THE ROLE OF EMPLOYEE ENGAGEMENT AND SELF-EFFICACY IN THE RELATIONSHIP BETWEEN HIGH-INVOLVEMENT WORK SYSTEMS AND

THE ROLE OF EMPLOYEE ENGAGEMENT AND SELF-EFFICACY IN THE RELATIONSHIP BETWEEN HIGH-INVOLVEMENT WORK SYSTEMS AND

INDIVIDUAL PERFORMANCE A mediated moderation analysis

Master Thesis, MScBA, specialization Human Resource Management University of Groningen, Faculty of Economics and Business

June 15, 2014

IRMA VISSERMAN Student number: S2394227

Archipelweg 60A 8924 AC Leeuwarden

Tel.: 06-16101009

E-mail: i.m.visserman@student.rug.nl

Supervisor

Prof. Dr. O. Janssen

1

THE ROLE OF EMPLOYEE ENGAGEMENT AND SELF-EFFICACY IN THE RELATIONSHIP BETWEEN HIGH-INVOLVEMENT WORK SYSTEMS AND

INDIVIDUAL PERFORMANCE A mediated moderation analysis

ABSTRACT

Prior research has focused on the relationship between high-involvement work systems (HIWSs) and performance, but the linking mechanisms between these variables remain unclear. In order to answer the question whether self-efficacy influences the mediated relationships between high-involvement work systems, employee engagement, and individual performance, a field study was conducted in a Dutch healthcare organization, using a sample of 81 employees. The results suggest that HIWSs are directly related to employee engagement. However, no positive indirect relationship was found between HIWSs and leader-ratings of individual performance through employee engagement, and self-efficacy interacts negatively in their effects on employee engagement. In contrary, a positive indirect relationship was found between HIWSs and self-ratings of individual performance through employee engagement.

Keywords: high-involvement work systems; employee engagement; self-efficacy; self- ratings of individual performance; leader-ratings of individual performance.

INTRODUCTION

In recent decades, many organizations have designed and introduced systems and practices of human resource management with the specific intention to enhance employees’

involvement in the company. High-involvement work systems (HIWSs) consist of meaningful and interrelated patterns of human resource (HR) practices that are focused on providing employees with decision-making power, access to information, development and training opportunities, and incentives for high performance (Benson, Young, & Lawler, 2006). They are designed to invest in people as resources, instead of costs to be controlled (Pfeffer, 1998), and assume to increase organizational performance (van Buren & Wemer, 1996).

Evidence of HIWSs effects on organizational performance is mixed. Research has

identified various positive outcomes of HIWSs for employees, organizations, and customers,

such as higher productivity (Arthur, 1994), profitability (Hoque, 1999), and lower turnover

(Arthur, 1994). However, the causal mechanisms linking HIWSs to performance remains

2

unclear (Edwards & Wright, 2001), in particular regarding the underlying processes that lead from HIWSs to increased motivation and performance for individual employees. It is assumed that individual employees work harder and spend more effort on their job because of HIWSs (Pfeffer & Veiga, 1999), but Cappelli, Bassi, Katz, Knoke, Osterman, and Useem (1997) argue that this could also be caused by, for example, the fear of job loss when implementing HIWSs. Also Guest (1998) argues that there is uncertainty about the relationship between systems of HR practices, employees’ attitudes, and organizational performance. He mentions that HIWSs might influence employees’ attitudes and behaviors in some way, but that it is still unclear what specific attitudes are influenced and how these attitudinal effects are related to individual work performance. Because the effects of HIWSs on individual attitudes and performance of employees are barely examined, it is not clear to what extent the practices of HIWSs influence or increase these attitudes and performance. Therefore, the current study will focus on the relationship between HIWSs and individual performance.

Some scholars suggest that HIWSs have a positive impact on workers’ individual performance. Lawler (1992) suggests that employees using the practices of HIWSs improve their individual performance because they allow them to participate in the company as a whole. More recently, Böckerman, Bryson, and Ilmakunnas (2013) also assume that HIWSs should lead to enhanced individual performance because employees do actually feel involved in the company and are willing to spend more time and effort on their tasks. Although some speculations are made about individual performance, most research has, however, been focused on organizational performance. Parker and Griffin (2011), therefore, argue that more research is needed to investigate the relationship between HR practices focused on HIWSs and individual performance.

Research has largely ignored the question whether HIWSs do indeed result in increased work engagement among individual employees, and whether this increased work engagement leads to enhanced individual performance levels. While several job attitudes as consequences of HIWSs were examined (e.g. job satisfaction, motivation; Mendelson, Turner,

& Barling, 2011), no research has yet been focused on examining the specific relationship between HIWSs and employee work engagement. This is surprising given the claims made by scholars that HIWSs may promote individual performance through enhancing employees’

engagement in the conduct of their work. Employee work engagement is “a positive, fulfilling

state of mind that is characterized by vigor, dedication, and absorption” (Schaufeli, Salanova,

González-Romá, & Bakker, 2002: 74). I propose that if the company focuses on facilitating

employees with decision-making power, access to information, development and training

3

opportunities, and incentives for high performance, they will feel more engaged in their work, because they are more enthusiastic and personally invested in the job (Alfes, Shantz, Truss, &

Soane, 2013). The intention of HIWSs is to invest in people as resources (Pfeffer, 1998), but it has not yet been examined whether these systems indeed result in elevated levels of employee work engagement. Furthermore, I propose that increased levels of work engagement induced by HIWSs would lead to enhanced individual performance. Engaged employees are expected to put more efforts and energy to their work, which should result in higher performance. Thus, I propose and test that individual perceptions of HIWSs will result in higher levels of individual performance through increased employee engagement.

However, employees may differ in the extent to which they will get engaged in response to HIWSs. Theories of person-situation interaction (e.g., Mischel & Shoda, 1995) suggest that contextual factors, such as HIWSs, do not affect individuals as if they are blank slates. Individuals have beliefs and attitudes that may filter the messages they receive from the enactment of HIWSs. The practices of HIWSs are rather challenging and could therefore be interpreted differently among employees. Some employees may consider themselves to be more capable of using HIWSs in an effective way than others do. This could be influenced by their level of self-efficacy, which is the belief of employees to be capable of doing their jobs (Bandura, 1997). I propose that employees with different levels of self-efficacy respond differently to HIWSs, and may consequently develop different levels of engagement. Ample research describes self-efficacy as a moderator in the relationship between stressors and strain (e.g. Salanova, Peiró, & Schaufeli, 2002; Stetz, Stetz, & Bliese, 2006). Less is known about the influence self-efficacy may have in how employees respond to HIWSs in terms of work engagement (Llorens, Schaufeli, Bakker, & Salanova, 2007). Self-efficacy causes employees to set themselves higher goals and acts as a self-motivating mechanism. Self-efficacious employees put more effort in their job to overcome obstacles and perceive higher levels of competence. Therefore, it is expected to play an important role in employee engagement (Llorens et al., 2007). I examine self-efficacy as a moderator in the relationship between HIWSs and employee engagement, because employees with high self-efficacy are expected to react more positively to HIWSs and, therefore, may feel more engaged in their job. Because of this enhanced engagement, employees will also perform to higher levels.

To contribute to the literature, the current study examines the relationship between HIWSs and individual performance. There is substantial debate in the literature about the presence and effectiveness of HIWSs at the individual level (Edgar & Gear, 2005).

Researchers examined HIWSs in relation to organizational performance (e.g. Appelbaum &

4

Berg, 2000), but there is a lack of information about individual employees’ perceptions of HIWSs and how these perceptions are related to employee engagement and individual performance. Thus, the present research fills this gap by examining the relationship between individual employees’ perceptions of HIWSs and their individual performance, and the mediating role of employee engagement in this relationship. Finally, more research that investigates the boundary conditions that may influence the mediated relationship between HIWSs, employee engagement, and individual performance is needed. Self-efficacy is expected to strengthen this mediated relationship. Thus, I examine a mediated moderation model in which the relationship between HIWS and individual performance is mediated by employee engagement, while self-efficacy moderated the first-stage path from HIWS to employee engagement of this mediated relationship. Figure 1 displays this mediated moderation model.

FIGURE 1 Research model

THEORY High-involvement work systems

Zatzick and Iverson (2011: 3462) define HIWSs as “systems that consist of synergistic and mutually reinforcing human resource practices that are designed to increase employees’

ability, motivation, and opportunity to contribute to the organization, which in turn is thought to positively impact performance.” The practices in HIWSs consist of providing employees with decision-making power, access to information, development and training opportunities, and incentives for high performance (Benson et al., 2006).

Employees’ decision-making power is used to let employees experience more autonomy and control in their jobs. This can be achieved by practices like employee

High-involvement work systems

Self-efficacy

Individual Performance Employee work

engagement

5

participation and using self-directed work teams in which employees have the power to make their own decisions about planning and staffing (Lawler, 1992). Furthermore, information must be accessible for employees to have the possibility to make decisions. The information should be about key processes, events, and results and could consist of mailings and feedback towards employees (Lawler, 1992). Both practices are suggested by Lawler (1992) to improve organizational performance. Moreover, development and training opportunities provide employees with knowledge about the entire work system and improve their skills and abilities. Also job rotation and cross-functional work teams could increase knowledge of employees. It increases skills that employees need to succeed in the organization and therefore leads to higher performance (Lawler, 1992). Finally, incentives for high performance could be realized by rewards. They should be given to employees to ensure that their motivation is consistent with the organizational goals. The rewards could be linked to group performance as well as individual performance and could consist of HR practices like gain sharing and profit sharing (Lawler, 1992).

In order to examine the influence of HIWS on individual performance, it is important to find out how employees perceive the practices of HIWSs, rather than assuming that employees just follow the HR policies of the company’s HIWSs (Nishii, Lepak, & Schneider, 2008). Attitudes and behaviors of employees are influenced by the way employees experience or perceive HIWSs. Different interpretations of HIWSs could be influenced by past experiences, values, or beliefs (Gratton & Truss, 2003). To date, the effectiveness of HIWSs has been primarily examined in relation to organizational outcomes (Guest, 1997). Guest suggested that the systems of HIWSs (e.g. teamwork, participation, & rewards) influence employees’ attitudes and work behavior, which ultimately influence team performance and organizational performance. A limitation in this suggested mediation process is that the link between employee attitudes and their behavior has been hardly examined (Guest, 1998). He concludes that HIWS could influence performance by re-shaping employees attitudes. The relationship between HIWSs, employees’ attitudes and behaviors, and individual performance is barely examined. Therefore, the process that leads from HIWSs to individual performance will be examined in this study.

Many researchers have examined the relationships of HR practices with different

organizational outcomes like turnover and organizational performance (Edwards & Scullion,

1982; Edwards & Wright, 2001). Also mediators were used to clarify this relationship. Job

satisfaction and motivation were examined as mediators by Allen, Shore, and Griffeth (2003),

but the outcomes were not consistent. The intention of HIWSs is to involve employees. As

6

such, employee engagement might be a mechanism that can explain the effect of HIWSs on individual performance. Employee engagement is a motivational construct that is definite, unique and valid according to empirical evidence (Schaufeli, Bakker, & Salanova, 2006). The purpose of this paper is to examine the potential mediating role of employee engagement, because employees who are challenged in their jobs by HIWSs, are more likely to be personally engaged in their job (Kahn, 1990), which may elevate their job performance.

Therefore, I will examine the mediating role of employee engagement.

High-involvement work systems and employee engagement

HIWSs specific intention is to enhance employees’ involvement in the company.

Since employee involvement is typically conceptualized as a “practice” and employee engagement as an attitude or behavior, I focus on examining employee engagement as a mediator. As Lawler (1992) stated, HIWSs could influence employee engagement, but there is still a gap in the literature concerning the question whether employees in organizations who work with HIWSs, actually experience more engagement, and whether they also perform better because of this engagement. Also Guest (1998) concludes that research is needed to examine the link between employees attitudes aroused by HIWSs and their work performance. Therefore, the current study examines whether HIWSs lead to an engaged attitude of employees and if this elevated work engagement, in its turn, results in better individual performance.

Employee engagement is defined as “a positive, fulfilling state of mind that is characterized by vigor, dedication, and absorption” (Schaufeli et al., 2002: 74). Vigor means that employees experience high levels of energy, while they are working. They experience also high energy when their tasks are challenging. Dedication means that employees experience strong involvement in their work, and a sense of significance, enthusiasm, inspiration, pride and challenge. Absorption means that employees are fully concentrated on their tasks and that time rises quickly.

To clarify the relationship between HIWSs and employee engagement, I provide

reasons why the different practices of HIWSs are linked to employee engagement. The first

practice to be examined is decision-making power. With this practice, employees should have

autonomy and control in their jobs by making their own decisions about planning and

regulating their own task activities (Lawler, 1992). Kahn (1990) found evidence that

employees feel more engaged in their work if their tasks are challenging and varied. If

employees have the power to make their own decisions about planning and regulating, their

7

jobs will be more challenging and varied. The job will indeed be more extensive, because employees do not only focus on executing their own tasks, but are also challenged to make their own planning, monitor the progress of their work, and make corrections if necessary.

This varies their jobs (vigor) and is expected to cause absorption because the job will become more challenging. Moreover, autonomy has been found to be positively related to employee engagement (Van den Broeck, Vansteenkiste, De Witte, & Lens, 2008), because employees who experience autonomy are more endorsed to the decisions they make by themselves. They have the feeling that their decisions directly influence the outcomes of their work. They are therefore more willing to contribute to the organization and are expected to experience a sense of significance (dedication).

The second practice is providing access to information about key processes, events, and results. Information could be made accessible by mailings and feedback (Lawler, 1992).

By getting feedback and information from supervisors, employees are better able to learn about their work and how it is embedded in the whole organization and they will in turn feel engaged in the organization (Schaufeli & Bakker, 2004). They could better focus and concentrate on their tasks, because they have the right information (absorption). Even if they receive information or feedback that is negative, employees could use this to improve their performance. Furthermore, the likelihood of achieving organizational goals will be higher if employees are provided with useful feedback. It is expected that employees will feel proud and inspired if they have more knowledge about the organization (dedication). If they receive information about the organizational outcomes, they know what they are working for and are expected to try harder to influence these results. This also leads to higher employee work engagement (Schaufeli & Bakker, 2004).

The third practice is the opportunity for development and training, which increases knowledge and skills of employees (Lawler, 1992). Studies about employee engagement suggest that employees will experience more engagement if they invest in intellectual effort, which could be obtained by training (Alfes, Truss, Soane, Rees, & Gatenby, 2010). Their competences will improve, and the feeling of high competences is one of the basic needs of employee engagement. It is expected that employees experience higher levels of energy if they have the possibility to develop themselves (vigor), because they keep learning and are better able to perform their jobs. Furthermore, Alfes et al. (2010) state that employees will be more engaged in the organization if they experience meaningful connections with others.

These connections could increase if the organization provides training within work teams.

Employees have a need for affiliation which plays an important role in employee engagement.

8

If employees are provided with team training, they could inspire each other and make other team members more enthusiastic (dedication).

The final practice of HIWSs is the incentive for high performance which consists of rewards for individual or group performance. Incentives are highly used to increase employee engagement. Employees receiving financial incentives experience more participation and feel more engaged in their jobs (Anderson, Grossmeier, Seaverson, & Snyder, 2008), because they know they will be rewarded if they perform well. Besides, Lawler (1986) suggests that if reward incentives are closely related to performance, employees will experience shared goals and objectives of the company and in turn higher engagement. Thus, it is expected that through the use of incentive research systems, employees feel more involved and challenged by these goals and objectives (vigor, dedication).

Gooderham, Parry, and Ringdal (2008) state that the HR practices of HIWSs have additive effects, meaning that the different HIWSs dimensions complement one another in their effects on outcome variables. When employees are provided with more dimensions of HIWSs, the overall effect on work engagement will be stronger.

Hypothesis 1. High-involvement work systems are positively related to employee work engagement.

Employee work engagement and individual performance

In sum, I have provided theoretical underpinning for the hypothesis that the components of HIWSs including decision-making power, access to information, development and training opportunities, and incentives for high performance induce employee engagement.

Furthermore, empirical evidence shows that there exists a positive relationship between employee engagement and individual performance (Rich, LePine, & Crawford, 2010). This relationship could be explained by the positive emotions of engaged employees which broaden people’s ‘thought action repertoire’. This leads them to become more attentive and absorbed in their work (Shantz, Alfes, Truss, & Soane, 2013).

Specifically, as mentioned before, employee engagement exists of vigour, dedication,

and absorption (Schaufeli et al., 2002). First of all, vigour is expected to increase individual

performance, because employees experience a high level of energy and are willing to invest

effort. This should result in the meeting of goals, because people work harder (Kahn, 1990). It

was indeed found that employees who work harder, have higher levels of performance

(Brown & Leigh, 1996). Second, dedication is predicted to increase individual performance,

9

because employees who are involved in their jobs, identify with their tasks, and have therefore more focus. They interpret more situations as opportunities and perform better (Hillman, Nicholson, & Shropshire, 2008; Kreiner, Hollensbe, & Sheep, 2006). Finally, absorption is predicted to increase individual performance, because people who experience absorption invest more energy into their work roles. They work with greater intensity, are more focused, and perform better because of that (Lepine & Crawford, 2010). Based on this theory and previous research, I hypothesize that employee work engagement is positively related to individual performance.

Hypothesis 2. Employee work engagement is positively related to individual performance.

High-involvement work systems, employee engagement and individual performance The first part of the theory describes evidence on the positive relationship between HIWSs and employee engagement. The practices of HIWSs, consisting of decision-making power, access to information, development and training opportunities, and incentives for high performance, were separately linked to employee work engagement. Evidence was found that all dimensions of HIWSs are positively related to employee work engagement. Moreover, it was stated that HR practices of HIWSs have additive effects, meaning that the different HIWSs dimensions complement one another in their effects on outcome variables (Gooderham et al., 2008). When employees are provided with more dimensions of HIWSs, the overall effect on work engagement will be stronger. Besides, in the previous paragraph, evidence was found on the relationship between employee work engagement and individual performance. The three dimensions of employee work engagement, consisting of vigor, dedication, and absorption were related to individual performance, and were predicted to have a positive influence. Therefore, employee work engagement may explain the positive effect of HIWSs on individual performance.

Hypothesis 3. Employee work engagement mediates the indirect relationship between high-involvement work systems and individual performance.

Self-efficacy

As mentioned before, employees may differ in the extent to which they will get

engaged in response to HIWSs. As the use of HIWSs might be challenging and demanding,

10

employees may differ in how they approach, interpret, and respond to these systems. To be able to use HIWSs effectively, high self-efficacy is necessary, because employees should put more effort in their jobs and could be confronted with obstacles. The basic notion is that employees with high self-efficacy produce greater effort and persistence and set higher goals, leading them to react better to HIWSs than employees with low self-efficacy (Bandura, 1997).

To my knowledge, no published study to date has examined whether and how self-efficacy may moderate the relationship between HIWSs and employees’ engagement.

Self-efficacy is defined as the “belief in one’s capabilities to organize and execute the course of action required to produce given attainments” (Bandura, 1997: 3). I have several reasons to suggest that employees with high self-efficacy will respond to HIWSs with higher work engagement levels than employees with low self-efficacy. First, decision-making power should increase employee engagement in particular if the employee has high self-efficacy.

The employees should be able to make their own decisions about planning and regulation and have more autonomy and control in their job. It could be challenging to do so, but self- efficacious employees have the core belief that they have the capacity and abilities to achieve their goals and will persevere if they experience difficulties reaching their goals (Salanova, Llorens, & Schaufeli, 2011). They will not give up and will be able to use this practice to experience more engagement in their job. Second, access to information about key processes, events and results should also increase employee engagement if the employee is self- efficacious. As mentioned before, this practice shares the organizational goals and objectives with their employees. Employees with high self-efficacy want to achieve desirable future outcomes. They want to pursue the organizational goals and will therefore be more engaged in their job and the organization (Howell, 2009). They have the willingness to spend energy and effort on completing a task and reaching the goals (Ouweneel, Le Blanc, & Schaufeli, 2011).

Third, development and training opportunities could be better used by employees with high

self-efficacy, because they try other options when they do not achieve their goals at first

(Bandura, 1997). They are therefore expected to participate in training and increase their

skills, because they want to develop themselves. Finally, incentives for high performance in

forms of rewards could increase engagement if self-efficacy is high. Self-efficacious

employees are persevering and remain confident in their tasks (Bandura, 1997), and are

therefore expected to react positively on rewards. They have the belief and confidence that

they will reach their goals and know that they will be rewarded for it. Employees with low

self-efficacy could react negatively towards rewards, because they have no confidence of

reaching their goals and if they fail they will not be rewarded. Overall, by using the practices

11

of HIWSs it is expected that employees with high self-efficacy will experience more employee engagement than employees with low self-efficacy.

Hypothesis 4. Self-efficacy moderates the relationship between high-involvement work systems and employee work engagement such that this relationship is more positive when self-efficacy is high rather than low.

For the final hypothesis, all previous hypotheses are combined to test whether there is an indirect relationship between HIWSs and individual performance through employee work engagement, and whether self-efficacy moderates this indirect relationship. The final hypothesis is:

Hypothesis 5. Self-efficacy moderates the indirect relationship between high- involvement work systems and individual performance through employee work engagement, such that this indirect relationship is more positive when self-efficacy is high rather than low.

METHOD Sample

One-hundred and ten employees in 15 teams who worked at a healthcare organization in the Netherlands were approached to participate. All information for the study was gathered through hard-copy questionnaires. There were two versions of the survey, one leader questionnaire and one employee questionnaire. The employees provided self-reports of their perceptions of HIWSs, employee work engagement, self-efficacy, and individual performance. In addition, the team leaders provided leader-ratings of employees’ individual performance. Each team member was assigned with a unique code to link the employee ratings with the leader-ratings of that employee. Participation was voluntary and confidential.

The final sample consisted of 81 respondents, resulting in an overall response rate of

73.3%. The teams ranged in size from 8 team members to 12 team members. The average age

of the employees was 43 years (s.d. = 11.25) and the average team tenure was 6 years (s.d. =

4.92). 94 % of the respondents were female. The response rate for the leader questionnaires

was 100 %.

12 Measures

High-involvement work systems were measured using eight items from the Veterans Affair Employee Survey (Harmon et al., 2003). Answers are given on a seven-point Likert- type scale, with 1 as strongly disagree and 7 as strongly agree. Sample items are: ‘Sufficient effort is made to get the opinions and thinking of people who work there’ (Decision-making power), ‘Managers let employees know how their work contributes to the organizations’

mission and goals’ (Information), ‘I am given a real opportunity to improve my skills in the organization’ (Training), and ‘Employees are rewarded for providing high-quality services to their clients’ (Incentives). The following two items of the original 10-item scale were not included in the questionnaire: 1. Employees have a feeling of personal empowerment and ownership of work processes, and 2. I feel encouraged to come up with new and better ways of doing things. These items were excluded because they do not measure high-involvement work systems but employees’ reactions to these systems in terms of their empowerment, feelings, and creativity behaviors.

Employee engagement was measured using the nine-item Utrecht Work Engagement Scale (Schaufeli et al., 2006). The nine items of the UWES-9 are grouped into three subscales reflecting the underlying dimensions of vigor (3 items), dedication (3 items), and absorption (3 items). Sample items are: ‘At my work, I feel bursting with energy’ (Vigor), ‘I am enthusiastic about my job’ (Dedication), and ‘I feel happy when I am working intensely’

(Absorption). Answers are given on a seven-point Likert scale from 1 (never) to 7 (always).

Results of previous studies showed that the UWES-9 constitutes a valid and reliable tool to assess work engagement (e.g., Schaufeli et al., 2006).

Self-efficacy was measured with an eight-item scale (Chen, Gully, & Eden, 2001). All items were scored on a seven-point Likert scale with 1 as strongly disagree and 7 as strongly agree. Sample items are: ‘I will be able to achieve most of the goals that I have set for myself’

and ‘When facing difficult tasks, I am certain that I will accomplish them.’

Individual performance was measured using Denison, Hooijberg, and Quinn’s (1995)

five-item general performance scale. The performance scale was completed by the

respondents (self-rating) and by the leaders (leader-rating). I chose for self-ratings as well as

leader-ratings for three reasons. First of all, leaders usually have less opportunities to observe

employees than the employees themselves have. This makes it hard for leaders to reflect true

performance levels (Latham & Wexley, 1982; Lawler, 1967). Second, self-raters attribute

good performance to their own behavior and poor performance on environmental attributes,

whereas leaders attribute good performance to organizational factors and poor performance to

13

behavior of the employee (DeVader, Bateson, & Lord, 1986; Jones & Nisbett, 1972). Finally, leader-raters could be subjective, and therefore they may not be able to rate individuals fairly on their true performance (Blum & Naylor, 1968). Although, employees have more opportunities to observe their own work, and there is a chance that leaders are subjective in their judgment, self-ratings and leader-ratings should somehow be gathered or combined to meet their potential (Blum & Naylor, 1968). Therefore, employees provided self-ratings on individual performance, and were rated by their direct leader on the five items of the general performance scale. Sample items of the leader-ratings are: ‘To which degree does the employee meet the performance standards?’ (1 = far below standards, 7 = far above standards), and ‘How is the performance of the employee compared to his/her peers?’ (1 = much worse, 7 = much better). Items of the general performance scale were rewritten so that they could be answered by the employee himself. Sample items are: ‘To which degree do I meet the performance standards?’ (1 = far below standards, 7 = far above standards), and

‘How is my performance compared to my peers?’ (1 = much worse, 7 = much better).

Control variables. I included the following control variables: gender (1 = male, 2 = female), age, organizational tenure, team tenure, type of contract (1 = fixed contract, 2 = temporary contracts, 3 = temporary worker, 4 = internship, 5 = other), highest education (1 = secondary education, 2 = MBO 2, 3 = MBO 3, 4 = MBO 4, 5 = higher vocational education).

These control variables might be related to one of the study variables.

Data analysis

The independent variable, the mediator, and the moderator were standardized. To test

the reliability of the scales used, the Cronbach’s Alpha was calculated. Cronbach’s Alphas for

all scales were above .70. Descriptive statistics and the relationship between variables were

examined with the use of means, standard deviations, and correlation analysis. To test

Hypotheses 1 – 3, I used the mediation analysis procedure and accompanying PROCESS

model of Preacher, Rucker, and Hayes (2007). The fourth and fifth hypotheses are tested with

the mediated moderation analysis procedures and accompanying PROCESS model outlined

by Preacher et al. (2007). An additional bootstrapping procedure with percentile confidence

intervals was used (Selig & Preacher, 2008) to test the conditional indirect relationship

between HIWSs and individual performance through employee engagement at high, middle,

and low values of self-efficacy.

14 RESULTS Descriptive statistics and correlations

Table 1 represents the descriptive statistics which includes the means, standard deviations, Cronbach’s Alpha coefficients, and correlations among the study variables.

Regarding the demographic variables, the level of education was positively correlated with individual performance. Furthermore, employee engagement was positively correlated with HIWS and with self-efficacy. Finally, self-ratings of individual performance were positively correlated with self-efficacy and with employee engagement.

Hypotheses testing

Hypothesis 1 stated that HIWSs are positively related to employee engagement. After taking the control variables of gender, age, organizational tenure, team tenure, the type of contract, and the highest level of education into account, the results of the mediation analysis revealed a positive relationship between HIWSs and employee engagement (B = .34, p = .000;

see Table 2).

Hypothesis 2 predicted that employee engagement was positively related to individual performance. As shown in table 2, the relationship between employee engagement and leader- ratings of individual performance was found to be non-significant (B = 0.05, p = .640).

However, a significant relationship was found for self-ratings of individual performance (B = 0.29, p = .011). Therefore, the second hypothesis was accepted for self-ratings of individual performance but not for leader-ratings of individual performance.

Hypothesis 3 stated that employee work engagement should mediate the indirect relationship between high-involvement work systems and individual employee performance.

Because the effect of employee engagement on leader-ratings of individual performance was not significant, a mediation could not emerge. Indeed, the bootstrapping procedure with percentile confidence interval indicated that HIWSs and individual performance were not indirectly related through employee engagement (indirect effect = 0.02; standard error = 0.04;

95% confidence interval = -.070 to .092; see Table 2). Thus, no support for hypothesis 3 was

found for leader-ratings of individual performance. However, the bootstrapping procedure

indicated that there was a significant indirect relationship between HIWSs and individual

performance through employee engagement for self-ratings of individual performance

(indirect effect = 0.11; standard error = 0.05; 95% confidence interval = .042 to .222; see

Table 2). Therefore, support for hypothesis 3 was found for self-ratings of individual

performance.

15 TABLE 1

Descriptive statistics and correlations for study variables

Variables M SD 1 2 3 4 5 6 7 8 9 10 11

1. Gender 1.94 .24

2. Age 42.86 11.25 .05

3. Org. tenure 11.87 9.82 -.080 .472**

4. Team tenure 5.98 4.92 -.232* .332**

.386**

5. Type of contract 1.04 .19 .05 -.295** -.237* -.234*

6. Highest education 3.44 .90 -.221* -.245* -.178 -.099 .198

7. HIWS 4.80 1.03 -.095 .015 .020 .017 -.025 -.049 (.88)

8. Self-efficacy 5.27 .73 -.091 -.029 -.049 .023 -.05 .037 .176 (.91)

9. Employee engagement 5.71 .83 .035 .132 .085 -.021 -.099 -.153 .406** .630** (.92)

10. Leader-rat. Ind. Performance 4.67 .90 -.107 -.132 -.106 -.019 .118 .320** .124 .033 .058 (.94)

11. Self-rat. Ind. Performance 5.32 .71 .016 .076 .052 -.024 -.144 -.136 .183 .659** .445** .160 (.93)

Note: Cronbach’s alpha coefficients are on the diagonal in parentheses. N=81. *p < .05. **p < .01.

TABLE 2

Results of mediation analysis

Employee engagement

Predictor B SE t p

Constant 5.72 .87 65.38 .000

Gender .03 .09 .31 .758

Age .08 .11 .76 .447

Org. tenure .03 .11 .32 .748

Team tenure -.08 .10 -.78 .440

Type of contract -.05 .09 -.51 .610

Highest education -.08 .10 -.81 .418

HIWSs .34 .09 3.84 .000

R² .21

Leader-ratings of individual performance

Predictor B SE t p

Constant 4.69 .10 45.46 .000

Gender -.01 .11 -.14 .819

Age -.06 .12 -.46 .648

Org. tenure -.03 .12 -.25 .806

Team tenure .03 .12 .28 .777

Type of contract .06 .18 .32 .748

Highest education .27 .11 2.52 .014

HIWSs .12 .11 1.05 .298

Employee engagement .05 .11 .47 .640

R² .14

Indirect relationship between HIWSs and leader-ratings of individual performance

through employee engagement Indirect effect SE 95% confidence interval*

.02 .04 -.070, .092

*Based on 1000 bootstrap samples

Self-ratings of individual performance

Predictor B SE t p

Constant 5.32 .08 68.78 .000

Gender -.02 .08 -.23 .819

Age .00 .09 .03 .978

Org. tenure .00 .09 .01 .990

Team tenure -.04 .09 -.49 .627

Type of contract -.06 .09 -.64 .523

Highest education -.05 .08 -.64 .521

HIWSs .01 .08 .12 .905

Employee engagement .29 .09 3.41 .001

R² .21

Indirect relationship between HIWSs and self-ratings of individual performance

through employee engagement Indirect effect SE 95% confidence interval*

.11 .05 .042, .222

*Based on 1000 bootstrap samples

17

Hypothesis 4 predicted that self-efficacy moderates the relationship between HIWSs and employee engagement such that this relationship is more positive when self-efficacy is high rather than low. As presented in table 3, HIWSs and self-efficacy were found to significantly interact in their effects on employee engagement (B = -.26, p = .000). The interaction effect is shown in figure 2. Contrary to Hypothesis 4, the plot shows that employee engagement was relatively high for employees with high self-efficacy, regardless of HIWSs;

the simple slope for high self-efficacy is almost flat. However, the simple slope for employees with low levels of self-efficacy was positive and strong, indicating a positive relationship between HIWSs and engagement for employees with low self-efficacy. Therefore, although an interaction effect between HIWSs and self-efficacy in prediction of employee engagement emerged, Hypothesis 4 was rejected. I come back to this outcome in the discussion section.

FIGURE 2

Relationship between HIWSs and employee engagement moderated by self-efficacy

Finally, hypothesis 5 predicted that self-efficacy moderates the indirect relationship

between HIWSs and employee performance through employee engagement, such that this

indirect relationship is more positive when self-efficacy is high rather than low. When HIWSs

and self-efficacy were controlled for, employee engagement was not significantly related to

leader-ratings of individual performance (B = 0.05, p = .640; see table 3). A bootstrapping

18 TABLE 3

Results of mediated moderation analysis

Employee engagement

Predictor B SE t p

Constant .09 .07 1.23 .222

Gender .06 .08 .84 .404

Age .13 .09 1.50 .139

Org. tenure .03 .09 .29 .775

Team tenure -.09 .09 -1.06 .294

Type of contract .20 .13 1.61 .113

Highest education -.11 .08 -1.35 .180

HIWSs .25 .07 3.40 .001

Self-efficacy .52 .08 6.79 .000

HIWSs x self-efficacy -.26 .07 -3.86 .000

R² .64

Leader-ratings of individual performance

Predictor B SE t p

Constant 4.69 .10 45.46 .000

Gender -.01 .11 -.14 .819

Age -.06 .12 -.46 .648

Org. tenure -.03 .12 -.25 .806

Team tenure .03 .12 .28 .777

Type of contract .06 .18 .32 .748

Highest education .27 .11 2.52 .014

HIWSs .12 .11 1.05 .298

Employee engagement .05 .11 .47 .640

R² .14

Indirect relationship between HIWSs and

leader-ratings of individual performance through employee engagement at high, middle, and low values of self-efficacy.

Self-efficacy Indirect effect SE 95% confidence interval*

High (M + 1 SD) -.00 .01 -.034, .020

Middle .01 .03 -.040, .081

Low (M – 1 SD) .02 .05 -.082, .138

*Based on 1000 bootstrap samples

Self-ratings of individual performance

Predictor B SE t p

Constant 5.32 .07 68.78 .000

Gender -.02 .08 -.23 .849

Age .00 .09 .03 .978

Org. tenure .00 .09 .01 .990

Team tenure -.04 .09 -.49 .627

Type of contract -.06 .10 -.64 .523

Highest education -.05 .08 -.64 .521

HIWSs .01 .08 .12 .905

Employee engagement .29 .09 3.41 .011

R² .21

Indirect relationship between HIWSs and self-ratings of individual performance through employee engagement at high, middle, and low values of self-efficacy.

Self-efficacy Indirect effect SE 95% confidence interval*

High (M + 1 SD) -.00 .03 -.068, .064

Middle .07 .03 .022, .146

Low (M – 1 SD) .14 .05 .063, .272

*Based on 1000 bootstrap samples

19

procedure with percentile confidence interval indicated that HIWSs and leader-ratings of individual performance were not indirectly related through employee engagement at both high levels of self-efficacy (M + 1 SD; indirect effect = -.00; Standard Error = 0.01; 95%

confidence interval = -.034 to .020) and low levels of self-efficacy (M – 1 SD; indirect effect

= 0.02; Standard Error = 0.05; 95% confidence interval = -.082 to .138). Thus, there is no indirect relationship between HIWSs and leader-ratings of individual performance through employee engagement, regardless of the level of self-efficacy. However, when HIWSs and self-efficacy were controlled for, employee engagement was significantly related to self- ratings of individual performance (B = 0.29, p = .011; see table 3). The bootstrapping procedure suggested that HIWSs and self-ratings of individual performance were not indirectly related through employee engagement at high levels of self-efficacy (M +1 SD;

indirect effect = -.00; Standard Error = 0.03; 95% confidence interval = -.068 to .064).

Nevertheless, HIWSs and self-ratings of individual performance were directly related through employee engagement at low levels of self-efficacy (M – 1 SD; indirect effect = 0.14;

Standard Error = 0.05; 95% confidence interval = .063 to .272). Contrary to hypothesis 5, the bootstrapping procedure shows that self-efficacy does moderate the indirect relationship between HIWSs and self-ratings of individual performance through employee engagement, such that this indirect relationship is more positive when self-efficacy is low rather than high.

Therefore, there is no support for hypothesis 5, regardless the use of leader-ratings of individual performance or self-ratings of individual performance.

DISCUSSION

The goal of this research was to examine the relationship between HIWSs and

individual performance, and whether employee engagement would mediate and employee

self-efficacy would moderate this relationship. Individual performance was measured by

employees’ self-ratings as well as leader-ratings, because different raters have different

definitions of effective performance and, thus, the same individual performance may be

assessed differently (Landy & Farr, 1980). As can be seen in table 1, a low correlation of .16

can be found between self-rating and leader-rating. This means that there is low agreement

between the leader and the self on the performance of the employee. Despite of the low

correlation between self-rating and leader-rating, it is used in this study, because numerous

advantages of using multiple rates have been cited. First of all, greater reliability exists with

the use of multiple raters, there is more fairness for the rated employee, and the rated

employee will accept the outcomes better (Latham & Wexley, 1982). Finally, multiple raters

20

are used in this study, because raters at different levels of the organization could measure performance differently. The leader is in a different layer of the organization than the team- member, so they may weigh performance dimensions dissimilarly (Borman, 1974).

A mediated moderation model was used to examine the indirect relationship between HIWSs and individual performance through employee engagement, moderated by self- efficacy for the first stage that links HIWSs to employee engagement. Study results support the relationship between HIWSs and employee engagement. This means that individuals experience more engagement in jobs where employees are provided with decision-making power, access to information, development and training opportunities, and incentives for high performance (Benson et al., 2006). Nevertheless, it was not supported that employee engagement leads to higher individual performance when performance was rated by the leader. However, employee engagement did lead to higher performance, when performance was rated by employees’ self-reports. A reason why this relation was significant and the relationship between employee engagement and leader-rater performance was not, could be that leaders weigh performance dimensions differently than employees do. Self-raters attribute good performance to their own behavior and poor performance on environmental attributes, whereas leaders attribute good performance to organizational factors and poor performance to behavior of the employee (DeVader et al., 1986; Jones & Nisbett, 1972).

Besides, leaders usually have less opportunities to observe employees than employees themselves have. This makes it hard for leaders to reflect true performance levels (Latham &

Wexley, 1982; Lawler, 1967). On the contrary, employee defensiveness could have had an influence on the differences in ratings. Employees may inflate their rating, in order to enhance their evaluation. If they rate themselves high, they expect that their evaluation will be high as well (Holzbach, 1978). Therefore, also leader-ratings are used to measure individual performance. The mediating role of employee engagement was rejected with leader-ratings of individual performance, but accepted with self-ratings of individual performance.

It was proposed that self-efficacy operates as a boundary condition which moderates between HIWSs and employee engagement. It was found that HIWSs and self-efficacy interact negatively in their effects on employee engagement. Thus, the relationship between HIWSs and employee engagement is more positive when self-efficacy is low rather than high.

This means that employees with high self-efficacy experience relatively high levels of

employee engagement, even if HIWSs are present at a low level. Employees with low self-

efficacy feel more engaged in their jobs if they could use more HIWSs. There is a strong link

between self-efficacy and engagement. Higher self-efficacy leads to higher engagement, apart

21

from HIWSs. So employees with high self-efficacy do not need HIWSs to get engaged in their job. Self-efficacy itself is enough to meet high levels of vigor, dedication, and absorption. Employees with low self-efficacy are more dependent on sources they do not possess themselves, like HIWSs, to get more engaged in their job. Thus, HIWSs do in particular work well for employees with low self-efficacy.

Finally, it was proposed that the indirect relationship between HIWSs and individual performance through employee engagement would be more positive when self-efficacy was high rather than low. Findings show that HIWSs are not indirectly related to leader-ratings of individual performance through employee engagement when self-efficacy was high rather than low. However, an indirect relationship was found between HIWSs and self-ratings of individual performance through employee engagement when self-efficacy was low.

Theoretical implications

This study has several theoretical implications. The research was focused on the linking mechanisms between HIWSs and individual performance. The causal mechanisms linking HIWSs to performance was not clear in previous literature, and particularly the underlying processes that lead from HIWSs to increased individual performance was largely ignored. The results contribute to the literatures on HIWSs, employee engagement, and self- efficacy.

First of all, the study examined the relationship between HIWSs and employee engagement. Results show that HIWSs indeed enhanced employee engagement at the individual level. This finding contributes to literature on HIWSs and to literature on employee engagement, because it has not been found before. Moreover, employee engagement appeared to function as a mediator between HIWSs and self-ratings of individual performance.

However, this relation could not be confirmed when individual performance was rated by the leader. HIWSs may not directly relate to individual performance, they do indeed lead to positive work attitudes like employee engagement.

Second, this study examined the possibility that the indirect relationship between

HIWSs and employee engagement would be more positive when self-efficacy was high rather

than low. The results are different than expected, because it turned out that the relationship

between HIWSs and employee engagement is more positive when self-efficacy was low

rather than high. It is an interesting implication that high self-efficacy, as a feature of the

individual employee, works as a substitute for HIWSs to encourage employee engagement,

whereas employees with low self-efficacy actually need HIWSs to experience employee

22

engagement. Therefore this is an important contribution to the literature on HIWSs, self- efficacy, and employee engagement.

Finally, this study makes a contribution to the literature by examining individual performance with the use of leader-ratings as well as self-ratings. When employees had the opportunity to rate themselves, a positive relationship was found between HIWSs and individual performance, and between employee engagement and individual performance, whereas this relationships were not found when individual performance was rated by the leader. A low correlation of .16 was found between self-ratings and leader-ratings. Also previous research shows low self-leader correlations of .05 (Klimoski & London, 1974) and .02 (Ferris, Yates, Gilmore, & Rowland, 1985) on ratings of individual performance.

Nevertheless, also higher correlations of .56 (Pym & Auld, 1965), and .60 (Williams & Seiler, 1973) were found, which means that there is a wide range of findings. Blum and Naylor (1968) argue that leader-ratings are a rich source of data that cannot just be abandoned.

Despite the fact that leader-raters could be subjective, they should somehow be gathered to meet their potential. Also self-ratings are important, because they offer a range of information that can be provided by the self (Mabe & West, 1982). Moreover, evidence shows that self- rating shows less halo error than leader-rating (Baird, 1977), which means that leaders do not rate their employees on the different dimensions of performance, but on an overall judgment (Holzbach, 1978). However, self-raters could be biased, while leader-raters share a set of common perceptions (Holzbach, 1978). Moreover, as mentioned before, different raters may define performance differently (Borman, 1974). Therefore, implications are made to literature on leader-ratings and self-ratings of individual performance.

Limitations and future directions for research

There are a number of limitations in this research worth noting. First, a limitation of this research concerns the representativeness of the sample. The sample consisted of predominantly female respondents (94%) from a healthcare organization. Perceptions of HIWSs and employee engagement might vary with gender. Moreover, it remains unclear to what extend similar results would be obtained in organizations in different sectors. Therefore, further examination with generalization of the present results to men, and other organizations, is needed to figure if perceptions of HIWSs and employee engagement are differently.

Second, results about HIWSs might be disfigured, because employee perceptions of

HIWSs were drawn from a single source. Blunt conclusions might be a result, because of

23

common method bias (Harmon et al., 2003). However, according to Judge, Locke, Durham, and Kluger (1998), this may not be as much of a limitation as is often thought.

Third, and in relation to the previous limitation, HIWSs were only assessed by measuring perceptions of the employee. More objective indicators could be used to reduce assessors’ distortions. However, Cordes and Dougherty (1993) argue that the measurement of HIWSs is most valid when using the perceptions of employees, it cannot be confirmed with this study if the employees’ perceptions reflect reality. Future research might develop a model to measure leaders’ perceptions of HIWSs to assess similarities and differences.

A fourth limitation in this study is the broad concept of HIWSs. According to Harmon et al. (2003) HIWSs receive little support because of their difficulty and uncertainty. To make the practices of HIWSs more obvious the different parts of the system could be measured separately to examine to which practices employees respond better. Future research can take advantage of this, by measuring differential relationships between the different dimensions of HIWSs with different outcomes.

Fifth, this study uses self-ratings as well as leader-ratings to measure performance.

Results are disconcerting, because most hypothesis are accepted when using self-ratings of performance, but are rejected when using leader-ratings of performance. This is caused by common method bias, because there is only a low correlation of .16 between self-ratings and leader-ratings. A disadvantage of using self-ratings is that there is inconsistency in self- evaluation validity, meaning that one individual could interpret questions and rating scales differently than other individuals (Mabe & West, 1982). Future research could incorporate a third party to evaluate performance. Peer-ratings could be used to decrease common method bias.

Practical implications

Despite the limitations, the results of this study have several practical implications for

healthcare organizations. First of all, the results add to the growing evidence that HIWSs

enhance employee engagement. Findings are extended to the healthcare sector, where

relatively little evidence existed. Information and results of the current study could be used by

managers to create work environments for employees of healthcare organizations, by using

HIWSs to enhance employee engagement. An advantage of HIWSs is that they are controlled

by the organization. When organizations implement more HIWSs and provide employees

with more decision-making power, access to information, development and training

24

opportunities, and incentives for high performance (Benson et al., 2006), they are more likely to enhance their engagement.

Second, the study examined the moderating effect of self-efficacy in the relationship between HIWSs and employee engagement. Despite the fact that the results are different than expected, because the relationship between HIWSs and employee engagement is more positive when self-efficacy was low rather than high, it is an important implication for practitioners. High self-efficacy turns out to be a substitute for HIWSs to encourage employee engagement, whereas low self-efficacious employees actually need HIWSs to experience employee engagement. This means that HIWSs are positive predictors for employee engagement, even when self-efficacy of the employee is low. Therefore, it is recommended to invest in HIWSs, because it causes employees to feel engaged in their jobs.

Finally, results of the self-ratings and leader-ratings show that there is a low correlation between these two types of performance appraisal. Practitioners should be aware that there may be much disagreement between members of the organization when using self- ratings (Harris & Schaubroeck, 1988). However, it turned out that self-ratings also have advantages. For example, self-ratings can improve personal development, they attribute to better communication between the self and the leader, and they clarify different opinions of the self and the leader (Carroll & Schneier, 1982; Cummings & Schwab, 1973). Therefore it is recommended to managers to use self-ratings of individual performance for purposes like performance evaluation for administration, identification of training needs, or to test evaluation of a training (Thornton, 1980).

CONCLUSION

In conclusion, the present results show that HIWSs are directly related to employee engagement. Nevertheless, no indirect relationship was found between HIWSs and leader- ratings of individual performance, through employee engagement. However, a positive indirect relationship was found between HIWSs and self-ratings of individual performance.

Moreover, self-efficacy interacts negatively in their effects on employee engagement,

meaning that the relationship between HIWSs and employee engagement is more positive

when self-efficacy is low rather than high. The results may support future research related to

HIWSs and employee engagement. Finally, in terms of practical implications, it is

recommended to invest in HIWSs, because high self-efficacy turns out to be a substitute for

HIWSs to encourage employee engagement, whereas low self-efficacious employees actually

need HIWSs to experience employee engagement.

25

REFERENCES

Alfes, K., Shantz, A. D., Truss, C., & Soane, E. C. 2013. The link between perceived human resource management practices, engagement and employee behavior: a moderated mediation model. The International Journal of Human Resource Management. 24(2):

330–351.

Alfes, K., Truss, C., Soane, E.C., Rees, C., & Gatenby, M. 2010. Creating an engaged workforce. Wimbledon: CIPD.

Allen, D.G., Shore, L.M., & Griffeth, R.W. 2003. The role of perceived organizational support and supportive human resource practices in the turnover process. Journal of Management. 29: 99–118.

Anderson, D.R., Grossmeier, J., Seaverson, E.L.D. &, Snyder, D. 2008. The role of financial incentives in driving employee engagement in health management. ACSM’s Health &

Fitness Journal. 12(4): 18-22.

Appelbaum, E., & Berg, P. 2000. High performance work systems: giving workers a stake. In Blair, M., & Kochan, T. (Eds), The new relationship: human capital in the American corporation. 102-44. Washington, DC: Brookings Institution Press.

Arthur, J.B. 1994. Effects of human resource systems on manufacturing performance and turnover. Academy of Management Journal. 37(3): 670-687.

Baird, L.S. 1977. Self and supervisor ratings of performance: as related to self-esteem and satisfaction with supervision. Academy of Management Journal. 20: 291 – 300.

Bandura, A. 1997. Self-efficacy: the exercise of control. New York: Freeman.

Benson, G.S., Young, S.M., & Lawler, E.E. 2006. High-involvement work practices and analysts’ forecasts of corporate earnings. Human Resource Management. 45(4): 519–537.

Blum, D., & Naylor, J. 1968. Industrial psychology: Its theoretical and social foundation.

New York: Harper & Row.

Böckerman, P., Bryson, A., & Ilmakunnas, P. 2013. Does high involvement management lead to higher pay? Journal of the Royal Statistical Society. 176(4): 861 – 885.

Borman, W.C. 1974. The rating of individuals in organizations: an alternate approach.

Organizational Behavior and Human Performance. 12: 105-124.

Brown, S. P., & Leigh, T. W. 1996. A new look at psychological climate and its relationship to job involvement, effort, and performance. Journal of Applied Psychology, 81: 358–368.

Cappelli, P., Bassi, L., Katz, H., Knoke, D., Osterman, P., & Useem, M. 1997. Change at

Work. New York: Oxford University Press.