Exposure to versus participation in human resource management practices: what is important for employee job satisfaction and commitment?

Transcription

1 Exposure to versus participation in human resource management practices: what is important for employee job satisfaction and commitment? Laetitia Hauret (LISER, Luxembourg Institute of Socio-Economic Research) 1 Ludivine Martin (LISER, Luxembourg Institute of Socio-Economic Research & CREM UMR CNRS 6211) 2 Nessrine Omrani (PSB Paris School of Business) 3 Donald R. Williams (Kent State University) 4 Preliminary version February 2016 Abstract: This paper takes advantage of a recent employer-employee database from a service-based European economy to examine the relationships between the Human Resources Management (HRM) strategy of the employer and employees job satisfaction and organizational commitment. These attitudes are the key mediating variables between the HRM strategy of the firm and its performance. The paper identifies which is the more important link to employees attitudes, the exposure to the HRM strategy or the effective participation of employees in HRM practices. This paper is the first to combine the two factors in a single analysis. The main results show that a high exposure to HRM is not sufficient to obtain positive employee attitudes and that effective employees participation in HRM is required. By deepening the examination of differences in employees attitudes by comparing groups of employees who face different levels of HRM and who are involved in different levels of HRM, we find that differences in employees involvement in HRM practices influences the way they react to their personal, occupational and workplaces characteristics. Keywords: Human Resources Management; exposure; participation; employee; job satisfaction; organizational commitment JEL Codes: J28; M12; M5; R23 1 LISER - Luxembourg Institute of Socio-Economic Research, 11 Porte des sciences, L-4366 Esch-sur-Alzette Luxembourg; Tel.: ; laetitia.hauret@liser.lu 2 LISER - Luxembourg Institute of Socio-Economic Research, 11 Porte des sciences, L-4366 Esch-sur-Alzette Luxembourg ; CREM, Faculté des Sciences Economiques, 7, Place Hoche, Rennes Cedex France. Tel.: ; Fax: ; ludivine.martin@liser.lu corresponding author. 3 PSB Paris School of Business, 59 rue Nationale, Paris France. Tel: ; n.omrani@psbedu.paris 4 Kent State University, College of Business Administration, 475 Terrace Drive, Kent, Ohio United States; Tel: ; dwilliam@kent.edu 1

2 1 Introduction In recent years, a growing body of research has devoted attention to assessing the relationships between Human Resource Management (HRM) practices and firm performance (for a review, see Becker and Huselid, 2006). Various scholars emphasize that HRM practices enhance organizational performance across sectors, sample characteristics, HRM practices studied, and organizational performance measures used (Bryson et al., 2005; Capelli and Neumark, 2001; Huselid, 1995; Ichniowski et al., 1997; MacDuffie, 1995; Ramsay et al., 2000). The HRM strategies appear to be effective when organizations define a unique HRM architecture that builds a competitive advantage that in turn improves the performance of the firm, in line with the resource-based view (Barney, 2001). Scholars also put forth that HRM practices need to be adopted as a bundle to achieve positive results (Bowen and Ostroff, 2004; Combs et al., 2006; Godard, 2004; Wood, 1999). The assumed underlying process ranges from the HRM system, the resulting increase of workforce skills, knowledge, empowerment, employee engagement and commitment, to firm performance (Combs et al., 2006; Macky and Boxall, 2007). Some of the research focuses on the linkages between HRM practices and employees attitudes such as job satisfaction, organizational commitment or motivations. Generally speaking, the research takes two perspectives. On one hand, some researchers follow the organizational perspective and examine the exposure of employees to the HRM strategy through the adoption and the level of diffusion of practices inside the workplace (Ramsay et al., 2000; White and Bryson, 2013; Wood and De Menezes, 2011). In this stream of research, employees are neglected while human assets are the mediating variables between the HRM strategy and performance. On the other hand, several papers focus on the employee perspective and look at the participation or involvement of employees in HRM practices (Böckerman et al., 2012; Gallie et al., 2001; Godard, 2001; 2010; Guest, 1999; 2002; Kalmi and Kauhanen, 2008; Macky and Boxall, 2007; Martin, 2016a; 2016b; Mohr and Zoghi, 2008). Integrating the two approaches is useful to deepen the understanding about the black box that links the HRM strategy of the employers and subsequent organizational performance. As employers need to understand how to generate and sustain the expected positive returns for the firm a question remains unanswered: How does HRM contribute to firm performance? Essentially, is exposure sufficient or is effective participation of employees is required to obtain positive results? 2

3 To examine this research question, we use recent employer-employee data collected in a small open European country characterized by the predominance of its service sector, namely Luxembourg. The data constitute a representative sample of employees of the private sector working in workplaces of at least 15 employees. Due to the specificity of the Luxembourgish labor market characterized by a large proportion of foreign employees, the results apply not only to Luxembourgish workers but also French, Belgian, German, Portuguese and workers of other nationalities. This paper contributes to the literature in several ways. First, using more recent data than that used by other scholars in the HRM literature, we address the question of whether the link between HRM practices and employees attitudes is still relevant. Second, we investigate the relationships between the HRM strategy of the firm and employees attitudes using both the exposure and the participation perspectives made possible by the use of matched employeremployee data. The two attitudes examined in this paper are job satisfaction and organizational commitment. These are the two facets of overall job attitudes that are key mediating variables between the HRM strategy and firm performance (Harrison et al., 2006). Third, we deepen the examination of differences in employees attitudes, using the Oaxaca- Blinder decomposition method, by comparing groups who face different levels of HRM exposure and who have different levels of HRM participation. More precisely, we identify how much of the difference in employees attitudes between groups is due to the fact (i) that they have different characteristics (personal, occupational and workplace) or (ii) that they value job characteristics differently. The exposure and the participation in HRM can indeed modify the way that employees react to their work environment. The results support integrating the both organizational and employees approaches in the analysis of employees attitudes. The results indicate that a high level of exposure to HRM is not sufficient to improve employees attitudes; effective participation of employees is required. Moreover, our results show that the differences in employee involvement in HRM practices affect the way they react to their personal, occupational and workplace characteristics. The paper proceeds as follows. Section 2 reviews the literature and develops hypotheses. Section 3 describes the data, the variables and the estimation strategy. Section 4 presents and discusses the results. Section 5 concludes. 3

4 2 Related literature A growing body of research supports the existence of a positive relationship between the HRM architecture of the firm and its performance. This result holds across sectors, sample characteristics, HRM practices 5 studied, and organizational performance measures used (for a review, see Becker and Huselid, 2006). The positive relationships between HRM practices and firm performance are often explained by supposed positive linkages with employees responses. The HRM strategy is a way for managers to redistribute to employees firm-level productivity gains and to convince them that their work values and their contribution to the firm performance is recognized (Kalmi and Kauhanen, 2008; Osterman, 2000). By increasing employees skills and knowledge and empowering employees to make decisions, the HRM strategy contributes to the firm-level performance. As emphasized in the literature, HRM practices are introduced in order to obtain employees positive behaviors such as employee commitment, job satisfaction, task involvement or work motivation. Two perspectives of organizational and employees are equally examined in the literature. In the organizational perspective, analysts focus on the exposure of employees to the HRM strategy adopted in the firm. Most of these studies analyze directly the link between HRM employees exposure and firm performance without taking into account explicitly the reaction of employees to these practices. MacDuffie (1995) focuses on the automobile industry and studies 62 assembly plants worldwide in 16 countries in He tests the effect of HRM practices on organizational performance. Performance was measured by labor productivity and consumer-perceived quality. He shows that bundles of HRM practices are good predictors of performance. From the response of 968 US firms studied in the nineties, Huselid (1995) shows that his measure of the HRM architecture of the firm is positively related to labor productivity and negatively related to labor turnover. Ichniowski et al. (1997) show, through a sample of 36 homogeneous steel production lines owned by 17 companies, that innovative human resource management practices improve business productivity and new systems of participatory work practices have large, economically important effects on the performance of the business that adopt the new practices. Capelli and Neumark (2001), through a panel analysis of US data collected in 1994, study the relationship between HRM practices (Total Quality Management (TQM), autonomous teams, meeting to discuss work- 5 There is no consensus in the existing literature on the scope of the HRM architecture and scholars have not agreed on a common terminology of the included HRM practices (High-Performance Work Practices - HPWP, Alternative Work Practices - AWP, Innovative Work Practices - IWP, High-Involvement Management - HIM, etc.). 4

5 related problems, teamwork training, job rotation, cross-training, pay-for-skill, profit-sharing) and firm-level outcome. Firm-level performance was measured by productivity (sales per worker). Results reveal the HRM practices raise productivity. Bryson et al. (2005), based on about 1500 British workplaces surveyed in 1998, emphasize that exposing employees to two HRM practices or more is positively related to the organizational financial performance. Nonetheless, some studies that follow the organizational perspective concentrate their attention to the link between the exposure of employees to the HRM strategy (i.e. the adoption and the level of diffusion inside the workplace of HRM practices) and employees attitude. Ramsay et al. (2000), on employer-employee data collected in Britain in 1998, provide results about a score of 24 HRM practices. They find positive links between a score of HRM practices adopted by firms and commitment, job discretion, pay satisfaction, perceived security, and management relations, but also job strain. In contrast, Wood and De Menezes (2011), using a British dataset collected in 2004 on about employees, show that HRM exposure is not statistically related to employees job satisfaction and contentment. Nevertheless, there are significant and positive links between employees perception about supportive, consultative, informative management and job satisfaction and contentment. White and Bryson (2013), through a British dataset collected in 2004 of 2295 establishments and employees, study intrinsic job satisfaction and organizational commitment. HRM are measured by the sum of practices measured at the workplace level about participation, team working, development (training), selection (recruitment practices), and incentives (bonus only). They show that the relationship between the number of HRM and employee job attitudes is J-shaped. As seen above, the employee point of view is largely neglected in this stream of research while their attitudes are the mediating variables between the HRM strategy and the performance of the firm. A second stream of research focuses on the employee perspective and looks at the participation or involvement of employees in HRM practices. Guest (1999), on 1000 UK employees of the private sector collected in 1997, and Guest (2002), on 2000 UK employees of the whole economy collected in 2001, sustain that the more employees experienced HRM practices, the more satisfied they are and the better their psychological contract. Godard (2001), on 508 Canadian employees surveyed in , and Godard (2010), on 750 Canadian employees surveyed in , study various employees attitudes and HRM practices taken as a bundle. Their results support a positive link between the bundle of work practices and commitment, job satisfaction, empowerment 5

6 and a negative link with fatigue and coercion. Gallie et al. (2001), using a sample of 3469 British employees in 1992 and 2224 in 1997, conclude that task discretion (employee s scope for decision-making), control over work performance (supervision), forms of employee involvement and extrinsic reward (extra payment) are important determinants of commitment. Macky and Boxall (2007), on 424 employees of New Zealand, find a positive relationship between high performance work practices and the attitudinal variables of job satisfaction, trust in management and organizational commitment, stressing that innovative work practices can provide win-win outcomes for both employees and employers. Mohr and Zoghi (2008), on about employees working in Canada ( ), study the linkage between seven management practices and job satisfaction. Their results show that individual practices and the bundle of these practices are almost positively related to job satisfaction. Böckerman et al. (2012), on a Finish survey collected in 2003 and covering 3755 employees, examine the relationships between the four core High Involvement Practices (self-managed teams, information sharing, training and performance-related pay) and employees attitudes. Their results reveal positive and significant relationships of the four practices studied and various aspects of employee wellbeing and especially job satisfaction. Using the same dataset and the same HIM, and adopting a bundle approach, Kalmi and Kauhanen (2008) underline also these positives associations. Martin (2016a; 2016b) on Luxembourgish data collected in 2013 examines the links between seventeen HRM practices and work motivations (Martin, 2016a) and focuses on a bundle of the four HIM practices and its association with work motivation, job satisfaction and on-the-job search (Martin, 2016b). The main results sustain that HRM practices studied individually or on a bundle are almost positively linked with work motivations. Moreover, the bundle is significantly and positively related to work motivations, job satisfaction and significantly and negatively related to on-the-job search. Grounded on the existing literature, we formulate the following hypotheses: Hypothesis 1 HRM exposure should be positively associated with employees attitudes Hypothesis 2 Effective participation of employees should be positively associated with employees attitudes To our knowledge, there is no existing evidence on integrating both perspectives to examine the black box that linked the HRM architecture with firm performance. This paper adds to the literature by analyzing not only HRM exposure or effective involvement of employees in 6

7 HRM practices, but the two. A priori, we are not able to predict which of the two variables has a stronger relationship with employee attitudes. We focus on two employees attitudes, namely job satisfaction and organizational commitment that refer to attitudinal outcomes. These are the two facets of overall job attitudes that have been shown to be key mediating variables between the HRM architecture and firm performance (see the meta-analysis of Harrison et al., 2006). We examine a HRM bundle including various HRM practices that can be assumed to foster employees attitudes. The domains covered are participation in the organizational life, working in teams, development, and incentives. 3 Data and methodology 3.1 Data The data used in this paper come from a nationally representative linked employer-employee survey for Luxembourg conducted in The employer survey consists of a selfcompletion survey of the Human Resources Responsible of all workplaces with 15 or more employees of the private sector. The employee survey questionnaires were sent to a stratified random sample of employees working in all workplaces with 15 or more employees of the private sector. This survey consists of an online self-completion survey of employees selected from employees aged at least 15 years. The sample was drawn from the data register of the social security of Luxembourg and employees were contacted at their personal home address. Due to the specificity of the Luxembourgish labor market characterized by a large proportion of cross border employees 6, the employee survey was conducted in four countries (Luxembourg, France, Germany and Belgium) and three languages (French, German and English). To take into account sample attrition due to absence of linked employee or employer data in some cases, the effective samples used in this paper are 1,238 workplaces and 8,373 employees. The database includes weights to account for non-response and survey design probabilities. Each sample is representative at its own level and at the linked level. 6 The cross border employees represent 53% of the working population in the private sector in

8 3.2 Dependent variables Our measures of employees attitudes consist of two variables: job satisfaction and organizational commitment. Our overall measure of job satisfaction is close to that used in the literature (e.g. Clark et al., 1998; Clark, 2001). The variable is based on a question asked to employees: How satisfied are you with your work? with responses ranging on a scale from 0 ( completely dissatisfied ) to 10 ( completely satisfied ). The organizational commitment measure is based on two questions asked to employees: To what extent do you agree or disagree I feel committed to my company with a four point response scale (1-4) and To what extent do you dedicate yourself to your work because this job fulfils my career plans with responses ranging on a scale from 0 ( not at all for this reason ) to 10 ( exactly for this reason ) recode as 1-4 to be summed with the other item. To compute the organizational commitment measure, the two items were summed at the employee level. Descriptive statistics about the dependent variables are shown in Table 1. Table 1. Job satisfaction and organizational commitment measures Values taken Mean Standard deviation Job satisfaction 0,, Organizational 2,, commitment Observations 8, Measures of HRM In the literature, isolated individual practices or a system of practices referred as a bundle are examined and the HRM practices studied partially overlap. Based on existing research, it is becoming increasingly clear that HRM practices need to be adopted together as there is an additive advantage of adopting complementary and overlapping practices to achieve positive results. Moreover, looking at the HRM system as a whole allows taking into account the positive and negative complementarities between practices that is not possible when studying isolated individual HRM practices (Bowen and Ostroff, 2004; Combs et al., 2006; Godard, 2004; Wood, 1999). We use this bundle perspective to characterize the exposure of employees to HRM and the effective participation of employees to the HRM strategy of their employer (see Table 2 for details). The studied HRM practices included in our bundle are close to those used in the literature and cover participation, team working, development and incentives domains (e.g. 8

9 Macky and Boxall, 2007; Mohr and Zoghi, 2008; White and Bryson, 2013). Using the median values taken by the bundle, we created four groups of employees: Exposure low & participation low (LL); Exposure high & participation low (HL); Exposure low & participation high (LH) and Exposure high & participation high (HH) (see Table 3 for details). Table 2. HRM practices included in the HRM bundle Domain name Contents Workplace mean Employee mean Participation Meeting between management and the staff 79.59% 78.01% Changes with employees involved 87.59% 19.59% Attitude surveys 36.49% 61.18% Team working Team work 47.71% 30.92% Quality circle 54.07% 24.29% Job rotation 82.66% 51.26% Development Development included in the strategy 42.13% 53.67% Training 52.45% 42.14% Appraisal 61.71% 58.32% Incentives Individual pay incentive 65.14% 32.41% HRM bundle Sum of the 10 HRM practices 6.10 (2.33) [6] 4.52 (2.00) [5] Observations 1,238 8,373 Notes: Standard deviations are reported in parentheses for non-binary variables. Median values are reported in square brackets for non-binary variables. Table 3. Distribution of the exposure-participation groups Workplace Employee Exposure low & participation low (LL) Exposure high & participation low (HL) Exposure low & participation high (LH) Exposure high & participation high (HH) Observations 1,238 8, Control variables Control variables are included in all the analyses. For individual characteristics, they are: gender, age (3 categories: less than 30 years, 30-49, 50 and more), nationality (6 categories: Luxembourgish; French; German; Belgian; Portuguese; other nationalities), family situation (with partner (yes/no), child in the household (yes/no)), level of education (3 categories: less than secondary, secondary, higher than secondary) and commuting time (8 categories from less than 10 minutes to 1 hour and more). For occupational characteristics, they are: 9

10 occupations (7 categories: Professional and managers; Associate professionals; Administrative and Clerical; Sales and service personnel; Craft; Plant operatives; Nonqualified operatives), full-time job (yes/no), permanent contract (yes/no), tenure (in months), unionized (yes/no), hourly wage, intensity of information and communication technologies (ICT) used by employees (Internet, , ERP - Enterprise Resource Planning, workflow, Intranet, groupware), the sum of harmful working conditions (i.e. adverse factor that affects the employee for a large part of the work time (noise, vibrations, extreme temperatures; radiation, rays or chemical or biological agents; lifting or moving heavy loads; performing rapid, repetitive, monotonous movements; uncomfortable working position)). For the workplace characteristics, size (4 categories: employees; employees; employees; 250 employees and more), sector (7 sectors: Manufacturing; Construction; Trade, accommodation and food services; Transportation and storage; IT and communication; Finance; Other services), the fact that the workplace belongs to a foreign owned firm, to a multisite firm on the Luxembourgish territory, is in activity for more than 20 years, the number of concurrent the firm faces on its main market (coded as less than 6; 6-25; more than 25), the class of percentage of graduate employees (coded as: 0-5%; 6-24%; 25-49%; 50% and more), the percentage of colleagues with the same gender as the respondent and the percentage of colleagues with the same nationality as the respondent. Appendix Table A1 provides descriptive statistics and compares the characteristics of employees, their job and their workplace related to their exposure/participation group. Reported figures reveals that the mean values for the high-high (HH) category of workers differ from the others for several variables. First, some differences appear in their personal characteristics. High-high employees are more likely than the others to, for example, have a post-secondary level of education. They are also more likely to be German and less likely to be Portuguese. They spend more time, each day, commuting from home to work than the employees who work in workplaces that provide a low level of HRM exposure. In contrast, they commute less than the employees of the high-low group. Second, differences appear in the characteristics of their occupation. High-high employees occupy more often a top occupation (professional and managers) and they are less likely sellers, service personals, craftsmen or unskilled workers. Therefore, their jobs are better paid, they use more often ICT and they face fewer harmful working conditions. On average, they work for their organization since more time. They are more likely to work full time than the low-low group and less likely than the low-high group. Their contracts are more often permanent than the low-high 10

11 group. Third, they work in different types of firms. They work in larger organization (with at least 250 employees) and more often in finance sector and less often in construction and trade, accommodation and food service sectors. Their firm is more often foreign owned, most often in activity since 20 years, less often multisite, has a higher percentage of graduates and a lower percentage of colleagues with the same gender as the respondent. The differences highlighted above imply the need to carefully control for individual, occupational and workplace characteristics. 3.5 Estimation strategy The measures of job satisfaction and organizational commitment are treated as continuous variables. In a first step, we use linear regression with a robust variance estimator. Robust standard errors are clustered at the organization level to correct for the fact that some employees are employed by the same organization and therefore the observations may not be entirely independent. The parameters of the following model are estimated: Y! = LL! β! + HL! β! + LH! β! + X! β! + ε! Where Y! is the level of job satisfaction (or organizational commitment) of individual i, LL!, HL!, LH! are the exposure-participation group, (HH! is the reference group), X! a vector of the individual, occupational and workplace characteristics (and a constant), β the vector of coefficients and ε! is the random error term normally distributed. The high-high (HH) category is chosen as the reference group in accordance with the J-shaped relationship between the number of HRM and employees attitudes shown by White and Bryson (2013). Moreover, in our sample high-high is the most common category. In a second step, we conduct the analyses using the same method, but separately by exposureparticipation groups (keeping the high-high as the reference group). Y!" = X!" β! + ε!" where Y!" is the level of job satisfaction (or organizational commitment) of individual i of group j (j=ll, LH, HL, HH), X!" a vector containing the values of individual, occupational and workplace characteristics for individual i of group j and the intercept, β! the vector of coefficients for group j and ε!" is the random error term normally distributed. 11

12 The results from these models are used to decompose the employees differences in, on one hand, job satisfaction, and on the other hand, organizational commitment into three components following a variant of the Blinder (1973) Oaxaca (1973) decomposition done by Daymont and Adrisani (1984). The first (the endowments effects ) is the part of the difference that is attributable to differences in individual, occupational and workplace characteristics between the groups. The second (the coefficients effects ) is the part that is attributable to employees differences in the coefficients (β) on those characteristics. The third (the interaction) is the part that is attributable to the simultaneous effect of differences in endowments and coefficients. Taking the example of the low-low employees, the decomposition results from constructing the counterfactual asking, what would the level of job satisfaction (or organizational commitment) be for low-low employees if they had the same individual, occupational and workplace characteristics as high-high employees, and, what would the level of job satisfaction (or organizational commitment) be for low-low employees if they placed the same value on characteristics as high-high employees? The specification for the decomposition is the following: E Y!! E Y!! = E X!! E X!!! β!! + E X!!! β!! β!! + E X!! E X!! (β!! β!! ) The first term on the right hand side is the part attributable to differences in the outcome variable between the two groups that is due to differences in the covariates X. The second term is the part attributable to differences in the valuation of personal, occupational and workplace characteristics. The third part is the interaction term. We present estimates of these components for job satisfaction and organizational commitment by comparing each group with the reference group (high-high). Robust standard errors are clustered at the workplace level to correct for the fact that some employees are employed by the same workplace. Sensitivity analysis has been done by applying an alternative decomposition method. More precisely, a twofold decomposition method following Neumark (1988) has been done. The results are mostly the same, and are available from the authors upon request. It should be noted that the data are cross-sectional and so we are only able to identify the strength of conditional correlations not causal relationships. Satisfied and committed employees can indeed be the ones that participate the more in the HRM practices designed by their employer. But for the exposure level, it is not obvious that HRM practices are adopted only by firms with low satisfaction and commitment levels (see Table 3). Our analysis is also 12

13 not immune to a potential self-selection issue. Satisfied and committed employees could have been attracted to work in the current firm because of the information they had when they postulated about the availability of management practices in the desired job. The available data do not permit to solve perfectly this issue, but as observed with the descriptive statistics workplaces with a high exposure are half with a low participation of their employees and half with a high participation (see Table 3). Thus the self-selection issue may not be too serious. 4 Results First we show the results of the analyses obtained with linear regression models on the relationships between the combined workplace- and individual-level variables about the exposure/participation to HRM and the two measures of employees attitudes: job satisfaction and organizational commitment. We then report the results of analyses that decompose the employees differences between the exposure/participation groups. All analyses include the set of control variables described above. The results of the linear regressions are displayed in Table 4. Table 4. Exposure-participation to HRM, job satisfaction and organizational commitment Job satisfaction Organizational commitment (1) (2) Exposure low & participation low -1.11*** -0.73*** (0.09) (0.06) Exposure high & participation low -0.99*** -0.64*** (0.07) (0.04) Exposure low & participation high (0.09) (0.06) Male (0.08) (0.05) Age years -0.15* -0.16*** (0.09) (0.06) 50 years and more *** (0.11) (0.08) German -0.26** (0.12) (0.07) Belgian 0.42*** 0.17*** (0.10) (0.06) French

14 (0.10) (0.06) Portuguese (0.13) (0.08) Other nationality (0.12) (0.08) Living with partner (0.07) (0.05) Child (0.06) (0.04) Secondary education -0.29*** (0.09) (0.06) More than secondary -0.57*** -0.23*** (0.11) (0.07) Commuting time -0.03* (0.02) (0.01) Professional and managers 0.69*** 0.74*** (0.17) (0.11) Associate professionals 0.56*** 0.52*** (0.15) (0.10) Clerical 0.52*** 0.34*** (0.15) (0.10) Sales and service personnel 0.43*** 0.22** (0.16) (0.10) Craft 0.67*** 0.49*** (0.15) (0.10) Plant operatives 0.69*** 0.52*** (0.16) (0.11) Full time *** (0.11) (0.06) Permanent contract (0.12) (0.08) Tenure (0.00) (0.00) Unionized -0.32*** -0.11** (0.07) (0.04) Hourly wage 0.01*** 0.00* (0.00) (0.00) Sum of ICT 0.04* 0.07*** (0.02) (0.02) Sum of harm working conditions -0.29*** -0.08*** (0.03) (0.02) employees (0.10) (0.06) employees *** (0.10) (0.06) 250 employees and more *** 14

15 (0.09) (0.06) Construction 0.48*** 0.28*** (0.12) (0.08) Trade, accommodation and food service * (0.12) (0.08) Transportation and storage ** (0.14) (0.09) IT and communication (0.12) (0.09) Finance -0.37*** -0.13* (0.12) (0.08) Other services (0.11) (0.07) Foreign owned (0.06) (0.04) Multisite firm (0.07) (0.04) Number of concurrent on the market (0.04) (0.02) More than 20 years of activity (0.07) (0.05) Percentage of graduate employees (0.04) (0.02) Percentage of colleagues with the same gender (0.16) (0.09) Percentage of colleagues with the same nationality 0.29*** 0.06 (0.11) (0.08) Constant 6.56*** 5.29*** (0.32) (0.19) F-test 20.9*** 25.97*** R-squared Observations 8,373 8,373 Notes: Beta coefficients * significant at 10%; ** significant at 5%; *** significant at 1%. Robust standard errors adjusted for 1268 clusters (working in the same workplace) in parentheses. Weighted estimations. The results show that, compared with high-high employees (high exposure and high participation), a low participation in HRM irrespective of the level of exposure is negatively associated with employee job satisfaction (column 1) and organizational commitment (column 2). Conversely, a high participation in HRM while the workplace does not provide a high exposure is not statistically different from the high-high employees both in the job satisfaction and organizational commitment regressions. These results show that it is the participation rather than the exposure to HRM that matters in the employees attitude. 15

16 Employees need to participate in especially training, exercise voice during meetings between the staff and managers, participate in the decision making process at the team level in order to support an improvement of their skills, knowledge and beyond to be convinced to respond positively to the HRM system in place no matter the level of HRM exposure they face. Regarding personal characteristics, the results reveal that the level of education is negatively linked with job satisfaction and organizational commitment. These results are identical to those found in the literature for job satisfaction by Clark et al. (1996). Even if the link between education and organizational commitment is often not significant in the literature (DeCotiis and Summers, 1987; Madsen et al., 2005) our result is in line with the one obtained by Mowday et al. (1982). As found in literature (DeCotiis and Summers, 1987; Gallie et al., 2001; Madsen et al., 2005), being male is not linked with organizational commitment. Being male appears also not linked with job satisfaction and on this point the literature shows mixed results (Clark, 1997; Sousa-Poza and Sousa-Poza, 2000; Hauret and Williams, 2016). Our results show that age is negatively linked with organizational commitment whereas Madsen et al. (2005) conclude to a positive correlation. However, in our study, in line with previous literature (Clark et al., 1996), job satisfaction appears to be U shaped in age. Living with a partner, contrary to Clark et al. (1996) and Mathieu and Zajac (1990), is not linked with job satisfaction and organizational commitment. In line with Stutzer and Frey (2008) having longer commuting time is negatively linked with job satisfaction. In the specific labor market context of Luxembourg, we control for nationality and find that being Belgian rather Luxembourgish is significantly associated with increases in employees attitudes while being German is negatively linked with job satisfaction. Regarding occupational variables, the results show, in line with Clark (1996) or Gallie et al. (2001), that being unionized or being a non-qualified operatives decreases job satisfaction and organizational commitment. Conversely, the hourly wage positively matters, as found in the literature (Clark, 1997), for job satisfaction. Working full-time, having good working conditions and intensively use ICT are positively linked with organizational commitment. In literature, only the links between employees attitudes and a narrow range of ICT (i.e. Computer and Internet use) has been studied (Gallie et al., 2001; Martin and Omrani, 2015) and computer use reveal non-significant links with organizational commitment or job satisfaction whereas Internet use appears to be positively linked with job satisfaction. In our 16

17 study, like Gazioglu and Tansel (2006), having a permanent contract is not linked with job satisfaction. Regarding workplace variables, according to existing evidence (Gallie et al., 2001) we see that working in a large firm tends to decrease organizational commitment. However, contrary to Clark (1996), the relation between the firm s size and job satisfaction appears not significant in our study. As Gazioglu and Tansel (2006), the employees working in construction s sector are more job satisfied. In line with Frijters et al. (2006), we find that the nationality composition of the workplace is linked to job satisfaction: job satisfaction increases with the percentage of colleagues who have the same nationality as the respondent. As we have described above, the high-high group has some characteristics which are positively linked to job satisfaction and/or organizational commitment (top occupation, high paid job, ICT use) but also others which are negatively linked to these attitudes (commuting time, level of education, work in larger organization). Therefore, it s interesting to determine the extent to which the differences in the average level of job satisfaction and organizational commitment, between the different group and the high-high, can be explained by the differences of characteristics, and the extent to which they are explained by categories differences in the weight placed on those characteristics. The results of the Blinder-Oaxaca decompositions are presented in Table 5 for job satisfaction and for organizational commitment. The table presents the mean differential in the dependent variables and the percentages associated with the part of the differential that is due to group differences in the characteristics, the part that is due to differences in the coefficients and the part that is due to interaction. 17

18 Table 5. Blinder-Oaxaca decomposition results, by employees group HH versus LL HH versus HL HH versus LH contribution % contribution % contribution % Job satisfaction Diff 1.066*** 0.997*** ** Part diff means Part diff coeff 1.169*** *** Part diff inter Organizational commitment Diff 0.814*** 0.702*** Part diff means 0.246*** Part diff coeff 0.692*** *** Part diff inter Notes: * significant at 10%, ** significant at 5%, *** significant at 1%. Robust standard errors adjusted for 1268 clusters (working in the same workplace). Weighted estimations. The results are mixed according the categories of employees studied. On the one hand, for the low-low category and the high-low category, we find that it is not the differences of personal, occupational and workplace characteristics that explain the largest part of the job satisfaction or organizational commitment gap but the differences placed in the coefficients. Indeed, for these two categories of employees, the contribution of differences in coefficients explains between 85% and 110% of these gaps. This result suggests that the differential of employees involvement in HRM contribute to modify the way that they react to their personal, occupational and workplace characteristics. Applying the coefficients of the high-high to the employees who have a low participation in HRM, irrespective of the exposure s level, would increase significantly their job satisfaction and their organizational commitment. On the other hand, we find that the difference in job satisfaction between the high-high and the low-high is mainly due to a difference of characteristics: 85% of this gap is explained by a difference of characteristics and 36% by a difference of coefficients. Note that these contributions are, however, not significant at the 10% threshold. These results suggest that highly involved employees give almost the same values in the characteristics studied whatever the level of HRM exposure they face. No significant difference between the high-high and the low-high is highlighted in our results for organizational commitment. 18

19 The last step is to examine the extent to which each category of variables (personal, occupation, workplace) contribute to the group differential in job satisfaction and organizational commitment. This is done by summing over each category of variables the estimates of the contributions that each variable makes to the explained (differences in means), unexplained (differences in coefficients) and interaction components of the overall group differentials in job satisfaction or organizational commitment. The results are presented in Table 6. Table 6. Decomposition results by variable type Job satisfaction HH versus LL HH versus HL HH versus LH Contribution % Contribution % Contribution % Diff means Total Individual ** *** Occupational 0.482*** *** Workplace *** Diff coeff Total 1.169*** 0.980*** Individual Occupational ** Workplace *** * Constant 1.538* *** ** Diff Inter Total Individual * Occupational Workplace ** Organizational commitment HH versus LL HH versus HL HH versus LH Diff means Total 0.246*** Individual * Occupational 0.429*** ** *** Workplace ***

20 Diff coeff Total 0.692*** 0.633*** Individual Occupational * Workplace *** Constant 0.989** *** *** Diff Inter Total Individual Occupational *** Workplace Notes: * significant at 10%, ** significant at 5%, *** significant at 1%. Robust standard errors adjusted for 1268 clusters (working in the same workplace). Weighted estimations. Individual characteristics: gender, age, nationality, living with partner, child in the household, level of education and commuting time. Occupational characteristics: occupation, full-time job, permanent contract, tenure, unionized, hourly wage, intensity of ICT use and harmful working conditions. Workplace characteristics: size, sector, foreign owned, multisite, in activity for more than 20 years, the number of concurrent on its main market, percentage of graduate employees, percentage of colleagues with the same gender as the respondent and percentage of colleagues with the same nationality as the respondent. Referring to the part arising from differences in means, we observe that, in most cases, the main differences come from a difference in occupational characteristics. These include variables such as occupation, type of contract and hourly-wage. However, it is interesting to see that if for the low-low and the high-low, the differences in occupational characteristics contribute to increase the job satisfaction or the organizational commitment gap, for the lowhigh, these differences contribute to decrease the gaps. The differences in workplace characteristics play only a role on the job satisfaction (or organizational commitment) gap between groups of employees who have both a highly participation in HRM practices but not the same level of exposure. Referring to the part arising from differences in coefficients, we observe that the largest contributor to the unexplained part is the constant term. This result is unsatisfactory because the interpretation of the constant term in Blinder-Oaxaca decomposition gave rise to a criticism of the method (Jones, 1983). We also find, however, that the differences in workplace coefficients matter in the job satisfaction specification for the high-low and the low-high and in the organizational commitment specification for the high-low. Nevertheless, 20

21 if these differences contribute to increase the gap with the high-high for the low-high group, for the high-low they contribute to decrease the gap. 5 Conclusion Prior work has argued that the use of HRM practices is positively linked with firm performance (e.g. Capelli and Neumark, 2001; Huselid, 1995; Ramsay et al., 2000). Scholars also sustain that HRM practices need to be adopted as a system as there is an additive advantage of adopting complementary practices to achieve positive results (Bowen and Ostroff, 2004; Combs et al., 2006; Godard, 2004; Wood, 1999). It is also widely recognized that employees response to the HRM strategy designed by their employer play a mediating role between the HRM strategy and firm performance (e.g. Kalmi and Kauhanen, 2008; Martin, 2016a; 2016b; Mohr and Zoghi, 2008). This paper focuses on the links between the HRM strategy and the two key mediating variables between the HRM strategy and firm performance, namely job satisfaction and organizational commitment. More precisely, the paper seeks to identify if it is the exposure to the HRM strategy or the effective participation of employees that is more importantly linked with the two employees attitudes studied. The existing literature cannot provide an answer to this issue because the relationships between employees attitudes and the exposure of employees to the HRM strategy, on the one hand, or the involvement of employees in HRM practices, on the other hand, are studied separately. This paper takes advantage of a recent employer-employee data set collected in Luxembourg to reconcile these two perspectives. Due to the specificity of the Luxembourgish labor market characterized by a large proportion of foreign employees, the results relate not only to Luxembourgish but also French, Belgian, German, Portuguese and other workers. Our main results show that a high exposure to HRM is not sufficient to obtain positive employees attitudes and that a high effective employees participation in HRM is required. By deepening the examination of differences in employees attitudes by comparing groups of employees who face different levels of HRM and who are involved in different levels of HRM, we find that the differential of employees involvement in HRM practices contributes to modify the way that they react to their personal, occupational and workplaces characteristics. 21

22 Our results provide practical managerial implications for employers as employees attitudes are the key mediating variables between the HRM strategy and firm performance. Even if all employees do not have the same role in the value creation process of the firm, the results sustain that managers have to favor a high involvement of the staff in the HRM system. Employees need especially to participate in trainings, to exercise voice during meetings with managers, to take decisions at the team level to strengthen their skills, knowledge and empowerment and beyond to respond positively to the HRM system in place no matter the level of HRM exposure they face. A potential shortcoming of this paper is that the data are cross-sectional. Due to the fact that they are only measured one time (2013) they do not allow us to identify a causal relationship between exposure-participation to HRM and employees attitudes. Only additional research using employer-employee panel data will permit to overcome this issue. References Barney, J. (2001) Resource-based theories of competitive advantage: A ten-year retrospective on the resource-based view. Journal of Management 27(6), Becker, B.E., and M.A. Huselid (2006) Strategic human resources management: where do we go from here? Journal of Management 32(6), Blinder, A.S. (1973) Wage discrimination: Reduced Form and Structural Estimates. The Journal of Human Resources 8(4), Böckerman, P., Bryson, A. and P. Ilmakunnas (2012) Does high involvement management improve worker wellbeing? Journal of Economic Behavior & Organization 84(2), Bowen, D.E., and C. Ostroff (2004) Understanding HRM-performance linkages: The role of the strength of the HRM system. Academy of Management Review 29(2), Bryson, A., Forth, J., and S. Kirby (2005) High-involvement management practices, trade unions representation and workplace performance in Britain. Scottish Journal of Political Economy 52(3), Cappelli, P., and D. Neumark (2001) Do high-performance work practices improve establishment-level outcomes? Industrial and Labor Relations Review 54(4),