HRM Practices, Organizational Citizenship Behaviour, and Performance: A Multi-Level Analysis


  • The authors have contributed equally to this paper.

Tom Redman, Durham Business School, Durham University, Mill Hill Lane, Durham DH1 3LB, UK (


We examine the relationship between HRM practices, conceptualized at the workplace level, and individual employee attitudes and behaviour. We focus on two possible explanations for the relationship: social exchange and job influence/employee discretion. Findings from a study of employees in North-East England suggest that there is a positive impact of HRM practices on organizational citizenship behaviour, through an effect on perceived job influence/discretion. There was no such effect for perceived organizational support. These findings provide support for a job influence and opportunity explanation of HRM effects on employee attitudes and behaviour.


Much of the early work evaluating the impact of human resource management (HRM) practices on performance focused on the organizational level of analysis, and examined the effect of systems of HRM practices on organizational outcomes such as employee turnover, productivity, machine efficiency, scrap rates, customer alignment, customer satisfaction, and financial and perceptual measures of firm performance (Arthur, 1992, 1994; Becker and Gerhart, 1996; Delaney and Huselid, 1996; Delery and Doty, 1996; Huselid, 1995; Rogg et al., 2001; Youndt et al., 1996). This body of work is now extensive, with a meta-analysis of the HRM–organizational performance relationship drawing on 92 studies conducted between 1990 and 2005 (Combs et al., 2006).

More recently, attention has turned to the effects of systems of HRM practices on individual employee attitudes and behaviours (Allen et al., 2003; Kuvaas, 2008; Wright et al., 2003; Zacharatos et al., 2005). Allen et al. (2003) show that the positive relationship between ‘supportive human resource practices’ and organizational commitment, job satisfaction, and employee turnover is mediated by perceived organizational support, whilst Zacharatos et al. (2005) show that a ‘high-performance work system’ is associated with trust in management and safety climate, which mediate relationships with personal safety orientation and safety incidents. Kuvaas (2008) found that the positive relationship between ‘developmental human resource practices’ and work performance was not mediated by perceived organizational support, affective commitment, or organizational justice (procedural or interactional).

Most of the studies adopt a single unit of analysis, either individual employees or the organization/business unit. For example, Allen et al.'s (2003) two studies were conducted entirely at the individual level of analysis, using single-organization samples, with individual-level perceptual measures of HRM practices. Zacharatos et al.'s (2005) employee-level study involved respondents from two organizations, treated as a single sample, and again with individual-level perceptual measures of HRM practices. Kuvaas (2008) gathered data across 64 savings banks, but this was an employee-level study, using perceptual measures of HRM.

There are a small number of multi-level studies of HRM (Holman et al., 2009; Liao and Chuang, 2004; Sun et al., 2007; Takeuchi et al., 2009; Whitener, 2001; Wu and Chaturvedi, 2009). Most of these examine the effect of organization- or establishment-level HRM practices on individual employee outcomes, focusing on the effects of HRM on individual attitudes such as job satisfaction (Takeuchi et al., 2009; Wu and Chaturvedi, 2009), trust in management (Whitener, 2001), and organizational commitment (Takeuchi et al., 2009; Whitener, 2001; Wu and Chaturvedi, 2009). To our knowledge, no study has examined the effect of organizational or establishment-level HRM on individual organizational citizenship behaviour and in-role performance. This is an important omission, since evidence suggests that employee behaviours may have implications for organizational performance (e.g. Organ et al., 2006), and can therefore provide an explanation for the relationship between HRM and organizational performance.

In this paper, we are concerned with the relationship between HRM practices, and employees' organizational citizenship behaviour (OCB) and in-role performance. We explore the nature of the HRM practices–behaviours linkage. In addition to the social exchange explanation implied by earlier HRM studies (e.g. Allen et al., 2003; Takeuchi et al., 2009; Takeuchi et al., 2007; Zacharatos et al., 2005), we examine an additional explanation for the effects of HRM: that HRM practices enhance employee performance by providing greater intrinsic motivation and opportunity to perform through higher levels of perceived job influence and discretion. In so doing, we address the widely recognized need for a richer understanding of the linkages between HRM and outcomes (Evans and Davis, 2005; Wright and Nishii, 2007), and the suggestion that research examining the intermediate-linkages in HRM–performance research ‘should be given a high-priority by HRM scholars’ (Ferris et al., 1999, p. 394). We are also responding to calls for researchers to diversify their attention away from social exchange in analysing the antecedents of OCB (e.g. Restubog et al., 2008; Zellars and Tepper, 2003), and to suggestions that employees' job influence and discretion may be a significant element in the link between high-commitment HRM practices and performance (Delbridge and Whitfield, 2001; MacDuffie, 1995).

In the extant literature, HRM–outcomes relationships have been under-theorized (Guest, 1997). They have been addressed in terms of social exchange theory, but recently this framework in general has been criticized (Coyle-Shapiro and Conway, 2004; Cropanzano and Mitchell, 2005; Kiewitz et al., 2009). We provide a complementary perspective by examining job characteristics theory explanations of HRM–outcomes linkages. In so doing, we contribute towards a deeper understanding of the specific psychological mechanisms that may mediate the relationship between HRM and individual employee performance. Furthermore, given that employee performance has been shown to be associated with organizational performance (Ilgen and Pulakos, 1999), we believe that our analysis provides additional insights into the underlying processes and mechanisms involved in HRM effects, thereby addressing the so called ‘black box’ problem (Paauwe, 2009; Ramsay et al., 2000).

In an organizational context, discretion has been defined as an individual's right to make choices based on an authoritative assessment of the situation (Feldman, 2001, p. 164). Following this approach, we define perceived job influence/discretion as the amount of freedom of choice employees perceive they have over important aspects of their work, such as the range of tasks undertaken, the pace of work, how the job is done, working hours, when breaks are taken, and such like. According to job characteristics theory the core characteristic of autonomy/discretion produces a ‘critical psychological state’ of experienced responsibility for the work, which in turn leads to improved work effectiveness (Hackman and Oldham, 1980). Job discretion is argued to enhance employees' sense of responsibility for work outcomes and increases their willingness to go the ‘extra mile’ to complete tasks. Conversely, low job discretion may foster a sense of ‘learned helplessness’, ‘reduced industriousness’, or ‘going through the motions’ (Eisenberger, 1992; Seligman, 1975), which is likely to decrease employee performance and participation in citizenship behaviours. There is evidence suggesting a strong association between job characteristics and employee attitudes and behaviours (Humphrey et al., 2007), and we take this further by examining the mediating role of job influence/discretion in the HRM–performance relationship, so that our paper also contributes to a more complete understanding of the antecedents of OCBs.

In this study, we sample across a range of workplaces, using managers' ratings of HRM practices at the workplace level to predict employees' individual-level attitudes and behaviours. Such an approach allows us to go beyond individual perceptions of HRM and specifies practices at the workplace level of analysis. We begin by discussing the issues associated with conceptualizing and measuring HRM. We then review the relatively small number of HRM studies that have used multi-level designs. We argue that there is a need to extend this research to consider organizational citizenship and performance behaviours as outcomes and also to assess alternative mediating mechanisms in the HRM–employee outcomes relationship. Next, we examine social exchange and perceived job influence explanations for HRM effects, which provides the basis for the development of our hypotheses. We then present our methodology and findings, concluding the paper with a discussion of the implications and limitations of the study.


The specification of the ‘bundle’ of HRM practices varies considerably across studies (Boselie et al., 2005; Dyer and Reeves, 1995; Wood and Wall, 2007). To some extent this reflects different conceptualizations of the underlying work system. Work systems have been conceptualized as ‘high involvement’ (Lawler, 1986), with information sharing the key characteristic, and as ‘high commitment’ (Wood and De Menzies, 1998), with the essence being a work system that aims to encourage employees to identify with the goals of the organization and to motivate them to work hard to accomplish those goals. More recently, a ‘high performance work system’ approach is beginning to dominate HRM research. This is conceptualized as a system of interconnected HR activities, designed to ensure that employees have a broad range of superior skills and abilities, which are utilized to achieve the organization's goals, and thereby provide sustainable competitive advantage (Way, 2002; Wood and Wall, 2002).

It appears that consensus is emerging on the nature of the high performance work system construct. Wright and Boswell (2002) identify three broad conceptual categories of HRM practices. First, employee skills, with HR activities aimed at attracting talented employees and developing their skills. Second, motivation, with practices such as performance related pay aimed at eliciting high levels of effort. Third, the use of empowerment programmes to enable employee voice and influence. There appears to be growing empirical support for the impact of HR programmes conceptualized along these lines (Combs et al., 2006). Furthermore, these HR activities are seen as interdependent ‘bundles’, such that the use of one HR activity often necessitates the inclusion of others, such as performance-contingent rewards and performance appraisal. In this paper we follow this high performance work system conceptualization and define an HRM programme as a formal integrated system of HR activities that includes selective recruitment and selection, extensive training and development, regular performance appraisal, performance-contingent rewards, and high levels of employee involvement (Becker and Huselid, 1998; Zacharatos et al., 2005). Given our focus on social rather than economic exchange (see below), we examine collective performance-contingent rewards, based on work group, department, team, or organization performance, rather than on individual performance.

There have been suggestions that individual employee-rated measures of HRM may be better predictors of employees' attitudinal outcomes than are manager/employer-level HRM measures (Edgar and Geare, 2005; Khilji and Wang, 2006). However, any suggestion that employee-rated, rather than employer-rated, HRM measures account for significant variance in employee attitudes may be attributed to common method bias. In fact, some authors have found little difference between manager ratings and employee ratings in assessing the impact of HRM on attitudinal outcomes (e.g. Takeuchi et al., 2007). In this study, given the need to minimize concerns about common method bias, we use an HR manager rating of HR practices at the workplace level of analysis.


The great majority of studies of the impact of HRM tend to focus on a single level of analysis, most often the organization level (Combs et al., 2006), but with a growing number of individual-level studies (Allen et al., 2003; Kuvaas, 2008; Zacharatos et al., 2005). Despite some theorizing on multi-level HRM models (Arthur and Boyles, 2007; Bowen and Ostroff, 2004; Ostroff and Bowen, 2000; Wright and Nishii, 2007), there is relatively little empirical work adopting a multi-level approach. In part, this may reflect the practical difficulties of gaining research access to a large number of internal respondents across multiple organizations or workplaces (Wright and Boswell, 2002, p. 266).

What few multi-level studies there are have typically specified HRM as an organization- or business-unit-level variable. Sun et al.'s (2007) study of People's Republic of China hotels found that high performance HR practices, evaluated by managers at the hotel level, were positively associated with service-oriented OCB, evaluated by supervisors but aggregated to the hotel level, and that this mediated the relationships between HR practices and hotel-level voluntary labour turnover and productivity. This was a multi-level study, with 81 hotels nested within 12 cities, but there was no analysis at the employee level. Other studies have, however, involved the analysis of employee data nested within higher-level units. For example, in a study of 257 employees in 25 units of a US restaurant chain, Liao and Chuang (2004) examined the effects of three manager-rated store-level HR practices, employee involvement, training, and performance incentives, on individual employee service performance. They found that only employee involvement practices were positively associated with employee performance. In a study of 522 employees in 76 Japanese establishments, Takeuchi et al. (2009) reported that ‘concern for employees climate’, an aggregated establishment-level measure of the extent to which employees feel that the establishment values and shows concern for them, fully mediated the relationships between manager-rated establishment-level high performance work systems (HPWS) and employee's individual-level job satisfaction and organizational commitment. Wu and Chaturvedi's (2009) cross-level study of 1383 employees in 23 Asian manufacturing and service firms found that individual-level procedural justice mediated the relationship between an aggregated employee-rated measure of establishment-level HPWS and individual employee job satisfaction and organizational commitment. Finally, Whitener's (2001) study of 1689 employees in 180 US credit unions found that organizational-level (HR manager-rated) HRM practices of internally equitable rewards moderated the individual-level perceived organizational support–organizational commitment relationship, such that it was stronger in organizations with high internal equity of rewards. In addition, the relationship between perceived organizational support and trust in management was stronger in organizations with highly developmental appraisal practices, but weaker in organizations with highly comprehensive training opportunities.

These multi-level studies have focused mainly on the effects of organizational- or establishment-level HRM on individual employee attitudes. Of the studies cited above, only Sun et al. (2007) and Liao and Chuang (2004) examine employee behaviours. In the former, this is specified at the hotel- rather than the individual-level of analysis and in the latter this was ‘employee service performance’, an employee self-rating of specific customer service-related behaviours. There is clearly a need to extend the multi-level research to consider mainstream organizational citizenship and performance behaviours. This is a key aim of the present study.

In addition, we go beyond the direct OCB linkages identified by Sun et al. (2007), and we test potential mediators other than concern for employees and social exchange (Sun et al., 2007; Takeuchi et al., 2007, 2009). We follow the earlier studies by specifying HRM at the unit level of analysis. In line with our interest in the impact of HRM on employee attitudes and behaviours, we specify our mediator and outcome variables at the individual level. This raises the issue of the theoretical bases for the HRM–outcomes linkage, to which we now turn.


The relationship between an organization and its employees may be conceptualized as involving economic or social exchange. Economic exchange is based exclusively on a specific contractual relationship, requiring specific performance of contractual obligations, with no expectation of performance beyond the specified terms of the contract (Blau, 1964). Social exchange, however, involves imperfectly specified terms and a norm of reciprocity, such that discretionary benefits provided to the exchange partner are returned in a discretionary way in the longer term (Blau, 1964; Eisenberger et al., 1986). Obligations are non-specific, and trust is essential to the long-term viability of the relationship.

Employment relationships may be seen as having the characteristics of social exchange (Blau, 1964). For example, organizational justice has been seen as providing the employer's side of such an exchange, with employees reciprocating through high levels of discretionary OCB (Moorman, 1991). Furthermore, consistent with the social exchange view, the justice–OCB relationship has been shown to be mediated by trust and perceived organizational support (Konovsky and Pugh, 1994; Moorman et al., 1998), and the latter in particular has been used as an indicator of employees' perceptions of a favourable social exchange (e.g. Allen et al., 2003; Eisenberger et al., 1986; Wayne et al., 1997).

Other organizational inputs into the employment relationship have also been considered in a social exchange context. Wayne et al. (1997) considered the quality of leader–member exchange as an input into a social exchange, and terms of employment may also be seen in this way. Thus, Van Dyne and Ang (1998) found that contingent workers have lower levels of OCB, which they interpret as evidence that ‘. . . when individuals like contingent workers feel that organizations view them as short-term, temporary, or dispensable, they reciprocate by performing only required duties and minimizing citizenship behaviors’ (p. 694).

There is strong evidence that individual HRM practices impact on employee perceptions of organizational support. Providing training and development and the investment of managerial time in appraising the performance and training needs of employees sends strong messages that they are valued organizational assets (Rhoades and Eisenberger, 2002; Tansky and Cohen, 2001). Reward strategies, for example pay for performance which often pay above market rates (Pfeffer, 1998), can increase feelings of being supported and valued by the organization. Employee involvement practices can also signal that employee contributions are valued and that the organization seeks to build a social exchange relationship with employees (Allen et al., 2003).

Human resource management policies in general may be seen as an input into the social exchange process, as the evidence of positive effects of bundles of ‘high performance’ or ‘high commitment’ work practices on employee attitudes, behaviour, and turnover suggests. More specifically, HRM practices which demonstrate that the organization is committed to employees in the long term, wishes to invest in them, and is concerned about their welfare and development are likely to result in employees feeling that the organization is being supportive, and so be positively associated with OCB. Thus, Allen et al. (2003) found that ‘supportive human resource practices’ (i.e. participation in decision making, fairness of rewards, and growth opportunities) contributed to employees' perceived organizational support, which mediated the relationships between HRM practices and organizational commitment, job satisfaction, and employee turnover. They argue that this is because perceived organizational support creates feelings of obligations to repay the benefits and support received from the organization, based on the norm of reciprocity (Eisenberger et al., 1986). In other words, perceived organizational support reflects the strength of the employee's perceived social exchange relationship with the organization.

There is evidence at the business establishment level that high performance work systems are associated with employee perceptions of social exchange. In a Japanese study, Takeuchi et al. (2007) found that high performance work practices, rated by both employees and managers, were positively associated with collective human capital and establishment-level social exchange, which mediated the relationship with a subjective measure of establishment performance. Similarly, in their hotel-level study, Sun et al. (2007) argue that the mediating role of service-oriented citizenship behaviours in the relationship between high performance human resource practices and both productivity and turnover reflects a relational view of the employment relationship, with employees reciprocating the organization's favourable treatment.

Social exchange envisages the employee reciprocating the employee's supportive treatment, and one possible form of reciprocation is organizational citizenship behaviour. Two of the most commonly-cited dimensions of OCB are compliance and altruism (Organ et al., 2006). Compliance involves cooperative behaviours which help increased efficiency, such as volunteering for things that are not absolutely required by the job. Such behaviours go beyond the basic performance of job requirements, to include discretionary behaviours which reflect a cooperative adherence to the spirit as well as the letter of organizational requirements. Altruism refers to discretionary helping behaviours, such as assisting others with their work when they have been absent or are overloaded. Such behaviours benefit specific individuals, although they may also indirectly benefit the organization as a whole by increasing the effectiveness of the work group.

A review of the literature reports over 30 dimensions of OCBs, with considerable variation in the nature of behaviours (Podsakoff et al., 2000). However, ‘helping’ behaviours have been described by Podsakoff et al. (2000, p. 516) as being an important form of citizenship behaviour by ‘virtually everyone who has worked in this area’. Our focus on compliance and altruism is consistent with this ‘helping’ conceptualization of OCB, helping the organization in general (compliance) and specific individuals such as co-workers (altruism). Furthermore, in a study of the consequences of HRM, helping behaviours targeted at the organization and individuals who work there are of particular interest, given the strong linkage of these dimensions to organizational performance (Organ et al., 2006).

Based on the above, we offer the following hypotheses:

Hypothesis 1a: Perceived organizational support will mediate the positive relationship between HRM practices and OCB (compliance).

Hypothesis 1b: Perceived organizational support will mediate the positive relationship between HRM practices and OCB (altruism).

In contrast to OCB, in-role behaviour has been characterized as non-discretionary and thus as not providing currency in a social exchange relationship. However, recent debates have questioned the degree to which the discretionary–non-discretionary distinction can be neatly drawn when differentiating OCB and in-role behaviours. Harrison et al.'s (2006) meta-analytic findings suggest that behaviours typically seen as in-role, such as absence and lateness, may be better interpreted as controllable forms of input reduction and thus subject to the same motivations as OCB. Therefore, consistent with this, we suggest that Hypothesis 1 will also apply in the case of in-role behaviour.

Hypothesis 1c: Perceived organizational support will mediate the positive relationship between HRM practices and in-role behaviour.


Social exchange theory has become a strong theme in the OCB literature. However, there have recently been calls for researchers to diversify their attention away from social exchange in analysing the antecedents of OCB (Zellars and Tepper, 2003). Perceived job characteristics may provide an explanation for OCB, which lies outside the traditional social exchange explanation.

The job influence/discretion construct has been used in a wide range of studies. Job influence has been found to be an antecedent of organizational commitment (Sherer and Morishima, 1989; Snape and Chan, 2000), job and pay satisfaction (Cappelli and Sherer, 1988), job demands, stimulating work, skill development, and involvement (Petterson et al., 1995). It has also been used as a measure of ‘psychological workload’ in predicting weight gain (Overgaard et al., 2004). Holman et al. (2009) conducted a multi-level analysis (establishments nested within countries) of work design variation in call centres, finding that job discretion was negatively associated with quit rates and labour costs. Closely related concepts, such as job autonomy, task control, and decision latitude (Hackman and Oldham, 1980; Karasek, 1979), have been found to be positively related to employee wellbeing and motivation (Parker and Wall, 1998), and meta-analytic studies report relationships between job autonomy and behaviour (e.g. performance, absenteeism, and turnover intent), attitudes (e.g. job satisfaction and organizational commitment), role perceptions (e.g. role conflict and role ambiguity), and wellbeing (e.g. stress, burnout, and overload) (Fried and Ferris, 1987; Humphrey et al., 2007). In addition, job influence has been studied as an outcome variable, for example of employee involvement and participation programmes (Delbridge and Whitfield, 2001), and of family-friendly management practices (Ortega, 2009). In these studies, perceived job influence/discretion is typically seen as a job autonomy-type variable, tapping the amount of influence employees perceive they have over aspects of their work, such as the range of tasks undertaken, the pace of work, how the job is done, working hours, when breaks are taken, etc.

In this paper, we are interested in the link between HRM practices and job influence/discretion. There are suggestions that high levels of job influence are necessary for such practices to impact positively on organizational performance. For example, MacDuffie (1995, p. 198) argues that flexible production systems require an ‘enriched’, ‘motivated, skilled and adaptable’ workforce, whilst Appelbaum and colleagues' manufacturing studies draw on sociotechnical theory and job characteristics theory to argue that worker self regulation is the primary mechanism through which job design influences outcomes (Appelbaum et al., 2000; Berg et al., 1996).

Most recently, Ortega (2009) has examined the relationship between employees' ratings of their involvement in performance pay, job rotation, continuous improvement, teamwork, and vertical communication on the one hand, and employee discretion on the other. The latter was measured by asking employees whether they could choose the order in which they conduct tasks, the methods with which they work, the speed of work, timing of breaks, and working hours. Findings suggested that performance pay and vertical communications were positively associated with discretion, but other practices, such as teamwork and job rotation, were either unrelated or even negatively associated with some aspects of discretion (Ortega, 2009, p. 21).

The literature on high-commitment HRM practices has suggested that employees' degree of job influence and discretion is a significant element in the causal link between such practices and performance (Delbridge and Whitfield, 2001; MacDuffie, 1995). HRM practices which build employees' skills and involvement provide them with the ability and opportunity to exercise a higher degree of perceived influence over decision making in the job. This in turn may bring a heightened sense of self efficacy and intrinsic motivation to perform (e.g. Hackman and Oldham, 1980). In addition, job holders with higher levels of job influence may have greater potential for task-related interaction with supervisors and co-workers, and thus have a work environment more conducive to performing discretionary OCB. There is some support for the latter in Settoon and Mossholder's (2002) finding that relationship quality is associated with interpersonal citizenship behaviour.

The literature on individual HR practices and employee influence and discretion reinforces these arguments. Thus, skill development and employee involvement are positively associated with job influence (Petterson et al., 1995), and employee involvement practices provide employees with increased ‘role-making’ opportunities to expand and enrich their job roles and thus increase their positive perceptions of job influence (Evans and Davis, 2005; Wood and Wall, 2007). There is also evidence of a link between job characteristics and OCB. Studies have reported that task feedback, task routinization, intrinsically satisfying tasks, and task scope all have significant correlations with OCB (Podsakoff and Mackenzie, 1995; Podsakoff et al., 1996), and Podsakoff et al.'s (2000, p. 551) review of the literature demonstrates a consistent relationship between task variables and OCB. The suggestion is that employees who are able to exercise a high degree of influence in their jobs are motivated and enabled to perform.

The above arguments may be expected to apply to performance in general, both OCBs and in-role behaviours. As we noted in our discussion of Hypothesis 2c, we may question the extent to which OCBs are entirely and unambiguously discretionary and in-role behaviours non-discretionary. It is more likely that both aspects of performance will be seen by employees as having discretionary and non-discretionary elements, which opens up the possibility for motivational factors, as well as social exchange, to influence both. The argument developed above suggests that job discretion provides the ability, opportunity, and motivation to perform in general. We hypothesize as follows:

Hypothesis 2a: Perceived job influence will mediate the positive relationship between HRM practices and OCB (compliance).

Hypothesis 2b: Perceived job influence will mediate the positive relationship between HRM practices and OCB (altruism).

Hypothesis 2c: Perceived job influence will mediate the positive relationship between HRM practices and in-role behaviour.

The hypothesized model is shown in Figure 1. We included control variables at the individual (organizational tenure, gender, and managerial and professional job status) and workplace (age of workplace, public or private sector, and natural logarithm of total workplace employment) levels.

Figure 1.

Hypothesized model
Note: We also included control variables at the individual (workplace tenure, gender, managerial and professional job status) and workplace levels (sector, log of employment, and age of workplace).



The data came from a study of human resource management in North-East England. The sample was identified through the Arbitration, Conciliation and Advisory Service (ACAS). We used the ACAS North-East officers' contacts list to approach HR officers or senior managers to request participation in the study, and 114 workplaces initially agreed to participate. A survey facilitator was identified in each establishment, typically located in the HR department, and provided with guidance on how to distribute the surveys. Each workplace was sent a pack consisting of a questionnaire to be completed by the on-site manager responsible for HRM, a questionnaire for the senior general manager on site, and 50 questionnaires to be distributed to a sample of employees. Questionnaires had a workplace identifier and were returned direct to the university. We received at least partial responses from 60 workplaces, a 53 per cent response rate. The respondents are broadly representative of the main employers in North-East England, with an emphasis on manufacturing (for example, light engineering, defence industries, processed food, brewing, pharmaceuticals, commodity and specialist chemicals, and steel) and the public sector (NHS, local government, universities, civil service, and uniformed services – police, fire, and ambulance), and rather less emphasis on private services (e.g. TV media, distribution, privatized utilities, legal, and accountancy); small companies were underrepresented. We received 867 responses from employees in these workplaces, representing a 29 per cent response rate.

In this paper, we used data only from those workplaces for which we had responses from an on-site manager responsible for HRM and a general manager. We also excluded those workplaces for which we had fewer than three individual employee respondents (Ambrose and Schminke, 2003). Along with missing values, this reduced the sample for analysis to 28 workplaces, and 519 employees. In this sample, the number of employee respondents per workplace ranged from 7 to 40. The analysis reported is based on data from the employee, HR, and general manager questionnaires.

In this employee sample, mean organizational tenure was 12.15 years, average age was 40.25 years, 38 per cent were female, 76 per cent were married or living as married, 17 per cent were in managerial or supervisory jobs, and 20 per cent were in professional-level jobs. Looking at the workplaces, average total employment was 639, average establishment age was 47 years, and 75 per cent were in the private sector, of which 16 were in manufacturing.


The measures of perceived job influence, perceived organizational support, OCB, in-role behaviour, and the individual-level control variable (managerial status) were included in the employee questionnaire. The measure of HRM practices and the workplace-level controls (sector and employment) were taken from the HRM manager questionnaire.

Perceived organizational support was measured with four items from Eisenberger et al. (1986), responding on a seven-point scale from ‘strongly disagree’ (= 1) to ‘strongly agree’ (= 7). In selecting the items for perceived organizational support we followed the recommendations of Rhoades and Eisenberger (2002) that short forms of POS should capture the key facets of the definition of POS. Thus items 8, 9, 20, and 25 from the original scale (Eisenberger et al., 1986) were used. These are high loadings items (all between 0.72 and 0.83), which have been used in the various short forms of the scale and are drawn from the core facets of employee wellbeing (e.g. ‘the organization really cares about my well being’) and contribution (e.g. ‘the organization cares about my opinions’).

Perceived job influence/discretion was based on Magneau and Hunt's (1996) scale, responding to the question ‘Thinking about your present job, how much say do you actually have in the following decisions?’ (‘No say’ = 1 to ‘A lot of say’ = 4), with four items covering the tasks to be performed, the amount of work to be performed each day, establishing work rules and procedures, and deciding how exceptional issues are to be dealt with.

Organizational citizenship behaviour (OCB) was measured with nine items. Five items, based on Podsakoff et al. (1990), represented altruism (e.g. ‘Help others who have heavy workloads’), and four items, drawn from Smith et al. (1983), represented compliance (e.g. ‘Volunteer for things that are not absolutely required’). In-role behaviour (IRB) was measured with three items from Williams and Anderson's (1991) scale, selecting based on their high factor loadings (e.g. perform all the tasks that are expected of you [0.87]). Responses for OCB and in-role behaviour were on a five-point scale, reflecting the frequency of engagement in the activity (‘never’[= 1] to ‘always’[= 5]).

Whilst self-reported OCBs and IRBs are quite common in the literature (Podsakoff et al., 2000), it is more usual to use supervisor reports. However, in the present study, we already had a complex research design – gathering data from employees, HR managers, and general managers across multiple establishments – and it was neither feasible nor cost-effective for us to gather supervisor ratings of individual employee behaviours. In particular, our initial discussion with employers suggested that research access and response rates were likely to be compromised by requesting supervisor ratings. On balance, we felt that it was preferable to use self reports of employee behaviour, as was done in at least one earlier multi-level study of employee performance (Liao and Chuang, 2004). It should be noted that questionnaires were completely anonymous and were mailed back directly to the university by respondents, which may reduce the incentive for individuals to overstate their own citizenship and in-role behaviours. Furthermore, our concern was with the variance in such behaviours across individuals and workplaces, rather than with their overall mean levels in the sample, so that any tendency for self reports to systematically overstate such behaviours is unlikely to affect our findings.

We estimated a measurement model for the employee sample, including all the scales measured in the employee questionnaire: perceived job influence, perceived organizational support, compliance, altruism, and in-role behaviour. The hypothesized five-factor model provided a reasonable overall fit (χ2 = 748.589; d.f. = 160; GFI = 0.886; AGFI = 0.850; CFI = 0.908; RMSEA = 0.078). All indicators loaded significantly (p < 0.001) on their latent variables. A single-factor model provided a poor fit (χ2 = 4102.531; d.f. = 170; GFI = 0.511; AGFI = 0.396; CFI = 0.386; RMSEA = 0.197), with a significant deterioration in chi-square relative to the hypothesized model (change in χ2 = 3353.942; change in d.f. = 10; p < 0.01). We also found a significant deterioration in chi-square relative to the hypothesized model for 4-factor models which combined altruism and compliance (change in χ2 = 695.234; change in d.f. = 4; p < 0.01), and perceived job influence and organizational support (change in χ2 = 922.122; change in d.f. = 4; p < 0.01), and also for a 3-factor model combining altruism, compliance, and in-role behaviour (change in χ2 = 1019.116; change in d.f. = 7; p < 0.01).

Table I shows the item loadings for the 5-factor model, along with the variance extracted, construct reliabilities, and squared correlations amongst constructs. All factor loadings exceeded 0.5, apart from one in-role behaviour item. The variance extracted was at least 0.5, apart from in-role behaviour, which is close to that level, and construct reliability estimates all exceeded 0.7. These findings suggest convergent validity. There is also evidence of discriminant validity. We examined the modification indices associated with the 5-factor model, and none of these indicated standardized cross loadings in excess of 0.3. Furthermore, as shown in the bottom section of Table I, for all constructs the variance extracted substantially exceeded the squared correlations with other constructs.

Table I.  Standardized factor loadings, variance extracted, reliabilities, and squared correlations for the 5-factor measurement model (employee-rated variables)
ItemPerceived org. supportPerceived job influenceComplianceAltruismIn-role behaviour
PJI1 0.81   
PJI2 0.84   
PJI3 0.83   
PJI4 0.88   
COMP1  0.83  
COMP2  0.92  
COMP3  0.88  
COMP4  0.50  
ALT1   0.65 
ALT2   0.72 
ALT3   0.74 
ALT4   0.74 
ALT5   0.69 
IRB1    0.42
IRB2    0.87
IRB3    0.64
Variance extracted0.650.700.640.500.45
Construct reliability0.720.890.840.790.77
Interconstruct squared correlations

Ratings of HRM practices in the workplace were collected in the HR manager questionnaire. Respondents were asked to assess the percentage of employees or jobs in their workplace which were covered by a range of specified HRM practices. The specific practices included in the system of HRM practices have varied between studies (Dyer and Reeves, 1995). However, the common theme is that the practices address the recruitment, development, motivation, and involvement of employees. Accordingly, we include multiple practices which are likely to address each of these areas, as shown in the Appendix. The items describe a HPWS approach which selects staff carefully, appraises their performance regularly, motivates them with performance related rewards, and provides them with formal induction, training, and promotion opportunities, and with information and opportunities to give their views. The HR manager was asked to provide a rating for managers and professionals as one group and also for all other employees. Separate HRM practices ratings by the HR manager for the job groups of managers and professionals and for other employees were computed by averaging the 12 items, and an overall workplace rating was computed from these, weighted by the proportion of managers and professionals and other employees.

We had only a small sample of HR managers (one per workplace), with 50 workplaces for which all HR data was available, so that with 12 items in the HRM measure, the ratio of subjects to variables (STV) was only 4.17. This is less than the minimum recommended STV ratio of 5:1 often cited for exploratory factor analysis (e.g. Hair et al., 2006). Nevertheless, we performed a factor analysis, and the results are summarized in the appendix. The first four factors each had eigenvalues greater than 1 and accounted for 59 per cent of variance. A fifth factor had an eigenvalue of 1.003, but this fell below an ‘elbow’ point in the scree test and did not add to the interpretability of the solution. The rotated factor loadings for the four-factor solution are shown in the Appendix. Factor 1 related to the provision of off-the-job training, performance appraisal, quality circles, attitude surveys and formal induction for new recruits, and so appeared to represent a developmental approach to HRM (‘Development’). The second factor related primarily to the use of a rigorous selection process (‘Selection’), although the provision of employee newsletters and briefings also loaded on this factor. Factor 3 related to the use of group and organization-wide incentive pay (‘Rewards’). The final factor was concerned with on-the-job training and internal promotion, representing an internal labour market approach to HRM (‘Internal labour market’).

These findings, although they should be treated with caution given the small sample size, provide some evidence for multidimensionality. However, where separate dimensions have been identified, it has been common to combine them into a single HRM practices dimension (e.g. Bae and Lawler, 2000; Sun et al., 2007). Such an approach is consistent with the argument in the literature that it is the system of practices as a whole that constitutes a strategic asset for the organization (Becker and Huselid, 1998; Datta et al., 2005). In the case of our measure, it is significant that careful selection, employee development and involvement, internal development and promotion of employees (‘make’ rather than ‘buy’), and group-based incentives have all been identified as characteristic of the high performance work system approach to HRM (Combs et al., 2006), providing some theoretical rationale for combining them into a single measure. Finally, using a unitary measure has the advantage of model parsimony.

Following on from such arguments, and in order to provide for comparison with previous findings, we initially followed the single-dimension approach, aggregating across all 12 items to produce a single measure of HRM practices. This provides the basis for our core analysis. However, in light of the findings of the factor analysis and mindful of the possibility that job characteristics and social exchange might be differentially related to our four dimensions, we also conducted a post hoc analysis using the four HRM dimensions separately: development, selection, rewards, and internal labour market. Given the exploratory nature of this dimensional analysis, we offer no specific hypotheses here.

The level 1 control variables were organizational tenure (measured in years), gender (female = 1; male = 0), and managerial or professional job status (= 1; other job = 0), all taken from the employee questionnaire. The level 2 controls were sector (public services = 1; private sector = 0), the natural logarithm of total workplace employment, both taken from the HR manager questionnaire, and establishment age, taken from the general manager questionnaire.


Means, standard deviations, and correlations for the individual-level variables are shown in Table II. We evaluated the impact of HRM practices on employees' organizational citizenship and in-role behaviours using hierarchical linear modelling (HLM). Compliance, altruism, and in-role behaviours were modelled separately as individual-level dependent variables, with HRM practices as the workplace-level independent variable.

Table II.  Means, standard deviations, correlations, and reliabilities (individual-level variables)
 MeanS. D.12345678
  • Note: 2-tailed tests. n = 519.

  • * 

    p < 0.05,

  • ** 

    p < 0.01,

  • *** 

    p < 0.001.

Perceived organizational support4.061.44−0.11*0.16***0.15**0.05    
Perceived job influence2.790.960.010.030.35***0.19***0.42***   
In-role behaviour4.750.54−0.050.15**−0.01−0.17***0.02−0.070.22***0.19***

HRM Practices, Perceived Job Influence, and Organizational Support

A condition for mediation by perceived organization support and job influence is that HR practices are positively associated with these variables. Therefore, before estimating the models for compliance, altruism, and in-role behaviour, we estimated models with perceived organization support and perceived job influence as dependent variables. We adopted a staged approach to the HLM analysis, as shown in Tables III and IV. We began with a null model, with no level 1 or level 2 predictors. The ratio of between-group to total variance provided an intra-class correlation coefficient (ICC) of 0.128 for perceived job influence and 0.136 for perceived organizational support, suggesting that 12.8 and 13.6 per cent of the variance in perceived job influence and perceived organizational support resides between groups. Although the ICCs were small, there was statistically significant between-group variance in perceived job influence and perceived organizational support, suggesting that it was appropriate to examine level 2 predictors.

Table III.  Results of HLM analysis for the antecedents of perceived job influence
Independent variableNull modelModel 2Model 3Model 4Model 4Model 6Model 7
  1. Notes: Unstandardized coefficient estimates with robust standard errors. Estimates of the random error variance components are in parentheses. n = 519 for individual-level variables. n = 28 for group-level variables.

  2.  p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.

Level 1       
Constant2.78*** (0.12***)2.42*** (0.13**)2.37*** (0.11*)2.45*** (0.11*)2.75*** (0.14**)2.76*** (0.13**)2.22*** (0.12**)
Tenure 0.00 (0.00)0.00 (0.00)0.00 (0.00)0.00 (0.00)0.00 (0.00)0.00 (0.00)
Female (= 1; male = 0) 0.13* (0.00)0.13* (0.01)0.13* (0.01)0.13* (0.01)0.13* (0.01)0.13* (0.01)
Manager (= 1; other = 0) 1.07*** (0.08)1.05*** (0.06)1.07*** (0.06)1.08*** (0.06)1.06*** (0.06)1.05*** (0.06)
Professional (= 1; other = 0) 0.68*** (0.01)0.67*** (0.01)0.67*** (0.01)0.69*** (0.01)0.68*** (0.01)0.69*** (0.01)
Level 2       
Sector (private = 0; public = 1)  0.150.06−
Log employment  −0.07−0.06−0.03−0.07−0.04
Age of establishment
HRM practices  0.01*    
HRM dimensions       
 Development   0.01*   
 Selection    −0.00  
 Rewards     0.00 
 Internal labour market      0.01*
Within-group residual variance0.810.650.660.660.660.660.66
Model deviance1400.301300.881323.841324.451326.301325.761323.85
Table IV.  Results of HLM analysis for the antecedents of perceived organizational support
Independent variableNull modelModel 2Model 3Model 4Model 4Model 6Model 7
  • Notes: Unstandardized coefficient estimates with robust standard errors. Estimates of the random error variance components are in parentheses. n = 519 for individual-level variables. n = 28 for group-level variables.

  • † 

    p < 0.10,

  • * 

    p < 0.05,

  • ** 

    p < 0.01,

  • *** 

    p < 0.001.

Level 1       
Constant3.98*** (0.29***)3.66*** (0.40**)3.00*** (0.39*)3.22*** (0.38*)3.32*** (0.45**)3.84*** (0.39*)3.04*** (0.41**)
Tenure −0.01 (0.00)−0.01 (0.00)−0.01 (0.00)−0.01 (0.00)−0.01 (0.00)−0.01 (0.00)
Female (= 1; male = 0) 0.39* (0.06)0.35* (0.04)0.36* (0.04)0.37* (0.05)0.36* (0.11)0.36* (0.04)
Manager (= 1; other = 0) 0.70** (0.30)0.62** (0.32)0.64** (0.30)0.66** (0.26)0.59** (0.39)0.64** (0.28)
Professional (= 1; other = 0) 0.36* (0.19)0.30 (0.20)0.30 (0.18)0.32 (0.17)0.26 (0.21)0.32 (0.18)
Level 2       
Sector (private = 0; public = 1)  0.60**0.400.54*0.76**0.40
Log employment  −−0.080.02
Age of establishment  −0.00−0.00−0.00−0.00−0.00
HRM practices  0.01*    
HRM dimensions       
 Development   0.01   
 Selection    0.00  
 Rewards     0.01** 
 Internal labour market      0.01
Within-group residual variance1.811.651.651.651.651.641.65
Model deviance1819.581796.801811.921814.601815.101810.511814.31

We then estimated model 2, including the level 1 control variables only. The results suggest that managers, professionals, and women had higher levels of perceived job influence (Table III) and perceived organizational support (Table IV), with organizational tenure negatively associated with perceived organizational support only. The random error variance on the constant was significant for both dependent variables. We then proceeded with model 3, which included the level 2 predictors. According to these results, none of the level 2 controls was significant in the perceived job influence analysis (Table III), but perceived organizational support was higher in public sector workplaces (Table IV). HRM practices was positively associated with both perceived job influence (Table III), and perceived organizational support (Table IV).

HRM Practices and Employee Behaviour

Turning now to the analyses for compliance, altruism, and in-role behaviour, we used a similar staged approach, as shown in Tables V, VI, and VII. Beginning with the null models, we calculated ICCs of 0.055 for compliance, 0.065 for altruism, and 0.053 for in-role behaviour. Although the ICCs were smaller than for perceived job influence and organizational support, there was still significant between-group variance in compliance, altruism, and in-role behaviour.

Table V.  Results of HLM analysis for the antecedents of OCB (compliance)
Independent variableNull modelModel 2Model 3Model 4
  • Notes: Unstandardized coefficient estimates with robust standard errors. Estimates of the random error variance components are in parentheses. n = 519 for individual-level variables. n = 28 for group-level variables.

  • † 

    p < 0.10,

  • * 

    p < 0.05,

  • ** 

    p < 0.01,

  • *** 

    p < 0.001.

Level 1    
Constant3.01*** (0.05**)2.80*** (0.09*)2.23*** (0.06)2.68*** (0.02)
Tenure −0.00 (0.00)−0.00 (0.00)−0.00 (0.00)
Female (= 1; male = 0) 0.09 (0.03)0.08 (0.02)0.04 (0.02)
Manager (= 1; other = 0) 0.68*** (0.03)0.64*** (0.01)0.32*** (0.03)
Professional (= 1; other = 0) 0.29* (0.11)0.24* (0.11)0.05 (0.04)
Perceived organizational support   0.01 (0.00)
Perceived job influence   0.30*** (0.08)
Level 2    
Sector (private = 0; public = 1)  −0.06−0.15
Log employment  0.060.04
Age of establishment  −0.00−0.00
HRM practices  0.010.00
Within-group residual variance0.790.730.730.62
Model deviance1370.941345.341366.021313.88
Table VI.  Results of HLM analysis for the antecedents of OCB (altruism)
Independent variableNull modelModel 2Model 3
  1. Notes: Unstandardized coefficient estimates with robust standard errors. Estimates of the random error variance components are in parentheses. n = 519 for individual-level variables. n = 28 for group-level variables.

  2.  p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001.

Level 1   
Constant3.29*** (0.05***)3.11*** (0.07)3.17*** (0.05)
Tenure 0.00 (0.00)0.00 (0.00)
Female (= 1; male = 0) 0.35*** (0.01)0.34*** (0.01)
Manager (= 1; other = 0) 0.27** (0.07)0.11 (0.07)
Professional (= 1; other = 0) −0.05 (0.22*)−0.16 (0.16)
Perceived organizational support  −0.03 (0.00)
Perceived job influence  0.15** (0.02)
Level 2   
Sector (private = 0; public = 1)   
Log Employment   
Age of establishment   
HRM practices   
Within-group residual variance0.710.650.62
Model deviance1319.211301.731296.02
Table VII.  Results of HLM analysis for the antecedents of in-role behaviour
Independent variableNull modelModel 2Model 3Model 4
  • Notes: Unstandardized coefficient estimates with robust standard errors. Estimates of the random error variance components are in parentheses. n = 519 for individual-level variables. n = 28 for group-level variables.

  • † 

    p < 0.10,

  • * 

    p < 0.05,

  • ** 

    p < 0.01,

  • *** 

    p < 0.001.

Level 1    
Constant4.76*** (0.02**)4.72*** (0.05**)4.54*** (0.04**)4.51*** (0.05**)
Tenure −0.00 (0.00)−0.00 (0.00)−0.00 (0.00)
Female (= 1; male = 0) 0.19** (0.05*)0.19** (0.05*)0.17** (0.05)
Manager (= 1; other = 0) −0.01 (0.04)−0.00 (0.06)−0.02 (0.12)
Professional (= 1; other = 0) −0.19* (0.09)−0.19* (0.10)−0.21* (0.10)
Perceived organizational support   0.02 (0.00*)
Perceived job influence   −0.00 (0.01)
Level 2    
Sector (private = 0; public = 1)  −0.16*−0.24**
Log Employment  0.050.06
Age of establishment  0.000.00
HRM practices  −0.00−0.00
Within-group residual variance0.
Model deviance820.48807.40834.23836.01

We then estimated model 2, with level 1 control variables only (compliance, altruism, and in-role behaviour –Tables V, VI, and VII, respectively). These results suggest that compliance was higher for managers and professionals, altruism was higher for managers and for women, and in-role behaviour was for higher for women and lower for professionals. For compliance and in-role behaviour, but not for altruism, the model 2 results showed significant random variance components for the intercept, so that we proceeded to include the level 2 controls and HRM practices in model 3 for compliance and in-role behaviour only. According to these results (Table V and VII), of the level 2 controls only sector was significant. There was a positive direct relationship of HRM practices on compliance, but no such direct HRM relationship for in-role behaviour.

Moving to model 4 for compliance and in-role behaviour, we added the hypothesized mediators, perceived job influence, and organizational support. In the case of compliance (Table V), perceived job influence was significant, with a positive coefficient, and the coefficient for HRM practices was no longer significant. Perceived organizational support was non-significant. Taken along with the earlier results on the associations between HRM and the potential mediators, these findings suggest that perceived job influence, but not perceived organizational support, fully mediated the relationship between HRM practices and compliance, providing support for Hypothesis 2a, but not for Hypothesis 1a. For in-role behaviour (Table VII), there was no direct HRM practices effect and neither perceived organizational support nor perceived job influence were significant, providing no support for Hypotheses 1c or 2c.

Although it was inappropriate to include level 2 predictors for altruism, we did examine the effects of the mediators, perceived organizational support and perceived job influence (model 3 of Table VI). There was no significant effect for perceived organizational support, providing no support for Hypothesis 1b, but there was a significant positive effect of perceived job influence on altruism. According to Baron and Kenny (1986), the lack of a direct effect from HRM practices to altruism in the previous model rules out mediation. However, Seibert et al. (2004) argue that a significant direct relationship between the independent and dependent variables is not required for mediation, and that only relationships between the independent and mediator, and between the mediator and dependent variables are necessary, referring to this as an ‘indirect relationship’. Mathieu and Taylor (2006) suggest that the lack of a direct relationship between independent and dependent variable rules out mediation, because ‘[i]f no such relationship exists, then there is nothing to be mediated’ (p. 1038), but they also accept that an ‘indirect relationship’ does not require an initial relationship between independent and dependent variable. They recommend a sequence of decision rules, beginning with a test for full mediation and then proceeding to assess partial mediation and an indirect relationship. We followed their recommended sequence, according to which perceived job influence was not a mediator, but it did intervene in the indirect relationship between HRM practices and altruism. HRM practices were significantly associated with perceived job influence, which was significantly associated with altruism; the Sobel test of this indirect relationship was significant (Sobel test statistic = 1.95; two-tailed p = 0.05). Whilst not strictly supporting the mediation Hypothesis 2b, this finding suggests that job influence plays a role in transmitting the effects of HRM practices to altruism.

Post Hoc Analysis of the HRM Dimensions

Finally, as explained earlier, we also conducted a post hoc analysis using the four HRM dimensions identified in the factor analysis: development, selection, rewards, and internal labour market. As shown in Table III, the development and internal labour market dimensions were positively associated with perceived job influence, the selection and rewards dimensions having no significant association. As shown in Table IV, only the rewards and internal labour market dimensions were positively associated with perceived organizational support. These results suggest that job influence was associated with a developmental approach to HRM, with an emphasis on employee training, appraisal, involvement, and internal career development. In contrast, perceived organizational support was associated with collective incentive pay and internal career development, implying that such HRM practices in particular are interpreted by employees as inputs into a social exchange process.

The results for HRM dimensions were not included in Tables V, VI, and VII for reasons of space. The only significant direct association between HRM dimensions and employee behaviours was for the internal labour market dimension in the case of compliance, and this was fully mediated by perceived job influence.


Our findings suggest that HRM practices had a positive association with compliance, mediated by perceived job influence, and that perceived job influence intervened in a significant indirect association between HRM practices and altruism. HRM practices were significantly associated with perceived organization support, suggesting that such practices are seen by employees as demonstrating that the organization is concerned about their welfare and values their contribution. However, there was no association between support and OCB/IRB, so that there was no evidence that perceived organizational support mediated the relationship between HRM and employee behaviour. The implication is that the impact of HRM practices on compliance and altruism is transmitted via perceived job influence only, providing support for an intrinsic motivation and opportunity view of HRM effects on employees' organizational citizenship behaviour.

The study has implications for theory and research. Our findings on the importance of perceived job influence in the HRM practices–compliance relationship support the view that it is time to look beyond purely social exchange explanations of OCB (Zellars and Tepper, 2003). HRM practices are significant, not just as currency in a social exchange relationship with employees, but also for their role in enhancing employees' sense of job influence. The latter may provide intrinsic motivation, a sense of self confidence, and the opportunity to perform OCB. We have not examined intrinsic motivation, a sense of self confidence, and the opportunity to perform OCB, separately in the current study, but our findings suggest that the role of these issues in the HRM practices–performance linkage represents a fruitful area for further research. Studies which can examine these links in more detail would be useful.

Recent years have seen a decline in the amount of research on work design. According to Humphrey et al. (2007, p. 1332) this reflects a ‘case closed’ attitude amongst researchers, in that the motivational approach is so widely accepted that there seems to be little need for more research in this area. Wood and Wall (2007) have also emphasized that despite the prominence given to work design in seminal accounts of HRM (e.g. Walton, 1985), work design items are not common in studies of HPWS. Our findings point to the importance of work design, represented by the degree of perceived job influence and discretion, in transmitting the effects of HRM practices to individual employee behaviour and performance. There is a view that jobs in the new knowledge economies have become less uni-dimensional, less tightly defined, and less routinized (e.g. Cascio, 1998), with a greater need for employees to exert influence and use their discretion. Given such developments, along with our findings, we suggest that it would be premature to neglect the role of work design.

Our findings suggest that high performance work practices provide workers with the autonomy and discretion needed to meet to the demands of the modern workplace. Job influence and discretion may remove the need for workers to constantly check with managers for permission to act, something that is likely to be important in more complex organizational contexts. High performance work practices may have the potential to reduce the ‘under-utilization’ of employees in high-discretion work environments by tapping employees' citizenship behaviours (Huselid, 1995, p. 637). Jobs that are structured so that employee discretion and influence is limited run the risk of squeezing out employees' capacity to use their skills and abilities to the full in engaging in behaviours that help both the organization and their colleagues to find better ways of working.

Multi-level designs have the potential to bridge the gap between the hitherto largely separate research traditions of strategic human resource management (SHRM), with its emphasis on organizational level analysis of HRM system and performance outcomes, and micro-level organizational behaviour, which focuses on individuals' attitudinal and behavioural responses. Reviews of the SHRM literature have recently called for more multi-level studies (Paauwe, 2009, p. 133), and HRM researchers are now beginning to respond (Takeuchi et al., 2009; Wu and Chaturvedi, 2009). Multi-level research designs have the potential to open up the ‘black-box’ of the HRM performance linkage, since HRM is essentially a unit-level management intervention that must surely transmit to organizational-level outcomes via some kind of effect on employees' attitudes and behaviours. In this paper, we have demonstrated that company-level HRM may have an effect through an association with individual employee behaviour, and that perceived job influence is an intervening variable in this process. Multi-level studies that examine other potential mediators of the HRM–outcomes linkage would also be valuable. For example, it would be useful to consider other job characteristics as potential mediators, including feedback from the job and task identity. In addition, there are other potential alternatives to social exchange, aside from job characteristics, including the possibility that HRM has its effects through work intensification (Ramsay et al., 2000), and organizational trust and identification (Restubog et al., 2008). The former promises a stronger link with the industrial relations research tradition, stepping away from the unitarist assumptions of much of the HRM research and emphasizing the implications for employee wellbeing, whilst the latter introduces social identity perspectives as an alternative to exchange-based explanations of HRM.

Studies examining potential moderators of the relationships between HRM, job characteristics, and outcomes would also be of interest, and again such studies are likely to need to examine cross-level effects. Aside from early work on HR ‘fit’ with business strategy, research on contextual moderators is in the relatively early stages, although with some encouraging results to date (e.g. Datta et al., 2005; Sun et al., 2007). We need to learn more about the potential boundary conditions of HRM effects, and organizational or environmental moderators such as environmental uncertainty or change turbulence (Herold et al., 2007; Waldman et al., 2001), would be worthy of investigation by HRM researchers. There would also be merit in investigating potential individual-level moderators of HRM effects, including variables such growth needs strength, for example.

Our post hoc analysis of the dimensions of HRM suggested that the development and internal labour market dimensions were positively associated with perceived job influence, whilst the rewards and internal labour market dimensions were positively associated with perceived organizational support. These findings suggest that the job characteristics effects of HRM are especially associated with an approach that emphasizes internal careers and employee development, which may be seen as involving a ‘make’ rather than ‘buy’ approach to staffing, whilst social exchange is associated with financial rewards and internal careers, which may be seen as employer inputs into a social exchange process. This analysis was exploratory and post hoc rather than theory-driven, but these findings suggest that future studies should pay more attention to the potentially differential effects of specific HRM dimensions. We are not calling for the ‘unbundling’ of HRM practices, but rather a recognition that HRM may consist of more than a unitary bundle, something which was recognized in Huselid's (1995) pioneering study, but which has perhaps been neglected in some of the more recent work on SHRM.

Our findings should be interpreted in light of the limitations of the study. First, the sample is based on one English region, which may limit the potential generalizability of the findings. North-East England was traditionally an area of heavy industry with high levels of unionization and industrial militancy, and in recent years it has been characterized by structural change and a relatively high level of unemployment. Whether similar results, particularly on the non-significance of perceived organizational support, would be found in an area with a different industrial relations tradition is an interesting area for further research.

Second, the use of HRM managers' ratings of HRM practices rules out common method bias in the prediction of self-reported employee attitudes and behaviour. However, our mediator variables (perceived organizational support and job influence) came from the same employee questionnaire as OCB and in-role behaviour, raising the possibility of common method bias in the relationships between these variables. However, our measurement model (reported above) provided evidence for the discriminant validity of the employee attitudes and behaviours, including perceived organizational support and job influence, which suggests that common method bias did not account entirely for the observed relationships amongst the employee-reported constructs.

Third, we have not measured social exchange perceptions directly and studies that do so may provide a fruitful line of further research in the SHRM field (e.g. Takeuchi et al., 2007). Fourth, given that our study was conducted at the establishment level and included a mixed sample of public and private organizations, as well as manufacturing and services based firms, it was not possible to gather common establishment-level performance data of an objective nature. Future studies could usefully attempt this, perhaps in less diverse samples. Finally, our study had a cross-sectional research design, which means that we cannot draw firm conclusions about causality. Future studies might usefully adopt a longitudinal element, but given the problems of research access for multi-level studies, we must recognize that this will be very demanding.

The question of how HRM practices impact on employee attitudes and behaviour is important for management. Our conclusion, that such effects are transmitted through perceived job influence, suggests that issues of job influence and discretion are key to designing effective HRM strategy. Managers need to think beyond providing HRM practices aimed at providing benefits and support, and should consider the effect of such practices on the degree of influence employees may exert in their daily work. This will require HRM practices that build employees' skills and knowledge via training and development, provide promotion opportunities to higher level jobs, and give employees deeper and more frequent opportunities to exercise discretion in their work through employee involvement practices such as problem solving and quality improvement groups and through the design of the work itself.



What percentage of jobs/employees. . . ?1234
  1. Note: The extraction method was principal components analysis. Varimax rotation with Kaiser normalization.

Involve off-the job training, arranged and financed by the organization0.76   
Are covered by a regular (e.g. annual or 6-monthly) formal performance appraisal0.72   
Are involved in regular quality circles or similar problem solving groups discussing quality and/or workflow issues0.66   
Are asked to complete an employee attitude survey on a regular basis (e.g. annually)0.57   
Have a formal induction programme for new recruits0.57 0.31 
Are covered by an information sharing programme (e.g. employee newsletter or briefings) 0.74 0.32
Involve a sequence of two or more interviews before recruitment 0.63  
Involve a formal psychological selection test before recruitment0.340.59 −0.43
Are covered by a bonus scheme based on the performance of the work group, department, or team  0.76 
Are covered by a bonus scheme based on the performance of the establishment or organization as a whole  0.76 
Involve on-the-job training   0.79
Are normally (in more than half the cases) filled by internal promotion from within the organization rather than by recruiting from outside  0.380.58