How work redesign interventions affect performance: An evidence-based model from a systematic review

Performance

Performance is conceptualized and operationalized differently across diverse theoretical perspectives (Waggoner et al., 1999). In our review, we consider performance at three different levels: the individual, the team and the organization.

At the individual level, we define job performance as individual behaviour that contributes to organizational goals and adds value to organizations (Campbell et al., 1993). We draw on the model of positive individual work behaviors (Griffin et al., 2007), which proposes a three by three matrix combining forms of positive performance that vary according to uncertainty (proficiency, adaptivity and proactivity), and level of performance contributions that vary according to interdependence (individual, team and organization). Carpini et al. (2017) recently showed that 40 performance constructs, covering most of those that exist in the literature, can be mapped onto this model. We also consider individual-level withdrawal behaviours (such as intention to leave and individual work absenteeism) as key performance constructs. Measures may be subjective (e.g. obtained using self-report surveys), or involve researcher observation (e.g. Welsh et al., 1993), other-ratings (e.g. supervisor ratings of employees’ performance) or company data (e.g. absenteeism, accidents).

At the team level, we consider team effectiveness as a key team performance outcome (McGrath, 1964). Team effectiveness outcomes are ‘results and by-products of team activity that are valued by one or more consitituencies’ (Mathieu et al., 2000: 273). ¹ Team performance thus includes aspects such as the quality of team results and the quantity of team results. In the work design literature, team performance indicators include manager ratings of team effectiveness (Hall et al., 2008) and objective measures of team productivity (e.g. Pearson, 1992). In the production and operations literature, productivity may be indicated by the amount of product produced or work completed within a given time frame (e.g. Locke et al., 1976; Orpen, 1979). A further important team performance indicator is team safety, which has been particularly emphasized in human factor research (e.g. Guimaraes et al., 2015; Parenmark et al., 1993).

At the organization level, economists and management accountant theorists have tended to define performance in terms of positive financial outcomes, such as revenue, profit and sales volume (e.g. Welsh et al., 1993). More recently, other indicators have been considered. For example, the balanced scorecard (Kaplan and Norton, 2001) also includes customer outcomes such as customer satisfaction, internal business processes such as turnover and sickness absence, and learning and growth, important indicators of future success.

Work design theories and performance

Work design theories propose that, because of the effect of individual and group work characteristics on motivation, as well as on other mechanisms such as learning and quick response, work design can affect individual and team work performance (Andrei and Parker, 2017).

In terms of intrinsic motivation, job characteristics theory suggests that jobs high in characteristics such as autonomy, task variety and job feedback intrinsically motivate individuals to perform well (Hackman and Oldham, 1975). Such characteristics can also stimulate proactivity through building self-efficacy, promoting broader role orientations and enhancing activated positive affect, which Parker et al. (2010) referred to as ‘can do’, ‘reason to’ and ‘energized to’ forms of proactive motivation. Job characteristics theory has been expanded to incorporate social and relational factors that also drive performance through motivation. Social support can fulfil basic human needs by increasing individuals’ sense of belonging to a team or department, in accordance with self-determination theory (Deci and Ryan, 2000), or promoting psychological empowerment (e.g. Spreitzer, 1995). Need satisfaction and empowerment have both been linked to performance (e.g. Xanthopoulou et al., 2009; Yeatts and Cready, 2007). Broader relational perspectives have recently emphasized the importance of task significance and beneficiary contact for increasing individuals’ perceptions of the positive impact their work has on others, fostering prosocial motivation (Grant and Parker, 2009).

The job demands–resources model (JD-R; Bakker and Demerouti, 2007) builds on these earlier models as well as others, such as the job demands–control model (Karasek, 1979), to suggest that when resources (job characteristics) are high, individuals are able to cope with strain from high workload and cognitive and emotional demands, promoting well-being and subsequent performance.

Work design is also theorized to lead to performance through learning mechanisms. Leach et al. (2003) noted that timely feedback allowed individuals to learn, problem-solve and more efficiently complete tasks in the future, over and above the effect of empowerment alone on job knowledge and subsequent performance. Social interaction can also allow the transfer of knowledge between employees (e.g. Parker et al., 2017b).

The ‘quick-response’ mechanism (Wall et al., 1992) suggests that autonomy allows individuals to deal with problems when they arise rather than deferring them to higher levels, resulting in faster decision making. More directly, individual specialization may increase how fast work tasks are performed (Humphrey et al., 2007).

Most of the above theorized individual-level mechanisms have also been applied to the team level of analysis. For example, sociotechnical systems theory proposes that creating autonomous, self-managing, work groups or teams leads to team empowerment, where groups of individuals can take control and decide together how to carry out work tasks and meet work goals (Goodman et al., 1988). Such team work designs are also expected to improve individual and team performance through motivation, learning and quick responses.

Research suggests the above relationships are contingent on individual and contextual factors. Growth need strength is an important individual factor, and refers to the degree to which individuals desire to be satisfied at work through ‘growing’ (Hackman and Oldham, 1975). Individuals with higher growth need strength are predicted to respond more strongly and positively to more motivating jobs, leading to better outcomes, including performance. A key contextual contingency for performance is uncertainty: high uncertainty increases the need for individuals to take responsibility for work tasks and reactively adapt and problem-solve in response to the changing work environment (Wall et al., 2002). Wall et al. (2002) suggested that under uncertain conditions, quick response and effectiveness is more likely to be realized if timely feedback about performance is present, alongside autonomy. Timely feedback promotes faster learning, allowing individuals to reevaluate work methods, problem-solve and change their methods sooner rather than later, leading to optimal performance (see also Cherns, 1987).

Some evidence supports the above theories. For example, meta-analyses show positive relationships between several individual job characteristics (e.g. job autonomy) and performance (Fried and Ferris, 1987; Humphrey et al., 2007), and Wall et al. (2002) observed a positive link between empowerment, increased uncertainty and performance. Much evidence, however, focuses on correlational studies, with limited attention to intervention studies that help establish causality. Authors of reviews have acknowledged inconsistencies when this broader literature is considered (e.g. Oldham and Fried, 2016; Parker, 2014). In one of the only reviews that has focused exclusively on interventions, Kelly (1992) concluded that, though job redesign led to improved job satisfaction, it was not associated with improved performance. Kelly (1992) suggested a twin-track model in which the drivers of performance (extrinsic motivation, goal setting and more efficient work methods) are different from the drivers of job satisfaction.

A further challenge is that work design theories are concentrated at the individual and team levels, yet individual workers and teams are located within organizational systems. As such, wider factors may shape the effects of work design on performance, such as organizational policies. This highlights the need to take a broader perspective. Next, we discuss how human resource management theory (HRM) provides useful insights into how work design might shape performance.

Human resource management theory (HRM), work design and performance

Scholars have long been interested in the link between human resource work practices, or policies and procedures to develop employee skills, participation and motivation, and performance (Batt, 2002). The early focus of this research was on short-term productivity and efficiency, with a later, longer-term strategic view acknowledging employees as both a major source of human capital and an organizational cost (Beer et al., 2015). This longer-term strategic view stresses the importance of aligning high-performance work practices (HPWP) with each other (horizontal alignment), as well as with overall strategic business goals (vertical alignment), in order to drive performance (Christina et al., 2017; Guest, 1997). High-involvement management (HIM) practices are a type of HPWP that focus on direct employee partipation in decision-making, teamwork and consultation forums (Wood et al., 2012). As such, these practices have particular parallels with work redesigns that increase autonomy. Both HPWP and HIM primarily focus on organization-level outcomes, especially revenue, efficiency and productivity.

Horizontal alignment is particularly pertinent for organization-led work redesign, which can potentially have cross-level effects on team and individual performance (Christina et al., 2017). For example, employment practices such as reward structures and training systems can provide individual workers with work motivation and the necessary skills and abililities to carry out their work, but can also provide teams with the motivation and skills necessary to achieve team goals. If job design is aligned with these practices, such as by enabling participation in decision making and freedom over how to carry out work tasks, individual and team performance is more likely to improve. Despite this theorizing, we were unable to find any HRM intervention studies which have explicitly measured change in job design alongside other HRM changes and performance. For example, a recent randomized intervention in a multinational firm suggested that aligning HRM practices and job design influenced worker behaviour, as well as organizational and environmental outcomes, but job design was not actually measured in this study (Christina et al., 2017).

Summary

In sum, there is established theory to suggest that work design can enhance individual and team performance via various mechanisms, with some suggestion of stronger effects for particular individuals and contexts. There is also strong theory about the importance of aligning work design with other practices as well as overall business goals to achieve organizational performance. Figure 1 summarizes this theory, and indicates the key theoretical mechanisms and moderators that theory suggests are involved in the relationships between work design and individual, team and organizational performance. However, as we have discussed, the empirical evidence base is not clear. As we argue next, by systematically synthesizing findings from intervention studies, we can consolidate our knowledge of whether, why and when work design leads to performance.

OPEN IN VIEWER

Figure 1.

A diagram indicating the relationships between key theoretical perspectives of work design and theorized performance outcomes.

Research questions

We have seen that there is reasonable cross-sectional evidence linking work design with performance, with some suggestion from reviews that findings from intervention studies might be more mixed. Intervention designs offer greater ability to determine causal links, so it is important to consider evidence from such studies more systematically. Thus, investigating whether top-down work redesign interventions enhance performance, as theory suggests, is our first research question (RQ1). To do so, we assess the quality of studies, taking into account how research methods might affect the confidence we can place in the results (e.g. whether studies are randomized, whether a control or comparison group is present, what was measured, type of measures, number of measurement points, and sample size). Scholars regularly call for such considerations when evaluating the effectiveness of interventions (e.g. Briner and Walshe, 2015; Lamontagne et al., 2007; Nielsen and Randall, 2013; Pawson, 2013; Snape et al., 2016). As we discuss later, we assess a range of top work redesign interventions, including individual (e.g. job enrichment), team (e.g. autonomous work groups) and organizational (e.g. system changes) interventions.

We then seek to establish whether changes in work characteristics account for any relationship between top-down interventions and performance (RQ2). Work redesign interventions aim to change the nature and organization of work tasks, activities and responsibilities, which should be reflected in changes in employees’ perceptions of work charcteristics (Andrei and Parker, 2017), but to date it is unclear whether any effects of work redesign on performance can be explained by the theorized change in work characteristics. Showing the role of work characteristics clearly establishes the intervention as a ‘work redesign’ rather than reflecting some other process (e.g. Hawthorne effects).

Third, we identify the mechanisms that link changed work characteristics to performance (RQ3). Although many mechanisms have been theorized (e.g. motivation, learning, quick response), we seek to show which have been demonstrated empirically in intervention studies, as well as the quality of this evidence.

Fourth, beyond identifying whether and how work redesigns affect performance, boundary conditions, or when top-down work redesign interventions are effective, must also be considered (RQ4). We focus on three categories of boundary conditions: intervention implementation effectiveness, intervention context and person factors. Intervention implementation effectiveness includes, for example, whether interventions ran according to plan (e.g. whether all training sessions were conducted), degree of attrition, and resistance from managers and/or employees. Intervention context includes organizational factors such as setting, organizational design (e.g. alignment of pay, feedback, reward, information and communication systems) and organizational climate (e.g. economic and political instability such as redundancies, mergers, unrest over pay, policies or procedures). Person factors that can moderate work design effects include, for example, personality and work preferences. In sum, our research questions are as follows:

Findings for RQ1: Effectiveness of interventions on performance

Thirty-nine studies demonstrated a positive effect on performance, the majority of which used stronger research designs (e.g. randomized or quasi-experimental), with the overall quality of the evidence being considered ‘promising’ (Table 3, supplementary material). These results are consistent with theory and provide considerable evidence that organization-led work redesigns do positively influence performance (Figure 3), with the most promising results at the individual level.

Fourteen studies had mixed effects on performance (Figure 4), 11 of which involved randomized and/or controlled designs. Two studies demonstrated a negative effect. One non-randomized study observed higher turnover and absenteeism amongst employees in autonomous work groups than traditionally designed jobs, although work characteristics improved (Cordery et al., 1991) (Figure 5). The authors speculated that this was a result of factors external to the organization, such as the distance workers had to travel to the workplace, and differences between the intervention and control groups in terms of workload. Another study observed decreased productivity following increased employee responsibility for work scheduling (Powell and Schlacter, 1971), possibly owing to low autonomy needs amongst workers, change being undesired, or the change being insufficient to have the desired effects.

Our results suggest that work redesign interventions are worthy of investment for practitioners wishing to increase performance. We rated the evidence for this statement as promising, owing to the number of studies demonstrating a positive effect on performance, the number of studies with methodologically more rigorous research designs, and the decent sample sizes across most studies. We downgraded our rating from ‘strong evidence’ owing to some inconsistent results, with some studies observing positive effects on performance and others not (Figures 3–5). As we discuss later, it is likely there are boundary conditions that shape the performance effects of various work redesigns. In sum, we conclude:

Findings for RQ2: Mediating role of work characteristics

The answer to the question about whether the effects of work redesign interventions on performance are accounted for by a change in work characteristics is ‘most likely’. The majority of the studies showing positive effects on performance also demonstrated a positive effect on perceptions of work characteristics (k = 22; Figure 6). Only three of these studies tested (and confirmed) a mediation relationship between the intervention, changes in work design, and performance (Bellé, 2014; Holman and Axtell, 2016; Grant, 2008a), preventing stronger causal conclusions.

To unpack this evidence further, we examined the findings for each type of intervention in more detail. With respect to non-participative job enrichment and enlargement interventions, four studies reported a positive effect on work design and performance (Figure 6). Gard et al. (2002) measured individual perceptions of supervisory support and team autonomy, goal clarity, communication and collaboration, group work effectiveness and service quality. He et al. (2014) noted a positive effect on individual job characteristics (but did not clearly specify which) and organizational turnover. Michalos et al. (2013) noted an increase in individual task variety and a decrease in individual error probability, while Morse and Reimer (1956) reported increased individual autonomy alongside increased organizational productivity and decreased clerical costs.

One controlled job enrichment intervention had a mixed effect on work design yet an overall positive effect on manager-rated productivity and colleague-rated innovation (Keller and Holland, 1981). Following an individual’s role change or promotion to a higher job level, role ambiguity decreased and task variety, autonomy and performance at the individual level increased. Job feedback, contact with others, and supervision activities did not improve. The changes in work characteristics were sufficient, however, to improve performance.

One randomized study noted a mixed effect on work characteristics and a mixed effect on performance (Orpen, 1979). While individual perceptions of task variety, task identity and autonomy increased, perceptions of task significance and feedback did not; and while organizational turnover and absenteeism decreased, supervisor and group ratings of productivity did not change. This suggests that different aspects of work design may be differentially affected by a job enrichment intervention, and not all work redesigns should be expected to affect all work characteristics. Increases in autonomy and task variety were noted across most studies that measured these aspects, suggesting that these may be easier to increase via job enrichment than other work characteristics such as task significance or feedback, which might need a more focused intervention (such as a relational work redesign for task significance, or a feedback-oriented intervention for feedback). Although several job enrichment studies did seek to increase these aspects, it is not known how well the interventions were carried out, or whether enough was done to manipulate these specific attributes, and thus employees may not have perceived a change. Given the limitations to the studies, and the mixed effects, we conclude:

With respect to participative job enrichment studies, seven studies had positive effects on work design and performance (Figure 6). One intervention reported a mixed effect on individual perceptions of skill variety, task significance, job feedback and feedback from others, and a positive effect on organizational turnover (Griffeth, 1985; Figure 8). This study tested whether participation in job redesign was a contextual moderator that changed employee work design perceptions. A strong design with four groups was employed (participation and redesign; participation and no redesign; no participation and redesign; and no participation and no redesign). However, participation in brainstorming redesign ideas did not enhance the effects of enriched work, despite the fact that some of the ideas were used. The author suggests this was owing to the short nature of the participation (90 minutes), and the fact that actual job changes were limited.

In contrast, Holman and Axtell (2016) employed a more significant type of participation via scenario planning. Over two days, call centre employees participated in brainstorming and implementing ideas and solutions to problems with their current work design. Consistent with theory, statistical mediation ascertained that individually perceived job control and feedback mediated between the intervention and supervisor-rated job performance as well as individual outcomes such as employee well-being and psychological contract fulfilment. The success of this intervention demonstrates that participative job enrichment interventions have most positive effects when employees are able to significantly contribute throughout the redesign process, and when the ways in which the jobs are changed align with the aspects of work design targeted for change. This concurs with previous findings around participation (e.g. Daniels et al., 2017).

We did not upgrade our rating of the strength of the evidence owing to implementation issues in many studies (k = 9). For example, in one study, one department continued to operate in an authoritarian manner as opposed to changing to meet the planned redesign. Several studies experienced management resistance impacting the extent of implementation (e.g. Guimaraes et al., 2012, 2015; Jayawardana and O’Donnell, 2009; Wall et al., 1990), and Pearson’s (1992) study was characterized by economic volatility, budget cuts and a decrease in the amount of work available. These issues make it difficult to determine cause and effect. However, Holman and Axtell’s (2016) study gained manager confirmation that the study had been fully implemented. Coupled with the results of their mediation analyses, this study suggests that well-designed and effectively implemented top-down work redesign interventions can improve performance.

For relational interventions, five studies demonstrated a positive impact on both work design and performance (Figure 6). Bellé (2014) reported that individual productivity was higher when employees were exposed to a transformational leader as well as beneficiaries of their work, or a self-persuasion condition (in which employees persuaded themselves why their work was worthwhile). This effect was mediated by pro-social impact, the ‘degree to which employees feel that their actions benefit other people’ (Grant, 2008a: 110). Whereas traditional work design models consider that increasing task significance increases the meaningfulness of individuals’ jobs and thereby impacts performance, contemporary theories suggest that pro-social impact is an alternative mechanism (Bellé, 2014; Grant, 2008a). Grant (2008a) conducted three experiments which demonstrated that increased individual task significance positively impacted performance in terms of amount of money raised (Grant, 2008a: studies 1 and 2) and helping behaviour (2008a: study 3). In one of these studies, Grant found hypothesized mediation relationships between perceived social impact and social worth (the ‘degree to which employees feel that their contributions are valued by other people’; Grant 2008a: 110). Despite small sample sizes, Grant demonstrated effects across several occupations.

Showing a very different type of relational intervention, Parker et al.’s (2013) study found that increasing structural support in the presence of individually perceived role clarity increased individual job performance (proactivity) and lowered perceptions of role overload. This was the only intervention study which manipulated changes in actual support (in this case, a new nursing support role), which is a stronger design than focusing on perceived support.

Taken together, these five studies indicate robust results in the context of ‘helping’ industries, and suggest the importance of considering the relational aspects of job design in these contexts. Studies show the more directly connected an individual is to the significance or impact of his or her work on other people, the more likely that positive effects on performance will be observed owing to increased awareness of prosocial impact and sense of social worth. Grant (2008a) found that those displaying low levels of the personality trait conscientiousness, and strong prosocial values, were most responsive to an intervention in which fundraisers read positive stories from beneficiaries of the money raised. Grant hypothesized that this type of person is more likely to be motivated by external cues rather than an internal drive to work hard, and have life values congruent with the importance of positively impacting others. For these people, fulfilling their values of benefiting others through their work is likely to be particularly important, and thus a positive effect on performance is more likely to be observed following an intervention to increase perceived social impact and worth. Bellé (2014) showed that these effects of relational work redesign are enhanced by a transformational leader who is motivating and inspiring, and also highlighted the importance of pro-social motivation for the success of work design interventions to increase performance in the public sector. Overall, though, the relatively small number of studies, conducted by a rather small set of researchers, shows the relative dominance of research related to task characteristics rather than social characteristics (Hackman and Oldham, 1975).

With respect to the creation of autonomous work groups, three studies observed positive effects on both work design and performance (Figure 6). Cordery et al.’s (2010) study showed that team task uncertainty (in this case, the extent to which biological, chemical or mechanical methods formed part of water treatment) moderated the relationship between the intervention and team performance (Cordery et al., 2010). Allowing teams to make decisions to deal with this uncertainty buffered the impact of task uncertainty and increased performance. This highlights the importance of context for the success of interventions. The impact of context was also apparent in Pearson’s (1992) study, which found an increase in both work design (individually perceived decision making, job scope, role clarity) and performance (team productivity and safety) amongst autonomous teams compared to employees in traditionally organized jobs. This was in spite of economic volatility and organizational turbulence, including restructuring, budget restrictions, deregulation of the transport sector, increased competition, manager resistance and attrition.

Cordery et al. (1991) cited more favourable individual work attitudes yet increased turnover and absenteeism in autonomous work groups. Nevertheless, Cordery et al.’s (1991) results support Wall et al.’s (1986) findings, and also indicate the importance of factors beyond work design for the success of the intervention. For example, workers at one site had to travel further than workers at another site, had higher workloads (which may have contributed to absenteeism), and expressed dissatisfaction with progress in negotiating work design agreements with managers. Morgeson et al. (2006) also noted the influence of wider organizational factors. They reported increased autonomy in the intervention group but observed improvements in performance only when organizational reward, feedback and information systems were poor. They argued that creating autonomous work groups counteracted the effect of poor organizational systems on employees in traditionally designed jobs by improving work motivation. Other autonomous work group interventions also observed mixed results (e.g. Pasmore and King, 1978; Yeatts and Cready, 2007). Overall, the limited number of studies, mixed effects, and methodologically less rigorous research designs, led us to develop the following evidence statement:

With respect to system-wide changes, interventions that additionally involved strategies such as job enrichment and enlargement reported a positive effect on work characteristics (when measured), with three also reporting improved performance. Workman and Bommer’s (2004) randomized, controlled study found that high-involvement work processes (increasing employee perceived interdependence, knowledge sharing and participation via team-oriented structures and rewards, and structural alignment) had the biggest impact on performance (e.g customer service), compared with other intervention groups (an autonomous work group and a job design alignment).

Tregaskis et al. (2013) found that the implementation of high-performance work practices resulted in sustained increases in organizational productivity and safety performance, although they observed a mixed effect on individual work design. While individual perceptions of autonomy, skill development and reward increased, perceptions of workload and pressure also increased. Improvements in work characteristics may have mitigated the potential negative effects of increased demands, and contributed to improved outcomes. This interaction effect has been observed before in relation to well-being outcomes, with one study finding that worker well-being was maintained despite increased demands in a company undergoing downsizing owing to improved work characteristics (Parker et al., 1997). Smith et al. (2012) found that lean techniques, culture change and changes in work processes improved organizational pathology patient safety. Further, Rickard et al. (2012) noted positive effects on individually perceived work design, with improvements in adaptability and communication, and a reduction in job demands. Turnover decreased in only one of the two hospitals, despite reductions in stress in both hospitals, the reason for which is unclear but could be an artefact of the non-randomized and uncontrolled nature of the intervention. Gallegos and Phelan (1977) found that increasing employee responsibility alongside redesigning the feedback and reward system improved organizational productivity. Together, these five studies suggest that aligning system-wide changes with specific work design changes, such as enrichment and enlargement, is likely to be more effective than targeting system changes alone. This supports the findings of other studies that focused on work design changes rather than system-wide changes (e.g. Holman and Axtell, 2016).

In contrast, Narayanan and Nath’s (1982) randomized study described a positive effect of flexitime on individual autonomy and colleague/employee–manager relations, but no change in manager perceptions of work group productivity measured 3 months after the intervention, yet a decrease in absenteeism. It is possible that insufficient time was allowed for the effects of the intervention to be observed, or that individuals may have perceived increased performance, but this was not observed by managers. Welsh et al. (1993) observed increased individual job feedback, but no change in any other work design characteristics following the introduction of the shorter, full pay, voluntary street vending shift for bookstore employees. A positive effect on organizational sales volume and effectiveness was also observed. This could be owing to the Hawthorne Effect, the small sample size, or it may be that specific work design changes are needed alongside work practice change for perceived work characteristics to change.

Findings for RQ3: Mechanisms underlying interventions

Three key mechanisms emerged that explained why perceived changes in work characteristics might lead to changes in performance: (i) intrinsic work motivation; (ii) quick response; and (iii) employee learning. Studies did not always statistically test these mechanisms, and hence some evidence is circumstantial. Intrinsic work motivation was hypothesized to improve in accordance with classic job design theory (Hackman and Oldham, 1975), and this perspective was predominantly adopted by researchers carrying out participative and non-participative job enrichment and enlargement interventions. Relational interventions focused on prosocial work motivation. The social embeddedness of jobs, roles and tasks was taken into account, with an emphasis on influencing employees’ perceived social impact on beneficiaries and their social worth (e.g. Grant, 2008a). Quick-response and learning mechanisms through which work design impacts performance are both illustrated by Wall et al.’s (1992) study. These researchers described how, when operators on a production line were given the autonomy and responsibility to resolve faults on the operating line, there was a quicker response to issues that decreased downtime. Importantly, these operators worked under conditions of high technological uncertainty, lending support to the theory that the impact of increased autonomy on performance is contingent on the level of uncertainty. In addition, operators were also required to gain specialist knowledge which improved their understanding of how the faults arose and how to fix them, increasing their learning. Similarly, Griffin (1991) recounted the implementation of a computer system which automated error messages, providing faster responses from workers and enabling self-monitoring of sales performance. This study particularly highlights the benefit of timely feedback, which was supported by Leach et al.’s (2003) study. The opportunity to learn on the job commonly co-occurs with increased skill and task variety, job complexity, and social interactions which foster knowledge about interdependencies and promote technical knowledge. As well as improving quick response, this can improve motivation (Hackman and Oldham, 1975). Considering the evidence as a whole, we formulated the following statement:

Findings for RQ4: Boundary conditions impacting intervention effectiveness

Some types of intervention are most likely to be suited to certain contexts. For example, when operator autonomy was increased, Wall et al. (1990) reported increased performance, particularly for high-variance systems, with a smaller increase in performance for low-variance systems. This suggests a moderating effect of uncertainty, with a stronger effect on performance under high-uncertainty conditions. In high-uncertainty conditions, greater autonomy allowed operators to quickly respond to problems when they arose, decreasing the downtime that would have been inevitable if operators had to wait for specialists to respond. Similarly, Cordery et al. (2010) observed that task uncertainty (not knowing which operational problems will occur when) positively moderated the impact of increased autonomy on performance in the context of teams. This was proposed to be owing to non-routine decision-making abilities enabled in teams given greater autonomy, which allowed them to respond more quickly to unexpected events. Quick-response mechanisms were thus important in both Wall et al.’s (1990) and Cordery et al.’s (2010) studies.

Further, our results suggest that job enrichment and enlargement interventions have differential effects on performance under different conditions. Enrichment strategies to increase autonomy may be particularly effective when uncertainty is high owing to quick response and learning. However, enlargement strategies that increase the number or variety of tasks are unlikely to be moderated by uncertainty as they do not facilitate these mechanisms. These assertions are supported by the wider literature; Wall et al. (2002) theorized that empowering employees through job enrichment strategies which increase decision latitude and responsibilities promotes performance under conditions of high uncertainty, and that not empowering employees under conditions of high uncertainty causes performance losses. This study was not empirical, however, and while the first assertion has been borne out empirically, the second has not been shown: we found that increasing autonomy under conditions of uncertainty improves performance, but it was not clear whether a negative impact on performance occurs under high uncertainty if autonomy is not increased.

To conclude, Leach et al. (2013) stated that work redesign may fail to have desired effects on performance if uncertainty is not considered when strategies to increase control are introduced. Together, the evidence suggests that under high-uncertainty conditions, job-enrichment strategies which increase autonomy have differential effects on performance compared to job enlargement strategies. This is an important point, with practical implications. For example, organizations wishing to increase performance under conditions of high uncertainty would be better off increasing worker autonomy than adopting other strategies such as job enlargement. It is possible that increasing autonomy under conditions of low uncertainty is beneficial owing to motivational benefits (Hackman and Oldham, 1975), although the effects may be weaker. Further work is needed to test this assertion. Given our findings and the supporting literature, the following evidence statement was developed:

In terms of other contexts, all of the relational interventions designed to increase task significance were conducted with employees who worked with people in the community (e.g. fundraisers). It is not known whether relational interventions would be equally as effective for increasing performance in settings where beneficiaries are distal, or where beneficiary contact is already high. With respect to the former, for example, in a car factory, the effect of a production employee’s job on the beneficiary (car owner) is distal, and it is not clear whether contact is important in such contexts. On the other hand, in settings where employees are frequently in contact with beneficiaries, such as in a hospital, it may not be advisable to attempt to increase performance by increasing beneficiary contact. Doctors and nurses, for example, typically report high emotional demands owing to regularly dealing with chronically ill patients and their carers. In these circumstances, increasing beneficiary contact may exacerbate negative outcomes such as burnout, particularly if workload is high and organizational systems are not aligned to offer support (Grant and Parker, 2009). Interventions are likely to be most successful if they are designed with context in mind, so that they are appropriate for the setting.

In addition, across contexts, we noted that alignment of organizational design elements (such as feedback, pay and reward systems, and information and communication systems) appears important for the success of top-down work design interventions, reinforcing the findings of previous work design research (e.g. Daniels et al., 2017; Goodman et al., 1988; Grant and Parker, 2009). Without such alignment, changes to work design are unlikely to be impactful and sustainable. For example, changing from a traditional work design to a highly interdependent, autonomous work group typically involves changes in goals, task allocation and problem-solving strategies. This necessitates aligned pay and information and communication systems that may not have been needed previously (Goodman et al., 1988), and resonates with the HRM literature around horizontal and vertical alignment (Beer et al., 2015; Christina et al., 2017). Consistent with Daniels et al.’s (2017) review, we have also seen how organizational employment practices such as flexitime (e.g. Narayanan and Nath, 1982) and performance mangagement systems (e.g. Workman and Bommer, 2004) might positively affect work design and performance. Currenting an anomaly to this evidence, Morgeson et al.’s (2006) results suggest a more nuanced relationship, given their findings that creating autonomous work groups improved performance when organizational systems were ineffective but not when they were effective.

Equally, organization-wide work design changes need strong manager support to be fully implemented and effective, further supporting the alignment of work practices with job design. For example, an empowerment intervention is likely to fail if managers do not actively devolve autonomy to employees, encourage employee voice and participation, and act on employee suggestions and opinions (Danford et al., 2004). This supports the use of high-involvement strategies promoted by HRM (Wood et al., 2012). Employees may develop mistrust in management if intended changes are not realized, or changes are perceived to be geared towards organizational goals (e.g. profit) at the expense of employee well-being (Danford et al., 2004). For example, organizations themselves may have motives for work redesign changes, such as downsizing, merging, or trying out new ways of working which may be perceived negatively by employees. Management therefore has a key role to play in gaining employee motivation and buy-in to interventions. This could include reassuring employees that managers are committed to the change as well as educating and involving managers in work redesign changes. Amongst our studies, Pearson (1992) reported management recalcitrance that limited success, Guimaraes et al. (2012) reported that some managers were inflexible towards worker opinions, and Guimaraes et al. (2015) reported resistance from immediate supervisors. Jayawardana and O’Donnell (2009) noted that one production line retained an authoritarian style as opposed to devolving responsibility. Based on our evidence, we developed the following evidence statement:

Individual-level boundary conditions were also identified in our studies. Grant (2008a) observed that those low in conscientiousness or with strong prosocial values were most responsive to a task significance intervention. Parker et al. (2013) observed that where employees had higher clarity about others’ work roles, structural support had a greater effect on performance. Furthermore, they found that where negative work affect was lower, structural support was associated with higher perceived skill utilization and proactivity, whereas where negative work affect was higher, structural support reduced perceptions of role overload. Workman and Bommer (2004) found that those with higher group work preferences achieved greater satisfaction following group-orientated interventions (e.g. the creation of autonomous work groups) than those with lower group work preferences. Although limited, this initial research strongly suggests that there are important individual-level boundary conditions that need to be considered by job redesign interventions. Implications could include aligning selection and recruitment systems with job design, so that person–job fit is improved, or allowing managers the autonomy to craft jobs to better fit individual traits and strengths. Alternatively, training and development opportunities could allow individuals to develop personal aspects such as self-efficacy, increasing person–job fit. Based on our evidence, we suggest:

Limitations and future directions

The strength of our research lies in our systematic methods, including extensive searching and double-coding of a range of research, across varying research designs, variables and contexts. Despite our inclusiveness, we retained the quality of our review by focusing on longitudinal intervention studies conducted in the field, thereby promoting ecological validity and causal conclusions. We do, however, acknowledge a number of limitations. Some interventions could have been classified into several categories. Double-coding of our results ensured consistency. Furthermore, space constraints limited discussion of all observed relationships. We also noted that most of the included studies were conducted in western contexts, therefore the results of this review may not be applicable to other contexts. Finally, we acknowledge the limitations of two-dimensional diagrams for capturing the complexity of our findings.

We propose five key directions for future research. First, rather moderate evidence was found for the impact of work redesigns at the team and organizational level. Team- and organization-level research that statistically assesses the relationship between work design and performance, as well as mediation relationships, is needed. For example, investigating the impact of employment practices on work design and performance could contribute to developing evidence-based policy that protects workers whilst also maintaining high performance, enabling organizations to remain competitive.

Second, cross-level effects are not routinely investigated in work design intervention research. We recognize that it may be difficult to collect data at different levels in organizations, with some studies citing trust issues preventing data collection (e.g. MacDonald and Bodzak, 1999) and others being unable to match data to participants (Cordery et al., 1991; Morgeson et al., 2006). Nevertheless, exploring cross-level relationships would help researchers design interventions that evoke intended effects between, as well as within, individuals, teams and departments.

Third, few studies have explored boundary conditions beyond the role of uncertainty. Those which did (e.g. Grant, 2008a; Morgeson et al., 2006; Parker et al., 2013) suggested the importance of person factors, as well as intervention context and implementation. Even within a broad industry, different interventions appeared more appropriate in different contexts (e.g. relational interventions involving beneficiary contact in contexts where such contact is initially low). Research exploring boundary conditions could helping target interventions more appropriately.

Fourth, few interventions measured job demands; those which did focused on role overload (e.g. Parker et al., 2013), role ambiguity and role conflict (e.g. Weigl et al., 2013). Increasingly, jobs are characterized by a variety of different demands that employees must manage effectively in order to perform at a high level. These include job insecurity, technological demands such as interruptions from emails and downtime owing to technological failure (Braukmann et al., 2018). The job demands–resources model suggests that resources (job characteristcis) can buffer the impact of demands (Bakker and Demerouti, 2007). Future research could investigate the interaction between resources and a variety of contemporary job demands in intervention contexts. This echoes an earlier call to extend work characteristics in future work (Parker et al., 2017a).