Integrating and synthesising technostress research: a meta-analysis on technostress creators, outcomes, and IS usage contexts

ABSTRACT

The expansion of technostress research in the organisational and private IS usage contexts has generated substantial theoretical and empirical insights into the relationship between technostress creators and psychological and behavioural outcomes. However, we observe empirical inconsistencies in terms of effect sizes and conceptual inconsistencies regarding the aggregated and disaggregated treatment of technostress creators. Against this background, we argue that a fine-grained estimation and comparison of effect size strengths of technostress creators on outcomes can provide clarity on these essential matters. Using the Hunter and Schmidt method, we integrated and synthesised empirical data from 102 articles, encompassing 113 independent studies with a total of 49,955 observations. Our analysis offers four important contributions to the technostress literature. First, it confirms that technostress is meaningful in terms of its detrimental impact on both psychological and behavioural outcomes. Second, the results provide accurate effect size estimates for technostress creators on different outcomes in organisational and private usage contexts. Third, the results reveal that psychological outcomes are more immediate than behavioural outcomes. Fourth, the findings suggest that in certain contexts, a disaggregated account of technostress creators can reveal meaningful empirical information.

KEYWORDS:

• Technostress
• technostress creators
• IS usage context
• meta-analysis

1. Introduction

Technostress is often characterised as a key “dark side” phenomenon, emphasising the negative consequences of technology use, although it can also have positive effects (eustress) (e.g., Tarafdar et al., 2019). It is commonly understood as “a modern disease of adaptation caused by inability to cope with new computer technologies in a healthy manner” (Brod, 1984, p.16).

Since the first mention of technostress in practitioner writings in the 1980s and 1990s, a wealth of scholarly research has been conducted to understand its emergence and consequences (Tarafdar et al., 2007; Riedl, 2013; LaTorre et al., 2019; Grummeck-Braamt et al., 2021). In discussing this matter, the literature has largely relied on five conditions that create technostress, referred to as “technostress creators”: techno-complexity, techno-invasion, techno-insecurity, techno-overload, and techno-uncertainty (Ragu-Nathan et al., 2008). These conditions are linked to various psychological and behavioural outcomes, such as satisfaction, exhaustion, commitment, and productivity (Tarafdar et al., 2007, 2011, 2019; Maier et al., 2014; Maier, Laumer, Eckhardt, 2015; Pirkkalainen & Salo, 2016; Califf et al., 2020; Grummeck-Braamt et al., 2021).¹

Technostress creators and outcomes have been empirically investigated in a wide range of work-related settings, such as sales technologies (Tarafdar, Pullins, et al., 2015), telework (Suh & Lee, 2017), enterprise social media (Chen & Wei, 2019), and healthcare (Califf et al., 2020). In addition, a growing body of research has begun to link technostress to private domains, such as social media usage (Maier, Laumer, Weinert, et al., 2015; Tarafdar et al., 2020; Salo et al., 2022), in-vehicle dashboards (Nastjuk & Kolbe, 2015), and smartphones (Lee et al., 2014). Collectively, this substantial body of empirical research provides rich contextual insights that advance our understanding of the antecedents and consequences of technostress.

At the same time, we observe empirical and conceptual inconsistencies in technostress studies, presenting two important research opportunities. First, there are inconsistencies in terms of the reported magnitude of the effects (estimated effect sizes) between technostress creators and their consequences. For example, while some studies have estimated effects as high as −.7 (e.g., Maier, Laumer, Weinert, et al., 2015; Pirkkalainen et al., 2017; Alvarez-Risco et al., 2021), others have found effects smaller than −.1 (Brooks et al., 2017; Chandra et al., 2019) and some have even pointed to a positive effect direction (e.g., Hung et al., 2015; Califf et al., 2020). Although some degree of heterogeneity among studies can be explained by pure chance (i.e., sampling error, Hunter & Schmidt, 2004), the strength of heterogeneity here suggests potential differences in the contextual and outcome-related relevance of technostress creators. Therefore, knowledge about the importance of these technostress creators for different outcomes across and within different usage contexts – specifically, precise point estimators with high statistical power – can guide future studies in research problem formulation.

Second, empirical research has employed two different approaches for studying how technostress creators affect outcomes. In line with the original approach (Ragu-Nathan et al., 2008), most studies have conceptualised technostress creators as a multidimensional, superordinate construct that affects outcomes at an aggregate level (e.g., Tarafdar et al., 2007, 2010, 2011; Ragu-Nathan et al., 2008; Maier, Laumer, Weinert, et al., 2015; Srivastava et al., 2015). Other studies have investigated the impact of technostress creators at the individual (disaggregated) level (e.g., Maier, Laumer, Weinert, et al., 2015; Brooks et al., 2017; Chandra et al., 2019; Califf et al., 2020). Each perspective has its merits, as described in research on psychological stress (Edwards, 2001). On the one hand, in line with the principle of parsimony and generalization, the aggregated formulation of technostress creators reduces both empirical and nomological complexity. On the other hand, the disaggregated approach accounts for the empirical and nomological specificity of the individual technostress creators. For example, Chandra et al. (2019) relied on the disaggregated approach to investigate employee innovation, finding a negative relationship with techno-complexity, no relationship with techno-overload, and a positive relationship with techno-uncertainty. Thus, valid variance in one technostress creator might not be distinctively captured by an aggregate construct and nuances in terms of different technostress-related outcomes might get lost. Hence, empirical evidence identifying when the five key technostress creators should be investigated as an aggregate construct versus treated individually is crucial for guiding future studies. However, such evidence for an aggregated or disaggregated treatment of the technostress construct is scarce and mixed, thus there is a need for rigorous empirical research investigating this question across multiple studies.

Against this background, we believe that the time is right for a meta-analytical investigation that integrates and synthesises the body of empirical findings in the technostress literature and clarifies empirical and conceptual inconsistencies. Therefore, we undertake an analysis summarising the empirical findings of technostress creators at different levels (i.e., technostress as an aggregate vs. disaggregated construct), for different outcomes (psychological vs. behavioural), and in different contexts (private vs. organisational usage). To do so, we collected and analysed the empirical findings of 102 articles, comprising 113 independent studies with a total of 49,955 observations. As the first meta-analytic investigation of the technostress phenomenon,² our work contributes to the technostress literature in several meaningful ways.

First, our meta-analysis provides true effect size estimates and insights about the absolute relevance of technostress creators in the realm of dark-side effects. Unlike with typical significance testing, in which researchers seek empirical evidence for the existence of an effect (i.e., whether the effect is different from zero), estimating effect sizes involves determining the magnitude of an effect. These effect sizes are important for illustrating the effect’s practical relevance. In technostress research, empirical studies have been primarily concerned with uncovering cause-and-effect relationships and have therefore focused their analytical procedures on statistical inferences, for which smaller sample sizes are adequate. Meta-analytical integration into a large data set allows us to uncover precise effect size estimates with high statistical power. Knowledge about the absolute relevance of technostress creators across studies and for specific relationships is key for research problem formulation (Rai, 2017).

Second, our research complements existing technostress review studies that provide insights into the theories, methods, historical roots, and evolution of technostress gained from systematic literature reviews (Riedl, 2013; Fischer & Riedl, 2015, 2017; LaTorre et al., 2019; Tarafdar et al., 2019) and scientometric analyses (Bondanini et al., 2020; Grummeck-Braamt et al., 2021). By summarising existing empirical knowledge, we reveal effect size estimates for different outcomes (psychological and behavioural), distinct contexts (private and organisational), and five technostress creators. This allows the research community to better identify where and how to examine research opportunities in the technostress domain.

Third, our work provides comparative analyses for the effects of technostress creators on psychological and behavioural outcomes and sheds light on the relative importance. The results reveal that psychological outcomes are more strongly linked to technostress creators than behavioural outcomes. This not only emphasises the importance of psychological outcomes as having a potential mediating role on behavioural outcomes but also highlights the need to take a deeper look at technostress as a mental health challenge (Tarafdar, Gupta, et al., 2015) and investigate different and novel psychological outcomes.

Fourth, our study bears implications for a conceptual discussion on when technostress creators should be treated as an aggregate or disaggregated construct. With our meta-analytic review, we find the need to approach this with theoretical and contextual arguments.

2. Background

The description of the technostress phenomenon has a rich history in books (Brod, 1984; Weil & Rosen, 1997). It appeared in top-tier IS journals about a decade and a half ago, e.g., Tarafdar’s et al. (2007) examination of how the phenomenon affects role stress and productivity. While early studies focused on the organisational context (e.g., Tarafdar et al., 2007; Ragu-Nathan et al., 2008), more recent studies have extended technostress research to the private context (Maier, Laumer, Weinert, et al., 2015 Tarafdar et al., 2020; Salo et al., 2022). In the following, we present the main theoretical concepts underlying our meta-analysis, with a focus on the current state of research on technostress creators.

2.1. Theoretical foundations of technostress

The transactional model of stress (TMS) by Lazarus and Folkman (1984) is a widely adopted theoretical framework for studying technostress in the organisational and private usage contexts (e.g., Ragu-Nathan et al., 2008; Maier et al., 2019, 2022; Pirkkalainen et al., 2019; Tarafdar et al., 2019).³ In the TMS, stress results from a transaction between an individual and the environment, in which the environmental demands exceed the individual’s capacity to respond to it (McGrath, 1976; Lazarus & Folkman, 1984; Cooper et al., 2001). Applying the TMS to human interactions with technology, technostress can be understood as “a process that includes (1) the presence of ‘technology environmental conditions’; which are appraised as (2) demands or ‘techno-stressors’ that are taxing on the individual and require a change; which set into motion (3) ‘coping responses’; that lead to (4) psychological, physical, and behavioral ‘outcomes’ for the individual” (Tarafdar et al., 2019, p. 8).

Technology environmental conditions are the characteristics of technologies (e.g., complexity, reliability, usability) and technology-related events (e.g., system breakdown) that have the potential to create demands in the individual (e.g., Ayyagari et al., 2011; Galluch et al., 2015; Tarafdar et al., 2019). An unfavourable appraisal of such conditions results in factors referred to as technostress creators, or techno-stressors, which lead to stress. The degree to which technology environmental conditions act as stressors is subject to specific cognitive appraisal processes (Lazarus & Folkman, 1984; Califf et al., 2020). In primary appraisal, individuals assess and classify technology environmental conditions in terms of how they might affect their well-being. If a situation is deemed to be stressful, individuals engage in secondary appraisal to evaluate their available response resources (Lazarus & Folkman, 1984; Cooper et al., 2001). An individual experiences negative technostress if there are insufficient resources available to handle a situation.

To understand the impact of technostress creators on psychological and behavioural outcomes, it is important to note that the cognitive appraisal processes underlying the experience of technostress depend on the situational context (Lazarus & Folkman, 1984). In the following, we elaborate on technostress creators in the organisational and private usage contexts.

2.2. Organisational context: technostress creators and outcomes

The situational context for the organisational usage of technology comprises three main dimensions: (1) the IS usage domain (Brown & Venkatesh, 2005), (2) the purpose of IS (Davis, 1989), and (3) the degree of voluntary usage (Venkatesh et al., 2003).

As technology use has become the norm in virtually all organisations, technostress appears in a wide range of professional usage domains. For example, D’Arcy et al. (2014) showed that security-related stress in response to complex security requirements increases users’ susceptibility to engage in non-secure behaviour. For the use of technology in professional sales, Tarafdar, Pullins, et al. (2015) observed a negative link between technostress and performance outcomes that is moderated by staff members’ technology competence. In the context of healthcare, Califf et al. (2020) found a detrimental effect of technostress creators on nurses’ job satisfaction and turnover intentions.

In the organisational context, IS predominantly serves a (productivity oriented) utilitarian purpose (Wu & Lu, 2013), driven by the aim to increase performance and efficiency (van Der Heijden, 2004). Prescribed by implicit norms at work (Ayyagari et al., 2011), the organisational usage of technology is commonly characterised by strong social norms (Goodhue & Thompson, 1995). In this context, Brown et al. (2002) noted that “when individuals must perform specific behaviors, the importance of their beliefs and attitudes as antecedents to the performance of those behaviors is likely to be minimized” (p. 283).

Against this background, research typically identifies five main technostress creators in the organisational context: techno-complexity, techno-invasion, techno-insecurity, techno-overload, and techno-uncertainty (Tarafdar et al., 2019).⁴ Techno-complexity arises when users feel that the IS requires considerable effort to learn and understand (Tarafdar et al., 2007; Ragu-Nathan et al., 2008). As organisations strive to remain competitive by heavily investing in the latest IS, users’ existing knowledge can rapidly become obsolete. The new functionalities and associated technology jargon that come along with changes in IS can be intimidating. To ensure that they can efficiently apply the technology to accomplish their work tasks, users must constantly upskill. This necessary increase in job skills along with greater task difficulty can affect job performance and increase users’ stress levels. Common attributes of techno-complexity include users feeling they lack knowledge about IS, require a long time to understand and use IS, and lack time to upskill (Tarafdar et al., 2011). Techno-complexity applies not only to technologies but also to IS policies that contain technical language or are generally difficult to comprehend (D’Arcy et al., 2014).

Techno-invasion occurs when technologies invade users’ personal time such that they feel an expectation of non-stop availability, blurring the line between their work and private lives (Tarafdar et al., 2007; Ragu-Nathan et al., 2008). Because IS can provide permanent connectivity, allowing users to be contacted anywhere and at any time, individuals often feel pressure to respond to technology demands immediately (e.g., using a mobile phone on vacation to respond to an urgent work email). Such invasion of non-work time and space can cause stress because users feel that they are never free from technology. Common attributes of techno-invasion include users feeling constant connectivity to work, the need to sacrifice personal time for technology-related upskilling, a lack of time for family commitments, and concerns about surveillance and monitoring linked to a violation of personal privacy (e.g., Tarafdar et al., 2011, 2019; Day et al., 2012).

Techno-insecurity refers to individuals’ persistent concerns of job loss at the hands of innovative technologies (e.g., being replaced by technology or someone with stronger technological skills; Ragu-Nathan et al., 2008; Califf et al., 2020). Driven by constant changes in technology, the worry is that organisations will continually raise their expectations regarding both technology’s role in their business and the technological skills required of staff. The perception that one’s own role can be assumed by others or even rendered obsolete by technology (e.g., automation) can lead to feelings of job insecurity and stress. Typical indicators of techno-insecurity are related to users’ feelings of constant pressure to upskill to avoid being replaced, being threatened by co-workers with better technology skills, and working in an environment where employees limit their sharing of IS knowledge with others for fear of being replaced (Tarafdar et al., 2011).

Techno-overload describes circumstances in which individuals feel forced to work more and faster due to technology (Tarafdar et al., 2007; Ragu-Nathan et al., 2008). The capability of IS to provide a variety of information from different sources can cause information overload: a situation in which the user cannot efficiently handle the excessive amount of information (Savolainen, 2007; Ragu-Nathan et al., 2008). While technology can aid in multitasking and help users accomplish tasks faster, exceeding a person’s “healthy” multitasking capabilities can lead to frustration and stress. Techno-overload also encompasses technology-mediated interruptions (e.g., incoming email), which can put pressure on the user to attend to the incoming information immediately (Tarafdar et al., 2011). Such interruptions can distract the user from completing a work task, which in turn necessitates additional effort (more work) to return to the initial task (Galluch et al., 2015). Techno-overload has also been related to situations in which IS policies (e.g., security requirements) increase a user’s workload and create time pressure to complete a task (D’Arcy et al., 2014). Common attributes of techno-overload relate to users’ feelings of IS-triggered time pressure; changes in work habits to adapt to new technology; and greater workloads due to increased information, technology complexity, and interruptions, exceeding healthy multitasking capabilities (Tarafdar et al., 2011).

Techno-uncertainty emerges when users cannot build a “base of experience” with the technology due to relentless changes and updates (Tarafdar et al., 2007; Ragu-Nathan et al., 2008). Users feel pressured to constantly learn about new technologies to prevent their knowledge and experience from becoming obsolete. While users may initially feel motivated to do so, the permanent effort to refresh their skills can arouse negative emotions (Califf et al., 2020). Common attributes of techno-uncertainty include feeling a loss of control due to constant changes in the software, hardware, and networks (Tarafdar et al., 2011). This also extends to corresponding changes in IS policies (e.g., security requirements; D’Arcy et al., 2014); a lack of effective communication about such changes can further exacerbate techno-uncertainty (Califf et al., 2020).

Technostress research has delved extensively into psychological and behavioural outcomes in the organisational context (Fischer & Riedl, 2017; Tarafdar et al., 2019; Grummeck-Braamt et al., 2021). Psychological outcomes reflect the “state of mind at a conscious level” (Weinert et al., 2020, p. 1202) and include conditions such as burnout, exhaustion, role conflict, role overload, and reduced job satisfaction (Tarafdar et al., 2007, 2011; Ragu-Nathan et al., 2008; Ayyagari et al., 2011; Srivastava et al., 2015; Califf et al., 2020). Behavioral outcomes refer to “actions, consciously intended or not, that [individuals] engage in” (Morales et al., 2017, p. 466). In our study, this includes the motivation (or intention) to carry out a specific behaviour as an indicator for the degree of effort an individual plans to exert on its execution (Ajzen, 1991). In this context, Conner and Armitage (1998) emphasised that “intentions and behavior are held to be strongly related when measured at the same level of specificity in relation to the action, target, context, and time frame” (p. 1430).⁵ Cooper et al. (2001) noted regarding behavioural outcomes from stress, that “it is necessary to determine whether a direct relationship can be assumed […] or whether the causal pathway between stressors and strains is always mediated by some affective state” (p. 69). This highlights the importance of considering both psychological and behavioural responses when studying technostress. Relevant behavioural outcomes include decreased performance, productivity, and turnover (Ragu-Nathan et al., 2008; Tarafdar et al., 2010, 2011; Maier, 2014; Califf et al., 2020). However, before technostress creators result in behavioural outcomes, they might first be mitigated by psychological responses (Ragu-Nathan et al., 2008).

2.3. Private context: technostress creators and outcomes

In the private usage context, technostress research focuses on social networking services (SNSs; e.g., Maier et al., 2014; Xu et al., 2014; Tarafdar et al., 2020), IS in vehicles (e.g., Nastjuk & Kolbe, 2015), and private smartphone usage (e.g., Lee et al., 2014). For instance, Maier, Laumer, Weinert, et al. (2015) found that social overload and other SNS-related technostress creators drive user exhaustion and discontinuous usage intentions. Beyond the SNS context, Nastjuk & Kolbe (2015) reported evidence of range stress associated with dashboards in electric vehicles, while Lee et al. (2014) linked technostress and compulsive smartphone usage to psychological traits such as social interaction anxiety, locus of control, and materialism.

The dimensions for the private use of technologies differ from the organisational context. Here, the purpose of IS use is mostly hedonic (entertainment oriented), with individuals aiming to generate self-fulfilling value (Wu & Lu, 2013). Hedonic technologies are commonly associated with leisure activities and focus on fun aspects to promote prolonged use in the private context (van Der Heijden, 2004). In this setting, a greater degree of voluntary usage affects users’ beliefs and attitudes towards technologies, which are more likely to be translated into actual behaviour (Brown et al., 2002). In addition, users are highly motivated to adjust their behaviour (e.g., discontinue use) if a technology is perceived as demanding (Maier, Laumer, Weinert, et al., 2015).

While there is general consensus on the five main technostress creators in the organisational context, there has been no such set established for the private one. Recent studies discuss six technostress creators associated with SNS usage: pattern, disclosure, complexity, invasion, uncertainty, and social overload (Maier, Laumer, Weinert, et al., 2015; Tarafdar et al., 2020). Pattern is a stressor that SNS users experience when they are forced to adjust their habitual behaviours to conform to their friends’ SNS usage (e.g., permanently checking for updates; Maier, Laumer, Weinert, et al., 2015). Disclosure refers to negative feelings around disclosing one’s personal information and the pressure to remain up to date about others’ SNS statuses (Krasnova et al., 2010; Maier, Laumer, Weinert, et al., 2015). While these two technostress creators are inextricably linked to the SNS setting, their applicability to other private contexts is limited.

The most consistent set of technostress creators investigated in the private context comprises techno-complexity, techno-invasion, techno-overload, and techno-uncertainty (e.g., Maier, Laumer, Weinert, et al., 2015; Zhang et al., 2016; Brooks et al., 2017). Despite important differences between the organisational and private contexts, one can draw meaningful parallels in the conceptualisation of these four technostress creators. For example, D’Arcy et al. (2014) pointed to techno-complexity emerging from “security requirements [that] are viewed as complex and thereby force employees to expend time and effort in learning and understanding security” (p. 289). Similarly, complexity in the private context relates to a “negative perception that the SNS is difficult to handle” (Maier, Laumer, Weinert, et al., 2015, p. 282). Another example is techno-invasion, which is known to extend users’ workdays beyond regular business hours (Day et al., 2012); in the private context techno-invasion reduces the amount of time individuals spend with their families (Maier, Laumer, Weinert, et al., 2015). Finally, techno-overload links to a range of SNS stressors, including social overload, information overload, and system feature overload. Social overload relates to stress resulting from the excessive social demands of other SNS users (Maier et al., 2014). Users experience such overload when they feel that they worry too much about the well-being or problems of friends on SNS sites or pay too much attention to others’ posts. The communication demands linked to social overload require significant cognitive investment, which can be overwhelming (Zhang et al., 2016). Information overload arises when the amount of information on SNS sites exceeds a user’s processing capabilities (Fu et al., 2020). System feature overload occurs when users perceive the features of an SNS site to exceed their demands (Fu et al., 2020).

Our study builds on the four commonly studied technostress creators in the private usage context: techno-complexity, techno-invasion, techno-overload, and techno-uncertainty.⁶

2.4. Conceptualisation of technostress creators and their impact on outcomes

Technostress research commonly conceptualises technostress creators into a second-order construct with the aforementioned first-order dimensions. Higher-order aggregate constructs are prevalent in research fields such as psychology, where, for example, a meta-analysis suggested that the Big Five personality model’s five traits – which had previously been aggregated into two constructs (“getting ahead” and “getting along”; Digman, 1997)—could even be combined into a single one (Rushton & Irwing, 2008). Johnson et al. (2011) noted that “higher-order constructs may provide a more parsimonious solution to considering individual dimensions separately” (p. 244). They summarised the advantages of aggregate constructs as having the potential to (1) act as superior predictors, (2) overcome the jangle fallacy (measures with different labels assess the same construct), and (3) facilitate the analysis of latent factors.

These considerations are also reflected in technostress research models (e.g., Tarafdar et al., 2007; Ragu-Nathan et al., 2008). For example, Maier, Laumer, Weinert, et al. (2015), opted for a parsimonious research model linking technostress creators as a second-order construct to psychological and behavioural outcomes. Although various studies apply similar models, their findings differ in terms of how the aggregate construct affects each outcome category. For example, Tarafdar et al. (2010) found technostress creators to have a stronger effect on behavioural outcomes (end-user performance) than psychological ones (end-user satisfaction), whereas Maier et al. (2019) reported a greater impact on psychological outcomes (job burnout) than behavioural ones (user performance).

Providing a potential explanation for these conflicting findings, recent research has suggested that individual technostress creators may have different effects. For example, Chandra et al. (2019) found behavioural outcomes (employee innovation) to be negatively affected by techno-complexity but not by techno-overload. Furthermore, they observed that techno-uncertainty exhibited a positive link to employee innovation. In the context of healthcare IS, Califf et al. (2020) found techno-complexity to adversely affect behavioural outcomes (turnover intention) but not psychological outcomes (job satisfaction). Conversely, techno-uncertainty exhibited a positive relationship with job satisfaction but had no influence on turnover intention. Such divergent findings warrant a systematic empirical investigation that considers the impact of individual technostress creators on psychological and behavioural outcomes.

In examining the influence of technostress creators, it is important to consider the contexts in which technostress emerges. Technostress creators were originally conceptualised and validated in an organisational setting, with studies showing that they reduce job satisfaction (Ragu-Nathan et al., 2008) or user performance (Tarafdar et al., 2010). More recently, Maier, Laumer, Weinert, et al. (2015) extended this research to the SNS context. They found that technostress creators cause discontinuous usage intentions of SNS sites. A recent bibliometric review (Grummeck-Braamt et al., 2021) revealed that since Maier, Laumer, Weinert et al's (2015) study, the private usage context has received increasing attention from technostress scholars (e.g., Zhang et al., 2016; Fu et al., 2020; Tarafdar et al., 2020). However, there has not yet been an empirical investigation of the potential differences in the relationships between technostress creators and outcomes in both usage contexts. This is concerning not only from a research point of view but also from that of practitioners, considering that both managers in organisations and individuals must choose the most effective techniques to mitigate technostress.

Based on the above theoretical considerations, the objective of our meta-analysis is to investigate how technostress creators differ (1) in their relative importance for behavioural and psychological outcomes and (2) between the private and organisational usage contexts. We present our research model in Figure 1.

Figure 1. Research Model of Technostress.

* Given its inherent link to job security, techno-insecurity pertains only to the organisational context.

3. Research design

This study employs a random-effects meta-analysis to examine the effects of aggregated and disaggregated technostress creators on behavioural and psychological outcomes in both private and organisational usage contexts.

Meta-analyses aggregate the quantitative findings of primary research, allowing these findings to be statistically analysed at a higher level (King & He, 2005; Trang & Brendel, 2019). This is especially appropriate for our analysis because it enables us to incorporate existing findings quantitatively and analyse contextual moderators. Overall, this helps create an understanding and reconciliation of empirical and conceptual inconsistencies in the technostress literature. Our meta-analytical procedure comprises three stages: (1) collecting data from quantitative research on technostress, (2) coding constructs of interest, and (3) consolidating findings and measuring variables in our database as the basis for our random-effects model (Hunter & Schmidt, 2004).

3.1. Data collection

Our data-collection procedure was designed to identify the body of studies that examine the relationship between technostress creators and behavioural or psychological outcomes. We conducted an exhaustive database search to collect studies published until the end of April 2022. Our database search involved combing through various scientific repositories: AIS eLibrary, Scopus, Web of Science, and ProQuest Dissertations & Theses. As technostress is a phenomenon that also finds wide consideration outside the IS discipline (Tarafdar et al., 2019), we expanded our scope beyond the classic IS journals and conferences. Moreover, we included dissertations and conference proceedings to counteract the “file drawer effect”, describing the propensity of journals to favour publishing significant research findings, which then biases the results of further analyses (Rosenthal, 1979; Dickersin, 1990; Trang & Brendel, 2019).

Our keyword selection was borrowed from a recent review by Tarafdar et al. (2019) and includes (1) (techno AND stress) OR stress OR strain OR coping and (2) (techno OR ICT OR telework OR telecommut OR “e-mail” OR electronic OR “virtual work”) AND (stress OR strain OR coping). We also extended the search results using backward and forward search.

Based on our results, we created an initial database of 163 research articles that examine technostress with a quantitative, empirical design. Our final sample comprises studies with the following inclusion criteria, which are presented in Table 1. First, the study must examine the relationship between technostress creators and behavioural or psychological outcomes.⁷ Second, to ensure comparability of results, the study constructs must conceptualise and measure technostress creators and outcomes as per our definitions in Tables 2 and 3. We excluded, for example, studies that mix items from different technostress creators to measure a specific creator or where the items for measuring a specific outcome represent both psychological and behavioural aspects. Third, the study must provide sufficient empirical information to derive correlations between technostress creators and outcomes. Lastly, if the same data sets were used in multiple studies, we included only the article that either was published in the more prominent outlet (e.g., journal publication vs. conference proceedings) or contains richer empirical information (de Wit et al., 2012).

Table 1. Inclusion criteria.

Criterion	Excluded studies	Examples of excluded studies
Study must report the relationship between at least one technostress creator and one psychological or behavioural outcome.	26	Schmidt et al. (2021)
Study must conceptualise and measure technostress creators and outcomes based on their definitions as given in Tables 2 and 3.	20	Ninaus et al. (2021)
Study must report sufficient information to derive correlations between the technostress creators and psychological or behavioural outcomes.	24	Fieseler et al. (2014)
Study provides an independent data set.	4	The conference paper by Pflügner & Maier (2019) was excluded from our analysis because it was an earlier version of Pflügner’s et al. (2021) paper.

Table 2. Definition of technostress creators.

Independent variable
Technostress creator	Criteria for measures	Literature support
Techno-complexity (“too difficult”): Complexity associated with IS makes users feel their technology skills are insufficient and forces them to put effort into learning and understanding IS.	Insufficient knowledge about IS to handle a task efficiently Requires significant time investment to understand and use IS Requires constantly updating technological skills Lack of time to update technological skills Difficulty comprehending IS or IS policies	Ragu-Nathan et al. (2008); D’Arcy et al. (2014); Galluch et al. (2015); Tarafdar et al. (2007, 2011, 2019)
Techno-invasion (“always connected”): Users’ personal space and privacy is invaded through permanent connectivity and monitoring and the surveillance capabilities of technology.	Being permanently connected through technology Invasion of personal time by technology Sacrifice time with family due to technology Sacrifice personal time such as weekends or vacation to keep current on technologies Invasion of personal space and privacy through monitoring and surveillance	Ragu-Nathan et al. (2008); Day et al. (2012); Tarafdar et al. (2007, 2011); Barber & Santuzzi (2015)
Techno-insecurity (“being replaced”): Users worry about being replaced by technology or individuals with better technological skills.	Feeling constant threat to job security due to new technologies Pressure to permanently update IS skills for fear of being replaced Feeling threatened by technologies or others with better technological skills Reduced IS knowledge sharing among co-workers	Ragu-Nathan et al. (2008); Tarafdar et al. (2007, 2011); Califf et al. (2020)
Techno-overload (“too much”): Users work more or faster to meet requirements caused by technology or IS policies, handle excessive amounts of information or technology features, and meet expectations of others when using, for example, SNS applications.	Pressure to work faster or more due to technology Forced to work within tight time schedules due to technology Adjustment of work habits due to technology Exposed to vast amounts of information due to technology Perceive too many technology-related interruptions	Ragu-Nathan et al. (2008); Tarafdar et al. (2010, 2011); D’Arcy et al. (2014); Maier et al. (2014)
Techno-uncertainty (“too many changes”): Users do not have the chance to familiarise themselves with the technology due to constant updates and changes and as a result must update their skills.	New technological developments that create new demands and are perceived as “unsettling” Constant changes in software, hardware, and networks that require persistent upskilling Existing knowledge and experience with technologies become rapidly obsolete Perceived lack of effective communication on technology upgrades and changes	Ragu-Nathan et al. (2008); Tarafdar et al. (2011); Day et al. (2012); D’Arcy et al. (2014); Califf et al. (2020)

Table 3. Definition of technostress-related outcomes and contexts.

Dependent variables
Outcome	Examples	Literature support
Psychological: Cognitive and affective reactions to technostress creators	Burnout, exhaustion, role overload, role conflict	Tarafdar et al. (2007); Maier, Laumer, Weinert, et al. (2015); Srivastava et al. (2015)
Behavioural: Actions or intended responses to technostress creators	Technology usage, productivity, task performance	Tarafdar et al. (2007); Maier et al. (2014); Maier, Laumer, Weinert et al., (2015); Tarafdar, Pullins, et al. (2015)
Contextual moderators
Context	Examples	Literature support
Organisational: Users have limited (if any) control over the decision to use technology and it serves a predominantly utilitarian purpose.	Healthcare IT, telework, professional sales	Tarafdar, Pullins, et al. (2015); Suh & Lee (2017); Califf et al. (2020)
Private: Users have extended control over the decision to use technology and it serves a predominantly hedonic purpose.	Automotive, smartphones, SNS	Lee et al. (2014); Maier, Laumer, Weinert, et al. (2015); Nastjuk & Kolbe (2015)

We considered only the studies with enough information to directly extract or indirectly derive a correlation. More specifically, we converted other test statistics into correlations where possible (Wu & Lederer, 2009). Furthermore, we examined the independence of the studies considered. Publications covering multiple unique data sets were considered as multiple studies. For instance, we evaluated the two data sets reported in Maier et al. (2019) as two separate studies. 10 articles included multiple datasets. Likewise, if the same data set was used in multiple articles, the reported correlations were considered under one study.

Following the recommendations of Hunter & Schmidt (2004) and other meta-analyses in IS research (Mandrella et al., 2020), we also contacted all authors of the identified technostress papers.⁸ Our goal in doing so was twofold: First, to further counteract the file drawer effect and to increase our sample size, we asked authors to share additional unpublished research, research in progress, or published papers that were not yet included in our database. Second, to derive effect size estimates for studies published with insufficient information, we requested the necessary details.

Our final sample included a total of 102 articles comprising 113 studies and 49,955 independent observations (see complete list in supplementary material).

3.2. Coding

In our coding process, we collected correlational information on the relationship between technostress creators and psychological or behavioural outcomes. As independent variables, we considered the five technostress creators described earlier: techno-complexity, techno-invasion, techno-insecurity, techno-overload, and techno-uncertainty (see Table 2). The dependent variables considered were grouped under either behavioural or psychological outcomes (see Table 3).⁹ Furthermore, we collected information on whether the study was conducted in the private or organisational context.

3.3. Analysis

Our analytical procedure follows the Hunter & Schmidt (2004) method, which has been widely adopted for meta-analyses in IS research (e.g., Gerow et al., 2014; Mandrella et al., 2020). This approach is well suited for our study for at least two reasons: First, it provides a random-effects estimator. While fixed-effects estimators assume that there is one true effect size that is identical for all studies included in the meta-analysis, random-effects estimators allow for effect sizes to vary from study to study. The latter is to be expected in our case, as we observe heterogeneity in study characteristics such as technology, age, and gender, which are in turn likely to influence effect sizes. Second, study characteristics such as sample size or reliabilities can bias the estimation of an effect size (Hunter & Schmidt, 2004). Fortunately, the Hunter and Schmidt method provides an approach to identify, measure, and account for such study artefacts, thereby providing “true effect sizes”.

For our analysis, we estimated effect sizes in terms of correlations at different levels of aggregation (technostress as an aggregate or disaggregated construct), for different outcomes (psychological vs. behavioural), and in different contexts (private vs. organisational). To do so, we relied on a subset strategy and estimated effect sizes for each subset separately. For example, to estimate the aggregate effect of technostress creators on psychological and behavioural outcomes, we defined two subsets. We then allocated each effect size according to our coding to either the psychological or the behavioural outcomes subset. However, within each subset some studies contained multiple effect sizes. In such cases, we calculated composite correlations for each study within a subset to prevent bias caused by the dependencies between the considered effect sizes (Hunter & Schmidt, 2004).

To provide more precise effect size estimates, we regarded the studies’ samples sizes and construct reliabilities as individual study artefacts. While all studies report their final sample sizes, some offer no information on construct reliability. To circumvent these missing values and avoid having to omit the empirical information from our analysis, we used the distribution of the reliability measure within a subset to calculate an attenuation factor (Hunter & Schmidt, 2004).

The Hunter and Schmidt method provides three main parameters for evaluating effect sizes: mean effect size estimates, credibility intervals, and confidence intervals. The mean effect size ($\hat{\rho }$) is a sample-size weighted point estimator of the artefact-corrected mean true population correlations (of a subset). Accordingly, it provides insights in the average magnitude of technostress creators. Credibility intervals (CV) refer to the parameter value distributions in which (typically) 80% of the ρ values are considered. They are helpful in identifying whether moderators are operating. Moreover, CVs that do not include zero can be interpreted as evidence that the true population correlations are consistently positive or negative. Confidence intervals (CI) provide information on the accuracy of the mean corrected population correlation (Hunter & Schmidt, 2004); 95% CIs that do not include zero can be interpreted as a statistically significant mean effect (p < .05; Whitener, 1990; Gerow et al., 2014).

For the comparison of subsets, we used two-tailed t tests (Trang & Brendel, 2019). More specifically, we tested whether the estimated mean $\hat{\rho }$ effect of one subset significantly differs from that of another. A significant test result indicates that the mean population correlation of one subset is significantly larger than the other.

4. Results

4.1. Effects of technostress creators as an aggregate construct

Table 4 reports the estimates of the average impact of aggregated technostress creators on both psychological and behavioural outcomes in the private and organisational contexts. We classified the relationship strength according to Cohen (1988) into small, medium, and large effect sizes before testing for differences between the outcomes and contexts.

Table 4. Meta-analysis results on technostress creators as an aggregate construct in the organisational and private contexts.

Outcome	Predictor	Context	$\hat{\rho }$	Magnitude	k	N	SDρ	SDr	80% CV	95% CI	PVA	FS N
Psychological	Aggregated technostress creators	Org	−.33	Medium	69	25,143	.25	.23	[−.66, .01]	[−.40, −.27]	4.7	45
		Pri	−.51	Large	22	11,848	.23	.21	[−.81, −.22]	[−.61, −.42]	3.5	35
Behavioural		Org	−.18	Small	46	18,927	.18	.17	[−.41, .06]	[−.23, −.12]	8.1	5
		Pri	−.33	Medium	24	14,359	.14	.13	[−.51, −.16]	[−.39, −.27]	9.5	16

$\hat{\rho }$ = mean true score correlation; Magnitude = magnitude of the effect size according to Cohen (1988): small (<.30), medium (.30–.50), large (>.50); k = number of studies contributing to meta-analysis; N = total sample size; SD_ρ = residual standard deviation of ρ; SD_r = observed standard deviation of r; CV = credibility interval around true score correlation ρ; CI = confidence interval around true mean score correlation ρ; PVA = percentage of variance accounted for by sampling and measurement error; FS N = Orwin’s fail-safe N, computed with a criterion correlation of 0.2, Org = organisational, Pri = private.

In the organisational context, we observe a medium mean impact of technostress on psychological outcomes ($\hat{\rho }$ = −.33, CI < 0, sig.), whereas the average impact on behavioural outcomes is small ($\hat{\rho }$ = −.18, CI < 0, sig.). At the same time, the CVs include zero for both psychological (CV = [−.66, .01]) and behavioural outcomes (CV = [−.41, .06]). Even though the CIs indicate that the average effect of technostress creators is negative, the CVs show that the correlations in individual studies may have positive effects.

The magnitudes of the effect sizes are greater in the private context. We observe a large estimated mean impact of technostress on psychological outcomes ($\hat{\rho }$ = −.51, CI < 0, sig.) and a medium impact on behavioural outcomes ($\hat{\rho }$ = −.33, CI < 0, sig.). Neither the CVs for psychological (CV = [−.81, −.22]) nor the CVs for behavioural outcomes (CV = [−.51, −.16]) include zero. This not only suggests that the mean effects are negative, but also provides evidence that the correlations at the population level (i.e., effects in different studies) are consistently negative.

We then tested for differences between outcomes and contexts. As already indicated by the non-overlapping CIs, the results of a series of t tests confirm that the negative impact of technostress is significantly higher for psychological outcomes than for behavioural outcomes in both the organisational (diff. = .15, t = 3.506, df = 113, p < .01) and the private context (diff. = .18, t = 3.237, df = 44, p < .01). We can also discern differences between the two usage contexts. The differential impact on both psychological (diff. = .18, t = 3.556, df = 68, p < .01) and behavioural outcomes (diff. = .15, t = 2.995, df = 89, p < .01) is significantly more pronounced for private usage as compared to organisational usage.

Finally, the estimated effects of the aggregated technostress creators on behavioural and psychological outcomes in the organisational and private contexts exhibit large CVs. This indicates heterogeneity in the population correlations and thus supports the need for a more nuanced investigation of the underlying individual technostress creators. This is in line with the small PVA (percentage of variance accounted for by sampling and measurement error) values, indicating that variation between studies can be explained only to a limited degree by sampling error and study artefacts.

Taken as a whole, our results suggest that technostress as an aggregate construct has negative effects on average (i.e., across individual study contexts) on both psychological and behavioural outcomes, with their influence on the former being stronger. Furthermore, their impact is more pronounced in the private context than in the organisational one.

In addition, the results of the heterogeneity analysis (large CVs and small PVA values) support the need for a more detailed assessment of the individual dimensions underlying the aggregate technostress creators construct. In the following, we therefore delve one level deeper and investigate the disaggregated effects of technostress creators.

4.2. Effects of individual technostress creators

In the second step, we focused on the disaggregated effects of the five technostress creators (see Table 5 for detailed results), allowing us to identify a differential pattern for the impact of individual technostress creators on psychological and behavioural outcomes depending on the usage context.

Table 5. Meta-analysis results on individual technostress creators in the organisational and private contexts.

Outcome	Predictor	Context	$\hat{\rho }$	Magnitude	k	N	SDρ	SDr	80% CV	95% CI	PVA	FS N
Psychological	Techno-complexity	Org	−.43	Medium	26	12,381	.23	.21	[−.73, −.14]	[−.52, −.34]	4.2	30
		Pri	−.40	Medium	5	1,555	.14	.14	[−.58, −.21]	[−.54, −.26]	14.1	5
	Techno-invasion	Org	−.35	Medium	43	16,833	.22	.20	[−.63, −.08]	[−.42, −.29]	5.8	33
		Pri	−.43	Medium	8	3,637	.27	.24	[−.77, −.09]	[−.62, −.24]	3.3	10
	Techno-insecurity	Org	−.35	Medium	26	12,768	.23	.21	[−.64, −.05]	[−.44, −.25]	4.2	20
	Techno-overload	Org	−.41	Medium	56	20,928	.23	.21	[−.70, −.11]	[−.47, −.35]	5.2	59
		Pri	−.49	Medium	22	11,848	.28	.25	[−.86, −.14]	[−.62, −.38]	2.5	32
	Techno-uncertainty	Org	−.14	Small	21	9,626	.17	.16	[−.37, .08]	[−.22, −.06]	8.2	7
		Pri	−.08	Small	5	2,337	.05	.06	[−.15, −.02]	[−.15, −.02]	52.1	3
Behavioural	Techno-complexity	Org	−.25	Small	20	10,309	.18	.16	[−.47, −.02]	[−.33, −.17]	7.1	5
		Pri	−.28	Small	9	6,307	.11	.10	[−.41, −.15]	[−.35, −.21]	13.7	4
	Techno-invasion	Org	−.12	Small	30	13,646	.18	.16	[−.35, .11]	[−.18, −.05]	8.0	12
		Pri	−.33	Medium	11	6,349	.20	.18	[−.59, −.08]	[−.45, −.21]	4.9	8
	Techno-insecurity	Org	−.13	Small	16	8,568	.12	.12	[−.28, .03]	[−.19, −.06]	13.9	6
	Techno-overload	Org	−.19	Small	41	17,355	.21	.20	[−.46, .09]	[−.26, −.12]	6.0	3
		Pri	−.35	Medium	22	13,553	.15	.14	[−.55, −.15]	[−.42, −.28]	7.4	17
	Techno-uncertainty	Org	.01	Small	18	8,594	.21	.19	[−.25, .28]	[−.09, .11]	5.9	19
		Pri	−.21	Small	7	5,171	.09	.09	[−.33, −.09]	[−.29, −.14]	16.6	1

$\hat{\rho }$ = mean true score correlation; Magnitude = magnitude of the effect size according to Cohen (1988): small (<.30), medium (.30–.50), large (>.50); k = number of studies contributing to meta-analysis; N = total sample size; SD_ρ = residual standard deviation of ρ; SD_r = observed standard deviation of r; CV = credibility interval around true score correlation ρ;| CI = confidence interval around true score correlation ρ; PVA = percentage of variance accounted for by sampling and measurement error; FS N = Orwin’s fail-safe N, computed with a criterion correlation of 0.2, Org = organisational, Pri = private.

For the organisational context, we found that four of the five technostress creators (all except techno-uncertainty) significantly influence psychological ($\hat{\rho }$= −.35 to −.43, CI < 0, sig.) and behavioural ($\hat{\rho }$= −.12 to −.25, CI < 0, sig.) outcomes, exhibiting similar patterns in terms of effect size with overlapping CIs. In contrast to the other technostress creators, techno-uncertainty appears to be distinct in terms of its effect size. The estimated mean correlations of techno-uncertainty on both psychological ($\hat{\rho }$ = −.14, CI < 0, sig.) and behavioural outcomes ($\hat{\rho }$ = .01, n.s.) are comparably small, with the latter being positive and insignificant. In addition, a series of t tests reveal a significantly smaller effect of techno-uncertainty on psychological outcomes in comparison with the effects of techno-complexity (diff. = .29, t = 4.810, df = 45, p < .01), techno-invasion (diff. = .21 t = 3.844, df = 62, p < .01), techno-insecurity (diff. = .21, t = 3.483, df = 45, p < .01), and techno-overload (diff. = .27, t = 4.893, df = 75, p < .01). The same can be discerned for the relationship of techno-uncertainty and behavioural outcomes, with significantly smaller effects as compared to those of techno-complexity (diff. = .26, t = 4.109, df = 36, p < .01), techno-invasion (diff. = .13, t = 2.275, df = 46, p < .05), techno-insecurity (diff. = .14, t = 2.346, df = 32, p < .05), and techno-overload (diff. = .20, t = 3.368, df = 57, p < .01). The t tests also reveal that each technostress creator has a significantly stronger effect on psychological outcomes than behavioural ones (techno-complexity [diff. = .18, t = 2.884, df = 44, p < .01], techno-invasion [diff. = .23, t = 4.725, df = 71, p < .01], techno-insecurity [diff. = .22, t = 3.530, df = 40, p < .01], techno-overload [diff. = .22, t = 4.826, df = 95, p < .01], techno-uncertainty [diff. = .15, t = 2.465, df = 37, p < .05].

For the private context, we found a similar but less clear pattern for the estimated mean correlations. All technostress creators other than techno-uncertainty display significant negative effects on both psychological ($\hat{\rho }$= −.40 to −.49, CI < 0, sig.) and behavioural outcomes ($\hat{\rho }$= −.28 to −.35, CI < 0, sig.). Again, the mean effect size of these three technostress creators are similar, with overlapping CIs. In turn, techno-uncertainty exhibits the smallest effect on both psychological ($\hat{\rho }$= −.08, CI < 0, sig.) and behavioural outcomes ($\hat{\rho }$= −.21, CI < 0, sig.). Again, t tests confirm a significant difference between the effect of techno-uncertainty and the other technostress creators on psychological outcomes (techno-complexity [diff. = .32, t = 4.813, df = 8, p < .01], techno-invasion [diff. = .35, t = 2.823, df = 11, p < .05], techno-overload [diff. = .41, t = 3.215, df = 25, p < .01]).

While the difference between the effect size of techno-uncertainty and the other technostress creators on behavioural outcomes is significant for techno-invasion (diff. = .12, t = 2.108, df = 16, p < .05) and techno-overload (diff. = .14, t = 2.322, df = 27, p < .05, p < .05), it is non-significant for techno-complexity (diff. = .07, t = 1.363, n.s.). Techno-overload has a significantly stronger effect on psychological outcomes than behavioural ones (diff. = .14, t = 2.067, df = 42, p < .05), while our results indicate non-significant differences for techno-complexity (diff. = .12, t = 1.781, df = 12, n.s.) and techno-invasion (diff. = .10, t = .930, df = 17, n.s.). Interestingly, the effect of techno-uncertainty on behavioural outcomes is significantly stronger than on psychological outcomes (diff. = .13, t = 2.900, df = 10, p < .05); this is in contrast with the estimations of other individual technostress creators.

At this point, it should be noted that the analysis of effects in the private domain has limited statistical power in terms of the number of studies used and total observations. The individual effects of techno-complexity, techno-invasion, and techno-uncertainty on psychological outcomes and the effects of techno-complexity and techno-uncertainty on behavioural outcomes only involve five to nine studies each and 1,555 to 6,307 total observations. Hence, the empirical results in the private domain must be interpreted with particular caution.

To sum up the second part of our analysis, we identified important differences in how individual technostress creators affect psychological and behavioural outcomes. The disaggregation of technostress creators provides more nuanced insights into their individual roles, revealing that they can behave differently in terms of outcomes and in different contexts.

5. Discussion

Motivated by inconsistent empirical findings in the technostress literature, we undertook a systematic meta-analysis synthesising and integrating their quantitative results. Through this meta-analysis, we enhance the understanding of how technostress creators differ in their absolute and relative importance for behavioural and psychological outcomes. Moreover, our findings reveal a more nuanced perspective regarding the widely adopted conceptualisation of technostress creators as an aggregate second-order construct. We suggest that the varying effects of technostress creators on psychological and behavioural outcomes signals the need for additional theorising about the individual technostress creators.

In the following sub-sections, we discuss how our meta-analytical findings advance the technostress research field. We first address the overall effects of technostress creators as an aggregate construct and then focus on the relationship between individual technostress creators and outcomes. We conclude the discussion by outlining the contributions and limitations of our study.

5.1. Summary of findings: technostress creators as an aggregate construct

5.1.1. Technostress is a meaningful dark side phenomenon

Our first main finding pertains to the relevance of technostress as a dark-side effect. Overall, our results indicate that technostress creators are a substantial psychological concern while also having relevant behavioural consequences. The estimated negative psychological effects are medium and large in the organisational and private contexts, respectively, with the magnitude of the effects smaller for behavioural outcomes. Compared to other influential non-technological stressors in the work and non-work settings (Podsakoff et al., 2007; Gilboa et al., 2008), the effect of technostress creators on behaviours can be regarded as meaningful.

The technostress literature has conceptualised technostress creators as stress-creating conditions that result in negative psychological and behavioural outcomes. By quantifying the overall mean effect that technostress creators have on outcomes, our results indicate that the demands from the technology environment are overwhelmingly appraised as threatening, highlighting the dark side of technostress (Tarafdar et al., 2011). While the mean effects are substantial, we also found large heterogeneity between individual effects of different studies, especially in the organisational domain. We note that recent and nascent research indicates that there are both positive and negative sides of stress (Tarafdar et al., 2019; Califf et al., 2020; Shirish et al., 2021). Under this view, the demands imposed by the technology environment are perceived as an opportunity to learn and grow, which leads to positive outcomes (Cooper et al., 2001). We suggest that when empirically more mature, this is an area for future meta-analytical studies.

5.1.2. Psychological outcomes emanating from technostress are more salient

Second, we identified in both usage contexts that aggregated technostress creators affect psychological outcomes more strongly than behavioural outcomes.

General models of stress, such as the stressor–strain–outcome model, consider psychological outcomes as a mediator between stress-creating conditions and behavioural outcomes (Koeske & Koeske, 1993). Under this perspective, behavioural outcomes result from psychological strain. This view is also supported by occupational stress models that characterise stress as a process that leads to predominantly affective outcomes which under certain conditions may translate into performance outcomes (Motowidlo et al., 1986; Jex et al., 1991) In this context, Motowidlo et al. (1986) found empirical support for the causal sequence of stress variables “from job conditions and individual characteristics (job experience, fear of negative evaluation […]) as exogenous variables, to perceptions of stressful events (their frequency and intensity), to subjective stress, to affect (anxiety, hostility, and depression), and, finally, to job performance” (p. 624).

This logical link between stressors, psychological outcomes, and behavioural outcomes is also reflected in technostress research models (e.g., Ragu-Nathan et al., 2008; Maier, Laumer, Weinert, et al., 2015; Califf et al., 2020). For example, Ragu-Nathan et al. (2008) found a direct impact of technostress creators on psychological outcomes (job satisfaction), which then affects behavioural outcomes (organisational commitment). However, it should be emphasised that the underlying processes that translate psychological into behavioural outcomes are complex (Cooper et al., 2001).

5.1.3. Technostress is more pronounced in the private usage context

Our third main finding is that aggregated technostress creators have a greater impact on psychological and behavioural outcomes in the private as compared to the organisational context. In the private usage context, the decision to use a technology is generally not prescribed by implicit norms (Maier, 2014). In other words, the user’s perceived ability to replace a certain technology by switching to another is greater. This decision is influenced by so-called “switching costs”, which are defined as the “perceived disutility a user would incur in switching from the status quo to the new IS” (Kim & Kankanhalli, 2009, p. 572). Two types of switching costs have been discussed in the IS literature: sunk costs and transition costs (Kim & Kankanhalli, 2009; Polites & Karahanna, 2012; Maier, Laumer, Weinert, et al., 2015). In this context, sunk costs refer to the time and effort an individual has invested in acquiring and learning how to use an existing IS, while transition costs refer to the time and effort an individual must invest to acquire and learn how to use an alternative IS. Switching costs can be perceived as stressful and represent an important determinant of how individuals react to technostress, especially in the private context (Maier, Laumer, Weinert, et al., 2015).

Organisational technology usage is characterised by strong social norms, utilitarian usage objectives, and often the mandatory use of specific technologies. Switching is seldom a viable option for users of organisational technology (Maier, Laumer, Weinert, et al., 2015). Users might therefore react to technostress through adaptation strategies that can mitigate its effects, such as building expertise to maintain work performance (Beaudry & Pinsonneault, 2005). By contrast, individuals in the private context are free to choose what technology they use. They might switch from one technology to another as a response to technostress, which could in turn lead to further stress (Maier, Laumer, Weinert, et al., 2015). SNSs are the most commonly examined application for technostress in the private context; their use is largely hedonic and actually involves continued use despite technostress-creating conditions (Tarafdar et al., 2020), which potentially sets users up for ongoing technostress. This could explain why individuals’ psychological and behavioural outcomes are more strongly affected by technostress creators in the private than in the organisational context.

Another possible explanation for this difference is that organisations invest in end-user support services (Weinert et al., 2020), which can reduce the effects of technostress. Examples of support resources include technical support provision, literacy facilitation, involvement facilitation, and innovation support (Ragu-Nathan et al., 2008; Tarafdar et al., 2011; Fuglseth & Sørebø, 2014). Technical support provides guidance on how to use technology within an organisation (Weinert et al., 2020). Literacy facilitation describes mechanisms to encourage the sharing of IS-related knowledge, whereas involvement facilitation involve including users when planning and implementing technology (Ragu-Nathan et al., 2008). Innovation support describes mechanisms that facilitate learning and experimenting with new technologies (Tarafdar et al., 2011). Such technostress inhibitors are relatively widespread in the organisational context and have the ability to not only reduce the perception of technostress creators but also decrease the negative effects of technostress (e.g., job and end-user satisfaction; Ragu-Nathan et al., 2008; Tarafdar et al., 2011). These support resources are not necessarily available to users in the private context.

Recent research has studied social support as an additional organisational technostress inhibitor (Weinert et al., 2020); such support is also relevant in the private usage context (Lo, 2019). Weinert et al. (2020) found that emotional and instrumental social support in the face of technostress creators increased positive outcomes (end-user performance) and decreased negative ones (techno-exhaustion and physiological arousal). However, previous research on technostress has indicated that social support might be a double-edged sword. Maier (2014) noted that “users are more strongly connected to their private social environment through IT than while using IT for work purposes, which might be experienced as [an] additional stressor of using such IT” (p. 29) because they feel compelled to respond to social support requests. Hence, social support can act as both a technostress creator and inhibitor.

5.2. Summary of findings: individual technostress creators

5.2.1. Techno-uncertainty differs in its relative importance for predicting outcomes

All technostress creators except techno-uncertainty exhibit a negative impact on psychological and behavioural outcomes, with similar effect sizes in both contexts. For both organisational and private usage, the negative effect of techno-uncertainty on psychological and behavioural outcomes is significantly smaller than the effects of the other technostress creators, indicating that techno-uncertainty differs in its relative importance in explaining outcomes.

Individuals experience techno-uncertainty as a result of technological changes that require a constant updating of their skills. However, such changes might not always be appraised as negative stress-creating conditions (Lazarus & Folkman, 1984). The credibility intervals of techno-uncertainty for psychological and behavioural outcomes include zero, indicating that the population correlations have not only negative but also positive effects on both outcomes. Our study is the first to provide substantiated, context-independent support for this rather unique role of techno-uncertainty.

5.2.2. The patterns for the estimated mean correlations differ between usage contexts

Our results indicate a relatively harmonious impact of individual technostress creators on psychological and behavioural outcomes in the organisational context, with each one having a greater impact on psychological outcomes than on behavioural ones. This is in alignment with findings regarding the relationship between technostress creators as an aggregate construct and outcomes in the organisational context (see Section 5.1.2).

The picture is more complex for the private usage context, with only techno-overload affecting psychological outcomes more strongly than behavioural outcomes. Techno-overload plays a central role in the private usage context and has been identified as the main source of stress when using SNSs (Maier et al., 2014; Salo et al., 2019; Fu et al., 2020). Techno-overload in this context arises from circumstances such as the platform providing an overwhelming amount of information and features or having too many social interactions with other users (Zhang et al., 2016; Cao, Sun, 2018). For example, SNS users often feel a loss of control over the social situation when they receive too many requests or feel pressure to provide excessive social support to other users. This feeling translates first to strong psychological reactions, such as feelings of exhaustion or dissatisfaction (Maier et al., 2014). Exhaustion caused by overload is generally considered to be an antecedent of changes in usage behaviour (Ravindran et al., 2014; Zhang et al., 2016). Perceived overload from using SNS sites can exceed a user’s cognitive capacity; this is more likely to occur when the SNS is used excessively (Maier, 2014). Once this happens and users feel excessive exhaustion, one can expect to see behavioural changes such as discontinued usage as the user tries to restore emotional stability (Cao, Sun, 2018; Fu et al., 2020).

The data shows a less clear picture regarding the differences in how the other technostress creators affect outcomes. Techno-invasion and techno-complexity do not statistically differ in their impacts on psychological and behavioural outcomes, while techno-uncertainty is the only technostress creator that has a greater effect on behavioural outcomes. A possible explanation for these results relates to the characteristics of the environmental conditions from which technostress creators arise. Such conditions include technological characteristics, the tasks associated with using a technology, and the social environment (Fischer & Riedl, 2017). Most technostress studies in the private usage context focus on SNS sites, where techno-overload accounts for demands that arise from the social environment (e.g., meeting the expectations of other users; Maier et al., 2014) and can therefore lead to strong psychological outcomes. However, unlike techno-overload, the other technostress creators are not directly entwined with social relationships. For example, users can switch platforms if they find the current one to be difficult to use (techno-complexity). As a result, direct behavioural responses such as discontinuous usage can be expected (Maier, Laumer, Weinert, et al., 2015).

5.3. Implications for research and future research opportunities

Empirical studies on technostress have been primarily concerned with identifying and validating technostress creator-outcome relationships (i.e., effects are different from zero). This paper is the first to shed light on effect sizes, which refer to the magnitude of the effect of technostress creators on outcomes. We estimated effect sizes with high statistical power by integrating and synthesising the empirical technostress literature into a data set comprising 49,955 independent observations gleaned from 102 articles that include a collective 113 individual studies. In doing so, we contribute to the technostress literature in four meaningful ways.

First, our meta-analysis provides strong empirical evidence that technostress is a relevant dark side phenomenon. While existing studies (see supplementary material) provide valuable insights, they are limited to their specific samples, the samples’ limited statistical power, and their contexts. Our study sets itself apart by investigating the technostress phenomenon in terms of effect sizes as pertain to the influence that technostress-creating conditions have on relevant outcomes. A key finding from our research is that the detrimental effects of technostress on psychological outcomes are greater than its effects on behavioural outcomes, with the difference being more pronounced for the private context. Such knowledge regarding the problem relevance of technostress is critical for both research problem formulation and practical implications of future research endeavours (Rai, 2017).

Our database (see supplementary material) reveals that technostress research has drawn on different theoretical lenses to examine different parts of the technostress process. Most articles in our meta-analysis applied the TMS (Lazarus & Folkman, 1984) to conceptualise the overarching technostress process in terms of how it evolves from a transaction between the individual and the environment through primary and secondary appraisal (Tarafdar et al., 2019). Within this process, several articles used the person–environment fit model (Edwards & Cooper, 1988; Edwards, 1991; Cooper et al., 2001) to frame primary appraisal (Ayyagari et al., 2011). Other articles applied the stressor–strain–outcome model (Koeske & Koeske, 1993) to distinguish between psychological and behavioural outcomes. Overall, there is agreement on the general mechanism governing how technostress emerges and its impact on psychological and behavioural outcomes. Yet, there are important differences with respect to the theoretical underpinnings for the stressors and outcomes in the respective domains.

In the organisational context, studies have built on role theory (Kahn et al., 1964) to provide context for role- and task-related stressors (Rangarajan et al., 2005; Tarafdar et al., 2010) and the job demands–resources model (Demerouti et al., 2001) to relate technostress creators to general job demands and outcomes (Day et al., 2012; Reinke & Chamorro-Premuzic, 2014). In the private context, studies have referred primarily to theories originating in social psychology. For example, social cognitive theory (Bandura, 2001) has been applied to explain technology usage patterns and effects on family and personal life (Zheng & Lee, 2016), while social support theory (Caplan, 1974; Cassel, 1976; Cobb, 1976) has aided in understanding how too much social support can lead to detrimental outcomes (Maier et al., 2014). However, the line between individuals’ private and professional lives is becoming increasingly blurred, and future research could examine how the theories applied in the organisational and private contexts link to one another. For example, using social support theory, Maier et al. (2014) theorised social overload based on the specific social interactions among SNS site users in the private context. It would be interesting to investigate whether social support theory provides a relevant theoretical lens for understanding how social interactions through organisational technologies (e.g., email or other workplace communication applications) affect psychological and work-related behavioural outcomes. In addition, our findings regarding the unique role of techno-uncertainty suggest that more theorising is needed to capture the specific characteristics of each technostress creator (Maier et al., 2014).

Second, by offering precise effect size estimates for different outcomes, contexts, and technostress creators, we help the research community identify where and how to explore research opportunities in the technostress realm. Examining the findings of single empirical studies and simply comparing their results (e.g., estimated correlations or path coefficients) can lead to inaccurate conclusions, as they can suffer from study-specific artefacts in terms of measurement instrument reliability, and low statistical power (Hunter & Schmidt, 2004). Based on the synthesis of an exhaustive and systematically generated list of empirical studies, our meta-analysis provides a comprehensive account of the relationships between technostress creators and different outcomes in different contexts. It complements existing systematic literature reviews and scientometric analyses that have provided guidance for technostress research by structuring the body of knowledge (LaTorre et al., 2019; Dragano & Lunau, 2020; Nisafani et al., 2020) identifying key areas for future research (Tarafdar et al., 2019); reviewing the methods employed in the field (Riedl, 2013; Fischer & Riedl, 2017) and unpacking the evolutionary patterns, trends, common publication outlets, or most productive authors and their institutional affiliations (Bondanini et al., 2020; Grummeck-Braamt et al., 2021).

Insight into the effect sizes of specific relationships can guide further studies in investigating the technostress phenomenon. For example, knowing about the substantial effect of techno-overload on psychological outcomes in the private domain can make an investigation of remedies a fruitful endeavour. In addition, the average effect for the techno-uncertainty relationship with behavioural outcomes being close to zero, together with broad credibility intervals, signals the need for detailed research examining when this technostress creator unfolds its negative potential or may even lead to positive outcomes. Moreover, our comparative analysis of effect sizes reveals the importance of the private context for the technostress phenomenon. Research has largely focused on the organisational context to study technostress (Grummeck-Braamt et al., 2021). By pointing out that the negative consequences are more pronounced in the private domain, we highlight the potential for future research herein.

Third, the comparative analysis of outcomes suggests that technostress creators are more strongly linked to psychological outcomes than to behavioural outcomes, which aligns with theoretical perspectives on psychological stress that propose psychological outcomes to be antecedents of behavioural outcomes (e.g., stressor–strain outcome model (Koeske & Koeske, 1993) and the holistic organisational stress process (Simmons & Nelson, 2007; Nelson & Simmons, 2011; Califf et al., 2020). This finding reinforces the importance of considering different types of psychological outcomes and their mediating effects on other relevant outcomes, such as turnover behaviour and task productivity. Moreover, by highlighting the powerful psychological consequences of technostress, we emphasise the need to explore additional outcomes such as a potential societal health impact.

Finally, our study bears implications for the discussion of when technostress creators should be examined in an aggregated or disaggregated formulation (Ragu-Nathan et al., 2008; Chandra et al., 2019). A large share of research has focused on technostress creators and outcomes at the aggregate level. However, we also found evidence that individual technostress creators differentially affect outcomes. As shown for the case of techno-uncertainty, for example, an aggregation may overshadow its individual effects. This has implications for the theorising and empirical modelling of technostress creators. The aggregated approach is relevant when a broad range of effects must be captured parsimoniously and the expected direction of effect for each individual technostress creator is anticipated to have the same direction, as may be true when considering a broad range of applications (e.g., general office applications). When considering this approach, such effects should be empirically verified. The disaggregated approach may be appropriate when only one or two stressors need to be studied. We note here that most studies that have adopted a disaggregated treatment have done so without prior theorising. Therefore, we suggest future studies adopting a disaggregated approach to take a more nuanced perspective on the individual roles of technostress creators (see Maier et al., 2014 for an example) by theorising about which of them are important for the study and their potential relationships with outcomes.

5.4. Implications for practice and policy

We offer several insights for managerial practice. First, organisations should be aware that technostress has significant psychological outcomes. While behavioural indicators of stress are more tangible and commonly tracked (e.g., employee turnover rates or individual productivity; Sullivan & Bhagat, 1992; Weinberg et al., 2015), psychological outcomes may remain unnoticed. Hence, it is important to identify relevant psychological indicators that might lead to such outcomes. This enables managers to implement timely interventions to prevent psychological health risks.

Second, by providing accurate estimates for each technostress creator, our study helps managers identify tailored countermeasures for technostress. Mitigation strategies discussed in previous technostress studies, such as technical support provision, literacy facilitation, involvement facilitation, and innovation support (Ragu-Nathan et al., 2008; Tarafdar et al., 2011; Fuglseth & Sørebø, 2014), require financial investment and therefore careful planning and prioritisation. To effectively implement mitigation mechanisms, managers could focus on the technostress creators with the strongest impact on outcomes.

Third, our findings provide empirical evidence for the harm that technologies can cause in the private usage context. A key strategic aspect for technology platform providers is to maintain a large user base and incentivise regular usage. When users lose interest in, for example, a SNS site, its providers experience significant financial damages (Chiu & Huang, 2015; Maier, Laumer, Weinert, et al., 2015). Technology providers should take our findings into consideration when designing technologies or guidelines around technology usage to create a user experience that minimises technostress. At the same time, users should critically reflect on how their use of specific applications can damage their psychological well-being (Tarafdar et al., 2020).

Finally, our findings reaffirm that technostress is a ubiquitous concern in both the organisational and private usage contexts. It has the potential to impair societal health and well-being and to cause sizeable burdens for healthcare systems (Hassard et al., 2018; Brunner et al., 2019). There are two important implications here for policy makers. One, with respect to technology policy, introducing mandatory technology design elements that aim to minimise or warn users of the dangers of technostress could be an effective legislative initiative to manage such stress. This is particularly important considering that worldwide, 4.6 billion people use SNS sites and the number of smartphone subscriptions is predicted to exceed 7.6 billion users by 2027 (Statista, 2022a, 2022b). Moreover, technology providers tend to incorporate addictive features in their technologies to maximise their user base and prolong usage (Montag et al., 2019). Two, with respect to healthcare policy, both public and private healthcare systems should consider investing in resources to combat technostress.

5.5. Limitations and boundary conditions

Inevitably, this study is also subject to limitations. One limitation is that for some analyses, the number of studies is below the recommended level of 10 (Switzer et al., 1992); additional studies would improve the stability of our results. This limitation is particularly relevant for the private usage context. Furthermore, some analyses exhibit high heterogeneity. While our moderator analyses explain some of the observed variance, they also suggest that further moderators may be at play.

A second limitation relates to the conceptualisation and measurement of outcomes, as we assessed both psychological and behavioural outcomes as aggregate measures. Future meta-analyses should investigate the relationship between technostress creators on more specific outcomes. At this point would also like to note that most studies in our database primarily rely on self-reported measurement instruments. Only three studies in our database utilise objective measures for the investigated relationships. Other meta-analyses have found evidence of inflated effect sizes when utilising self-reported measurement instruments (e.g., Sharma et al., 2009; Gerow et al., 2014; Mandrella et al., 2020). We find no evidence for such an effect in our database.¹⁰ However, to learn more about potential biases in the technostress literature, we believe that the effects of research designs such as longitudinal data collection, matched pair-designs, or objective sources for performance measurement should get specific attention in further research endeavours.

Thirdly, we did not account for the interrelationships between the psychological and behavioural outcomes of technostress creators. Thus, a potential avenue for future research is to investigate relationships between the different outcome types. Stress research shows that such relationships are complex (Cooper et al., 2001). Future research could explore the relationship between specific psychological outcomes (e.g., exhaustion or satisfaction) and behavioural outcomes (e.g., turnover or productivity) to understand how and to what extent specific psychological outcomes translate into specific behaviour. Future research could consider these links using meta-analytic structural equation modelling (Gerow et al., 2016). Further, given the limited data points on physiological outcomes, we had to discard this important link from our meta-analysis. This should be reconsidered by future research once there is a sufficient number of studies available for a meta-analysis of technostress creator’s impact on physiological outcomes.

Finally, a boundary condition of our study is that it is based on an empirical integration and synthesis of the five most commonly examined technostress creators. Some studies have developed qualitative descriptions of additional technostress creators (e.g., Salo et al., 2019), but no instruments for quantitative examination. Empirical insights imparted by such studies was thus not considered in our analysis. Once empirical data on these technostress creators increase, we see a great opportunity for future research to examine their effects in a meta-analytical study.

6. Conclusion

The substantial and growing body of technostress research calls for scholarly effort to take stock of its empirical findings to guide future research efforts. Therefore, this study investigates the effects of technostress-creating conditions on behavioural and psychological outcomes in the private and organisational IS contexts. To the best of our knowledge, this work presents the first meta-analysis on the phenomenon of technostress. We advance its theoretical development by systematically analysing its empirical foundations and identifying directions for future conceptual and empirical work.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplemental material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/0960085X.2022.2154712.

Additional information

Notes on contributors

Ilja Nastjuk

Ilja Nastjuk ([email protected]) is a Postdoctoral Researcher at the University of Goettingen, Germany. He earned a Ph.D. in Information Systems from the University of Goettingen in 2017 and a Ph.D. in Accounting and Corporate Governance from Macquarie University in 2018. His research interests span the influence of technology on stress and human behavior, the adoption of self-driving cars, and information security management. His work has been published in numerous peer-reviewed journals and conference proceedings, such as Technological Forecasting and Social Change, Electronic Markets, Computers & Security, Transportation Research Part D: Transport and Environment, Transportation Research Part F: Traffic Psychology, and International Conference on Information Systems

Simon Trang

Simon Trang ([email protected]) is an Assistant Professor and holds the Chair of Information Security and Compliance at the Department of Business Administration, University of Goettingen. He received his Ph.D. in management science, specializing in management information systems, from the University of Goettingen. His work focuses on information security management, privacy, and sustainable IS. His research has been published in outlets such as the Journal of the Association for Information Systems, European Journal of Information Systems, Information Systems Frontiers, and others

Julius-Viktor Grummeck-Braamt

Julius-Viktor Grummeck-Braamt ([email protected]) is a Postgraduate Researcher at the School of Information and Physical Sciences at the University of Newcastle, Australia. He holds one master’s degree in Management and one master’s degree in Information Systems. His research focuses on human-computer interaction, specifically technology-induced stress. His research has been published in international conference proceedings such as the Australian and New Zealand Academy of Management Conference and the Hawaii International Conference on System Sciences.

Marc T. P. Adam

Marc T. P. Adam ([email protected]) is an Associate Professor in Computing and Information Technology at the University of Newcastle, Australia. He received an undergraduate degree in computer science from the University of Applied Sciences Würzburg, Germany, and a PhD in information systems from the Karlsruhe Institute of Technology, Germany. His research investigates the interplay of cognition and affect in human-computer interaction. He is a founding member of the Society for NeuroIS. His research has been published in top international outlets such as Business & Information Systems Engineering, Communications of the Association for Information Systems, IEEE Journal on Biomedical and Health Informatics, IEEE Transactions on Affective Computing, Journal of the Association for Information Systems, Journal of Management Information Systems, and Journal of Retailing.

Monideepa Tarafdar

Monideepa Tarafdar ([email protected]) is Charles J. Dockendorff Professor at Isenberg School of Management, University of Massachusetts Amherst. She has published extensively in the area of technostress and related phenomena. She is Scientific Adviser to a Dublin start-up that designs programs in wellbeing-oriented use of technology. She has been an invited member of the policy sub-group on Digital Skills of the UK Government’s Department of Culture, Media and Sports. She has held appointments as Visiting Scholar at MIT Sloan CISR, Visiting Professor at Indian Institute of Management Calcutta, and Senior Research Fellow at Weizenbaum Internet Institute, Berlin. Her research is/has been funded by the Leverhulme Trust (UK) and the Economic and Social Science Research Council (ESRC-UK), as Principal Investigator of secured funding of over 1.5 million USD. She serves as Senior Editor at Information Systems Research and Information Systems Journal, and as Editorial Review Board member at Journal of MIS, Journal of AIS and Journal of Strategic Information Systems

Notes

1. Technostress also leads to physiological outcomes. However, as these outcomes are unfortunately underrepresented in technostress research, leaving us without the data points necessary for a meta-analysis, we focus on only psychological and behavioural outcomes. We revisit this issue in Section 5.3.

2. To the best of our knowledge, there exists one meta-analysis on technostress, which is available only as an abstract (Gerdiken et al., 2021); its findings are unpublished and not available to the broader research community. Based on the information provided in the abstract, it does not consider the private usage context, as it focuses exclusively on technostress at work. Furthermore, there is no investigation into the effects of individual technostress creators as compared to technostress as an aggregate second-order construct.

3. While the TMS is widely recognised as the main theoretical underpinning the process of technostress, it is important to note that researchers have used a number of different but complementary theories to understand the different stages of the technostress process. These include the person–environment fit model (Edwards et al., 1998; Edwards & Cooper, 2013), the stressor–strain–outcome model (Koeske & Koeske, 1993), and the job demands–resources model (Demerouti et al., 2001). We return to a discussion of this in Section 5.3.

4. Some studies have discussed techno-unreliability as a separate technostress creator, describing situations when technology is unreliable or behaves inconsistently (Adam et al., 2016; Califf et al., 2020; Weinert et al., 2020). Other research has considered such unreliability to be part of techno-overload. For instance, Tarafdar et al. (2011) noted that techno-overload captures situations in which a technology’s inconsistent behaviour (e.g., interruptions) forces the user to attend to it immediately and experience an information overload. We follow the latter in conceptualising unreliability-related stress as part of techno-overload.

5. It should be noted that intentions do not always translate into actual behaviour. This is also well known as the “intention–behaviour gap” (see Sheeran & Webb, 2016, for a detailed discussion).

6. Given its inherent link to job security, techno-insecurity is only a concern in the organisational context, and thus will be excluded from further analysis in the private usage context.

7. We identified 22 studies that focus on physiological outcomes. Of these 22 studies, 10 studies did not meet our criteria for the measurement of technostress creators (e.g., the stress-inducing condition represents a stimulus such as an implementation of technology) (e.g., Korunka et al., 1996). 9 studies did not provide usable empirical data on the relationship between technostress creators and outcomes (e.g., Fischer et al., 2019). Owing to the small number of studies with empirical information useful for a meta-analysis (Hunter & Schmidt, 2004), we decided to drop these articles and omit relationships with physiological outcomes as objects of analysis.

8. In total, we contacted 231 authors via email and ResearchGate to request further articles or missing correlations; the response rate was 21.21% (49 authors). These authors sent us 32 additional studies, 13 of which we integrated (of the other 19 studies, 3 were already included in our database, 6 did not include information useful for deriving correlations, 1 did not include outcomes, and 9 did not conceptualise technostress creators or outcomes as per our definition).

9. We conceptualised outcomes as negative consequences of technostress (e.g., decreased satisfaction and increased turnover intention). In other words, a positive relationship between technostress and outcomes means that technostress increases negative consequences.

10. We coded our database for studies that relied on self-reported and objective measurement instruments for the behavioural outcome variable (all psychological outcomes were measured based on self-reported measurement instruments). We then conducted a subgroup analysis for self-reported vs. objective measurement and estimated effect sizes between aggregated technostress creators and behavioural outcomes. The difference is small (diff _behavioural = .01) and insignificant (behavioural > .05).