Structural empowerment and organisational performance: the mediating role of employees’ well-being in Spanish local governments

Abstract We extend the ‘black box’ picture of public management and the ‘balanced view’ of HRM literature in the Spanish public context. Specifically, we explore the link between structural empowerment and organisational performance and the mediating role of three employee outcomes: job satisfaction and affective commitment as attitudinal variables related to eudaimonic well-being, and job anxiety as a variable related to hedonic well-being. Using a multilevel methodology with data on employees from 103 local authorities for 2016, results support the mutual gains perspective, but not as intensely as expected (empowerment does not affect attitudinal variables), since empowerment contributes to reducing job anxiety in Spanish local governments. Specifically, information and rewards are the structural empowerment dimensions that help to reduce job anxiety levels in employees and, thus, enhance performance. These results suggest that local government managers could usefully apply techniques such as disseminating information on the organisation’s aims among employees or paying bonuses for meeting targets.


Introduction
Debate has grown in the past decade over the contribution of human resource management (HRM) to organisational performance and employee well-being, corresponding to two different perspectives: mutual gains and conflicting outcomes (e.g. Van De Voorde et al., 2012. On one hand, the dominant mutual gains perspective states that HRM is beneficial for both organisational performance and employee well-being. This view holds that implementing HRM practices for employees has benefits for them, which therefore also enhances organisational performance. On the other hand, the conflicting outcomes perspective advocates that HRM has positive effects for organisational results, but has no effect on, or even negative outcomes for employee well-being. In this context, balanced studies that present arguments and empirical evidence taking into account both perspectives can help to unravel the complexity of the HRM-performance link (e.g. Godard, 2001;Ramsay et al., 2000).
Since different HRM practices can be associated with different employee and organisational outcomes (Jiang et al., 2012), sub-dimensions of HRM or specific HRM practices, such as empowerment practices, must be examined to understand such relationships (Van De Voorde et al., 2012). Empowerment practices have often been examined within bundles of other HR practices (e.g. Bashir et al., 2011;Patel & Conklin, 2012;Raineri, 2017), rather than individually, which prevents differentiation and masks the effect on results. Accordingly, in the present study we focus specifically on HRM empowerment practices. The definition of empowerment practices, or in this case, structural empowerment, is a set of practices allowing the transfer of power and authority from higher to lower levels of an organisation by sharing decision-making power, information, knowledge, and rewards (Bowen & Lawler, 1992). From a mutual gains perspective, at the organisational level these practices can lead to a more resourceful, rewarding and meaningful work atmosphere (Van De Voorde et al., 2016), thus contributing to better performance, as also suggested by social capital theory (Seibert et al., 2001). At the individual level, empowerment enhances employees' job satisfaction and affective commitment (e.g. Kirkman & Rosen, 1999). This claim is supported by social exchange theory (Blau, 1964) and the norm of reciprocity (Cropanzano & Mitchell, 2005), since employees interpret structural empowerment as indicative of organisational support and concern for their well-being (Van De Voorde et al., 2012), and reciprocate by showing greater satisfaction and commitment (Allen et al., 2003). However, from a conflicting outcomes perspective, empowerment may also make work more demanding, with added responsibilities and work intensification that can increase stress and job anxiety, as proposed by labour process theory (Ramsay et al., 2000).
This dilemma also appears within the New Public Management (NPM) approach, which first appeared in public organisations in OECD countries in the 1980s. One central idea of NPM is to improve effectiveness of public services through HRM practices that shift from uniform rules to more employee discretion, team working, recognition of employees' contributions and, in general, endeavours to stimulate involvement among employees by way of high commitment human resource management, where empowerment plays an important part (e.g. Bach & Givan, 2011;Pitts, 2005;Stanton & Manning, 2013). From the perspective of these NPM ideals, further study of structural empowerment is needed. Structural empowerment is considered to enhance services and performance at the organisational level by improving the communication processes that, according to social capital theory, can help build effective relationships and structures to solve day-to-day problems easily and deliver better services (Brunetto & Farr-Wharton, 2008). In the same line, the 'black box' arguments of the public management literature claim that structural empowerment makes organisations more adaptable, efficient and effective by building structures that match citizens' demands (Burgess, 1975). However, at the employee level, some researchers have warned of the human cost associated with NPM. Monitoring, pressure and intensified accountability of public staff can lead to alienation, fear, stress and anxiety states in employees (Chandler et al., 2002;Diefenbach, 2009). NPM therefore has its 'dark side' in that it places additional demands on the workforce (Diefenbach, 2009).
Calls have been made for a more balanced approach (Paauwe, 2004(Paauwe, , 2009), grounded on a wider and contextually based perspective of HRM, which takes into account performance aspects as well as factors involving related actors (such as employees). Despite these calls, previous empirical research has generally analysed the contribution of structural empowerment to organisational performance (e.g. Logan & Ganster, 2007), or has focused on positive consequences for employees, such as satisfaction or commitment (e.g. Kim, 2002;Kirkman & Rosen, 1999), while neglecting the possible negative consequences for employees' health. These studies therefore offer only partial frames (focusing on one kind of outcome variable, organisational or individual, and on positive employee outcomes), which hamper a more comprehensive view of the consequences of structural empowerment. Moreover, this partial view prevents analysis of the indirect or mediating path of work-related outcomes and, hence, impedes exploration of the 'black box' between structural empowerment and organisational performance. This comprehensive view has received even less research attention in the public sector context (Fernandez & Moldogaziev, 2013;Park & Rainey, 2007).
To address these research needs, we study how structural empowerment practices directly affect organisational performance in a public context and how this association is mediated by different forms of wellbeing (see Figure 1). We refer to two forms of well-being. First, job satisfaction and affective commitment are eudaimonic well-being variables, a view of well-being that entails engagement in meaningful, self-actualising and growth producing behaviours; that is, a view related to growth and purpose in life. The second form is job anxiety, as a hedonic well-being variable, since this view considers positive emotions and mood, and is therefore related to attaining pleasure and avoiding pain (Robertson & Cooper, 2010;Ryan & Deci, 2001;Ryff & Singer, 2008). Based on a sample of 103 Spanish local authorities, we use a multilevel mediation model to integrate the individual and organisational levels. This multilevel approach not only extends and refines single-level models but also represents a significant departure from them (Peccei & Van De Voorde, 2019). This paper makes several important contributions. First, it considers structural empowerment as a construct in itself, inspired by Bowen and Lawler (1992) multidimensional model, and not as a part of a bundle of other HR practices (e.g. Raineri, 2017). This approach allows us to determine its composition and consequences more accurately (Van De Voorde et al., 2016). Second, given the need for a more balanced view of HRM, we build on the analysis of employee satisfaction and commitment basic, desired attitudes in achieving public organisations' success (Park & Rainey, 2007) and examine a harmful aspect, job anxiety, in response to the need to test the association between HRM and health variables (Van De Voorde et al., 2012). We attempt to answer a timely question: is structural empowerment beneficial or detrimental, or does it have no link to these well-being variables? Discerning such associations can contribute to knowledge on the dominant mutual gains and the conflicting outcomes perspectives of HRM. Third, considering that NPM changes have been introduced in different ways and speeds in different countries (Pollitt, 2001), we study empowerment in the context of public services in Spain, where the relationships proposed may differ from other contexts. Spain has been slow to apply mainstream NPM techniques (Garcia, 2007). Furthermore, the Spanish public administration culture is still grounded in administrative law (Torres et al., 2011) and most public sector employees are civil servants, which complicates the introduction of some reforms based on employee participation. New human resource management techniques do not seem to fit well with the rigidity, structures and processes of Spanish administrations (Serna, 2008). Hence, the contextualisation of this study in the Spanish public sector contributes to the understanding and adaptation of empowerment practices to this reality.

Literature review and hypotheses development
The direct relationship between structural empowerment and organisational performance Bowen and Lawler (1992) developed the most well-known depiction of structural empowerment in services as an 'approach to service delivery'. This approach includes practices that increase employees' access to information and resources by: giving them information on their firm's operations; providing training that enables them to contribute at work; giving them power to make decisions that influence organisational activities; and providing performance-based rewards designed to encourage initiative. According to this definition, employees are able to respond faster to customer needs and improve performance by working more efficiently and effectively, thus making the service more reliable. Social capital theory supports these ideas, claiming that resources and communication processes embedded in an organisational structure improve coordination and activities. It therefore implies cost reductions and increased efficiency, and ultimately better organisational performance (Seibert et al., 2001). Previous empirical research in private sector organisations has found positive relationships between practices related to structural empowerment and quality, service and sales, and overall performance (e.g. Patterson et al., 2004;Seibert et al., 2004;Wallace et al., 2011).
Regarding the public sector, 'black box' approaches to public management claim that structural empowerment has important effects on the delivery of better performance (Ingraham et al., 2003), making organisations more adaptable, efficient, and effective (Burgess, 1975). Such approaches state that delegating power, training, and motivating employees with rewards enhances employee morale, which leads to behaviours and attitudinal dispositions towards the public suited to the particular demands of each service encounter (Chebat & Kollias, 2000), thus improving overall performance. Earlier contributions support this claim, showing a positive contribution of empowerment to performance in quality of work and accomplishment in federal organisations in the U.S. (Fernandez & Moldogaziev, 2013) and in Spanish local authorities (Barba & Serrano, 2015). From these arguments, we hypothesise that: Hypothesis 1: Structural empowerment is positively associated with perceived organisational performance in local governments.
Structural empowerment and employee well-being outcomes Job satisfaction and affective commitment Locke (1969, p. 316) defined job satisfaction as 'the pleasurable emotional state resulting from the appraisal of one's job as achieving or facilitating the achievement of one's job values'. Affective commitment refers to the emotional attachment to the organisation characterised by acceptance of the organisation's culture and values and by a desire to remain part of that organisation (Mowday et al., 1982). Both can be related to eudaimonic well-being, since they refer to individuals' feelings about the alliance between their true self and values, and their job reality (Ryan & Deci, 2001).
From the mutual gains perspective, Bowen and Lawler (1992) state that empowerment leads to more satisfied employees because they feel decision making is in their hands and perceive increased control over their jobs. Social exchange theory (Blau, 1964) provides clear theoretical support for such arguments, holding that employees interpret organisational actions, structural empowerment for instance, as indicative of organisational support (Van De Voorde et al., 2012). Such perceptions lead employees to reciprocate and show greater satisfaction and commitment (Allen et al., 2003). The norm of reciprocity (Cropanzano & Mitchell, 2005) also supports this relationship, as it refers to the socially accepted norm of returning a favour in exchange for help or resources. Employees are likely to feel more satisfied and committed to organisations that support them and give them resources through empowerment practices. Previous empirical research has found positive associations between empowerment practices, and job satisfaction and affective commitment (Holland et al., 2011;Kim, 2002;Kirkman & Rosen, 1999).
In the public sector, especially in health-care contexts, some positive relationships have also been found between structural empowerment and job satisfaction (Sarmiento et al., 2004), as well as affective commitment (Park & Rainey, 2007). In these contexts, as Kanter's (1993) theory of organisational empowerment explains, empowerment structures allow employees to mobilise the necessary resources to 'get things done', which raises their job satisfaction. Likewise, when employees have the opportunity to increase their competence and skills and are rewarded for contributing to organisational aims, they invest in the organisation and may demonstrate this by seeing themselves as part of the organisation. These same arguments can be translated to employees in local government where NPM reforms have been introduced. We therefore hypothesise that: Hypothesis 2a: Structural empowerment is positively associated with job satisfaction in local governments.
Hypothesis 2b: Structural empowerment is positively associated with affective commitment in local governments.

Job anxiety
Anxiety is defined as an emotional state of perceived apprehension and heightened agitation (Spector et al., 1988) and constitutes a measure of general mental health. Job anxiety is linked to a specific stimulus: the workplace. In contrast to job satisfaction and affective commitment, it is associated with hedonic well-being as it reflects the presence or absence of pleasure or pain (Ryan & Deci, 2001).
From a conflicting outcomes perspective, labour process theory (Ramsay et al., 2000) holds that empowerment practices promote not only discretion, but also added responsibility and work intensification, which may increase stress. Employees may feel that managers expect greater work effort (Van De Voorde et al., 2016), thus perceiving more pressure to perform and less overall control over their working lives (Orlitzky & Frenkel, 2005), which may cause anxiety.
In the NPM field and its documented 'dark side', the proactive attitudes that empowerment entails, together with established measurable standards of performance and an attempt to monitor and reward employees according to these measures, may generate higher pressure, and intensification of labour and accountability of public staff, all of which can lead to states of resentment, fear, stress and anxiety (Chandler et al., 2002;Diefenbach, 2009). Some authors (e.g. Clark, 1999;Vidal, 2007) take a critical view of empowerment, claiming that many workers have no desire for empowerment because they associate it with 'too much work', and prefer to remain comfortable within the old authority structure where they know the rules and their predictable work arrangements make their jobs feel more secure. Under conditions of empowerment such employees could experience higher levels of anxiety because of the mismatch between their desires at work and the implications of empowerment practices.
In addition, public sector workers are generally more likely to show worse levels of mental health (McHugh, 1998). It is well documented that working with the public (attending customers, for instance) is a major factor in the risk of developing anxiety (Wieclaw et al., 2008). Local governments provide services, so their employees are expected to deal with citizens in order to meet their requirements, which could increase the likelihood of their suffering anxiety at work. Therefore, we propose the following hypothesis: Hypothesis 2c: Structural empowerment is positively associated with job anxiety in local governments.

Employee well-being outcomes and organisational performance
Job satisfaction and affective commitment Job satisfaction and affective commitment are expected to contribute to organisational performance. Social exchange theory and the norm of reciprocity support such relationships, as employees tend to reciprocate the way they are treated in their organisation. Therefore, feeling satisfied with one's job and affective commitment resulting from empowerment may stimulate employees to respond with improved performance, contributing to the overall performance of the organisation (Eisenberger et al., 1990;Nica, 2016). The general consensus in the HRM literature is that employee attitudes are vital to achieving good organisational performance (Jiang et al., 2012).
Satisfied employees seek to work and perform their tasks to a high standard and to achieve customer satisfaction, and are more willing to adopt non-compulsory behaviours aimed at realising their objectives (Ogbonnaya & Valizade, 2018). Wood et al. (2012) studied a sample comprising industries from the private and public sectors, demonstrating that job satisfaction was positively associated with financial performance, labour productivity, quality and lower levels of absenteeism.
Regarding affective commitment, employees with high levels of commitment may strive to achieve more success for the company and show behaviours that benefit the organisation, thus contributing to improved productivity (Elorza et al., 2011;Kim et al., 2018). This relationship can also be transferred to public organisations, as affective organisational commitment is seen as a vital element in maintaining output in both public and private sectors (Perry, 2004). The few studies conducted in the public sector (e.g. Zhu & Wu, 2016) have found associations between affective commitment and organisational performance measures such as managerial accountability, work performance and organisational growth. Hence, the following hypotheses are proposed: Hypothesis 3a. Job satisfaction is positively associated with perceived organisational performance in local governments. Hypothesis 3b. Affective commitment is positively associated with perceived organisational performance in local governments.

Job anxiety
According to Sackey and Sanda (2009), symptoms of stress and strain, such as anxiety at work, make people less communicative, and increase tension, tiredness and low energy. These states are likely to result in lower levels of performance, therefore negatively affecting organisational performance. In a sample of companies from different sectors, Bakker et al. (2004) found that exhaustion (one of the dimensions of burnout) was negatively related to in-role and extra-role performance. As both inrole and extra-role performance are related to effective organisational operations, a negative link might also be found between job anxiety and organisational performance. Similarly, Ramsay et al. (2000) demonstrated that job strain was related negatively to labour productivity and positively to absence rate as measures of organisational performance. The explanation is that this psychological state related to stress at work affects behavioural outcomes like job performance because it reduces employees' energy levels and their efforts at work, leading to poorer performance (Singh et al., 1994). We therefore hypothesise that: Hypothesis 3c. Job anxiety is negatively associated with perceived organisational performance in local governments.

The mediation role
The above arguments lead us to expect mediating relationships. Both affective commitment and job satisfaction may facilitate an indirect relationship between structural empowerment and organisational performance, thus reflecting the mutual gains perspective of HRM and its positive outcomes for both organisations and employees (Ogbonnaya & Valizade, 2018;Van De Voorde et al., 2012). In fact, Wood and Ogbonnaya (2018) identify employee outcomes such as job satisfaction and commitment as the most notable intermediate variables in studies on the link between HRM and organisational performance. Hence, we propose the following hypotheses: Hypothesis 4a. Job satisfaction positively mediates the relationship between structural empowerment and perceived organisational performance in local governments.
Hypothesis 4b. Affective commitment positively mediates the relationship between structural empowerment and perceived organisational performance in local governments.
Likewise, and taking into account the aforementioned labour process theory (e.g. Godard, 2001), the benefits for organisational performance may be reduced because of the higher anxiety derived from structural empowerment. A mediating relationship can thus be hypothesised as follows: Hypothesis 4c. Job anxiety negatively mediates the relationship between structural empowerment and perceived organisational performance in local governments.

Procedure and sample
The empirical work was carried out in Spain, which offers an illustrative example of NPM implementation. In the Spanish context, a legal regulationthe Estatuto B asico del Empleado P ublico (2007) (Basic Statute for Public Employees)embodies the ideas guiding mainstream NPM. This statute endeavours to enhance employees' involvement through principles and techniques related to structural empowerment (e.g. training, performance-linked remuneration), highlighting the importance of improvement in the effectiveness, efficiency and quality of services delivered to the public.
The population of firms for our sample was selected from the Federaci on Española de Municipios y Provincias (Spanish Federation of Municipalities and Provinces) database, which lists all Spanish city councils and their contact information. We selected only large municipalities (more than 20,000 inhabitants) because they implement more strategic management practices. The application of this criterion yielded a population of 399 city councils.
The units of analysis are the local authority (organisational level) and the employees (individual level). Thus, two online questionnaires were prepared: one for local government managers (human resource managers, or in their absence, clerks) and a second for other public employees. Following Dillman et al. (2009), we carried out a pretest in which the managers' questionnaire was reviewed by four local government managers, and two employees in each of these local authorities were interviewed to obtain feedback on the questionnaire prepared for the other staff. This pretest confirmed that the instructions and the questions were clear and that the planned administration procedure would be effective. The next step was to contact all the city councils in the population by telephone to identify the human resource managers and request city council participation. Likewise, they were informed of the purpose and relevance of the research project and the confidentiality of the responses.
The managers were asked to complete the questionnaire addressed to them (on structural empowerment and performance), and were invited to send the employees' questionnaire (on employee outcomes) at random to a minimum of four employees.
City councils with fewer than four employee responses were removed, following previous contributions that established a similar minimum number of employees (Seibert et al., 2004). After this step, we equalised the number of respondents from each organisation by randomly sampling observations from city councils with more than six respondents (Schneider et al., 2003). As a result, some responses were deleted from these city councils, and the number of employees per organisation ranged between four and six, yielding a sample of 103 manager questionnaires and 461 employee questionnaires. The sample error for the organisational level sample was ±8.33 at the 5% significance level. The average number of employees per local authority was 4.48 (SD ¼ 0.7). The participants (employee level) were predominantly women (62.7%) and civil servants (79%), reporting an average of 17.52 years' experience in their organisations (SD ¼ 9.72), and an average age of 46.9 (SD ¼ 7.52).

Measures (see Appendices A and B)
At the organisational level, we used 22 items adapted from Lawler et al. (2001) scale to measure structural empowerment. Local authority managers were asked about the dimensions of decision-making power, information sharing, rewards, and knowledge and training. Guided by previous studies (e.g. Datta et al., 2005), a single index was created by taking the mean of the four subscales constructed from the survey items. Cronbach's alpha (a) for this empowerment scale was 0.93. We evaluated perceived organisational performance using the eight items from Walker and Boyne (2006) scaledesigned specifically for the public sectormeasuring output and efficiency, responsiveness, and service outcomes. Responses from local authority managers to the three dimensions were averaged to form an overall organisational performance score (a ¼ 0.90).
At the employee level, job satisfaction was measured with Warr and Inceoglu (2012) single item. The single-item measure is widely accepted in the literature (e.g. Warr & Inceoglu, 2012). We used the three items of affective commitment from Gellatly et al. (2006) organisational commitment scale (a ¼ 0.86). Job anxiety was assessed with Jensen et al. (2013) five items (a ¼ 0.89). Finally, we controlled for gender at the employee level, and for local authority size at the organisational level, in line with previous studies (e.g. Jensen et al., 2013). Similarly to previous research (e.g. Basit, 2019;Chen & Tang, 2018;Semerci, 2019) which only included control variables that were significantly related to the outcome variables in their structural models, we do not incorporate other demographic variables such as tenure and age as control variables to test the hypotheses because they were not significantly related to the well-being variables in the preliminary analyses.

Test of the measurement
As recommended by Ahire and Devaraj (2001) and Bagozzi and Yi (2012), confirmatory factor analysis (CFA) was used to assess the properties of the measurement scales. CFAs were performed for each scale to assess the dimensionality of the study variables (e.g. Mueller, 1996). The goodness-of-fit indices revealed an adequate fit to the data for all the models, except for the job anxiety scale, which showed a poor fit. As recommended in the literature (Nunnally, 1978;Sellin & Keeves, 1997), we removed one item with a low loading (see Appendix A), after which the model showed acceptable fit values. Table 1 displays the information regarding these CFAs and Appendix B shows the loadings. Then, since it is advisable to demonstrate the discriminant validity and reliability of all the measures where the substantive links of the model are tested (Mathieu & Taylor, 2007), an overall CFA from the organisational level was conducted. Its results showed an adequate fit (see Table 1).
Regarding the convergent validity of the measures, as displayed in Table  1 and Appendix B, the Bentler-Bonett normed fit index (BBNFI) values above 0.9 (Ahire & Devaraj, 2001), and the average variance extracted (AVE) above 0.5 (Fornell & Larcker, 1981), show satisfactory average communality. Furthermore, we used two approaches to test for discriminant validity. First, the adequate fit we found for the overall CFA ensures that each criterion really does represent a different concept (note that job satisfaction was excluded from the CFA because it was assessed using a single item). Second, to test this overall four-factor CFA more rigorously, we compared competing models, as shown in Table 2. Thus, we followed the four steps of a comparison test, and found that the one common factor model, the two-factor model, and the three-factor model did not fit the data better than the hypothesised four-factor model.

Common method bias
Some procedural techniques were applied to minimise the risk of common method bias (CMB). These techniques focused on the design of the data collection procedure as a step prior to initiating research, as recommended by Podsakoff et al. (2012) and Brannick et al. (2010). First, two types of informants (one manager and several employees) were approached in each city council, representing the two levels of the organisation, so as to lessen the risk of same source common method variance (CMV) (Shen, 2016;Snape & Redman, 2010;Vermeeren, 2014). Since the links to be examined involved relationships with aggregated grouplevel well-being variables (which were measured with multiple sources in each local government), concerns about bias decrease (Jong & Ford, 2016). We gave respondents an appropriate explanation and instructions, assured them that there were no right or wrong answers, and asked them to rate items as they considered fitting and as honestly as possible.
We also labelled each section of the questionnaires with titles for the variables of interest. Finally, and as noted earlier, we gathered data through online questionnaires rather than face-to-face interviews. We also tested for CMB by means of two statistical procedures . Given that our model is multilevel, we tested CMB for variables with responses, first at the organisational level, and then at the employee level, since the aim was to demonstrate that having the same data source does not hamper reliability. First, we applied the latent method factor approach by estimating a measurement model in which the items of structural empowerment and organisational performance were allowed to load on their theoretical constructs as well as on a latent common method factor. The variance extracted by the common Note: Dv 2 represents the difference between the v 2 value of the alternative measurement models and the hypothesised measurement model (four-factor model). aThe items of affective commitment and job anxiety were loaded on a single factor. bThe items of structural empowerment and organisational performance were loaded on one factor, and items from affective commitment and job anxiety on another one. cAll the items were loaded on one factor. factor was 0.078, which is much lower than the recommended 0.5 cut-off (Fornell & Larcker, 1981), and the factorial structure of the theoretical constructs was the same after introducing the latent common factor. Then, in order to reinforce the conclusion of the absence of CMB and following authors such as Ketkar andSett (2009), Craighead et al. (2011) and Oppenauer and Van De Voorde (2018), we applied a second technique: the chi-square difference test of two models, that is, the one-factor model versus the multifactor model (CFA) with structural empowerment and organisational performance. A comparison of the two models suggested that the two-factor CFA had a significantly better fit to the data than the single-factor model (v 2 difference ¼ 21.95, p < 0.001). We followed the same procedure for the employee level (affective commitment and job anxiety variables). For the latent method factor approach, the variance extracted by the common factor is 0.345. Regarding the chisquare difference test, the two-factor model (CFA) fit significantly better to the data than the single-factor model (v 2 difference ¼ 426.28, p < 0.001). Therefore, in light of the results obtained, we considered that CMB is not a concern in the present study.

Analytical strategy
We conducted multilevel structural equation modelling with the robust maximum likelihood estimator to analyse the hypotheses using MPlus software. To assess whether multilevel analysis was appropriate, we calculated the variation between group levels, estimating the intraclass correlation coefficient (ICC1) for job satisfaction, affective commitment and job anxiety. The results of these ICC1 showed values (0.076, 0.115 and 0.052, respectively), substantially greater than the recommended cut-off of 0.05, and comparable with ICC1s reported in previous organisational studies (e.g. Beham et al., 2015;Oppenauer & Van De Voorde, 2018;Vermeeren, 2014), thus supporting the suitability of multilevel analysis (Heck, 2001;Muth en & Satorra, 1995). Therefore, we considered that the multilevel procedure was appropriate. In order to obtain a more parsimonious model we estimated a path analysis using the composite measures (the mean of the corresponding items) of each variable studied. To examine the mediation role, we computed the indirect path based on the product of coefficients. The statistical significance of the indirect path was then further assessed with the RMediation application (Tofighi & MacKinnon, 2011), which computes the 95% confidence intervals for the indirect paths on the basis of the distribution of the product method, and thus overcomes problems traditionally associated with the Sobel method (MacKinnon et al., 2007). Table 3 presents the means, standard deviations and bivariate correlations of the research measures at the two levels of analysis. Table 4 shows the parameter estimates for the model. The proposed model has  an adequate fit, as shown by the value of the indices (v 2 (7) ¼13.318, pvalue ¼ 0.06; CFI ¼ 0.95; RMSEA ¼ 0.04). Table 4 shows the significant positive association between structural empowerment and organisational performance (b ¼ 0.26; p < 0.05), thus supporting Hypothesis 1. These findings indicate that the more structural empowerment implemented in a local authority, the better its organisational performance. Structural empowerment was hypothesised to be positively related to job satisfaction, affective commitment, and job anxiety (Hypotheses 2a, 2 b, and 2c, respectively). However, we found that structural empowerment is not significantly related to either job satisfaction or affective commitment; Hypotheses 2a and 2 b are therefore rejected. Similarly, Hypothesis 2c is rejected because while structural empowerment does have a significant relationship with job anxiety, it is contrary to that posited (b ¼ À0.49; p < 0.05).

Results
The results related to Hypotheses 3a-c were as follows. No significant relationship was found between job satisfaction and organisational performance; Hypothesis 3a was therefore rejected. Hypothesis 3 b was supported at the 10% confidence level (b ¼ 0.49; p < 0.10), confirming that affective commitment to a local authority is positively associated with organisational performance. Job anxiety was negatively and significantly related to organisational performance (b ¼ À0.33; p < 0.05), supporting Hypothesis 3c. The higher the levels of job anxiety in local authorities, the lower the level of organisational performance.
Hypotheses 4a-c posited that job satisfaction, affective commitment and job anxiety mediate the relationships between structural empowerment and organisational performance. As noted earlier, we found no statistically significant link between structural empowerment and job satisfaction or between job satisfaction and organisational performance; Hypothesis 4a was therefore not supported. Similarly, regarding Hypothesis 4 b, structural empowerment is not related to affective commitment, and therefore mediation is not confirmed. Structural empowerment is significantly related to job anxiety, and job anxiety is significantly associated with organisational performance, thereby meeting the requirements for mediation. The results of additional indirect path tests (b ¼ 0.16; p < 0.10) and the 95% confidence interval (CI) [0.007, 0.211] support a positive mediation of job anxiety in the structural empowerment-organisational performance relationship. Only job anxiety appears as a mediator variable, because the 95% confidence interval excludes the zero value. However, this result is contrary to what was posited in Hypothesis 4c, which supported the conflicting outcomes perspective. Empowerment thus contributes to improving performance by helping to reduce anxiety levels.

Supplemental analyses
To further examine the relationships between structural empowerment, employees' well-being variables and performance, we re-estimated the multilevel mediational model by performing four models, one for each of the four dimensions of structural empowerment (information, training, rewards and decision-making power) instead of using the single index. Given that structural empowerment consists of a sub-set of practices, aggregation in a single index may mask the possible existence of competing effects of the different dimensions, so that some could compensate others, leading to the loss of important information.
The results show that only the model in which the information dimension was introduced presented a significant direct influence on organisational performance (b ¼ 0.36; p < 0.01). Regarding the connection with well-being variables, only job anxiety was negatively associated with information (b ¼ À0.55; p < 0.01) and rewards (b ¼ À0.37; p < 0.05) in the corresponding models. Consequently, for the indirect link, the information (b ¼ 0.14; p < 0.10) [95% CI ¼ 0.003, 0.133] and rewards (b ¼ 0.27; p < 0.10) [95% CI ¼ 0.001, 0.240] dimensions are related to organisational performance through job anxiety. While there is a partial mediation in the model with information, the model with rewards shows full mediation, since the rewards dimension does not directly affect organisational performance. These findings mean that giving employees information and, to a lesser extent, providing performance-based rewards, are likely to be positively associated with performance on their own. However, the other two practices do not show any link with performance. In sum, the pattern when practices are individually considered is not very different from their aggregation in an index, as in both types of analyses (aggregated in a single index and disaggregated by dimensions) the mutual gains perspective is reflected and no competing effect is found.

Discussion
The rationale for this research was to examine the extent to which structural empowerment benefits both organisations and employees, or whether there is a trade-off in terms of which outcomes to prioritise. To this end, we studied the influence of structural empowerment on organisational performance and the mediating role of various forms of employee well-beingjob satisfaction, affective commitment and job anxietyin the context of local authorities. Our findings partially confirm the hypotheses put forward. The study demonstrates that structural empowerment is linked to organisational performance through job anxiety. At the individual level, structural empowerment does not appear to increase workforce anxiety; indeed, it is associated with lower levels of anxiety. Nevertheless, it was not found to be positive and significantly linked to job satisfaction or to affective commitment. Concerning the influence of the three forms of well-being on organisational performance, the results only confirmed the power of job anxiety and, less significantly, the role of affective commitment. In what follows, we discuss the main research implications.

Research implications
Mediating role of well-being: mutual gains versus conflicting outcomes perspective In consonance with the mutual gains perspective commonly developed in the HR literature (e.g. Peccei et al., 2013), the findings demonstrate the vital importance of structural empowerment practices in improving performance in organisations, and contrary to our expectations, in reducing employees' job anxiety levels. This deviates from the 'dark side' arguments in the NPM literature and the conflicting outcomes perspective of HRM (Chandler et al., 2002;Jensen et al., 2013). The supplemental analysis suggests that the information dimension in particular has considerable capacity to lower job anxiety levels, which could mean that employees appreciate knowing the city council's plans, their performance and other information that might affect them.
At the individual level, since structural empowerment does not appear to increase workforce anxiety, it could be inferred that the slow pace at which Spanish organisations are implementing NPM changes might mean that employees do not feel under pressure or strain. The job demand and control model (Karasek, 1979) lends support to this suggestion. This model claims that control over potential stressorssuch as tasks, activities and work decisionsand feelings of autonomy help employees cope better with the demands of their jobs and reduce perceived strain. Structural empowerment transfers decision-making power, and therefore more control and discretion, to employees. Consequently, they may feel less anxious in the work context because they feel they can face the demands of their jobs with greater autonomy, supported by training, information, and rewards. Taking into account that in the sovereign model of governance characteristic of the Spanish public administration (Torres et al., 2011), most employees are civil servants with permanent contracts, which may mean they perceive greater job security and less uncertainty and as a result, they are not afraid of challenges. In such a context, our supplemental analyses may also provide some possible explanations. Increased information may mean employees feel less worried about coming changes, and receiving rewards might help to ease any concerns they may have.
Despite the mutual benefits reported, our findings also are consistent with predictions of the conflicting outcomes perspective. We found confirmation for the sceptical view of HRM (no link between HRM and employees' well-being) and, partially, the 'dark side' of NPM, since despite the beneficial connection with performance, structural empowerment was not found to be positive and significantly linked to job satisfaction or to affective commitment, in contrast to previous studies (e.g. Kirkman & Rosen, 1999). Specifically, our findings reinforce the view that structural empowerment does not have beneficial consequences for employees in all situations or for all types of employees. This supports the idea that staff morale is not increased by empowerment practices in this public sector context.
The differences in the significance of the three well-being variables in our model may possibly be due to their distinct nature. As explained previously, job satisfaction and affective commitment can be related to eudaimonic well-being (work conditions-desires fit). By contrast, job anxiety is more related to hedonic (pain-pleasure) emotions. Because of this, job satisfaction and affective commitment in our study may have a different behavioural pattern from that of job anxiety. The person-organisation fit, related to the eudaimonic view of well-being, offers a suitable explanation. Vidal (2007) showed that empowerment does not necessarily increase satisfaction, since individuals' work orientation may mediate the association between empowerment and job satisfaction. Thus, perhaps if employees' orientation does not fit with the empowerment 'trend', they will neither experience more satisfaction nor feel more committed. This makes sense when the profile of tenure and age of the workforce in our sample is considered, since it may be representative of people who are less willing to change and grow. For instance, Rhodes (1983) argues that age is negatively related to the need for self-actualisation and growth.
Overall, these findings contribute to enrich the scarce research that considers a more balanced view of HRM (Paauwe, 2004(Paauwe, , 2009), is contextually based, and takes into account a wider perspective of the functioning of organisations.

Empowerment and performance
This paper also contributes to the social capital theory in that it illustrates the precepts of this approach (Seibert et al., 2001) in the context of local government. As previous contributions have contended (e.g. Brunetto & Farr-Wharton, 2008), the implementation of NPM practices, such as empowerment, seems capable of creating an organisational network in local government that provides a flow of relevant information and resources, or even emotional support, that arm employees with better tools to undertake their work, serve clients, and in turn, positively affect organisational outcomes. Likewise, as asserted by recent 'black box' arguments of public management, establishing a system of practices such as delegating power or training helps to build a more general structure in the organisation, providing support for achieving better performance (Burgess, 1975).

Employee well-being and performance
These findings enrich the growing body of research on how to stimulate performance in public organisations, an issue that remains highly topical. This research holds that symptoms of stress and strain lead to a state of tension and low energy that negatively affects performance (Sackey & Sanda, 2009). However, the positive influence of job satisfaction on organisational performance remains unconfirmed. This might be due to possible measurement effects that could have influenced the results. Judge et al. (2001) suggest that the concept and measurement of job satisfaction should perhaps be closer to emotions than to attitudes. According to some authors (e.g. Brief & Roberson, 1989), job satisfaction fails to anticipate performance because the current job satisfaction measure reflects cognitive evaluation more than affective tendency. In our case, the item used refers to a general job satisfaction assessment, so it does not specifically capture affect. This could explain the absence of a link between job satisfaction and organisational performance, even more so if we consider that affective commitment (which is specifically a measure of affect) does have a connection with performance.

Practical implications
From a practical perspective, our findings suggest two types of actions that may be valuable to city councils. First, in order to improve organisational performance local authority managers should strive to implement structural programmes including training, rewards based on productivity, dissemination of information, and tools for decision-making participation. Given the need to achieve efficiency, effectiveness, and public satisfaction, strategies are needed to face the new challenges. However, since workforce development tends to be a minor HR function in which many local governments have little experience (that is, HR departments deal mostly with payroll administration and organising training courses), it could be advisable to seek outsourced support from people development specialists to implement the structural programme.
Specifically, given the role that rewards and information play, managers should make use of these tools to compensate the increase in responsibilities and new challenges related to the implementation of structural empowerment. For instance, they could disseminate information on the organisation's aims among employees, or pay bonuses for achieving targets. Moreover, perhaps exploring the differences between, for instance, employees who receive rewards and those who do not would demonstrate the importance of such a technique in actual day-to-day practice.
Second, and turning to employee-related actions, public managers should aim to reduce employee anxiety at work. Detrimental consequences of anxiety disorders are a major problem in the public sector (McHugh, 1998), leading to high costs for organisations if they result in time off work due to sickness. Therefore, anxiety can be reduced by developing structural empowerment practices, but also by promoting employees' perceptions of control over potential stressors and feelings of autonomy (Karasek, 1979). On a practical level, local government managers could foster activities aimed at helping employees to cope with their demanding situations and, thus, balance their levels of job anxiety. Mindfulness sessions (Fortney et al., 2013), for instance, may be one way of helping workers reduce their anxiety and stress levels. Providing resources such as access to emotional support, physical assistance or sports activities would also be useful (Kouchaki & Desai, 2015;Witt & Carlson, 2006). Likewise, promoting perceived organisational support could also help employees to cope with anxiety (Sadatsafavi et al., 2015), since it enhances their sense of control (Johnson & Hall, 1988). Specifically, social support through colleagues is highly beneficial (Muschalla et al., 2010) and can be fostered by management through organised group activities and teamwork. Moreover, other strategies focused on employees' daily job routines, such as flexibility in work arrangements (e.g. flexible work hours) or personalised work surroundings (e.g. choice of office design) can also help to alleviate feelings of anxiety (Mellifont et al., 2016).

Limitations and future research
The first limitation of the study is the sample, which covers only local governments. Future work in other public settings is recommended in order to generalise our model. Specifically, given the particularities of Spanish organisations, such as the slow rate at which they implement changes, it would be advisable to compare our findings with a different EU country or in settings with lower job security to see whether the results vary. It would also be interesting to explore what the main causes of anxiety are in such settings, given its connection to organisational performance. A different sample would also overcome the effects of the specific characteristics of the workforce in our study in terms of age and tenure and its corresponding reticence to change. Furthermore, the cross-sectional design of the study does not allow inference of causality, so a longitudinal design may be useful to test causal links. In addition, as explained earlier, in the context studied employees' individual orientations may play a role in job satisfaction levels. Future studies should analyse whether the fit of employees' orientation with the values embedded in structural empowerment may have explanatory power. In a similar line, as mentioned earlier, job satisfaction could be better measured by trying to infer affect and emotion (Judge et al., 2001). Our study covered job satisfaction at a general level, which limits knowledge about its behaviour. Future studies could usefully examine how structural empowerment relates to different facets of job satisfaction. Also in terms of employee variables, it would be interesting to examine other outcomes, beyond job satisfaction, such as the occurrence of organisational citizenship behaviours and the absence of negative emotions. Furthermore, the concept of eudaimonic well-being (tested here with job satisfaction) could be examined in greater depth by analysing the three dimensions suggested by McMahan and Estes (2011): experience of a sense of purpose, development of personal strengths, and contribution to society. Likewise, virtue (Huta & Ryan, 2010) could also be a valuable construct to address eudaimonia. On the other hand, the study of the link between structural empowerment and well-being outcomes could be enriched by considering whether or not employees want to be empowered, since a difference in preferences could alter the connection between the two variables.
At the organisational level, future research could explore in greater depth which specific information and reward strategies are most useful in lowering job anxiety levels and enhancing organisational performance. This could be valuable in further elucidating the 'clue' techniques of the general policy of structural empowerment.
It is also important to note the perceptual nature of the organisational performance measure. We were unable to measure real performance given the absence of such data. Although performance measurement requirements are in place in Spain, they do not have specific indicators and are defined and introduced at the discretion of each local government. As a result, there is no register of performance for all Spanish local authorities. External tools, such as the Comprehensive Performance Assessment implemented in the United Kingdom as an external measure of the performance of local municipalities (Lockwood & Porcelli, 2013;Walker & Boyne, 2006), involve simulating models of explicit motivation for such organisations and collecting objective indicators for distinct areas of local government activity. If a comparable tool is ever introduced in the Spanish local government system, future empirical studies could make use of objective data to assess the links in this hypothesised model in a more rigorous way.
In addition, and because different employees in a local government could experience structural empowerment practices in different ways (Liao et al., 2009), future research could gather employees' perceptions of structural empowerment in order to obtain data from respondents at the two levels and to monitor possible divergences.
Another interesting line would be to explore curvilinear effects. Given that a new stream of psychological well-being studies has identified the limits of positive experiences (e.g. Grant & Schwartz, 2011), research could usefully examine the optimal levels of structural empowerment for a positive association with well-being variables.
One final suggestion for future research would be a qualitative study. Petter et al. (2002, p. 397) state that 'empowerment is both locally defined and individually valued', so interviews and feedback in a specific context could add valuable information to further understanding of its consequences.

Conclusion
The analysis of the mediating role of well-being suggests that in Spanish city councils structural empowerment acts as a 'reassuring' mechanism for employees, but not as a catalyst of well-being. Thus, our findings support the integration of the optimistic and sceptical perspectives on HRM, as considered in Peccei et al. (2013) and illustrated by Van De Voorde et al. (2016). Regarding the NPM discussion, the results do not completely confirm its 'dark side' (Diefenbach, 2009) of increased stress and anxiety, although it is noteworthy that the absence of influence of empowerment on employees' job satisfaction and affective commitment could be a sign of the inability of the new HRM within the NPM paradigm to accompany enhanced organisational performance by motivating the workforce.

Data availability statement
The data are not publicly available due to their containing information that could compromise the privacy of research participants; at the beginning of the data collection process the confidentiality of the data was assured to the participants. Therefore, the data that support the findings of this study are available from the corresponding author, upon reasonable request.