Employee ambidexterity, high performance work systems and innovative work behaviour: How much balance do we need?

Abstract

This study examines whether and how innovative work behaviour is related to explorative and exploitative activities. Polynomial regression analyses are used to test the relationship between ambidexterity (being engaged in explorative and exploitative activities in equal amounts) and innovative work behaviour, as well as between specialisation (being engaged in either explorative or exploitative activities) and innovative work behaviour. Furthermore, we use moderated polynomial regression analyses to examine a possible moderating effect of high-performance work systems (HPWS) on these relationships. Results indicate that balance at a high level, as well as specialisation, are conducive to innovative work behaviour. A moderating effect of HPWS was not supported by our data.

Keywords:

• Ambidexterity
• innovative work behaviour
• high-performance work systems
• polynomial regression analysis

Introduction

Surviving in today’s challenging economy drives firms to continuously innovate (Anderson, Potočnik, & Zhou, 2014). Firms attempt to find a balance between exploring new ideas and exploiting existing competences in an effort to satisfy existing customers, while still aiming to be future oriented and spot potential changes in customer bases, or emerging markets. This balancing process is referred to as organisational ambidexterity, and consists of simultaneously pursuing both exploratory and exploitative activities (March, 1991). Organisational ambidexterity has been identified as an important antecedent of technological progress (Gibson & Birkinshaw, 2004; He & Wong, 2004; Junni, Sarala, Taras, & Tarba, 2013; Raisch & Birkinshaw, 2008).

Within the literature, considerable research efforts have been devoted to identify and describe the factors that lead to organisational ambidexterity (He & Wong, 2004; Raisch & Birkinshaw, 2008). For example, in a survey among US high-tech firms, Beckman (2006) showed that team composition is an important antecedent of firm ambidexterity. Other studies have focused on performance outcomes of ambidexterity, in particular in terms of customer satisfaction (Gibson & Birkinshaw, 2004), and sales growth (He & Wong, 2004). However, a limitation of extant research is that we know very little about the outcomes of ambidexterity with respect to innovation. For example, it is unclear whether organisations should try to strive for high levels of exploration and exploitation simultaneously or emphasise either exploration or exploitation in order to maximise innovative performance.

A second limitation of current empirical research is that ambidexterity is nearly exclusively investigated at the organisation level (Junni et al., 2013; Zacher, Robinson, & Rosing, 2014). Hence, we lack in-depth insights about how ambidexterity at the employee level is related to specific employee behaviour, such as innovative work behaviour, i.e. ‘the intentional creation, introduction and application of new ideas within a work role, group or organisation, in order to benefit performance’ (Janssen, 2000, p. 288). Increasingly, there are calls for research that explores ambidexterity at the employee level (Birkinshaw & Gupta, 2013; Zacher et al., 2014). In this study, we therefore focus on employee-level ambidexterity rather than at the organisational level.

Theoretically, scholars have argued that in order to be ambidextrous, employees should be able to be explorative and exploitative at the same time in equal amounts. Yet, studies in the field of psychology (e.g. Eysenck, 1967) and economics (e.g. Reagans, Argote, & Brooks, 2005) suggest that employees should specialise and capitalise on activities in which they excel. Empirically, we know very little about the impact of specialisation at the employee level in either explorative or exploitative activities on innovative work behaviour (Zacher et al., 2014). This is an important research gap to address, given that the employees are the ones who have to perform explorative and/or exploitative activities (Gibson & Birkinshaw, 2004; Kang & Snell, 2009).

Furthermore, the question arises whether and how the organisational context can reinforce the impact of ambidextrous behaviour among employees. For example, would organisations that employ high-performance work systems (HPWS) be more successful in generating benefits from having ambidextrous (or highly specialised) employees? Insights in these organisational circumstances are currently inadequately developed (Junni et al., 2013; Prieto & Pilar Pérez Santana, 2012).

In this study, we aim to shed light on whether and how explorative and exploitative activities are related to innovative work behaviour. After gathering data in a survey among 160 employees we use polynomial regression to test the relationship between ambidexterity (being engaged in explorative and exploitative activities in equal amounts) and innovative work behaviour, as well as between specialisation (being engaged in either explorative or exploitative activities) and innovative work behaviour. Furthermore, we use moderated polynomial regression analysis to examine whether and how HPWS can moderate the effect of the (im)balance of employee explorative and exploitative behaviour on innovative work behaviour.

Our study contributes to current literature in several ways. First, we contribute to studies that call for more ambidexterity research at the employee level, as compared to the organisational level (Junni et al., 2013; Prieto & Pilar Pérez Santana, 2012; Zacher et al., 2014). Specifically, we increase current understanding about whether and to what extent employee ambidexterity improves employee innovative work behaviour. Second, we extend current understanding about the impact of human resource practices on employee ambidexterity. By studying the moderating effect of HPWS on the ambidexterity-innovative work behaviour relationship, we can differentiate between employees that may be more innovative as a result of engaging in ambidextrous (or specialised) activities. Third, we use polynomial regression analysis to analyse our data. Whereas the traditional method to analyse ambidexterity with difference scores collapses explorative and exploitative activities into one single ambidexterity score (see for instance Mom, Fourné, & Jansen, 2015; Patel, Messersmith, & Lepak, 2013), polynomial regressions view explorative and exploitative activities, and ambidexterity as three distinct constructs with separate measures. To the best of our knowledge this technique has not yet been widely applied in ambidexterity research (see for an exception Fu, Ma, Bosak, & Flood, 2015), yet it is specifically suitable in this context.

Literature review

Ambidexterity at the employee level

We define organisational ambidexterity as an organisation’s ability to pursue exploitative and explorative activities at the same time. Exploitative activities are associated with aspects such as improving efficiency, implementation and execution. Explorative activities are aimed at changing the existing models, experimentation and radically impacting organisational routines (March, 1991). At the employee level, exploitative activities consist of using present knowledge and skills to make short-term improvements in terms of efficiency and efficacy (Gibson & Birkinshaw, 2004; Kang & Snell, 2009). In contrast, explorative activities consist of behaviours such as searching for new product and process innovation, searching for competitive solutions and behaviours that require the employee to learn new skills or knowledge and require the employee to adapt current routines (Gibson & Birkinshaw, 2004; Kang & Snell, 2009).

Employee ambidexterity is defined as the behavioural orientation of employees towards combining exploitation- and exploration-related activities within a certain period of time (Mom, van den Bosch, & Volberda, 2009). Two main theoretical views on organisational ambidexterity can be distinguished in current literature, which can be translated to the employee level. First, several studies have adopted a contextual view on ambidexterity (e.g. Gibson & Birkinshaw, 2004). Contextual ambidexterity suggests that the organisational context should allow for undertaking explorative as well as exploitative activities simultaneously. This view on organisational ambidexterity implies a need for high levels of both exploration and exploitation. Exploration and exploitation are expected to reinforce each other. The higher the level of both activities, i.e. the higher the level of true ambidexterity, the better it is for organisational performance. Analogously, individual employees who engage in ambidextrous activities are expected to achieve an optimum performance in terms of innovative work behaviour, as innovation requires both exploration and exploitation (Rosing, Frese, & Bausch, 2011).

Second, several studies have adopted a structural view on ambidexterity (e.g. Benner & Tushman, 2003; Lavie, Stettner, & Tushman, 2010). Structural ambidexterity suggests that explorative and exploitative activities should be undertaken independently, because exploration and exploitation constitute competing goals, fight for the same resources and require different organisational capabilities (e.g. Bledow, Frese, Anderson, Erez, & Farr, 2009; March, 1991; Smith & Tushman, 2005) and incompatible organisational structures (Benner & Tushman, 2003). Several scholars have refined this idea by suggesting the need for a temporal sequencing of explorative and exploitative activities (e.g. Puranam, Singh, & Zollo, 2006). Short bursts of exploration may interrupt longer periods of exploitation (Levinthal & March, 1993). At the employee level, this view implies specialisation of employees in either explorative or exploitative activities. Individual employees who engage in either exploration or exploitation are expected to achieve an optimum performance in terms of innovative work behaviour.

When combining explorative and exploitative activities, three situations can emerge: (1) a balanced situation, where explorative and exploitative activities are equally present or absent; (2) an unbalanced situation, where explorative activities outweigh exploitative activities; and (3) an unbalanced situation, where exploitative activities dominate explorative activities.

We first consider the balanced situation. Starting from a contextual view on employee ambidexterity as outlined above, we expect that employees who engage in equally high levels of exploration and exploitation will have high innovative work behaviour, whereas employees who engage in equally low levels of exploration and exploitation will have low innovative work behaviour. This reasoning is also grounded in studies about ambidexterity in the context of innovation. Several studies in this line pose that it is unfeasible to separate exploration and exploitation, as it would forsake the opportunity to benefit from synergy effects of exploration and exploitation (Bledow et al., 2009; Rosing et al., 2011; Zacher et al., 2014). For example, pressing problems and distress in exploitation can drive an active approach to goal-oriented idea generation, i.e. exploration (Bledow et al., 2009). Hence, individual employees are expected to take advantage from these synergies between explorative and exploitative activities, and by doing both they are expected to perform well in terms of innovative work behaviour.

Hypothesis 1: In a balanced situation, ambidexterity is positively related to employee innovative work behaviour.

We now consider the unbalanced situations where an employee’s explorative activities dominate the exploitative activities, or where exploitative activities outweigh explorative activities. Starting from a structural view on employee ambidexterity, we expect that employees who engage in either exploration or exploitation are expected to achieve an optimum performance. Studies in the field of psychology point out that individuals differ in their desire to face situations of uncertainty or improvise solutions to challenges. Theorising in the psychology literature (e.g. Eysenck, 1967) suggests that large individual differences exist in people’s desire to take risks or to avoid possible hazards. Apter’s reversal theory (Apter, 1982) proposes a distinction between two alternative and reversible motivational states, the ‘telic’ and the ‘paratelic’. Some individuals are drawn towards activities that are directed towards reaching some long-term goal (telic state), whereas others prefer activities that are playful and challenging (paratelic). The perceived enjoyment in the telic state comes from the anticipation of accomplishment. In the paratelic state, enjoyment is derived from the activity itself and the pleasure and excitement that the activity induces. Following this line of thought, we argue that individual employees may be drawn to engage in either explorative or exploitative activities. This is in accordance with conceptual studies of Kang and Snell (2009) and Kang, Morris, and Snell (2007) who posed that generalists tend to prefer exploratory activities, while specialists are inclined to favour exploitative activities. Individuals differ in the way in which they are motivated, and hence they may differ in their preferred mode of working to fulfil their motivational goals. Consequently, either exploitative or explorative activities may be their preferred work mode to achieve innovative goals. Similarly, it can be argued that employees who need to be explorative may require different personal traits than employees who are expected to be engaged in exploitation. Explorative employees must be curious, crave newness and have a willingness to explore, whereas employees engaged in exploitation must be able to close their minds off from distractions and follow the existing routine (Bledow et al., 2009). Both types of employees can be innovative though.

This reasoning can also be linked to related literature in the field of economics. Economic theory has for long championed specialisation as a way to enhance employee learning and productivity (e.g. Reagans et al., 2005). The learning curve shows that greater experience at a certain task, i.e. either exploitative or explorative activities, yields higher productivity at that task. Following economic theory again leads to the expectation that for employees to specialise in either exploitative or explorative activities may be beneficial to employee performance on various accounts, including innovative work behaviour. Employees who are attracted to, and specialise in, exploitative activities are motivated to excel in exploitation, and hence are likely to display innovative behaviour that concentrates on streamlining the exploitation process. Similarly, employees who are attracted to, and specialise in, explorative activities are motivated to excel in exploration, and hence are likely to display innovative behaviour that concentrates on generating new ideas. Based on psychological and economic theories that suggest positive effects of situations of specialisation in activities at the employee level, we hypothesise a similar effect related to innovative work behaviour:

Hypothesis 2a: The more an employee’s exploitative activities dominate their explorative activities, the higher the employee’s innovative work behaviour will be.

Hypothesis 2b: The more an employee’s explorative activities dominate their exploitative activities, the higher the employee’s innovative work behaviour will be.

The moderating effect of HPWS

Prior research has shown that immediate work contexts shape the relationship between ambidexterity and various performance measures (e.g. Fu et al., 2015; Garaus et al., 2015). In a seminal paper Patel et al. (2013) suggested that HPWS may assist organisations in establishing an organisational context that promotes ambidexterity among employees. A HPWS contains several HR practices that work in concert to develop an organisational context that benefits the abilities and the motivation of employees (Garaus et al., 2015; Patel et al., 2013), and creates opportunities for their further development (Boxall & Purcell, 2011; Fu et al., 2015).

Until now, studies, including Patel et al. (2013) and Fu et al. (2015), have investigated the direct relationship between HPWS and organisational ambidexterity. However, when examining the relationship between ambidexterity and innovative work behaviour, it is likely that HPWS acts as a moderator. Specifically, we pose that the effects of (im)balance in explorative and exploitative activities on innovative work behaviour are moderated by HPWS. With ability, motivation and opportunities in place, i.e. in the presence of HPWS, employees are likely to be better able to convert their exploitative and explorative activities into innovative work behaviour (Garaus et al., 2015). The reasoning being that high levels of HPWS may tease out the best from employees and hence they become more successful in generating benefits (i.e. displaying innovative work behaviour) from being ambidextrous, than would have been the case with low levels of HPWS.

With respect to how this moderating effect changes across the different configurations of exploration and exploitation, we first discuss the relationship between exploration–exploitation balance and innovative work behaviour. High levels of HPWS may stimulate employees to excel. If organisations signal their commitment toward employees by offering HPWS, these employees are likely to reciprocate with higher levels of extra role behaviours (McClean & Collins, 2011; Veenendaal, 2015, p. 132). ‘Extra role behaviour’ refers to desired behaviour that goes beyond the formal job requirements (Organ, Podsakoff, & MacKenzie, 2005). There are three effects. Firstly, at high levels of HPWS, employees become more devoted to generating benefits from being ambidextrous and accordingly displaying innovative work behaviour, than would have been the case with low levels of HPWS. Secondly, because of the HPWS employees also become more successful in generating benefits from being ambidextrous, than would have been the case with low levels of HPWS. Thirdly, HPWS offer job resources such as, for example, autonomy and learning and development opportunities (Patel et al., 2013). Research by De Spiegelaere, Van Gyes, and Van Hootegem (2012) and Abstein and Spieth (2014) has shown that increasing the opportunities of employees to use and develop their professional skills has a strong relation with employee innovativeness. Building on these studies, we pose that HPWS will help employees to translate a certain exploration–exploitation balance into enhanced innovative work behaviour. Hence, a specific exploration–exploitation constellation will generate higher levels of innovative work behaviour relative to the situation where there are low levels of HPWS. For example, a constellation of zero exploration vs zero exploitation will be translated into zero innovative work behaviour in the presence as well as the absence of high HPWS. In contrast, a constellation of high exploration vs high exploitation is expected to be translated into extra high innovative work behaviour in the presence compared to the absence of high levels of HPWS.

Based on this reasoning, we expect that ambidextrous behaviour may translate into more innovative work behaviour for employees who perceive HR practices via HPWS compared to employees that do not experience such an organisational context. In particular, we expect that the relationship between balance in explorative and exploitative activities (i.e. ambidexterity), and innovative work behaviour, is strengthened when employees perceive more HPWS practices. We hypothesise:

Hypothesis 3a: The relationship between the exploration-exploitation balance and employee innovative work behaviour is moderated by HPWS in such a way that the positive relationship between exploration and exploitation balance and innovative work behaviour is strengthened when employees perceive more HPWS, and weakened when employees perceive less HPWS.

Alternatively, it may also be the case that specifically explorative (or exploitative) activities lead to more innovative work behaviour for employees that experience HPWS compared to those who do not. As the mind-sets needed for exploration differ substantially from those needed for exploitation (Gupta, Smith, & Shalley, 2006, p. 695), HPWS may entice employees to outdo themselves, give their absolute best and thereby intensify their preferred mode of working. Furthermore, HPWS may provide opportunities of employees to use and develop specific (explorative or exploitative) professional skills, thereby enabling them to be specifically successful in delivering effective innovative work behaviour (De Spiegelaere et al., 2012). We pose, that employees who perceive HR practices via HPWS are more successful in generating benefits of being highly specialised. At high levels of HPWS, employees are expected to display more innovative work behaviour at the same configuration of exploration and exploitation. In other words, when we investigate situations of specialisation of employees in either exploration or exploitation, we expect to find that the positive relationship to innovative work behaviour is strengthened. Hence, the relationship between imbalance in exploration and exploitation, and innovative work behaviour, is expected to be fortified when employees perceive more HPWS practices. We hypothesise:

Hypothesis 3b: The relationship between the exploration-exploitation imbalance and employee innovative work behaviour is moderated by HPWS in such a way that the relationship between exploration and exploitation imbalance and innovative work behaviour is strengthened when employees perceive more HPWS, and weakened when employees perceive less HPWS.

Method

Sample and procedure

Data for this study were collected in 2015 using an online survey among employees working in an organisational support unit of the Dutch Defence organisation. The survey was accompanied by a cover letter stating the purpose of the study and an assurance of confidentiality and anonymity. Prior to the distribution of the survey, two subject matter experts and one professional provided relevant remarks on the survey layout and the clarity of the survey items. Based on their feedback, the exact phrasing of the cover letter and the layout of the survey were slightly adapted to enhance ease of use and readability.

The survey was emailed to 210 employees, generating 160 valid responses (response rate of 77%). Of the respondents, 89% were male, 68% civilian and 84% was employed for over 10 years at the Defence organisation. Most of the employees (52%) have a higher education or university degree and 75% are over 40 years old.

We undertook several procedural remedies to minimise the usual risk of various biases. We limited respondents’ evaluation apprehension by ensuring and protecting respondents’ anonymity and asking respondents to answer the questions as honestly as possible (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). This procedure reduces the risk of social desirability bias (Podsakoff et al., 2003). The complexity of the research model, which contains interactions, reduces the risk of respondents ‘guessing’ the desirable answers (Malhotra, Kim, & Patil, 2006), and ensures that respondents cannot easily combine related items and produce the correlation needed to produce common method variance-biased pattern of responses (Chang, van Witteloostuijn, & Eden, 2010). By providing verbal labels for the scales and avoiding the use of bipolar numerical scale values (e.g. −2 to +2) acquiescence bias was even further reduced (Kulas, Stachowski, & Haynes, 2008).

Measures

Multiple-item scales, closely following previous studies, were used to measure each construct. The survey covered the following construct variables.

Exploration and exploitation. The explorative and exploitative work-related activities of employees were measured using, respectively, a five-item and six-item scale developed by Mom, Van Den Bosch, and Volberda (2007). Items were rated on a five-point scale, ranging from 1 (to a very small extent) to 5 (to a very large extent). An example item for exploration is ‘In the past year I was engaged in activities that focused on strong renewal of services, activities or processes’. An example item for exploitation is ‘In the past year I was engaged in activities that primarily focused on achieving short-term goals’. We conducted a two-factor confirmatory factor analysis (CFA) to examine the distinctiveness of the scales for exploration and exploitation. The results of the CFA showed that one item of the exploitation scale (i.e. ‘activities of which a lot of experience has been accumulated by yourself’) is highly correlated with several items of the exploration scale. After removing this item we are left with two distinct factors, with an acceptable model fit (χ²/df = 2.408; comparative fit index [CFI] = .94; non-normed fit index [NNFI] = .92; root mean square error of approximation [RMSEA] = .09; and standardised root mean square residual [SRMR] = .07). The estimated reliability was: exploration α = .89 and exploitation α = .81.

Innovative Work Behaviour was measured by adopting the nine-item measure from Janssen (2000), who based it on the scale for individual innovative behaviour in the workplace created by Scott and Bruce (1994). Employees rated how often they performed innovative activities from 1 (never) to 5 (always). An example item is ‘I create new ideas for improvements’. The estimated reliability was α = .92.

For HPWS, we adopted the 27-item scale of Patel et al. (2013). This scale contains items on participation, mobility, training, staffing, job description, appraisal, job security and incentive reward. All items were used, except the incentive reward items, as there is little inter- and intra-organisational variation in incentive reward practices in Dutch non-profit organisations. Employees indicated to what extent they experienced these practices themselves on a scale of 1 (strongly disagree) to 5 (strongly agree). An example item is ‘Extensive training programs are provided to me’. In line with Patel et al. (2013), we started with exploratory factor analysis using principal axis factoring to uncover the underlying factor structure of the high-performance HR practices.¹ The results of this analysis indicated a solution with eight factors having an eigenvalue higher than 1. The scree plot showed a bend at both one and nine factors, indicating a clear break in eigenvalues between the first and the second and the eight and the ninth component. We computed the eight-factor solution (explained variance: 73%). This eight-factor solution was easily interpretable, except for the mobility items. We decided to remove the mobility items, as these items are less relevant for our research context. We were left with a six-factor solution (explained variance: 72%), which was easily interpretable and in accordance with the original scales by Patel et al. (2013): participation, training, staffing, job description, appraisal and job security.

The subsequent confirmatory factor analysis (CFA) tested the factor structure of the six high-performance work practice dimensions and showed good model fit (χ²/df = .954; CFI = 1.00; TLI = 1.00; RMSEA .00, SRMR = .03). In line with the traditional approach for operationalising HPWS, we used an additive index to reflect a single comprehensive measure of the perceived HPWS (Huselid, 1995). A high score on this measure indicates that employees perceive a relatively intense investment in HPWS-practices. The estimated reliability of this construct was α = .87.

Control variables. We controlled for age (measured in years) and educational level (higher scoresreflect higher educational levels), as both might have an impact on innovativework behaviour (Janssen, 2000).

Analytical strategy

The primary interest of this study was testing our hypotheses at the individual level of analysis. However, employees were nested within departments, and we therefore controlled for the nested structure of our data by means of calculating intraclass correlations (ICC1) for the main variables in our model. The ICC1 scores were close to zero (i.e. exploration = .01; exploitation = .00; HPWS = .03; innovative work behaviour = .03). Hence, we decided to use conventional single-level analyses, rather than hierarchical linear modelling.

Polynomial regression with response surface analysis (Edwards, 1994) was conducted in order to test the effect of (im)balance between exploration and exploitation on innovative work behaviour. This approach allows investigating both balance and discrepancy between exploration and exploitation in predicting ambidexterity, and has significant advantages compared to ‘difference scores’, ‘average score’ or ‘summing’ methods, which are traditionally used in ambidexterity research (Junni et al., 2013; Patel et al., 2013). These traditional methods consist of calculating the algebraic difference between two measures (e.g. exploration minus exploitation in order to measure ambidexterity), or calculate the average or the sum of two measures. However, Edwards (1994) stressed the severe methodological problems related to these approaches, as these methods collapse two predictor variables, i.e. exploration and exploitation, into one single score, i.e. ambidexterity (Cronbach, 1958). Consequently, the relationship between the predictor and dependent variables from which the scores are derived remains hidden. In order to overcome this problem, polynomial regressions can be used in which both predictor variables and the outcome are viewed as three distinct constructs with separate measures (Edwards & Parry, 1993).

In line with the suggestions by Shanock and colleagues (Shanock, Baran, Gentry, Pattison, & Heggestad, 2010), we first checked how many respondents demonstrated discrepancies between exploration and exploitation. Next, we centred the exploration and exploitation around the midpoint of their respective scales, in order to reduce the potential risk of multicollinearity (Cohen, Cohen, West, & Aiken, 2013; Edwards, 1994). Subsequently, we regressed innovative work behaviour on the control variables (M1), the centred scores of exploration and exploitation (M2), the product of centred exploration and exploitation, the centred exploration squared and the centred exploitation squared (M3).

In order to test the moderating effect of HPWS (Hypotheses 3a and 3b), we conducted moderated polynomial regression analyses. After testing the first three models as described above, HPWS was entered (M4), and subsequently all moderation terms of HPWS were added (M5). Note that a necessary condition for our results being considered in line with our moderation hypothesis is that the moderated model yields a significant increment in explained variance (cf., Devloo, Anseel, & De Beuckelaer, 2011).

The results of the polynomial regression analyses were further qualified by response surface analysis (Edwards, 2002). In order to visualise the relationships, we used the Excel spreadsheet program from Shanock et al. (2010) to create a three dimensional image that represents the combined relationship of exploration and exploitation on innovative work behaviour.

Results

Means, standard deviations, reliability estimates and correlations for all measures are reported in Table 1. Both exploration and exploitation are significantly and positively related to innovative work behaviour. Only exploitation is significantly related with HPWS. We found no significant correlation between HPWS and innovative work behaviour.

Table 1. Descriptive statistics, reliability estimates and correlations of study variables.

Variable	Mean	SD	1	2	3	4	5	6
Age	47.23	9.53	–
Educational level	1.69	.76	−.40**	–
Exploration	3.21	.91	−.16*	.21*	(.89)
Exploitation	3.53	.72	−.03	−.08	.19*	(.81)
HPWS	2.67	.57	.12	−.15	.14	.20*	(.87)
Innovative work behaviour	3.30	.78	−.24**	.25**	.41*	.24**	.00	(.92)

Notes: Alpha-reliability coefficients are on the diagonal.

** p < .01;

*** p < .001.

Testing the joint effect of exploration and exploitation on innovative work behaviour

Hypotheses 1 and 2a–2b propose that balance and imbalance situations between exploration and exploitation would have an impact on the innovative work behaviour of employees. Table 2 reports the results of the polynomial regression analyses. A model consisting only of control variables (i.e. age and educational level) predicts only 8% of the variance in employee-innovative working behaviour (M1). Adding independent variables (i.e. exploration and exploitation) significantly increases explanatory power to 24% (M2). Exploration contributed about as strongly to innovative work behaviour as exploitation, as shown in the magnitude of the coefficients (B = .28 and B = .20, respectively). The regression analysis revealed a significant increase in R² and a significant higher order effect (M3), allowing us to use surface analysis to further investigate our hypotheses.

Table 2. Polynomial regression examining the impact of exploration and exploitation on innovative work behaviour.

Variables	Innovative work behaviour
	M1	M2	M3	M4	M5
	B (se)	B (se)	B (se)	B (se)	B (se)
Step 1
Age	−.13 (.07)*	−.10 (.06)	−.09 (.06)	−.09 (.06)	−.08 (.06)
Educational level	.14 (.07)*	.11 (.06)	.12 (.06)*	.12 (.06)*	.12 (.06)*
Step 2
Exploration (R)		.28 (.06)***	.34 (.07)***	.34 (.07)***	.33 (.11)**
Exploitation (T)		.20 (.08)*	.19 (.08)*	.19 (.08)*	.19 (.12)
Step 3
R²			.14 (.01)*	.15 (.06)*	.14 (.08)
T²			.06 (.07)	.06 (.07)	−.03 (.10)
R* T			−.15 (.06)*	−.16 (.06)*	−.07 (.13)
Step 4
HPWS				.05 (10)	.16 (.16)
Step 5
R*HPWS					.03 (.12)
T*HPWS					−.05 (.14)
R²*HPWS					.00 (.11)
T²*HPWS					−.14 (.11)
R*T*HPWS					.07 (.12)
Adjusted R²	.07	.22	.27	.27	.25
R² change	.08**	.16***	.06**	.00	.01

Notes: B denotes unstandardised regression coefficients are reported, (se) refers to standard error.

N = 160.

* p < .05;

** p < .01;

*** p < .001.

Figure 1 presents the three-dimensional surfaces of the polynomial regression. We differentiate between a balance situation (Hypothesis 1), and imbalance situations of exploitative activities dominating explorative activities (Hypothesis 2a), and explorative activities dominating exploitative activities (Hypothesis 2b). As recommended by Edwards (1994), the interpretation of this graph is based on four surface test values: a₁ (= b₁ + b₂), a₂ (= b₃ + b₄ + b₅), a₃ (= b₁ − b₂) and a₄ (= b₃ − b₄ + b₅), which represent the slope (a₁ and a₃) and the curvature (a₂ and a₄) of the surface.

Figure 1. Estimated surface relating exploration and exploitation to innovative work behaviour.

We investigate the balance situation by drawing an imaginary line of perfect balance between exploration and exploitation (R = T). The slope of this line (a₁) indicates how balance between exploration and exploitation relates to innovative work behaviour, while the curvature (a₂) expresses whether this relation is linear or curvilinear. Hypothesis 1 proposes that balance between exploration and exploitation will be positively related to employee’s innovative work behaviour, i.e. the higher the score on both exploration and exploitation, the higher innovative work behaviour will be. Consistent with this hypothesis, innovative work behaviour increases as values increase along the R = T line. Figure 1 shows that the lowest level of innovative work behaviour is found at the front corner at the right of the graph where both exploration and exploitation are low. Increasingly, higher levels of innovative working behaviour are found toward the (left) back of the graph, where exploration and exploitation are in balance and high. The results from the slope analysis (Table 3) confirm what the visual diagram represents; the R = T line has a significant linear shape (a₁ = .57, p < .01) and an insignificant curvilinear shape (a₂ = .03, p > .05). Hence, these results support Hypothesis 1: balance between exploration and exploitation is positively related to employee innovative work behaviour.

Table 3. Analysis of slopes and curvatures, effects as related to innovative work behaviour.

	Shape along balance line: exploration = exploitation (R = T)		Shape along imbalance line: exploration = −exploitation (R = −T)
Slope	a1 = b1 + b2	.57**	a3 = b1 – b2	.21
Curvature	a2 = b3 + b4 + b5	.03	a4 = b3 – b4 + b5	.34**

Notes: a₁ and a₂ represent the slope of each surface along the R = T line, while a₃ and a₄ represent the slope of each surface along the R = −T line, where b₁, b₂, b₃, b₄ and b₅ are the unstandardised coefficients on R, T, R², R × T and T², respectively.

*p < .05;

**p < .01;

***p < .001.

We investigated the imbalance situation by drawing an imaginary line of perfect imbalance between exploration and exploitation (R = −T). Again we investigate the slope (a₃) and curvature (a₄). Hypothesis 2a suggests that when exploitative activities exceed explorative activities, innovative work behaviour will be high. Similarly, Hypothesis 2b suggests that when explorative activities exceed exploitative activities, innovative work behaviour will be high as well. Innovative work behaviour is expected to be lower for mid-range results where the discrepancy is small. When we follow along the R = −T line from the left hand (upper) corner to the midpoint of the response surface, the degree of specialisation into exploitative activities diminishes and we move towards a point where exploration matches exploitation. We observe that along this stretch of the R = −T line innovative work behaviour decreases as was expected from Hypothesis 2a. Thus, when employees specialise in doing exploitative activities, innovative work behaviour is higher than when employees undertake both explorative and exploitative activities in equal amounts.

When we continue and move along the R = −T line from the centre to the right-hand (upper) corner of the response surface, the amount of explorative activities is surpassing the amount of exploitative activities. In other words, employees are increasingly specialising in explorative activities. In accordance with what was hypothesised in Hypothesis 2b, we found that innovative work behaviour continued to increase. These results indicate that innovative work behaviour is higher when exploration exceeds exploitation. Slope analysis confirms the visual diagram (Table 3): the R = −T line has a significant convex shape (a₄ = .34, p < .01) and an insignificant linear shape (a₃ = .21, p > .05). We conclude form these observations that Hypotheses 2a and 2b are supported.

Results for testing the interaction effect of HPWS

Models 4 and 5 in Table 2 represent the hierarchical steps for testing the moderating effect of HPWS, as described in Hypotheses 3a and 3b. As can be seen in Table 2, the explained variance does not increase significantly after entering the interaction terms of HPWS. Hence, for our sample Hypotheses 3a and 3b were not supported, as there was no significant interaction effect of HPWS on the relationship between exploration–exploitation (im)balance and innovative work behaviour.

Discussion and conclusion

Theoretical contribution

Ambidexterity studies, and especially empirical studies in this line, predominantly focus on organisational ambidexterity, conflate exploration and exploitation into one measure for ambidexterity, and are limited to investigating the direct effects of HR practices on ambidexterity. The aim of our study was to investigate ambidexterity at the employee level. With the help of advanced statistical methods, such as polynomial regression, we were able to provide a nuanced picture of how exploration and exploitation work together in influencing innovative work behaviour. In particular, we were able to distinguish three situations: (1) a balanced situation, where explorative and exploitative activities are equally present or absent; (2) an unbalanced situation, where exploitative activities dominate explorative activities; and finally (3) an unbalanced situation, where explorative activities outweigh exploitative activities. Furthermore, we investigated the moderating effect of HPWS on the ambidexterity-innovative work behaviour relationship. Our study goes beyond examining organisation level implications of HPWS, and reveals how HR practices impact on individual employees who undertake explorative and exploitative activities.

The results of the response surface analysis demonstrate that there is a significant linear relationship along the line of perfect balance (between exploration and exploitation) as it relates to innovative work behaviour (Hypothesis 1). This indicates that a combination of high levels of exploration and exploitation is positively related to innovative work behaviour. These findings are in line with a contextual view on ambidexterity (e.g. Gibson & Birkinshaw, 2004), in which it is suggested that both types of behaviour should be undertaken simultaneously (Rosing et al., 2011; Bledow et al., 2009). Furthermore, these findings are in keeping with findings from the study by Zacher et al. (2014). This study adopts a multiplicative combination of exploration and exploitation and demonstrates that innovative performance is highest when both exploration and exploitation are high.

When we focus on the imbalance between exploration and exploitation, the results of the response surface analysis indicate that there is a U-shaped curvature along the line of imbalance. That is, innovative work behaviour is relatively low when exploration and exploitation are equally present, and increases as the degree of discrepancy between exploration and exploitation increases (Hypotheses 2a and 2b). Moreover, the results suggest that the direction of the discrepancy (i.e. whether exploration or exploitation is higher) is not significantly related to innovative work behaviour. In other words, specialising in either explorative or exploitative behaviour is beneficial for innovative work behaviour. This is an important finding, which advances current knowledge on the ambidexterity-innovation relationship in at least two ways. First, whereas balance in exploration and exploitation may be beneficial at the organisation level (Junni et al., 2013), investigating the individual level indicated that at the individual level specialisation is beneficial as well. Although specialising in one type of behaviour might not be seen as real ambidexterity at the individual level (i.e. a situation in which both types of behaviour are combined), it does have important implications for the organisational ambidexterity literature. As organisational ambidexterity is dependent on the explorative and exploitative behaviours of individual employees, organisations can still be ambidextrous when individual employees specialise in either exploration or exploitation, as long as both types of behaviour are present at the organisational level. More research on this multilevel process is needed. Second, using polynomial regression we were able to refine the findings of studies that rely on difference scores or adopt product terms. For example, the multiplicative combination of exploration and exploitation of Zacher et al. (2014) is less suitable for demonstrating the effect of specialisation in either exploration or exploitation in relation to innovative work behaviour, and consequently the merits of specialisation were not identified in their study.

Whereas the ability to be ambidextrous resides in individual employees, it is likely to be assisted by the HR practices that are employed by an organisation (Ahammad, Lee, Malul, & Shoham, 2015; Patel et al., 2013). In contrast with our initial expectations, our moderated polynomial regression analyses did not indicate a significant moderating effect of HPWS on the relation between exploration–exploitation (im)balance and innovative work behaviour (Hypotheses 3a and 3b). This finding may result from the power issue at play. For moderated polynomial regression many interaction terms are included in the analysis, which increases the required sample size to detect an effect of a given size with a given degree of confidence. If our sample would have been larger we may have found significant results as the direction of the effects that we find is in line with our expectations.

Managerial implications

This study has some important implications for practice. The results of this study indicate that when employees are equally adept in undertaking explorative and exploitative activities, then this is beneficial for innovative work behaviour. Furthermore, employee specialisation in either explorative or exploitative behaviour is positively related to innovative work behaviour. Although it is often suggested in the literature that innovation requires that individuals need to show both explorative and exploitative behaviour (March, 1991; Rosing et al., 2011), the results in this study suggest that specialising in either exploitation or exploration is beneficial for innovative work behaviour as well. Hence, it might be a fruitful strategy for organisations to investigate personal preferences of employees and if these favour either exploration or exploitation it may be wise not to force employees to become more balanced. In contrast, if a preference exists, it may be beneficial to stimulate employees to (further) specialise in either exploration or exploitation. Employees who tend to demonstrate exploration behaviours, such as experimenting and challenging constraints, should be given the room to explore and advance upon hunches. Only when ideas are put to the test, it is possible to gather feedback on the success or failure of the idea itself and find the direction for further improvement (Caniëls & Rietzschel, 2015). Zacher et al. (2014) showed that managers can adopt a certain ‘opening’ leadership style, e.g. allowing for errors and stimulating risk taking, to encourage experimentation and exploration. Employees who prefer to engage in exploitation behaviours, such as adhering to procedures and focusing on goal achievement, should be encouraged in this behaviour too. One way to do this could be to adopt a ‘closing’ leadership style, e.g. establishing routines and monitoring goal attainment, which teases out the best from these employees (Rosing et al., 2011; Zacher et al., 2014). In that way, these employees will be motivated to continuously fine-tune processes to optimise performance.

Managers are already aware that at organisational level, exploration and exploitation are both needed. Our study suggests that managers may also realise that at an individual level it is important to connect to preferences and abilities of employees. Not every employee does necessarily have to excel in both exploration and exploitation.

Limitations and avenues for further research

Our study is subject to several limitations, each of which leads to opportunities for future research. First, when considering the generalisability of our findings, a potential limitation should be noted. Our sample was highly homogenous, pertaining to the members of one single organisation. Therefore, it is uncertain as to whether the results reported here would generalise to other samples as well, including other organisations and countries. Future studies should try to replicate this study in other settings to see if these findings apply to a broader population.

Second, similarly to many other studies of innovative work behaviour, we used cross-sectional data, which does not allow us to draw conclusions about the causality in our modelled relationships. Future studies could employ a longitudinal design to unravel the effects of exploration and exploitation on innovative work behaviour. Moreover, attention should be paid to the question how exploration and exploitation might have an impact on the different types of innovative work behaviour over time. Although we were able to show that specialisation in either exploration or exploitation is positively related to innovative work behaviour, it does not provide insight whether exploration and exploitation might have differential relationships with each of the dimensions of innovative work behaviour over time.

Third, we gathered data using self-reports, which may entail a risk for common method bias. There are well-known drawbacks of self-assessments, but there are also benefits. Individuals can continuously observe their own behaviour, and hence they can provide more examples of these than raters can (Parker & Collins, 2010). Individuals can potentially detect differences in their behaviour as a reaction to certain situations and they can do that better than raters (Lance, Teachout, & Donnelly, 1992). External raters often fall back on general impressions across all behaviours, and a halo effect manifests itself. Individual ratings overcome that drawback (Lance, LaPointe, & Fisicaro, 1994). Hence, we feel that we can justify the use of self-reports of innovative work behaviour. As reported in the method section, we undertook several procedural remedies to minimise the risk of common method bias, following recommendations of Podsakoff et al. (2003). Future research may want to investigate whether our findings are robust when adopting other indicators of innovation at the employee level, and including more than one data source.

Finally, we investigate the moderating effect of HPWS on the ambidexterity-innovation relationship. As prior research has identified HPWS as an important determinant of ambidexterity (Fu et al., 2015; Patel et al., 2013), it may be interesting for future research to investigate whether ambidexterity mediates the relationship between HPWS and innovation. HPWS may enhance different configurations of ambidexterity, which in turn may positively affect innovative behaviour.

In spite of these limitations, we believe that our study has extended our understanding about the role of exploration and exploitation for employee-innovative work behaviour.

Acknowledgement

The authors would like to thank Ilse Verdiesen who gathered the data and shared them with us.

Notes

1. The detailed results of the EFA are available upon request.