The term “telework” was introduced in 1976 (Nilles et al., 1976). It was widely spread primarily at the beginning of the 21st century as a new form of labor organization that could provide solutions to many individual, social and organizational problems (Bajzikova et al., 2016; Gálvez et al., 2020). teleworking is seen as a form of flexible work arrangement that entails working remotely from an employee for a large proportion of the employer’s time (Thompson and Vivien, 1998). Other researchers define the concept as “working from home by deploying information and communication technologies to keep in touch with colleagues and deal with allocated working tasks” (Nguyen and Armoogum, 2021, p. 4). Employees are free to work outside the office, usually facilitated by virtual communication tools such as teleconferences, videoconferences and intranets with remote log-in. Moreover, they can also decide when they work, defining their work schedules with flexible start and end times (Coenen and Wok, 2014). In essence, the focused literature analyzed teleworking from a wide variety of perspectives as researchers have provided many definitions of the concept. These are based on several major themes, such as organization, location and technology (Martino and Wirth, 1990).

Among the major transformations that have occurred over the recent decades in the world of work, the rapid expansion of flexible working practices has proved to be a valuable solution for many companies and institutions (OECD, 1995, 2016; Matli, 2020). Flexible working has become a concept that captures a plethora of working arrangements and a mantra for promoting the idea of working everywhere and anytime. In other words, it is designed to address both employer and employee needs in a mutually advantageous manner. There are different modalities of flexible working practices, such as flexible locations, flexible time, and flexible contracts, or a combination among them. Consequently, the following five main types of teleworking are encountered in the world of work (Morgan, 2004; Hislop and Axtell, 2009):

Telework intensity differs according to the amount of telework time that ranges from part-time to full-time telework (Perez Perez et al., 2003; Gajendran and Harrison, 2007). Part-time telework happens when a teleworker works partly from home, partly from the office or from a client site, while full-time telework occurs when he/she works from home or a place other than an office using telecommunication technologies all the time (Nakrošiene et al., 2019).

Telework use is moderate in teams where members worked outside the office once or twice per week or during up to eight working hours—the approximate equivalent of one working day—and extensive when team members worked outside the office more than twice per week or more than 8 h per week (Coenen and Wok, 2014). Some studies found that computer-mediated communication, especially email, could effectively support knowledge sharing in teleworking teams that lack the ability to engage in face-to-face communication (Lee et al., 2007).

There is a plethora of factors that influence teleworking, such as: individual factors, job and organizational factors, family/home factors (Baruch and Nicholson, 1997), and socio-demographic characteristics (e.g., gender, age, number of children, and marital status; Nakrošiene et al., 2019). On the other hand, the job demands-resources theory considers that working conditions can be divided into job demands (e.g., physical workload, time pressure, recipient contract, physical environment, and shift work) and job resources (e.g., rewards, feedback, participation, job control, job security, and supervisor’s support), and evaluates the effects of different teleworking impact factors on work outcomes (Demerouti et al., 2001). Consequently, higher job demands lead to strain and health impairment, and higher resources generate better performances (Parker et al., 2017).

According to Nilles (1991), technological factors increase organizational effectiveness and productivity, maximize the utilization of resources, improve wellbeing and job satisfaction within teleworking, increase equipment costs and raise technical issues (Pérez et al., 2007). Moreover, Haddon and Brynin (2005) consider that specific technologies may define different types of teleworking, and other authors determined the relevance of home office or satellite office location to workers’ measurable productivity levels and efficiencies (Garret and Danziger, 2007).

Teleworking has received substantial attention in the scientific community, with regard to the impact of its arrangements on individual teleworkers, including their social relationships, work–family conflict, job satisfaction, organizational commitment, and job performance (Kossek et al., 2006; Gajendran and Harrison, 2007; Martin and MacDonnell, 2012; Richardson and McKenna, 2014; Biron and van Veldhoven, 2016; Stripe and Zarraga-Oberty, 2017; Groen et al., 2018). As a consequence, teleworking is a relatively controversial concept in the literature, due to both its strengths and weaknesses (Gálvez et al., 2020; Nguyen and Armoogum, 2021). There are three viewpoints from which the advantages and disadvantages are presented: individual, organizational, and societal (Harpaz, 2002). From the individual point of view, the main benefits of teleworking are the potential to blend various aspects of people’s professional and personal lives (Dima et al., 2019), a better work-life balance (Baruch, 2000; Zhang et al., 2020). However, other authors reveal the pernicious impact of TW, portraying teleworkers as being overworked (Vega et al., 2015; Sorensen, 2017). From the organizational perspective, the main advantages of teleworking are the expansion of productivity, the increasing presence of the employees at work, and the establishment of a positive organizational image (Ward and Shabha, 2001). Nevertheless, the existing researches showed that the more teleworkers work from home, the less possibility they have of gaining work support from their supervisors (Lapierre et al., 2015) and lower visibility in the company (Cooper and Kurkland, 2002). Moreover, other disadvantages of teleworking are the decreasing time for conversations with colleagues (Wilson and Greenhill, 2004), the transformation of work processes, and the legal issues (Harpaz, 2002). Teleworkers may feel a state of loneliness (Bailey and Kurkland, 2002), a lack of cooperation and communication with colleagues, and insufficient social interaction. That leads to the decrease of their organizational identification (Ammons and Markham, 2004). Thus, trust is a necessary condition for interpersonal cooperation, and it can be diminished when employees interact less frequently (McAllister, 1995). From the societal viewpoint, the main advantages of teleworking are the reduction of environmental damage, the decrease of traffic congestion, the provision of solutions for special-needs populations and the decrease of energy consumption, while the main disadvantages are the unclear legal issues and the creation of a detached society (Harpaz, 2002).

In the literature, there are several studies that present and analyze the relationships between teleworking and the wellbeing (Azarbouyeh and Naini, 2014; Miron et al., 2021) and productivity (Ward and Shabha, 2001) of employees. In the last decades, more and more employers have focused their attention on the wellbeing of their employees by providing a healthy, pleasant, and supportive workplace (Weinberg and Doyle, 2017), and various stimulative benefits (Harter et al., 2003). The increase of the employees’ wellbeing makes them more engaged and, therefore, leads them to obtain a higher productivity (Beheshti, 2019). As the employees gain higher outputs, they feel a “heightened sense of wellbeing” (Bosua et al., 2013, p. 11.9).

During the time, various points of view from several domains (e.g., psychology, medicine, and sociology) have been expressed related to the concept of wellbeing. Thus, wellbeing constitutes not only a broad, complex, multi-dimensional, and multifaceted concept (Catană et al., 2021) but also a state (Dodge et al., 2012). However, two different perspectives seem to be predominant as they operationalize wellbeing as follows:

As a key human need, wellbeing is often associated and even considered synonym with concepts such as welfare, life satisfaction or quality of life, good health, autonomy, happiness, purpose in life, self-acceptance, comfort, prosperity, security, positive relationships with others or making contributions to the community (Ryff, 1989; Lindberg, 2002; Shah and Marks, 2004) and is deeply linked with a “good or satisfactory condition of existence” (Webster’s Encyclopedic Unabridged Dictionary of the English Language, 1996, p. 1620). It represents a “global assessment of a person’s quality of life according to his own chosen criteria” (Shin and Johnson, 1978, p. 478), a “balance point between an individual’s resource pool and the challenges faced” (Dodge et al., 2012, p. 230), and reflects “feelings about oneself in relation to the world” (Clements-Croome, 2005, p. 27). Several types of wellbeing have been found in the literature as follows:

At the organizational level, employee wellbeing refers to his/her psychological and physical health (Edwards, 1992). Thus, it integrates both his/her psychological wellbeing (e.g., depression, anxiety) and physical wellbeing (e.g., heart rate, blood pressure). Researches emphasize that a higher employee wellbeing is considerably associated with “better job performance, lower absenteeism, reduced probability of leaving an employer, and the occurrence of more discretionary work behaviors” (Warr, 1999, p. 392).

On its turn, productivity is an economic indicator, that measures “the efficiency of production, taking the form of a ratio of the output of goods and services to the input of factors of production” (The Chartered Management Institute, 2004, p. 342). Consequently, employee productivity expresses the rate of output per employee during his/her working time. In other words, it measures the individual employee’s output in a given amount of time. However, the employees’ perceptions of their productivity level may differ in comparison to their real value of productivity. In this respect, researchers assert that an accurate measurement of the employee productivity during teleworking is rather difficult to achieve without a careful review of how managers and employees perceive it (Bosua et al., 2017). Employee productivity is positively influenced by technological factors (e.g., modern equipment, information communication technologies), psychological factors (e.g., encourage and praise), social factors (e.g., flexible work schedule, workplace conditions), individual factors (e.g., knowledge, abilities) and/or managerial factors (e.g., participative management, quality circles). Based on all these findings, the authors summarize some of the main effects of the teleworking impact on wellbeing and productivity (Table 1).

Starting from the above-mentioned considerations, two research objectives were established as follows:

Objective 1 (O1): To identify some of the main teleworking impact factors on wellbeing and productivity and to present their items.

Objective 2 (O2): To benchmark the effects of teleworking impact factors on wellbeing and productivity of Romanian employees grouped in three clusters.

The authors have designed and empirically tested the theoretical model to show the impact of five TW factors on the wellbeing and productivity of graduate employees (Figure 1): individual and societal factors, organizational and work-related factors, technological factors, social factors at home, and social factors at work. Each factor is defined through a different number of items. The dependent variables are the wellbeing and productivity of the employees during teleworking and the independent variables are the five teleworking factors.

Starting from the above objectives, the following five research hypotheses were set up:

Against this background, this study aims to identify and analyze the above five teleworking impact factors that affect wellbeing and productivity of employees. To reach these objectives, the authors used a quantitative research method through a questionnaire applied to the Romanian employees who hold a Bachelor or Master degree.

The paper is structured as follows. Section 2 illustrates the materials and methods. Results and discussion are revealed in Sections 3, and 4, respectively. Section 5 displays the conclusions, along with their limitations and research perspectives.

The authors designed and elaborated the questionnaire, after reviewing the literature. It contains 50 items, out of which seven refer to socio-demographic information of the employees (gender, age, education level, income, location, marital, and parental status), nine questions regarding the advantages of teleworking, 13 questions regarding teleworking impact factors (Table 2), and nine questions related to their employment situation (Table 3). The others were not used as they did not show significant results. The respondents’ perceptions were measured by using the Likert scale from 1 (minimum) to 5 (maximum).

The authors used a quantitative research method, based on a survey. The surveying technique was carried on through the Computer-assisted web interviewing (CAWI). Some authors consider CAWI to be synonymous with conducting a web survey (Biffignandi and Bethlehem, 2021, p. 68) whereas others consider that the notion is not correct because “interviewing” implies the presence or involvement of an interviewer, which is not the case for web surveys (Callegaro et al., 2015, p. 51). The stratified sampling was used as a sampling method. It represents a data-gathering method in which participants are chosen based on predetermined criteria (employees with bachelor/master degree) so that the final sample has the same characteristics as the studied population (Taherdoost, 2016). The study was focused on Romanian people with at least graduade level of education, with a focus on the latter, who are employed in the areas of activity mentioned in Table 4. The information that are available from the Romanian National Institute of Statistics (2021a) refer to the types of occupation (Romanian National Institute of Statistics, 2021b; Romanian National Institute of Statistics, 2021c) specific to various fields of activity (CAEN, 2022).

The questionnaires were distributed by using the Google Forms platform, preserving the anonymity of the respondents. The data were collected between the 18th of January 2021 and the 17th of February 2021, through CAWI. After receiving, centralizing, and systematizing the answers gathered online, 327 questionnaires were validated from 373 responses (87.6%). The average completion time was 8 min. The questionnaire was applied to employees working in areas that are suitable for teleworking (e.g., education, public administration) especially due to the COVID-19 pandemic. Authors applied inclusion/exclusion criteria. The questionnaire was addressed only to teleworking graduate employees (Table 4).

The collected data were analyzed through the factor analysis that puts together “common variables into descriptive categories” (Yong and Pearce, 2013, p. 80). After identifying the effects of the five teleworking impact factors, the authors carried on cluster analysis in order to see how the respondents can be segmented and what are their defining characteristics. The cluster analysis aims to create homogeneous groups by assigning observations to bunches of people, so that these being similar, with respect to specific attributes of interest (Tryfos, 2005). Moreover, the authors handled a multiple linear regression in order to determine the way these factors influence the wellbeing and productivity of the employees. The analysis was conducted in R (version 4.0.5) using the RStudio (version 1.4.1106) user interface. The packages used in this analysis were tidyverse (1.3.0), psych (2.0.9), nFactors (2.4.1), psycho (0.5.0), psy (1.1), clustertend (1.5), factoextra (1.0.7), NbClust (3.0), and dendextend (1.14.0). A screeplot with the Eigen values and parallel analysis were performed. A factorial analysis was done in order to identify the factor loadings of the variables from the questionnaire. Promax rotation was used in order to maximize variance and to minimize items loading on all five factors (Table 2).

Cronbach’s ɑ was used to measure internal consistency of the questionnaire and the values were 0.88 for individual and societal factors, 0.74 for organizational and work-related factors, 0.85 for technological factors, 0.79 for social factors at home, and 0.83 for social factors at work. All the values are above 0.7 and dropping any of the items would lead to a lower value (Drost, 2011). The maximum likelihood method was used to extract the factors (Heeler et al., 1977).

The authors chose the cluster analysis in order to better analyze the data. Consequently, K-means clustering was preferred as it is one of the most popular clustering methods used (Towards data science, 2018). The authors computed Hopkins’ statistic, using the R package factoextra (The Comprehensive R Archive Network, 2021), which was determined to be 0.68, greater than 0.5 which showed there existed a clustering tendency among the 5 identified factors (Lawson and Jurs, 1990).

The authors used the package NbClust (Charrad et al., 2014) in order to identify the optimal number of clusters. By using various indexes and methods, including the popular average silhouette method, the elbow method, and the Euclidean distance method, the proper conclusion was the use of 3 clusters. The clusters descriptions are presented in the following section of the paper. Statistically significant differences are identified using Kruskal–Wallis analysis as the results do not respect the homogeneity or normality requirements of ANOVA test. The three clusters were benchmarked by taking into account the five teleworking impact factors, the socio-demographic and economic descriptors, and the employment teleworking conditions.

The authors identified three clusters of employees, by taking into account their specific homogeneity. In this respect, they focused on the values of the variables used in the cluster identification process (Table 5). The values are shown for the entire dataset as well as per cluster. Furthermore, the authors tested for statistically significant differences between clusters using Kruskal–Wallis one-way analysis of variance for variables that had ordered values and the Chi-squared test for nominal variables. The analysis of the results led to the identification of three clusters (Table 5).

The cluster differences were tested using Kruskal–Wallis analysis. All these differences were significant at the 95% level with the exception for clusters 2 and 3 with regard to technological factors.

Based on the above data, the results of our analysis show that clusters 2 and 3 are fairly similar with regard to the last 3 teleworking impact factors identified in the previous analysis whereas cluster 1 stands out quite a bite.

Technological factors, social factors at home and at work are of less concern for both clusters 2 and 3. The employees from cluster 2 stand out due to the fact that they consider that work from home is beneficial as it allows increased work productivity, more flexible schedule but also certain individual and societal gains like reduced pollution, and inclusiveness of people with disabilities. Individual and societal factors, and organizational and work-related factors are of very little importance for the employees of cluster 3.

Cluster 1 is markedly different in terms of the last three teleworking impact factors. The employees experienced significant technological issues, such as: insufficient IT skills, involvement in household activities, social isolation. Both individual and societal factors and organizational and work-related factors are of little importance.

The authors presented the socio-demographic and economic descriptors of the employees from each cluster (Table 6). The average age of all the respondents was 39.7 years old, with the following differences between the clusters: 39.65 for cluster 1, 38.71 for cluster 2, and 42.75 for cluster 3.

The analysis of these outcomes highlights the following:

It is worth noting that the authors used the Kruskal-Wallis analysis to check the significant differences between the various groups and they found almost no differences. Income was the only one where it was found a statistical difference (Chi-square = 14.014, df = 2 and p = 0.0009) and, therefore, the Pairwise Wilcoxon Rank Sum tests was used in order to identify the differences. The only statistically significant difference, at the 95% CI, is between cluster 1 (the one with the lowest incomes), and cluster 2 (the one with the highest incomes). The authors consider that this result is not surprising, because people with different income levels will have different perceptions on wellbeing and productivity during teleworking.

Moreover, the authors take into consideration the employment situation and teleworking conditions of the respondents (Table 3). The authors included a question related to COVID-19 pandemic and to what extent the employees consider that the decision of their employer to facilitate teleworking was influenced by the pandemic. The results are not presented because they are the same across all three clusters: the employees state that it had a significant impact (the average value was 4.48, on a Likert scale from 1—not important to 5—very important).

The results of our analysis demonstrated the following: