The Curvilinear Relation between Work Predictability and Creativity

ABSTRACT

In this study, we investigate how the level of work control predictability affects employee creativity. Specifically, we examine whether supervisor and coworker support moderate the predictability-creativity relationship. We use survey data from 128 employee–supervisor dyads from a governmental organization in Belgium. Multilevel analyses demonstrated a significant inverted U-curve relationship between predictability and leader-rated creativity, indicating that too little or too much work control predictability contributes less to employee creativity compared to moderate levels of work control predictability. Moreover, supervisor and coworker support moderate this relationship in such a way that for employees who receive low support the inverted U-curve relationship is significantly stronger compared to employees that receive high support. This is the first study that focuses on work control predictability in relation to creative performance.

Introduction

Employee creativity is an important source of organizational innovation, as organizations need novel and useful ideas to develop new products and services, and solve organizational problems (Amabile, 1996; Oldham & Cummings, 1996). Hence, organizations are increasingly interested in understanding the organizational context that is most conducive to employees’ creative performance (Hirst, Van Knippenberg, & Zhou, 2009; Windrum & Koch, 2008). In this respect, several studies have investigated the effect of work control on employee creativity (e.g. Chua & Iyengar, 2006). Work control is broadly conceptualized as the decision authority of employees to determine how a task should be carried out and which set of skills should be used (de Jong, Vegchel, Shimazu, Schaufeli, & Dormann, 2010; Karasek, 1979). Work control consists of several dimensions, including scheduling, method, and pacing control (Jackson, Wall, Martin, & Davids, 1993). Relatively little attention has been paid to the effects of work control predictability.

Work control predictability (WCP) can be seen as an umbrella concept that encompasses employees’ perceptions about work control as well as the extent to which the nature, onset and outcome of work events is known beforehand (Daniels & Guppy, 1994; Dywer & Ganster, 1991; Smith, Tisak, Hahn, & Schmieder, 1997). In short, WCP refers to a perception that work is driven by a confluence of externally imposed constraints. It refers to questions such as “Are daily work goals clear?” and “Can difficulties or disturbances to the daily tasks be foreseen?” To say that daily events are predictable implies only that employees can know something about the events that will occur during the day, apart from the question whether or not the employee can do anything to change it. Work control suggests that events are predictable (when control is exercised), since responding with controlling activities predicts the event’s occurrence and/or nature. However, a predictable event may or may not be controllable. By itself, information about the event (predictability) does not enable the employee to influence it (controllability), implying that work control and work control predictability are distinct concepts with potentially different effects on employee outcomes.

Investigating work control predictability in relation to creative performance is particularly relevant, because on basis of studies that relate work control to creativity, one may expect that predictability will decrease creativity in the same way as work control has been shown to decrease creativity (e.g. Du, Li, & Zhang, 2018; Joo, Yang, & McLean, 2014). However, it may be possible that some level of predictability of difficulties and disturbances to one’s work may actually create room for creative performance. Studies about job complexity (Shalley, Zhou, & Oldham, 2004) have shown that moderately complex jobs fuel motivation and creativity. These findings suggest that less predictable work may increase creativity because of the increased job complexity that comes along with less predictability. In contrast, both high and low levels of work control predictability may also have a straining effect, limiting creative performance. In line with studies examining enablers and inhibitors to creativity, which suggest a curvilinear relationship between constraints and creativity (Acar, Tarakci, & van Knippenberg, 2019; Ohly, Sonnentag, & Pluntke, 2006), we therefore suggest that there is an optimum level of work control predictability for creativity. Specifically, we argue that too little or too much job predictability could negatively associate with creativity.

The aim of the present study is to address the tension between work control predictability and creativity by proposing that there is an optimum level of work control predictability that is most beneficial to creativity. Hereby we integrate research on nonlinear associations between job design and critical employee outcomes (e.g. Chung-Yan, 2010), and the literature on determinants of employee creativity (e.g., Kremer, Villamor, & Aguinis, 2019). We answer to the call for studies to investigate nonlinear effects between organizational constraints and creativity (Acar et al., 2019). Furthermore, we propose that the relationship between work control predictability and creativity could be better understood if we pay close attention to the role of coworker and supervisor support. Recent studies indicate that work control and social support enables employees to engage in problem solving, and enhances job performance (e.g. Daniels, Glover, Beesley, Wimalasiri, & Cohen et al., 2013; Gordon, Demerouti, Bipp, & Le Blanc, 2015). With our study, we address the need for empirical evidence of how moderators, such as social support, may influence a curvilinear relationship between organizational constraints and creativity (Acar et al., 2019). Thereby, we contribute to studies that investigate the effects of specific job designs and a supportive work environment on employee performance (Janssen, 2001; Leung, Huang, & Lu, 2011) as well as to studies theorizing about creativity under constraints (Acar et al., 2019; Rosso, 2014).

Literature review and hypotheses

Job characteristics and creativity

Prior research about creative performance has increased our understanding about the relationship between work characteristics and employee creativity (Zhou & Hoever, 2014). A common theme in these studies is that not only workers’ creative ideas are needed, but also complementary organizational mechanisms have to be in place to facilitate implementation of these ideas (Woodman, Sawyer, & Griffin, 1993). Studies in this line have focused on interactions of various work characteristics with organizational support structures in relationship to creativity, such as time pressure (Baer & Oldham, 2006; Ohly et al., 2006), work control (Ohly et al., 2006), role stress (Leung et al., 2011), and job demands in general (under perceived fairness) (Janssen, 2001). Several types of relationships between work characteristics and creativity have been explored, including linear (Ohly et al., 2006) and curvilinear (Baer & Oldham, 2006; Janssen, 2001; Leung et al., 2011; Ohly et al., 2006) ones. All these studies have in common that they investigate job characteristics that somehow limit the work space of employees (e.g. time pressure, job demands, limited work control) in relation to creative or innovative output (e.g. innovative work behavior, new product innovation), and are moderated by some form of organizational support (e.g., perceived support for creativity).

The effect of job characteristics on individual creativity emerges from the motivation that follows from job characteristics. Job characteristics can enhance “individuals’ excitement about their work activities and their interest in completing these activities, and this excitement should foster creativity” (Shalley et al., 2004, p. 938). This reasoning is in line with the Job-Demands-Resources (JDR) model (Schaufeli & Bakker, 2004), which emphasizes a motivational potential as well as a straining potential of job resources. A job characteristic such as the level of work control predictability can act as a personal resource to employees, which can have a motivational effect, but also a straining effect. While previous research has expanded our knowledge about certain work characteristics in relation to creativity, it has not fully explained how work control predictability is related to individual-level creative performance, and further investigation is therefore warranted.

Work control predictability and creativity

Work control predictability pertains the clarity of daily work goals, the ability to foresee difficulties or disturbances that may affect one’s work tasks in the foreseeable future, work awareness (i.e. knowledge on how one’s decisions will impact the work of others), and the ability to determine the pace in which objectives are reached (Dwyer & Ganster, 1991; Smith et al., 1997). A job with a high level of work control predictability is characterized by clear daily work goals. In such a job an employee can determine the steps needed to fulfill a task and the pace in which daily work goals are reached. Difficulties or disturbances that affect one’s work tasks can easily be predicted and anticipatory actions can be taken. It is transparent how one’s decisions will impact the work of others. In contrast, a job with a low level of work control predictability is characterized by unforeseen events happening regularly. Prescheduled tasks cannot be finished as inputs are delayed or suddenly different work goals are prioritized by management. Organizational communication may be ambiguous and it may be difficult for employees to determine what exactly is expected from them.

Prior research has mainly focused on the effect of work control on overall employee performance (e.g. Gordon et al., 2015; Parker, Jimmieson, & Amiot, 2013). There is a lack of research that links work control predictability to employee creative performance. Drawing on Job Demands-Resources theory (Bakker, Demerouti, De Boer, & Schaufeli, 2003; Daniels et al., 2013; Gordon et al., 2015), we pose that the level of work control predictability can act as a personal job resource to employees, which can have a motivational effect as well as a straining effect with respect to creative performance.

Creativity typically needs tinker time. The creativity literature endorses organizational practices such as providing autonomy, enhancing empowerment, and defining broad work goals (e.g. Chang, Huang, & Choi, 2012; Foote, 2004). These practices ensure that employees have organizational backing in terms of time and resources, to explore, develop and test ideas. If work structures do not allow for trial and error, intrinsic motivation for creative activities decreases and individual employee creativity will be stifled (Amabile, 1996; Zhou, 2003). Following this line of reasoning one may expect that low (i.e., not too high) work control predictability will provide freedom and autonomy to employees, and hence will be beneficial for employee creativity. This reasoning aligns with the motivational effect in Job Demands-Resources theory (Bakker et al., 2003; Daniels et al., 2013; Gordon et al., 2015), which suggests that job resources, such as job control predictability, foster employee engagement and extra-role performance.

However, empirical evidence from health studies demonstrates that extremely low work control predictability tends to increase anxiety and stress (Noblet & Rodwell, 2009; Väänänen et al., 2008), which is detrimental to intrinsic motivation toward work tasks (Parker et al., 2013) and creativity (Baer & Oldham, 2006). Johns (2010) indicates that an unintended consequence of jobs with low structure and high work autonomy is emotional and mental exhaustion. A multitude of options for further activities can create stress and insecurity, is negative for well-being and health (Schaubroeck, Ganster, & Kemmerer, 2004; Schwartz, 2004), and dampens creativity (Chua & Iyengar, 2006). In terms of the Jobs Demands-Resources model, too low WCP provides a straining effect on employees as it generates psychological distress and insecurity, and hence it will be detrimental to employee creativity.

Too low work control predictability is not beneficial for creativity, but neither is too high predictability. Extremely high levels of work control predictability tend to go hand in hand with high external control and pressure, which reduces the interest in the task itself and inhibits the opportunity for employees to experiment at the workplace (Amabile, 1996; Parker et al., 2013; Zhou, 1998). Furthermore, highly predictable jobs typically consist of monotonous activities which decrease job satisfaction, increase psychological distress and absenteeism (Melamed et al., 1995; Schneider et al., 2011). In terms of JD-R, too high levels of WCP have a straining effect on employees. Moreover, highly predictable jobs may imply that there is no need to be creative (Finke, Ward, & Smith, 1992). Difficulties are predictable and solutions are readily available. Furthermore, daily output expectations are crystal clear, and there is no reason to engage in extra role behavior and be creative.

The combination of these two effects (motivational and straining) is likely to result in a non-linear relationship between work control predictability and creative performance. Specifically, we argue that too low work control predictability is not beneficial for creativity, as it increases insecurity about tasks and goals; but neither is too high work control predictability, as it stifles room for personal approaches to work. We propose that a job characterized by a moderate level of work control predictability enables employees to experiment and develop their own creative ideas (Amabile, 1996; Binnewies, 2008; Chang et al., 2012), without imposing too many, nor too few constraints. This intermediate, fitting level of structure and guidance prevents employees from being distracted by side issues (e.g. Chang et al., 2012; Foote, 2004). The provided clarity about what needs to be done and about what is expected from employees in terms of output leaves employees with slack time in which creativity can bloom. Hence, moderate (intermediate) levels of work control predictability are expected to be beneficial for creativity.

To further substantiate the expectation that moderate levels of work control predictability are superior in stimulating creativity to either low or high levels, we draw on studies in the field of organizational behavior (for an overview, see Acar et al., 2019) which have argued for, and found evidence of, an inverted U-shaped relationship between time pressure and creativity (e.g. Baer & Oldham, 2006; Ohly et al., 2006). Also studies investigating the impact of resource constraints on team creativity (Rosso, 2014) and creative problem solving (Meideros, Partlow, & Mumford, 2014; Biskjaer et al., 2020) find evidence for the existence of a “sweet spot,” in which moderate levels of constraints foster creativity, while too many or too few constraints obstruct creativity. In line with these studies, we argue that moderate levels of work control predictability generate the ideal circumstances for creativity to flourish, as employees enjoy clarity about tasks and work goals, as well as output expectations, while leaving sufficient room for exploring personal approaches.

Taken together, we hypothesize an inverted U-curve relationship between perceived work control predictability and employee creative performance, where moderate (intermediate) levels of work control predictability are most beneficial for creative performance. Hence,

H1. There is an inverted U-curve relationship between perceived work control predictability and employee creative performance

The moderating effect of social support

Over the last years, teamwork has become commonplace in organizations (Mathieu, Maynard, Rapp, & Gilson, 2008). Individual employees are mutually dependent when carrying out their activities and solving problems. This way of working requires interpersonal interaction, coordination and communication, i.e. supportive relationships with coworkers and supervisors at work (Bertucci, Conte, Johnson, & Johnson, 2010). Hence, individual creativity is fostered in a context of supportive interactions with coworkers and managers (Madjar, Oldham, & Pratt, 2002; Perry-Smith, 2006; Uzzi & Spiro, 2005; Cai, Lysova, Bossink, Khapova, & Wang, 2019). Support refers to the extent in which supervisors and coworkers provide input or information and encourage employees to develop and refine ideas (Madjar et al., 2002). We propose that the relationship between work control predictability and creativity may be influenced by differences in support structures, as support may influence employees’ ways to act upon different levels of work control predictability (Mohr & Wolfram, 2010). We argue that support will not only shift the inverted U-curve to a higher creativity level. We expect that the impact of support also changes the shape of the curve in the extreme ends. Specifically, we argue that the inverted U-curve relationship between perceived work control predictability and employee creative performance only holds for situations in which employees experience low supervisor and coworker support. When employees experience high support, the inverted U-curve is elevated and flattened out to some extent, see Figure 1.

Figure 1. Illustration of the hypothesized moderation effect.

Our argumentation draws on Job Demands-Resources theory (Bakker et al., 2003; Daniels et al., 2013; Gordon et al., 2015). We suggest that support can be conceived of as a job resource that moderates the relationship between work control predictability and creative performance by dampening the straining effect of work control predictability. In other words, in situations of high social support, negative effects of extreme levels of work control predictability on creativity are expected to be less detrimental to creativity as the coworker and supervisor support will take over and facilitate creativity. We justify this expectation as follows.

While situations that harbor moderate levels of work control predictability are expected to encourage creativity, situations with extremely high work control predictability limit freedom and autonomy of employees, and decrease employee creativity. We argue that supervisor and coworker support alleviates this negative association. Several empirical studies demonstrated that supervisor and coworker support increase employee motivation and mutual interaction, which in turn enhances problem solving capabilities and creative behavior (Amabile, 1996; Madjar et al., 2002; Rehman, Ahmad, Allen, Razig, & Riaz, 2019; Zhou, 2003). These empirical studies build on affect theory of social exchange (Lawler, 2001), which poses that supportive behavior of group members instigates interaction, cohesion, and feelings of accomplishment of group tasks. A social support structure may increase employees’ commitment to the group (Bishop, Scott, & Burroughs, 2000; Rehman et al., 2019), despite the monotony of predictable tasks. Coworker support has been found to boost employees’ self-efficacy and instigates positive assessments of the impact of their work (Bishop et al., 2000; Rehman et al., 2019), thereby relieving strain stemming from wearisomeness. Social interactions with coworkers may show how tasks interconnect and where there are possibilities for improvements, even if only in terms of processes and procedures (Perry-Smith, 2006; Uzzi & Spiro, 2005). Social support from supervisors may alleviate some of the rules and constraints that are of particular concern to an individual employee. Therefore, it is likely that when employees experience extremely high work control predictability, creativity will be higher for employees who receive supervisor and coworker support as compared to employees who are not supported. The negative effect of perceiving high work control predictability will be alleviated in the presence of social support structures. Hence, we argue that in the presence of support possible negative associations between (extremely high) work control predictability and creative performance are mitigated.

Similarly, extremely low work control predictability generates psychological distress and insecurity for employees, and decreases employee creativity. We argue that support alleviates the strain. In jobs with low work control predictability, it is unclear how one has to handle ever different disturbances and problems. Attempts to create something new often fail to bear fruit, which may result in frustration and uncertainty (Zhou & George, 2003). Supervisors and coworkers may provide emotional support to deal with setbacks (Madjar et al., 2002). Employees in jobs with low work predictability may even experience support as necessary to persist in their creative efforts. As such the negative association between work control predictability and creative performance, caused by the uncertainty and lack of structure that is associated with low work control predictability, will be alleviated by support from supervisors and coworkers. Furthermore, support from experienced coworkers or supervisors can provide ideas for the direction in which a solution can be found and may provide new inputs to the problem solving process of individual employees. Support generates ideas for possible solutions and further explorations, thereby enhancing employees’ creative output (Madjar et al., 2002; Zhou, 2003). Hence, we argue that in the presence of support possible negative associations between (extremely low) work control predictability and creative performance are mitigated.

As a result, it is likely that the inverted U-curve relationship between work control predictability and employee creative performance is elevated and flattened when employees experience high supervisor and co-worker support (Figure 1). In the presence of support, the straining effect of work control predictability on creative performance is expected to be buffered in the extremes of the curve. We propose that supervisor and coworker support interacts with work control predictability which affects employee creativity. Hence, we hypothesize the following moderating effects:

H2. Supervisor support moderates the inverted U-curve relationship between work control predictability and employee creative performance in such a way that for employees who receive high supervisor support the inverted U-curve relationship is significantly weaker, i.e. less pronounced, compared to employees who receive low supervisor support.

H3. Coworker support moderates the inverted U-curve relationship between work control predictability and employee creative performance in such a way that for employees who receive high coworker support the inverted U-curve relationship is significantly weaker, i.e. less pronounced, compared to employees who receive low coworker support.

Methods

Sample and procedures

We collected data using an internet-based survey among employees from a Belgian governmental organization. This organization provides services to civilians and has a broad societal impact. Employees in the organization support civilians with problems around tax payments, buying real estate, and immigration. The targeted respondents are typically general office workers and managers. They are required to be creative in finding solutions to very specific, mainly unprecedented, problems of civilians in a variety of topics. In every case unique individual circumstances have to be taken into account. As our objective is to gain insights in perceived coworker support and work control predictability, we measured these variables with self-assessment (Frese & Fay, 2001; Smith et al., 1997). Employee creativity was measured using supervisor ratings.

Prior to the distribution of the survey, a pilot study was conducted among eight participants who provided comments and suggestions on the clarity and readability of the questionnaire’s items. Based on their feedback, the content of the cover letter and the layout of the questionnaire were slightly adapted.

We distributed surveys to 1350 employees, of which 371 responded: 319 in the first round and 52 after a reminder (28% response rate). We then invited the supervisors of these employees to rate employee creativity. In total, 45 supervisors rated the creativity of their employees, with an average of 2.84 ratings per supervisor. In total 128 employee–supervisor dyads were established. The sample consisted of 57% male and 43% female employees. Typical for a governmental organization, the average age of employees in the sample is 51, and the average working experience is 26 years, indicating that most employees were knowledgeable about their organization’s working practices.¹

Measures

Multiple-item scales, closely following previous studies, were used to measure each construct. Unless otherwise reported, the construct variables were measured on seven-point Likert scales ranging from 1 (totally disagree) to 7 (totally agree).

Creative performance. Creativity studies have been criticized for failing to operationally delineate between innovation and creativity (e.g., West & Farr, 1990). We used eight items from Tierney, Farmer, and Graen (1999) to assess supervisors’ ratings of their employees’ creative problem-solving in their work on a scale ranging from 1 (never) to 6 (always). This scale measures the extent to which employees are directly involved in creative idea generation. The reliability of the eight-item scale was high (α = .95). An example item is “In the past year this employee tried out new ideas and approaches to problems.”

Social support for creativity can be provided by supervisors and/or coworkers. Typically, support consists of constructive feedback and suggestions to help and motivate employees to improve their creative ideas (Harrison & Rouse, 2015; Perry-Smith & Mannucci, 2017). To grasp this aspect of support, social support from supervisors for creativity was measured with the 3-item scale (α = .93) of Madjar et al. (2002). An example item is “My supervisor gives me useful feedback about my ideas concerning the workplace.” Social support from coworkers for creativity was measured with the 3-item scale (α = .87) of Madjar et al. (2002). An example item is “My co-workers other than my supervisor give me useful feedback about my ideas concerning the work.”

Work control predictability is an independent dimension of work control (e.g. Smith et al., 1997; Väänänen et al., 2008). To evaluate work control predictability we used the 5-item scale of Smith et al. (1997), who created the scale from Dwyer and Ganster’s (1991) work control scale, by using the five items with the highest average factor loadings on predictability. This scale specifically targets this predictability aspect of work control, i.e. the clarity of daily work goals, the ability to foresee disturbances, awareness of impact of the work of others, and the ability to determine the pace in which objectives are reached (α = .68). A sample item is “How much control do you have personally over how much work you get done?.”

As previous research has found age, tenure, and gender be either related to social support or to creative performance (Binnewies, 2008; Hopkins, 2002), these variables were considered as potential control variables. Age was measured in years. Gender was measured using a dichotomous variable (0 = male), and tenure was measured by years of experience in the current job.

We conducted a confirmatory factor analysis (CFA) to examine the distinctiveness of the scales for social support from supervisors, social support from coworkers, and work control predictability. The results show that a three-factor model fits the data better (χ2 = 74.20, df = 41, p = .001, CFI = .95, RMSEA = .08, SRMR = .07) compared to a two-factor model with social support from supervisors and social support from coworkers loading on one factor (χ2 = 240.96, df = 43, p < .000, CFI = .72, RMSEA = .19, SRMR = .12) and a one-factor model (χ2 = 316.30, df = 44, p < .000, CFI = .61, RMSEA = .22, SRMR = .15). These results show that the three self-reported variables have sufficient discriminant validity.

Analytical strategy

Our main focus is to test our hypotheses at the individual level of analysis. However, our data has a nested structure, as one supervisor supervises several employees. To control for this nested structure, we used multilevel analysis to examine the hypothesized effects. In nested samples, individual observations are not interdependent, violating the assumption of independence of observations used by standard statistical tests (Hox, 2010). The intra-class correlation of our dependent variable is .45, which means that 45% of the variance in the ratings of creativity is dependent on the supervisor. Therefore, we included a random intercept and a fixed slope for our independent variable work control-predictability to deal with the possibility that ratings of creativity are dependent on the nature of the supervisor. We centered all main variables around the group mean before including them in the interaction terms, as the support received from supervisors and the social support from coworkers can be dependent on the group or team in which the employee is working. Following recommendations by Kreft, De Leeuw, and Aiken (1995) for testing random intercept-models, we also included the group mean as a Level-2 predictor.

Following Aiken and West (1991), we used the following equation to test the hypotheses; Y = γ₀₀ + γ₁₀X + γ₂₀X² + γ₃₀Z + γ₄₀XZ + γ₅₀X²Z + μ₀₀ + μ₁₀(X) + r. Aiken and West (1991) propose adding a squared term of the independent variable (X²) – in our case WCP – to test for nonlinear associations. If this squared term is significant, this suggests that there is a nonlinear association between the independent and dependent variable. To test for a moderated nonlinear association, the squared term is multiplied by the moderator to test for a significant moderation effect of the moderator – in our case SS and CS – on the nonlinear association between the independent and dependent variable. In the equation above, γ₀₀ refers to the fixed intercept, γ₁₀X refers to the fixed effect of work control predictability (WCP), γ₂₀X² refers to the fixed effect of the squared term of WCP, γ₃₀Z refers to the fixed effect of either supervisor support (SS, in model 2) or coworker support (CS, model 3), γ₄₀XZ represents the fixed effect of the interaction between WCP and SS or CS, and γ₅₀X²Z stands for the fixed effect of the interaction between the squared term of WCP and SS or CS. Moreover, μ₀₀ refers to the random intercept of creativity and μ₁₀(X) stands for the random slope of WCP. Finally, r refers to the individual-level error.

Results

Table 1 reports means, standard deviations, and correlations between variables. Supervisor support and coworker support are significantly and positively correlated. Supervisor support is weakly negatively related to work control predictability. The weak negative correlation could mean that the lower work control predictability becomes, the more leader support is required to execute one’s job. We found no significant correlations between work control predictability and creative performance. Note that the control variables did not structurally associate with any of the main variables. We do see that the women in our sample are younger and receive more supervisor support than men. Over the last twenty years, increasingly more women have started to work for the sample organization which was originally male dominated. Because the final results do not change when the control variables are included in the regressions, we leave the control variables out of the final analyses to increase the power of our tests (cf. Becker, 2005). We do control for the group means of the main variables.

Table 1. Means (uncentered), standard deviations, and correlations (n = 128).

	Variables	Mean	S.D.	1	2	3	4	5	6
1.	Creative performance	3.41	1.19	(.95)
2.	Coworker support	4.81	1.23	−.02	(.87)
3.	Supervisor support	4.67	1.56	.01	.45**	(.93)
4.	Work control predictability	2.85	0.60	−.03	−.06	−.19*	(.68)
5.	Age	50.91	8.12	−.09	.13	.08	.01
6.	Tenure	13.18	5.75	.00	.18*	.05	.05	.56**
7	Gender^a	.43	.49	−.02	.04	.17*	.01	−.24**	−.04

Note. N = 128. Internal reliabilities (alpha coefficients) are given in parentheses on the diagonal. * p < .05, ** p < .01; ^a 0 = male

Table 2 presents the results of our multilevel analyses. First, we included the control variables and work control predictability, and the squared term of work control predictability (model 1). In models 2 and 3 we added the moderators supervisor support (model 2) and coworker support (model 3) as well as the interaction terms used to test the moderated curvilinear relationships.

Table 2. Results of multilevel analyses predicting employee creative performance (n = 128).

	Model 1	Model 2	Model 3
Intercept y00	4.22(.99)**	3.72(1.31)**	3.84(1.44)*
Controls
Group mean of WCP γ01	−.21(.34)	−.08(.35)	−.05(.33)
Group mean of SS γ02	-	.03(.13)	-
Group mean of CP γ03	-	-	−.02(.18)
Independent variable
Work Control Predictability γ10	.09(.21)	.19(.18)	.10(.21)
Work Control Predictability squared γ20	−.79(.39)*	−1.12(.40)**	−.70(.38)†
Moderaters
Supervisor Support γ30		−.05(.08)	-
Coworker Support γ30		-	−.15(.10)
Interactions
WCP x SS γ40		−.42(.20)*	-
WCP squared x SS γ50		.60(.31)†	-
WCP x CS γ40		-	−.30(.20)
WCP squared x CS γ50		-	.87(.36)*

Note. Main variables Work Control Predictability (WCP), Supervisor Support (SS) and coworker Support (CS) were group mean centered, Standard Errors are reported between brackets,† < .10; * p < .05, ** p < .01.

Hypothesis 1 proposes an inverted U-curve relationship between perceived work control predictability and employee creative performance. Our regressions in model 1 show that the main effect of WCP is not significant (γ₁₀ = .09(.21); ρ = ns), but the squared term of WCP does reach significance (γ₂₀ = −.79(.39); ρ < .05). This finding suggests a nonlinear relationship between WCP and employee creative performance, thereby supporting H1: moderate levels of work control predictability are related to the highest level of creative performance.

Hypothesis 2 and 3 propose that supervisor (H2) and coworker (H3) support moderate the inverted U-curve relationship between work control predictability and employee creative performance in such a way that for employees who receive high supervisor/coworker support the inverted U-curve relationship is significantly weaker, i.e. less pronounced, compared to employees who receive low supervisor/coworker support. Our results show that the interaction between WCP squared and supervisor support is significant at the .10-level (γ₅₀ = .60(.31); ρ < .10), meaning that further analysis of how the curvature changes in the presence of supervisor and coworker support is warranted.

We can make these results insightful by plotting the nonlinear relationship moderated by supervisor support (Figure 2) and by coworker support (Figure 3). Figure 2 shows that at high supervisor support levels, there is a weak negative relationship between WCP and employee creative performance with highest levels of employee creative performance at low (−2SD) and moderate (intermediate) levels of WCP. At low supervisor support, employee creative performance is higher at moderate levels of WCP compared to low (−2SD) and high (+2SD) levels of WCP.

Figure 2. Nonlinear relationship between WCP and employee creative performance moderated by supervisor support.

Figure 3. Nonlinear relationship between WCP and employee creative performance moderated by coworker support.

Simple slope tests were conducted to statistically assess the curves for both high and low supervisor support. We tested the simple slopes for low (−2SD), moderate (mean) and high levels of WCP (+2SD). Similar to Leung and colleagues (2011), we used two standard deviations from the mean, because the plots in Figures 2–3 suggest that the slopes tend to change to a larger extent at these values. For the curve of the low supervisor support-condition, the simple slope at low levels of WCP (−2SD) is positive and significant (t = 3.26; CI_low = 1.55; CI_high = 6.26), for medium levels of WCP it was also positive and significant (t = 2.16; CI_low = 0.06; CI_high = 1.33), and for high levels of WCP (+2SD) it was negative and significant (t = −2.23; CI_low = −4.72; CI_high = −0,29). This provides statistical evidence that the association between WCP and employee creative performance at low levels of coworker support has an inverted u-shaped curve. For high levels of supervisor support, the simple slope at low (t = −.54; CI_low = −1.24; CI_high = 2.16), medium (t = −1.02, CI_low = −.91; CI_high = .29). as well as high levels of WCP (t = −1.08; CI_low = −3.05; CI_high = .88) were all insignificant, which shows that there is no association between WCP and employee creative performance when supervisor support is high. This supports our predictions in H2.

Examining the moderating role of coworker support on the nonlinear relationship between WCP and employee creative performance, the results of the regressions show that the interaction term between WCP squared and coworker support is significant (γ₅₀ = −.87(.36); ρ < .05). Figure 3 indicates that for high coworker support, employee creative performance is highest at low levels of WCP (−2SD). At low coworker support, employee creative performance is higher at moderate levels of WCP compared to low (−2SD) and high (+2SD) levels of WCP. Like we did for the curves of low and high supervisor support, we used simple slope tests to evaluate the curves for low and high coworker support. For the curve of the low coworker support-condition, the simple slope at low levels of WCP (−2SD) is positive and significant (t = 3.39; CI_low = 1.37; CI_high = 5.13), for medium levels of WCP it was insignificant (t = 1.46; CI_low = −0.24; CI_high = 1.65), and for high levels of WCP (+2SD) it was negative and significant (t = −2.31; CI_low = −4.53; CI_high = −0.36). This provides statistical evidence that the association between WCP and employee creative performance at low levels of coworker support has an inverted u-shaped curve. For high levels of coworker support, the simple slope at low levels (t = −.39; CI_low = −2.35; CI_high = 1.56) at medium levels (t = −1.13, CI_low = −1.36; CI_high = .36) as well as high levels (t = 0.01; CI_low = −2.01; CI_high = 2.02) were all insignificant. This indicates that the association between WCP and creative performance is insignificant when coworker support is high, which supports H3.

Discussion

In this study, we developed and tested a model suggesting that supervisor and coworker support moderate the nonlinear relationship between work control predictability and employee creative performance. In our sample, we find support for the following relationships. We find that work control predictability demonstrates an inverted U-shaped relationship with creativity, indicating creativity is highest at intermediate levels of work control predictability (H1). Furthermore, our study showed that coworker/supervisor support moderate the inverted U-curve relationship between work control predictability and creativity (H2 and H3). We found that for employees that receive high support the inverted U-curve relationship is significantly weaker, i.e. less pronounced, compared to employees that receive low support. Specifically, we found that employees who experience low supervisor support have higher optimum levels of creativity under a moderate level of work control predictability than employees who experience high supervisor support. A similar relationship holds for coworker support, i.e. support lessens the negative effects of work control predictability on creativity in the extremes of the inverted U-curve. In other words, a supportive work environment buffers against the low creativity associated with very low (or high) work control predictability situations, but does not have a beneficial effect on creativity under moderate levels of work control predictability.

Theoretical contribution

Our results suggest that predictability of a work activity or task is important for employees as it affects their creativity. Specifically, an employee in a job characterized by moderate levels of predictability is evaluated by his/her supervisor as having higher creative performance compared to employees in jobs with low and high work control predictability. In other words, employees’ creativity will flourish when the nature, onset and outcome of work events is known beforehand to a certain extent, but is not pre-determined. Employees need room to determine the pace and method with which objectives are reached and have time and resources to develop and test ideas. These findings resonate with the results from studies about creativity under constraints (e.g. Acar et al., 2019; Rosso, 2014), which point to other constraining factors that show a curvilinear relationship with creativity such as time pressure (Baer & Oldham, 2006; Ohly et al., 2006), resource constraints (Shalley & Gilson, 2004), process constraints (Acar et al., 2019; Sethi & Iqbal, 2008) and output constraints (Berrone, Fosfuri, Gelabert, & Gomez-Mejia, 2013; Rosso, 2014). Constraints, including work control predictability, can be seen as enablers of creativity on the one hand, as constraints stimulate motivation and provide structure (Acar et al., 2019); and inhibitors on the other hand, as they discourage exploration and increase reliance on well-known approaches (Rosso, 2014). In terms of the JD-R model, the enabling (inhibiting) aspect of constraints in any form can be explained by increase (decrease) in motivation and a decrease (increase) in strain. In a review study, Acar and colleagues (2019) suggested an integrative framework linking constraints to creative outcomes via an inverted U-shape relationship. The findings of our study contribute evidence to this theoretical framework.

Our results show the moderating role of social support, thereby suggesting that the negative effects of constraints (such as work control predictability) on creativity can be damped in the extremes of the inverted U-curve. No studies have yet considered moderating factors for the curvilinear constraint-creativity relationship, despite the call for such research (Acar et al., 2019). Our findings show that a supportive climate buffers against low creativity especially in the extremes the curve (very low or high work control predictability situations). The dampening effect on the extremes of a boundary condition such as social support was line with our expectations and suggests that a boundary condition shapes employees’ perception about the constraint they are confronted with. When work control predictability is perceived to be more straining than it is motivational (as it is in the extremes of the inverted U-curve), the straining effect is alleviated to some extent in the presence of a social support structure. Following the argumentation of Acar et al. (2019), apart from influencing the perception about a constraining factor, moderators may also influence the capability of employees to overcome the constraining effect, especially in extreme situations. In other words, in situations of extremely high (or low) levels of organizational constraints (in our study proxied by work control predictability), social support may positively affect the capability to be creative. This is because social interactions can show how tasks and work processes interconnect and where tasks can be facilitated, uniformed or improved in other ways (e.g. made less tedious), even given extreme levels of work control predictability.

Our findings show that supervisor support does not have a beneficial effect on creativity under moderate levels of work control predictability. It could be the case that in these latter situations, support by supervisors is interpreted by employees as a signal of their incompetence in the perception of the supervisor. This feeling is likely to hamper their creativity as employees’ feelings of self-competency and self-determination are undermined (Chang et al., 2012; Zhou, 1998). This finding suggests that providing support to subordinates is a delicate task.

With our findings we advance several areas of current theorizing. First, we add to current knowledge about the predictability aspect of work control. Although several studies have suggested that predictability is an independent dimension of work control (e.g. Smith et al., 1997; Väänänen et al., 2008), few studies have explicitly investigated how work control predictability affects employee performance. Work predictability studies either focus on the predictability of work hours, specifically for part time workers and workers with on-call contracts (e.g. Tijdens & Dragstra, 2007), or predictability in terms of being informed about organizational changes (e.g. Pejtersen, Bjorner, & Hasle, 2010; Quist et al., 2013). These measures overlook the possible effect of the extent in which daily work events can be predicted by employees. Yet, stress researchers have consistently found the detrimental effects of not knowing what events will occur, under what conditions and what events will be like (e.g. Lin, Hsieh, Yeh, & Niddam, 2014; Miller et al., 1981). Our study highlights the role of work control predictability as a relevant job characteristic that influences employee creative performance.

We also add to studies about creative employee performance (Kremer et al., 2019). In today’s dynamic work environment, good performance not only involves being a good performer (Grant & Ashford, 2008) but also being proactive and creative (Zhou & Hoever, 2014). Whereas prior research has mainly focused on the effect of work control on overall employee performance (e.g. Gordon et al., 2015; Parker et al., 2013), we focused on creative performance. Predictability of work events may empower employees to make a well-timed response. For example, the employees in our sample may to some expect that during some peak hours in their day they may be confronted with an unrelenting stream of tough problems that require creative solutions. The predictability of the occurrence and timing of tough periods may engender a coping response, for instance by chatting with coworkers beforehand or getting a fresh cup of coffee to mentally prepare for the upcoming issues. By studying how work control predictability is related to creative performance, this study contributes to the creativity literature by increasing our understanding about the effects of contextual factors that may be encouraging or unsupportive for employee creativity (Zhou & Hoever, 2014). Moreover, we have taken a first step in considering a wider range of outcomes in the work control literature.

In addition, we add to the work control literature and studies that investigate the degree in which organizational constraints hinders the creative performance of employees (e.g. Acar et al., 2019; Rosso, 2014). Although theoretical arguments have been provided for a possible curvilinear effect of work control and employee performance (e.g. Baer & Oldham, 2006; Parker et al., 2013), our study is the first to empirically demonstrate that work control predictability has an inverted U-curve relation with employee creativity.

Finally, with these findings, we contribute to studies that investigate the effects of specific job designs and a supportive work environment on employee performance (Janssen, 2001; Leung et al., 2011). Recent studies indicate that work control and social support enables employees to engage in problem solving, and enhance job performance (e.g. Daniels et al., 2013; Gordon et al., 2015). However, previous research has concentrated on the interplay between individual traits and supportive contexts, and has ignored to some extent the interplay between job characteristics and the working environment in determining job performance. For instance, studies have investigated how employees with creative personalities are affected by different levels of feedback, support and control (e.g. Madjar et al., 2002; Zhou & George, 2001; Zhou & Hoever, 2014; Tse, To, & Chiu, 2018). In our study, we explicitly address the impact of job characteristics (i.e. work control predictability) and its interplay with the working environment (i.e. support structures) in influencing employee creativity, and we found that support influences employees’ ways to be creative in their job upon different levels of work control predictability.

Managerial implications

Practitioners should be aware of the way in which job and contextual factors interact and promote employee creativity. Based on the findings of our study that work control predictability shows a nonlinear relationship with employee creativity, and assuming that this relationship will be confirmed in future studies, we recommend designing jobs such that every job contains some degree of predictability. This will reduce unnecessary burdens that may arise due to lack of information and clarity. At the same time it leaves enough room for the employee to experiment with the task at hand which is likely to enhance intrinsic motivation. Managers should note that predictability differs from control. Although the nature of the work may be so that it is impossible to provide control over work events, it may still be possible to enhance predictability of the occurrence and nature of work events. For instance, work schedules may be designed in a way that an employee is only confronted with a certain category of difficulties or distractions during some dedicated hours. This allows employees to mentally prepare for these difficulties, even though they have no way to otherwise control them.

Ideally, managers implement a routine to assess the work control predictability. This may ensure that a balanced level is kept, maintaining a position at the optimum of the inverted U-curve. However, it may be difficult to regulate work control predictability at the organizational level. Our study shows that perceived support from the supervisor or coworkers buffers to a large extent against the negative effects of too low or too high work control predictability. In a fast changing work environment, managers should encourage communication and frequent knowledge sharing within and across teams in order to support employees to carry on their creative efforts. A supportive climate may cushion the negative effects of extreme levels of work control predictability.

Furthermore, our results suggest that when an organization provides low levels of support, low work control predictability is likely to suppress creativity. Managers may not be aware of this situation, and employees may not complain because they experience freedom and autonomy. Yet, especially in those situations a supportive climate could be beneficial for employee creativity. From a career development perspective, our findings may offer opportunities for human resource practices and career coaching activities (Verbruggen, Sels, & Forrier, 2007). Organizations can choose to offer coaching or mentoring to employees and supervisors in order to stimulate and facilitate support structures among coworkers as well as between supervisors and employees. For instance, active coaching of feedback seeking behaviors of employees may be fruitful (Ashford, Blatt, & VandeWalle, 2003).

Limitations and future research

Our study harbors limitations which provide avenues for future research. A first limitation resides in our use of a self-assessment questionnaire to examine perceived predictability and social support, as differences may exist between perceived support and predictability and the actual work conditions in an organization (Stroebe, 2000). Likewise, we used supervisor ratings to measure employee creativity, however, there may be differences between the actual creativity displayed at work and the possibilities that employees experience for displaying their creativity at work. In this line, several studies explore differences between self-rated creativity and self-rated potential for demonstrating creativity at work (Caniëls, Chini, & Ooms, 2015). Similarly, Caniëls and Rietzschel (2015) point toward differences between feeling creative and acting creatively at work. Hence, a promising avenue for future research is to investigate the consequences of these distinctions.

A second limitation concerns the measure of work control predictability. The scale shows an acceptable internal consistency (α = .68), but the scale includes two items (“How much control do you have personally over how much work you get done?” and “How much control do you have over how quickly or slowly you have to work?”) that can be interpreted as indications of work control rather than WCP. Taking out these items, however, lowers the internal consistency to an unacceptable level (α = .55). We also conducted an additional PCA-analysis, and found that the five items load on a single factor. This is in line with psychometric analyses by Smith and colleagues (1997) found that the two items loaded on one WCP-factor together with the three remaining items, which is why we chose this measure in our study. Future research, however, should carefully consider alternative measures, such as the scale part of the Occupational Stress Questionnaire (Elo, Leppänen, Lindström, & Roponen, 1992).

In addition, the relevance of our study is limited by the homogenous sample we used. Hamann and Foster (2014)show that profit and nonprofit workers have quite similar levels of work control/autonomy and support. Moreover, using a Belgian sample of profit and nonprofit workers, De Cooman and colleagues (De Cooman, De Gieter, Pepermans, & Jegers, 2011) found that there were no differences in decision authority and autonomy between the two groups. Nevertheless, we cannot determine whether the results reported here would also apply to other organizational settings, such as non-governmental organizations. Future research may employ more heterogeneous settings to assess whether the results of our study can be replicated.

Furthermore, it may be interesting to examine factors that influence work control predictability, as it could be enhanced by the organizational environment, such as level of expertise, or years of experience. Relatedly, it may be suitable to investigate job complexity instead of work control predictability, particularly for a more heterogeneous setting than the one we employed. The concept of job complexity is closely related to work control predictability, but is broader in nature. Job complexity is high when jobs are characterized by high levels of autonomy, feedback, significance, identity and variety (Hackman & Oldham, 1980). Future research that addresses multiple organizations, may want to assess the moderating role of job complexity for the support-creativity relationship (Li, Li, Shang, & Xi, 2015). Complex jobs may fuel intrinsic motivation by stimulating employees’ excitement about their tasks and activities, which may in turn increase their interest in completing those (Shalley et al., 2004).

Another promising avenue of future research comes from taking the learning orientation of employees into account. Employees are more creative when they have an innate orientation toward learning and development (Fuller & Marler, 2009). Moreover, they are more open to ideas of coworkers and supervisors and therefore they may react differently toward social support than employees who are less learning oriented. Hence, future research could look for a three-way interaction between individual characteristics (e.g. learning goal), job characteristics (e.g. control or predictability) and work environment (e.g. support) affecting employee creativity.

Finally, although data was gathered at two moments in time (first the independent and moderating variables by employees, and at a later moment in time, the dependent variable by leader-ratings of each employee), we cannot rule out that creative employees have different (more positive) perception of their work environment and to some extent craft it to become more predictable (yet not too much). A study by Salanova and colleagues (2010) showed that high performing employees may feel more supported by their organization than low performing employees, suggesting that they may perceive less work control predictability. Similarly, Kim, Im, and Qu (2018) found that creative self-efficacy positively influenced job crafting, suggesting that creative employees (who perceive themselves as being creative) will craft their work environment to improve it. Hence, replication of our study in a longitudinal design is needed to check the robustness of our findings.

Conclusion

The present study proposes that the relationship between work control predictability and creativity could be better understood if we pay attention to the role of supervisor support. Our study reveals that a lack of work control predictability can be compensated by a work environment with high supervisor support and employees who encounter a lack of support require intermediate levels of work predictability to be most creative. In this way, our findings refine current theorizing about the role of support in the creativity literature by showing how job and contextual factors interact, and affect creative performance of employees.

Acknowledgments

The authors wish to thank Wouter Gryson for gathering the data and sharing them with us.

Disclosure statement

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

Notes

1. We ran a number of t-tests to see if there were significant differences with respect to the total sample of 371 employees and the subsample of 128 employees concerning our key variables. The results show that there are no significant differences with respect to Work control predictability (t(497) = −.49, p = ns), Social support from supervisors for creativity (t(497) = 1.83, p = ns), and Social support from coworkers for creativity (t(497) = 1.85, p = ns).