acceptable model stability across the two data subsets. As can be seen by the path coefficients and significance levels reported in Figure 4, all of the estimated path coefficients in this simpler model were significant and in the direction hypothesized, with two exceptions. Neither the path from remote work self-efficacy to organizational commitment nor the path from remote work self-efficacy to job stress were significant.
Discussion
In this section we discuss, in turn, the overall predictive power of the model, the relationships between the antecedents of self-efficacy and self-efficacy, and the relationships between self-efficacy and the outcomes of self-efficacy. We conclude with a discussion of the study's limitations, implications for theory and future research, and implications for management practices.
Model Results
Predictive Power
The amount of variance explained in the remote work self-efficacy construct was approximately 36 percent in the initial analysis and 31 percent in the refined model. These levels of explained variance appear reasonable when compared with results from previous studies. For example, Silver et al. (1995) reported that previous experience explained roughly 30 to 35 percent of the variance in self-efficacy in two studies they conducted.
In contrast, the amount of variance explained in the IT self-efficacy construct in the initial model was considerably higher (approximately 82%). However, this latter finding is likely the result of the poor discriminant validity between the IT constructs, as already described. When the paths from IT experience and training to IT self-efficacy and from connectivity to IT self-efficacy were dropped from the initial model to create a more restricted version, the resulting IT self-efficacy R2 value was dramatically lower (0.16), while the rest of the paths in this restricted model remained largely unchanged. This provides further empirical evidence of discriminant validity problems among the IT-related constructs, but also suggests that the problem does not significantly affect the rest of the model. Results from the refined model analysis also support this conclusion, since most of the relationships changed little when the IT-related constructs were merged into a single, over-arching IT capabilities construct.
Antecedents of Self-Efficacy
The first six hypotheses addressed factors that could potentially influence an individual's judgments of self-efficacy. Hypotheses 1 to 4 were based directly on self-efficacy theory. The hypothesis dealing with the influence of previous experience and training on self-efficacy (H1) was supported in both the initial and the refined model analyses. This implies that employees with more experience and training at working remotely will have higher levels of remote work self-efficacy. This, in turn, will be positively related to performance (H8), job attitudes (H9 and H11) and behaviors (H10 and H12).
In both analyses, support was found for the influence of modeling on self-efficacy, as suggested by self-efficacy theory. The paths from the modeling by manager construct to the remote work self-efficacy construct and the IT self-efficacy construct (in the initial analysis) were significant (hypothesis 3). These results suggest that modeling activities by managers will be associated with higher levels of remote work job task self-efficacy, which in turn leads to higher levels of performance and more positive job attitudes. Many of the modeling activities were related to the manager being an effective communicator and will be discussed in more detail in the implications for management practice section.
The paths from computer anxiety to remote work self-efficacy and from computer anxiety to IT self-efficacy (hypothesis 4) were also significant, supporting self-efficacy theory. To the extent that computer anxiety provides a reliable indication of general IT related anxiety, the findings here support suggestions in the literature that having an ability to use information technology effectively (i.e., having lower levels of computer anxiety) is important in a remote work setting. Consistent with this, in the initial analysis, the path between IT self-efficacy and remote work self-efficacy was significant (hypothesis 7), and connectivity was found to be positively associated with IT self-efficacy (hypothesis 6). Analysis of the refined model also indicated a significant relationship between IT capabilities and remote work self-efficacy. Therefore, managers in virtual organizations should view IT as a key enabler of remote work management, indicating the importance of fulfilling their employees' IT needs.
Although connectivity was not significantly related to remote work self-efficacy, it would have an indirect effect on remote work self-efficacy through its effect on IT self-efficacy. The absence of a significant direct relationship may be the result of the way in which connectivity was defined here. Fulk, Flanagin, Kalman, Monge and Ryan (1996) suggest that there are two different types of connectivity: physical and social. The definition (and set of measures) of connectivity used in the current study dealt solely with the level of physical connectivity available to the respondents. However, individuals must also be willing and able to use such connectivity. While the IT self-efficacy construct assessed individuals' abilities to communicate, their willingness to do so was not measured here.
The expected influence of physical conditions, an environmental variable in the initial model, on self-efficacy (hypothesis 5) was not supported. Although non-significant, there was a positive relationship between physical conditions and remote work self-efficacy as hypothesized. Somewhat surprisingly, a negative (albeit non-significant) relationship was also noted between physical conditions and IT self-efficacy. This would be logical if IT was the source of distractions (e.g., arrival of faxes, e-mail, etc.). If the individual was very competent at using IT, and had correspondingly high IT self-efficacy, the level of use of IT would be high which could lead to more distractions.
Outcomes of Self-Efficacy
For all five of the outcome hypotheses, the paths leading from remote work self-efficacy to the outcome constructs in the initial model analysis were significant, substantive, and in the hypothesized direction. The expected relationships between remote work self-efficacy and the three behaviors examined in this study (i.e., performance, ability to cope, and stress) were all supported in the initial analysis (hypotheses 8, 10, and 12). The significant links between remote work self-efficacy and the two perceived performance constructs, overall productivity and remote work effectiveness, also demonstrate the critical importance of understanding and managing the self-efficacy construct. These findings are consistent with previous self-efficacy research which has demonstrated strong links between self-efficacy beliefs and performance (e.g., Gist & Mitchell, 1992). The positive relationship that was found between self-efficacy and both ability to cope and perceived stress is also consistent with previous self-efficacy studies.
The refined model analysis found similar support for the relationships between remote work self-efficacy and the two performance constructs and between remote work self-efficacy and ability to cope. However, the path between remote work self-efficacy and job stress was not significant in the refined model. Given the results of previous studies demonstrating a significant relationship between self-efficacy and stress and the results from the initial model analysis in the current study, we suggest that this non-significant finding may be specific to the data subset used to estimate the refined model.
Self-efficacy theory does not specifically address the impact of self-efficacy on attitudes. Two attitudes -- job satisfaction and organizational commitment -- were included in this study as a possible extension to self-efficacy theory. Self-efficacy was found to have a significant and positive impact on job satisfaction in both sets of analyses (hypothesis 9). However, the path from remote work self-efficacy to organizational commitment was only significant in the initial analysis (hypothesis 11). Overall, these results suggest that self-efficacy can be used to predict some attitudes as well as behaviors, at least in a remote work context.
We believe that the simpler, refined model (summarized in Figure 4) captures the key relationships found in the initial analysis, and maintains a similar level of predictive power while introducing greater parsimony. However, we also suggest that future researchers at least provisionally retain the organizational commitment and job stress constructs as remote work self-efficacy outcomes, since the evidence presented here is mixed. It seems premature to conclude definitively that self-efficacy has no impact on either of these outcomes.
Limitations and Implications for Future Research
The study described here is the first of its kind to develop and test a comprehensive model of remote management, using respondents who are employed in remote work across a wide range of occupations and geographical settings. Keen (1980) has suggested that it is better to borrow from a reference discipline rather than to invent an entirely new theory. Consequently, in developing our initial research model, we drew heavily on self-efficacy theory.
Our results clearly demonstrate that self-efficacy theory does indeed have substantial explanatory power in a remote management setting. Of course, such findings ultimately need to be replicated across other settings and over time before they can be fully accepted. For example, the current study has a cross-industry, cross-occupational perspective. Because the study does not control for the effects of specific tasks and industries, it cannot investigate potential remote work differences within a specific occupation or industry. Consequently, it must be left to future research work to determine, for example, whether substantial differences in remote work self-efficacy exist between high technology and non-technology workers, or whether remote work in the public and private sectors is fundamentally different.
The current study relies heavily on individual employees' perceptions collected through a large scale mail survey. One consequence of using self-report data, particularly in the case of the outcome constructs, is that such an approach likely introduces a common response bias across constructs. This may partially explain the significant relationships observed between self-efficacy and the various outcomes studied. Although our empirical results suggest that these constructs can be discriminated from one another empirically (e.g., see Table 2), we encourage the use of alternative methods of data collection in future studies. For example, assessments of performance and/or productivity could be obtained from respondents' managers and/or co-workers, or from more objective sources.
The current study adds to the external validity of self-efficacy theory by showing its applicability in a new research domain. With the exception of social persuasion, we were able to show that the sources of self-efficacy information (see Figure 1) did impact self-efficacy judgments, which in turn impacted perceived behavior. The model of remote work self-efficacy described here also represents an extension of existing self-efficacy theory in that it includes attitudinal as well as behavioral constructs.
While self-efficacy theory offers significant promise for remote management research, future researchers should consider competing theories that may also be relevant in a virtual work context (e.g., agency theory). We chose self-efficacy theory at this early exploratory stage of remote management research since we believed that a theory addressing important remote management factors, as identified by the literature, was desirable. However, future research might compare the two theories in areas where they lead to different predictions.
The cross-sectional nature of our survey design limits our ability to draw causal inferences. Although such a design is useful for identifying what set of relationships exist, it does not address why they exist. Future research efforts (e.g., in-depth case studies) will be needed to expand our current understanding of virtual work and its management along these lines.
In this study, employees were defined to be remotely managed if they worked in a different building than their manager, regardless of the distance between the buildings. Clearly, other definitions of remoteness could be used. For example, an alternative definition could be based on the frequency of face-to-face contact that remote employees have with their managers. Comparing the results obtained using these different definitions of remoteness would enhance our current limited understanding of remote work effectiveness.
Construct validity problems prevented us from including social persuasion activities in either of our research models. We attempted to measure social persuasion by the frequency of communication between the manager and the respondent. This was clearly too limiting a definition. Specific questions about social persuasion activities that are linked more directly to the specific tasks examined in the remote work self-efficacy construct could potentially improve measurement of this construct.
In a similar vein, we were not able to reliably measure modeling activities by people other than the respondent's manager, minimizing the influences of co-workers and team members in our models. The validity of this construct could be improved by developing a more comprehensive list (perhaps via focus groups or interviews) of modeling activities undertaken by co-workers and team members that are potentially important to remote workers.
Implications for Management Practice
The results of this study suggest that for remote workers to be effective, they need managers who are good communicators. The remote managers must have good listening skills, and need to be able to manage meetings and their employees’ time effectively. Being able to use information technology effectively to aid communication is also important, as is being available when the employees need coaching or other forms of help. Effective remote managers also support their employees' needs for information technology and support team building and social activities.
All of the practices mentioned above can be learned. Therefore, virtual organizations need to develop training courses and training materials that help their remote managers both learn about and implement effective remote management practices. The results of the current study suggest that these investments in training can result in higher levels of employee performance, job satisfaction, organizational commitment, and ability to cope, as well as lower levels of job stress.
The results of our study also indicate that remote work self-efficacy is positively associated with both higher perceived levels of performance and more positive work attitudes. Thus, training and supporting employees so that they are better able to carry out those tasks that were used to operationalize remote work self-efficacy will benefit virtual organizations. These tasks generally dealt with the employee's ability to set objectives, use time effectively, obtain access to managers and co-workers, use information technology, manage an office, and access information effectively. Actions by virtual organizations that help develop these abilities in their remote employees, as well as the employees' beliefs in their ability to carry out these tasks, are likely to be beneficial in terms of increased performance and positive work attitudes.
Work experience and training were also found to be positively related to self-efficacy. This implies that employees with more experience at being remotely managed will have higher levels of remote work self-efficacy, which in turn will be positively related to performance and job attitudes. Thus, attempts to increase the experience levels of remotely-managed employees should also benefit virtual organizations. For example, training employees in how to work effectively remotely would reduce the amount of time required to reach a certain experience level. Having experienced employees mentor inexperienced people would be another way of achieving this same goal.
Information technology appears to be a key enabler of remote management and remote work. Therefore, virtual organizations need to carefully consider and respond to their employees' IT needs, as well as the needs of their remote managers. Furthermore, appropriate computer training and on-going access to IT support/help staff should be provided to reduce computer anxiety levels. Reducing computer anxiety levels can be beneficial by leading to increased levels of remote work self-efficacy, which in turn increases job attitudes and behaviors.
Finally, the results of the current study can be used to assist managers in identifying workers who are suited to working in a virtual environment. The skills mentioned above could be incorporated into a diagnostic assessment tool that could then be used to identify individuals with characteristics and skills better suited to effective remote work. By selecting and developing employees for virtual work whom score highly on this diagnostic tool, the organization could improve its longer-term probability of success.
Conclusion
Virtual organizations are becoming an increasingly common organizational design. The employees in a virtual organization often work in locations remote from their manager. This remoteness creates many management and communication challenges. In order for organizations to adapt effectively to this new way of conducting work, our understanding of the relevant issues and key drivers must be increased. The current study helps to do this by using self-efficacy theory to predict relationships between the antecedents to remote work self-efficacy and the consequences of self-efficacy. These relationships were tested by surveying a diverse set of employees who worked remotely from their managers in a variety of virtual organizations.
Overall, the results obtained here indicate that remote employees' self-efficacy assessments play a critical role in influencing their remote work performance as well as their attitudes towards both remote work and their own organizations. Furthermore, strong relationships were observed between employees' remote work self-efficacy judgments and many of the antecedents included in the study. Because many of these antecedents can be controlled to some extent by managers, these findings suggest that it may be possible to enhance employees' work performance through management efforts to improve employees’ remote work self-efficacy. The current study demonstrates the validity of self-efficacy theory in a virtual work environment and also provides a basis for future research in the virtual work area through its development and testing of a remote management framework.
Acknowledgments
The authors gratefully acknowledge the insightful advice and suggestions from the reviewers, editors and Peter Seddon of The University of Melbourne, Australia. The second author would like to thank the Barford family for its financial support and encouragement of this work.
Footnotes
1 Attempts were made to measure and include the social persuasion construct in our research model. However, these attempts were not successful, and the social persuasion construct is consequently not discussed further in this paper.
2 We originally attempted to include a construct dealing with co-workers' modeling activities. However, this attempt was not successful, perhaps because, as one reviewer suggested, the geographically dispersed working relationships found in virtual organizations limit employees' opportunities to observe such activities. Thus, the modeling by others construct is not discussed further in this paper.
3 The questionnaire yeilded data that allowed us to identify remotely-managed respondents. In this way, participating organizations did not have to identify remotely-managed respondents specifically and we were also able to gather data from locally-managed employees for purposes not reported here.
References
Armstrong, J. S., & Overton, T. S. (1977). Estimating non-response bias in mail surveys. Journal of Marketing Research, 14 (3), 396-402.
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2), 191-215.
Bandura, A. (1978). Reflections on self-efficacy. Advances in Behavioral Research and Therapy, 1(4), 237-269.
Bandura, A. (1982). Self-efficacy mechanism in human agency. American Psychologist, 37(2), 122-147.
Bandura, A. (1986). Social foundations of thought and action. Englewood Cliffs, New Jersey: Prentice-Hall.
Bandura, A. (1988). Reflection on nonability determinants of competence. In R. J. Sternberg and J. Kolligian, Jr. (Eds.), Competence considered: Perceptions of competence and incompetence across the lifespan (pp. 315-362). Dordrecht, Netherlands: Kluwer Academic Publishers.
Bandura, A., & Cervone, D. (1986). Differential engagement of self-reactive mechanisms governing the motivational effects of goal systems. Organizational Behavior and Human Decision Processes, 38 (1), 92-113.
Bandura, A., & Schunk, D. H. (1981). Cultivating competence, self-efficacy, and intrinsic interest through proximal self-motivation. Journal of Personality and Social Psychology, 41 (3), 586-598.
Barclay, D., Higgins, C., & Thompson, R. (1995). The partial least squares (PLS) approach to causal modeling: Personal computer adoption and use as an illustration. Technology Studies, 2 (2), 285-309.
Barling, J., & Beattie, R. (1983). Self-efficacy beliefs and sales performance. Journal of Organizational Behavior and Human Performance, 5 (1), 41-51.
Birkinshaw, J., Morrison, A., & Hulland, J. (1995). Structural and competitive determinants of a global integration strategy. Strategic Management Journal, 16 (8), 637-655.
British Telecom (1984). Survey item bank. Bradford, West Yorkshire: England: MCB University Press.
Brown Jr., I., & Inouyne, D. K. (1978). Learned helplessness through modeling: The role of perceived similarity in competence. Journal of Personality and Social Psychology, 36 (8), 900-908.
Compeau, D. R. (1992). Individual reactions to computing technology: A social cognitive theory perspective. Unpublished doctoral thesis, The University of Western Ontario.
Davies, T. F. (1994). Predicting application software usage: A longitudinal study. Unpublished doctoral thesis, The University of Western Ontario.
DeSanctis, G. (1984). Attitudes toward telecommuting: Implications for work-at-home programs. Information & Management, 7 (3), 133-139.
Dillman, D. A. (1978). Mail and telephone surveys: The total design method. New York: Wiley.
Donaldson, B. (1990). Sales management: Theory and practice. London: MacMillan Education Ltd.
Duxbury, L., & Haines Jr., G. (1991). Predicting alternative work arrangements from salient attitudes: A study of decision makers in the public sector. Journal of Business Research, 23 (1), 83-97.
Duxbury, L. E., Higgins, C. A., & Irving, R. H. (1987). Attitudes of managers and employees to telecommuting. Infor, 25 (3), 273-285.
Fornell, C., & Barclay, D. W. (1983). Jackknifing: A Supplement to Lohmoller’s LVPLS Program. Ann Arbor, Michigan: Graduate School of Business Administration, The University of Michigan.
Fornell, C., & Larcker, D. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18, 39-50.
Frayne, C .A., & Latham, G. P. (1987). Application of social learning theory to employee self-management of attendance. Journal of Applied Psychology, 72 (3), 387-392.
Freedman, D. H. (1993). Quick change artists. CIO, 6 (18), 32-28.
Fulk, J., Flanagin, A. J., Kalman, M. E., Monge, P., & Ryan, T. (1996). Connective and communal public goods in interactive communication systems. Communication Theory, 6 (1), 60-87.
Gerber, B. (1995). Virtual teams. Training, 32 (4), 36-40.
Gist, M. E. (1989). The influence of training method on self-efficacy and idea generation among managers. Personnel Psychology, 42 (4), 787-805.
Gist, M. E., & Mitchell, T. R. (1992). Self-efficacy: A theoretical analysis of its determinants and malleability. Academy of Management Review, 17 (2), 183-211.
Gist, M. E., Schwoerer, C., & Rosen, B (1989). Effects of alternative training methods on self-efficacy and performance in computer software training. Journal of Applied Psychology, 74 (6), 884-891.
Green, D. H., Barclay, D. W. and Ryans, A. B. (1995). Entry strategy and long-Term performance: Conceptualization and empirical examination. Journal of Marketing, 59 (4), 1-16.
Greengard, S. (1994). Making the virtual office a reality. Personnel Journal, September, 66-79.
Handy, C. (1995). Trust and the virtual organization. Harvard Business Review (May-June), 40-50.
Hartman, F., & Guss, C. (1996). Virtual teams - constrained by technology or culture. Proceedings of IEMC 96. Managing virtual enterprises: A convergence of communications, computing and energy technologies (pp. 645-650). Vancouver, BC, Canada: IEEE.
Heinssen Jr., R. K., Glass, C. R., & Knight, L. A. (1987). Assessing computer anxiety: Development and validation of the computer anxiety rating scale. Computers in Human Behaviour, 3 (1), 49-59.
Higgins C., Duxbury, L., & Irving, R. (1992). Work-family conflict in the dual-career family. Organizational Behavior and Human Decision Processes, 51(1), 51-75.
Hill, T., Smith, N. D., & Mann, M. F. (1987). Role of efficacy expectations in predicting the decision to use advanced technologies. Journal of Applied Psychology, 72 (2), 307-314.
House, R. J., Schuler, R. S., & Levanoni, E. (1983). Role conflict and ambiguity scales: Reality or artifacts? Journal of Applied Psychology, 68 (2), 334-337.
Hulland, J. S. (1998). Use of partial least squares (PLS) in strategic management research: A review of four recent studies. Strategic Management Journal, in press.
Huws, U. (1990). New technology homeworkers. Employment Gazette (UK).
Illingworth, M. M. (1994). Virtual managers. InformationWeek (June 13), 42-58.
Keen, P. G. W. (1980). MIS research: Reference discipline and a cumulative tradition. Proceedings of the First International Conference on Information Systems (December), 9-18.
Lee, C., & Bobko, P. (1994). Self-efficacy beliefs: Comparison of five measures. Journal of Applied Psychology, 79 (3), 364-369.
Lent, R. W., Brown, S. D., & Larkin, K. C. (1987). Comparison of three theoretically derived variables in predicting career and academic behavior: Self-efficacy, interest congruence, and consequence thinking. Journal of Counseling Psychology, 34 (3), 293-298.
Locke, E. A. (1976). The nature and causes of job satisfaction. In M. D. Dunnette (Ed.), Handbook of industrial and organizational psychology (pp.1297-1349). New York: John Wiley & Sons.
Locke, E. A. (1991). The motivation sequence, the motivation hub, and the motivation core. Organizational Behavior and Human Decision Processes, 50 (2), 288-299.
Lucas Jr., H. C. (1996, December). The IT-based virtual organization. Panel presented at the International Conference on Information Systems, Cleveland, Ohio.
Lucas Jr., H. C., & Baroudi, J. (1994). The role of information technology in organization design. Journal of Management Information Systems, 10 (4), 9-23.
Mowday, R. T., Steers, R. M., & Porter L. W. (1979). The measurement of organizational commitment. Journal of Vocational Behavior, 14, 224-247.
Mowshowitz, A. (1994). Virtual organizations: A vision of management in the information age. The Information Society, 10 (4), 267-288.
Pedhazur, E. J. (1982). Multiple regression in behavioral research. New York: Holt, Rinehart and Winston.
Phelps, N. (1985). Mountain Bell: Program for managers. In Office Workstations in the Home. Washington, DC: National Research Council, National Academy Press.
Risman, B. J., & Tomaskovic-Devey., D. (1989). The social construction of technology: Microcomputers and the organization of work. Business Horizons, 32 (3), 71-75.
Rizzo, J., House, R., & Lirtzman, S. (1970). Role conflict and ambiguity in complex organizations. Administrative Sciences Quarterly, 15 (2), 150-163.
Roderick, J. C., & Jelley, H. M. (1991). Managerial perceptions of telecommuting in two large metropolitan cities. Southwest Journal of Business & Economics, 8 (1), 35-41.
Saks, A. M. (1995). Longitudinal field investigation of the moderating and mediating effects of self-efficacy on the relationship between training and newcomer adjustment. Journal of Applied Psychology, 80 (2), 211-225.
Schunk, D. H. (1981). Modeling and attributional effects on children’s achievement: A self-efficacy analysis. Journal of Educational Psychology, 73 (1), 93-105.
Silver, W. S., Mitchell, T. R., & Gist, M. E. (1995). Responses to successful and unsuccessful performance: The moderating effect of self-efficacy on the relationship between performance and attributions. Organizational Behavior and Human Decision Processes, 62 (3), 286-299.
Stanton, W. J., & Buskirk, R. H. (1987). Management of the sales force. Homewood, Illinois: Irwin
Staples, D. S. (1996). An investigation of some key information technology-enabled remote management and remote work issues. Conference Proceedings of the Australasian Conference on Information Systems. University of Tasmania, Hobart.
Staples, D. S. (1997). The management of remote workers: An information technology perspective. Unpublished doctoral thesis, University of Western Ontario, London, Canada.
Stumpf, S. A , Brief, A. P., & Hartman, K. (1987). Self-efficacy expectations and coping with career-related events. Journal of Vocational Behavior, 31 (2), 91-108.
Taylor, M. S., Locke, E. A., Lee, C., & Gist, M. E. (1984). Type A behavior and faculty research productivity: What are the mechanisms? Organizational Behavior and Human Decision Processes, 34 (3), 402-418.
Warr, P., Cook, J., & Wall, T. (1979). Scales for the measurement of some work attitudes and aspects of psychological well-being. Journal of Occupational Behavior, 52 (2), 129-148.
Weinberg, R. S., Gould, D., & Jackson, A. (1979). Expectations and performance: An empirical test of Bandura’s self-efficacy theory. Journal of Sport Psychology, 1 (4), 320-331.
Wold, H (1985). Systems analysis by partial least squares. In P. Nijkamp, L. Leitner, and N. Wrigley (Eds.), Measuring the unmeasurable (pp.221-251). Dordrecht: Marinus Nijhoff.
Wood, R., & Bandura, A. (1989). Social cognitive theory of organizational management. Academy of Management Review, 14 (3), 361-384.
Appendix A: Questionnaire Items
* after the item label designates reverse coding.
