Abstract
Empirical studies have found a positive relationship between public service motivation (PSM) and individual performance. However, it is unclear what public service motivated employees are doing in terms of behavior that makes them perform. Moreover, it is uncertain whether PSM inspires similar behaviors among employees in different contexts. Conceptualizing performance as a multidimensional construct, this study investigates the relationship between PSM and self-reported output, service outcome, responsiveness, and democratic outcome behaviors. Using structural equation modeling on survey data from 459 employees in people-changing (service production, aimed at changing the user) and 461 employees in people-processing (service regulation, categorizing, and processing users) organizations, the results show that PSM is related to all performance-related behaviors in the people-changing group, but neither to output nor responsiveness in the people-processing group. PSM’s relationship to behavior may thus differ between contexts.
Introduction
In recent decades, attention toward explaining why public employees shirk or act in their own interests has been balanced by studies trying to explain why they do their job correctly, work hard, and do good for society. Researchers have aimed to provide an alternative perspective to the “budget-maximizing, lazy” public employee by studying why firefighters, policemen, nurses, policymakers, and other public employees, despite the sometimes difficult circumstances, go above and beyond the call of duty and perform well (DiIulio, 1994; Perry & Wise, 1990). One explanation seems to be rooted in public service motivation (PSM), which drives employees in organizations or jobs with a public function to perform well (Brewer, 2008; Perry & Wise, 1990). Empirical research shows that PSM matters for whistle-blowing, ethical conduct, and performance (Andersen, Heinesen, & Pedersen, 2014; Bellé, 2013; Brewer, 2008; Brewer & Selden, 1998; Choi, 2004; Taylor & Taylor, 2011).
However, although studies generally show that those with high PSM perform better, this does not provide full insight into how public service motivated employees behave. Behavior is seen as a crucial intermediate between attitudes and performance in the human resource management (HRM) literature but has been underexposed within the public management literature (Boselie, Dietz, & Boon, 2005; P. M. Wright & Nishii, 2006). What performance, and performance-related behavior, actually is in public service providers is not easily captured: Public service providers have multiple goals and multiple stakeholders, and what they should do is politically determined (Boyne, 2002; Brewer, 2006; Brown, Potoski, & Van Slyke, 2006; Moynihan et al., 2011).
This multiplicity of interests makes it impossible to identify a single measure that accurately represents performance (Brewer, 2006). Boyne (2002) therefore conceptualized the performance of public service providers as multidimensional, consisting of output, efficiency, service outcomes, responsiveness, and democratic outcomes. If the desirable performance of public service providers is multidimensional, employees will have to show behaviors relevant to all those dimensions to perform well. In references to public employees, scholars have argued that multiple types of performance-related behavior are expected of them (Jorgensen & Bozeman, 2007; Moynihan et al., 2011). This study therefore follows Boyne’s (2002) multidimensional view—but applies it to the individual level.
Moreover, it is unclear whether the relationship between PSM and performance varies according to context (Ryu, 2014). When focusing on multiple dimensions of behavior, the institutional context becomes highly relevant because what is asked of the service providers depends on the service they provide. Liu and Perry (2014) found that the relationship between PSM and outcomes was mediated by organizational identification. Several authors have argued that the fit between employee PSM and the organization is important, and more attention should be paid to the context in which employee work (Bright, 2007; Ryu, 2014; Wright, 2007).
Regarding PSM, it may be highly influential what the dominant logic is regarding the task and purpose of the service. Kjeldsen (2014) distinguished between regulatory and production services and found that this distinction mattered for the relationship between PSM and job choice. If the primary goal of the service provider is to redistribute or regulate services, referred to as people-processing, other types of behaviors are seen as more appropriate than where the main purpose of the service is to produce services by changing people (Hasenfeld, 1972, 1983; March & Olsen, 1989). The institutional context, which determines whether this logic is predominantly people-processing or people-changing, provides guidelines for appropriate behaviors (March & Olsen, 1989; Scott, 2001; Thornton & Ocasio, 2008) and may thus also matter for the relationship between PSM and performance.
This article aims to further unravel the PSM–performance relationship by exploring the relationship between PSM and different types of behaviors in two types of service providers. This article contributes by enhancing knowledge on the context dependency of the PSM–performance relationship. Using structural equation modeling in Mplus v7 (Muthen & Muthen, 2010-2013) with survey data from public employees (n = 459 and n = 461), the relationships between PSM and various dimensions of self-perceived behavior are analyzed. The perceptions of employees can be seen as “one piece” of the performance puzzle (Andrews, Boyne, & Walker, 2006) given that performance is a multifaceted concept that is impossible to fully grasp in a single indicator (Brewer, 2006). As it is self-reported, this study only provides insight into the relationship between PSM and performance from the employee’s perspective. This article starts with a discussion of the relevant literature, from which several hypotheses are formulated. Then, the methods used are presented, followed by the results. In the final section, the results are discussed.
Theoretical Framework
PSM as an Institution-Based Motivation
Perry and Wise (1990) defined PSM as “an individual’s predisposition to respond to motives grounded primarily or uniquely in public institutions and organizations” (p. 368). Vandenabeele (2007), who placed PSM within an institutional framework, describes PSM as “the beliefs, values and attitudes that go beyond self-interest and organizational interest, that concern the interest of a larger political entity and that motivate individuals to act accordingly whenever appropriate” (p. 547). The latter definition specifically mentions that it is a motivational force to act accordingly whenever appropriate, which suggests that institutional context may play a role in determining how public service motivated employees behave, hence this definition is adopted here.
Vandenabeele (2013) empirically linked PSM to self-determination theory (Deci & Ryan, 2000). Self-determination theory posits that motivation is much more complex than intrinsic versus extrinsic types of motivation, and that it should instead be seen as on a continuum ranging from controlled to fully autonomous (Deci & Ryan, 2000). The more controlled the motivation is, the more it is influenced by external pressures and rewards. The more intrinsic, or autonomous, the motivation, the more internal drivers, as opposed to external pressures, determine behavior (Ryan & Deci, 2004; Vandenabeele, 2014). Vandenabeele (2013) found that PSM is mostly related to extrinsic but still autonomous types of motivation. Houston (2011) described PSM as an obligation-based type of motivation. Both these views position PSM as an intermediate type of motivation, one where both internal drivers and external pressures (the institutional context) play a role in determining behavior.
Significant relationships have been found between PSM and work effort (Frank & Lewis, 2004; Leisink & Steijn, 2009; Taylor & Taylor, 2011; B. E. Wright, 2007), job performance (Alonso & Lewis, 2001; Bright, 2007), organizational citizenship behavior (Kim, 2006), and organizational performance (Brewer & Selden, 1998; Kim, 2005). However, before the relationship between PSM and performance-related behaviors can be discussed, the concept of performance in service providers, and the related behaviors, needs to be discussed.
Behavior and Performance
What performance is, and what behaviors lead to performance, is complex and especially so for public service providers (Boyne, 2002). Public service providers have a multitude of stakeholders that they are expected to serve apart from their direct users. Each of these stakeholders has a different view on what constitutes good performance and may emphasize different aspects of performance (Andrews et al., 2006) with their interpretations of performance differing substantially (Andrews, et al., 2010). As all stakeholders, including political appointees, form an opinion on what is most important, performance in public service providers is an inherently subjective concept (Brewer, 2006). As such, there is no single criterion that can be used to accurately capture performance (Boyne, 2002).
Recognizing this, researchers have argued that in a public context a multidimensional view on performance is most appropriate. For instance, Brewer and Selden (2000) distinguished internal and external efficiencies, effectiveness, and fairness when studying perceived organizational performance. Boyne (2002) argued that performance consists of outputs (quality and quantity), efficiency, service outcomes (equity, value for money, impact), responsiveness (citizen and user satisfaction), and democratic outcome (fairness, participation, accountability), and this view is also used here.
This multidimensional view has been used in several studies (including Andrews, Boyne, Jae Moon, & Walker, 2010; Brewer & Walker, 2013; Walker, Boyne, Brewer, & Avellaneda, 2011). Although these studies focus on organizational performance, employees and their behaviors are seen as important factors in determining the performance of public service providers (Atwater, Ostroff, Yammarino, & Fleenor, 1998; Brewer & Selden, 2000; Delery & Shaw, 2001). Despite the HRM literature emphasizing that behavior is a crucial intermediate between attitudes and performance (Boselie et al., 2005; P. M. Wright & Nishii, 2006), behavior has received limited attention within the literature. This is surprising given that performance is mostly distant, and therefore hard to measure, whereas behavior represents a proximal outcome that can be linked to attitudes (Boselie et al., 2005; Guest, 1997).
Often on the individual level, the distinction is made between in-role and extra-role behavior (Kim, 2006; Podsakoff, Ahearne, & MacKenzie, 1997; Williams & Anderson, 1991). Although this distinction can show how employees behave regarding their task-specific role and extra—unpaid—tasks such as helping colleagues, this distinction is less sensitive to the specific context of public employees and the goals they are asked to strive for. In public service providers, policies come to life through the actions of the employees (Hasenfeld, 1983; Lipsky, 1980). If city hall employees are frustrated and therefore become less responsive to citizens, the public service provider as a whole will be evaluated negatively by citizens (Shingler, Van Loon, Alter, & Bridger, 2008). When looking at what behavior is expected of public employees, similarities can be seen to Boyne’s (2002) dimensions of organizational performance. For example, Jorgensen and Bozeman (2007) found multiple values describing how public service employees should behave, and argued that these employees are expected to “think and act” accordingly. Among these values were responsiveness, equity, accountability, reliability, and fairness (Jorgensen & Bozeman, 2007). Hood (1991) found three value clusters that can be emphasized: sigma (lean), theta (fair), and lambda (robust). It should thus be possible to distinguish various dimensions of behavior on the individual level.
In general, individual performance can be defined as an individual’s contribution to achieving the public mission of the organization. There is, however, no consensus on how individual performance can best be measured (Brewer, 2006; Meier & O’Toole, 2013). Most researchers agree that both subjective and objective measures have their weaknesses but also their values (Andrews, Boyne, & Walker, 2011; Brewer & Selden, 2000; Conway & Lance, 2010). As objective data are rarely available on a broad spectrum of performance aspects, these measures are often quite narrow and hard to compare across jobs (Andrews et al., 2006; Brewer & Selden, 2000). Although subjective data can be biased due to overestimation, research suggests that the assumption that individuals inflate reports on their own performance is overstated (Andrews et al., 2011; Conway & Lance, 2010; Spector, 2006).
Employees are seen as a valuable source of information because they have a good view on internal processes (Brewer, 2006; Vermeeren, Kuipers, & Steijn, 2014), “have a better all-round understanding of the challenges facing their organization” and their perceptions “provide more insight in performance measures on which organizational decisions are based” (Andrews et al., 2010, p. 109). This study uses self-reports from employees on performance-related behavior. When no other sources are available, employees can provide valuable insights. Moreover, as this study includes a range of different jobs and domains, subjective data are the most suitable as they are easier to compare.
Previous Findings on the PSM–Performance Relationship
When reviewing the empirical evidence on the relationship between PSM and performance, almost all survey studies have found a positive association (Andersen et al., 2014; Bellé, 2013; Kim, 2006; Leisink & Steijn, 2009; Vandenabeele, 2009), although Alonso and Lewis (2001) failed to find a relationship. Of these studies, only one analyzed more than a single-individual performance-related behavior. In a quasi-experimental setting, Bellé (2013) studied the relationship of PSM with persistence, output, productivity, and vigilance (behaviors) in voluntary tasks, and found PSM mattered for all four. An important caveat remains as no study has analyzed multiple dimensions of performance-related behaviors simultaneously. However, this does not mean that there is no evidence as to whether one could expect a relationship with each dimension.
First, regarding output, Bellé (2013) found that high PSM increased the quantity and quality of output on undertaking voluntary tasks for employees who perceived a high prosocial impact. Moreover, Park and Rainey (2008) presented evidence that PSM was indirectly related to the quality of work. However, there is also evidence that other norms, such as professional norms and standards, explain quality better than PSM (Andersen, 2009). This perhaps suggests that outputs are not the most salient dimension of performance for public service motivated employees.
Turning to efficiency, Ritz (2009) found that only the commitment to the public interest dimension was related to the internal efficiency of the organization, and other dimensions were not. Bellé (2013) found that PSM was related to efficiency measured as the time spent divided by the number of kits correctly prepared. Still, Petrovsky and Ritz (2014) found no relationship between PSM and internal efficiency after controlling for bias.
Three studies provide insight into how PSM might be related to service outcomes. First, Andersen and Serritzlew (2012) studied register data services of Danish physiotherapists. Although they found no differences in the number of services, having a high PSM did affect the proportion of disabled patients treated, suggesting that PSM contributes to attending to general well-being. Considering providing value for money, Moynihan (2013) debunked the idea that bureaucrats are budget maximizers by showing through a vignette study that PSM did not lead to budget maximization. Finally, Andersen et al. (2014) found that teachers’ PSM was positively related to student exam grades. Given that service outcome is a very public or community-oriented type of performance, the relationship with PSM may be strong.
Focusing more on the individual user, responsiveness then reflects a stakeholder entity other than societal service. Pedersen (2013) argued that it is not PSM but user-orientated motivation that relates to behavior aimed at individual clients. Nevertheless, one can assume that PSM is related to client satisfaction insofar as the latter overlaps with societal interests, whereas if these do not overlap, then service outcome will be placed above individual needs by public service motivated employees (Perry & Wise, 1990). For instance, a teacher may be as responsive as is possible to individual student needs, but not to such a degree that it disturbs teaching all the students.
Evidence of a relationship between PSM and democratic outcome is mostly linked to ethical conduct. For instance, Brewer and Selden (1998) found that those who blow the whistle were also highly motivated to serve society, and Choi (2004) found that the self-sacrifice dimension was related to more ethical conduct. Two studies have also found a positive relationship between PSM and the use of performance information, seen as indicative of accountability (Kroll & Vogel, 2014; Moynihan & Pandey, 2010). Finally, Kim (2006) related PSM to better compliance. Thus, there is evidence that PSM is related to democratic outcome.
What a public service motivated employee perceives as being appropriate behavior will to an extent be determined by the institutional context in which the work is done (March & Olsen, 1989). The institutional context matters for the behavior of employees because it provides employees with a set of guiding norms and may thus determine to what dimension PSM is related in which context. This is discussed next.
How Institutional Context Matters in the Relationship Between PSM and Behavior
Institutional theory emphasizes how institutions influence behavior by defining what is seen as appropriate and by providing individuals with social norms that they can use to make sense of a situation (Greenwood, Diaz, Li, & Lorente, 2010; Scott, 2001; Thornton & Ocasio, 2008). Institutions can impose restrictions or support certain behavior through rules and norms that define what behavior is appropriate (March & Olsen, 1989; Scott, 2001). For instance, in a private company, behavior aimed at getting the highest profit possible may well be highly appreciated, whereas in public service providers such behavior could be seen as inappropriate. Moreover, what is seen as appropriate behavior can change over time. For instance, market incentives and developments often placed under new public management such as output steering have led to a greater emphasis within public organizations on economic value as opposed to democratic values (Boyne, 2002; Bozeman, 2007; Moynihan, 2010), and this may influence employee behavior.
Institutions are present on multiple levels, and even carry through in the structures of organizations and job characteristics. Scott (2001) distinguished three pillars of institutions: coercive, normative, and cultural-cognitive. The first can be seen as a structural view of institutions: Rules and regulations can prevent individuals from acting in certain ways because they impose consequences on certain actions. Individuals also look at the norms and symbols in their environment for clues on how to act, and to deduce what is appropriate behavior (March & Olsen, 1989). Thus, institutions influence individuals not only through determining structures but also through determining a dominant normative logic on what is seen as appropriate behavior, and the meaning given to symbols and artifacts (Scott, 2001).
This implies that there are variations within institutional contexts that lead to different signals toward individuals. Vandenabeele (2011) empirically linked various institutional factors to PSM. According to Perry and Wise (1990), a public institutional logic incentivizes public service employees to do well. However, what this publicness is remains unclear (Rainey, 2003). First, it is unclear what public is and what it is not. Bozeman (1987), for instance, argued that all organizations are public to some degree because publicness is determined by ownership, financial resources, and political control. Recent bank takeovers in Europe illustrate how what was previously thought of as a private domain can suddenly become public. Second, even when concentrating on organizations with a public function, these organizations can differ substantially in their type and degree of publicness (Antonsen & Beck Jørgensen, 1997; Bozeman, 1987; Vandenabeele, 2008). Rather than public service providers being homogeneous, they actually form a complex web of organizations with different missions, stakeholders, tasks, and political control.
As this study focuses on PSM—the motivation to contribute to society—the type of work is highly relevant (Kjeldsen, 2014). Therefore, this study asks if differences in dominant logic regarding the task of the public organization matters for the relationship between PSM and performance. The present study argues that the primary process’s dominant service logic will influence what behavior employees with high PSM see as appropriate because the primary process leads to the PSM-desired outcome: a meaningful impact on society. Of course, other variations in the institutional context—for instance, regarding the potential impact on society of the job—can also be of importance for the relationship between PSM and performance. However, variation in the service logic relates to the mission and aim of the organization, and variation in the mission may mean that different behaviors are required to achieve the mission.
A fundamental distinction within public service providers regarding their primary process is whether services are produced that aim to change the users of the service or to regulate service by processing users (Hasenfeld, 1972, 1983). For instance, in a school, one might expect employees to build long-term connections with the students to be able to teach them, whereas city hall public employees are asked to refrain from too personal a contact, so as to stay as neutral as possible in assessing an application. Previous studies on PSM have found this distinction to be relevant (Kjeldsen, 2014; Van Loon, Leisink, & Vandenabeele, 2013).
Although Boyne’s (2002) dimensions are relevant in each public organization, the emphasis laid on them and the priority given to one dimension versus the other may differ between organizations with different logics. In people-changing organizations, the main purpose is to change the user and thus to provide a service (Hasenfeld, 1972). Kjeldsen (2014) called this “service production.” As examples, a student needs to learn new things and a patient needs to be cured. This type of service requires long-term and/or personal contacts, interactions with users, and a focus on being responsive toward the user. In people-changing service providers, one can expect employees who want to do good for society to be focused on responsiveness and treating all users properly, and reporting on progress as this is the key to delivering or creating the services. Without such cooperation, and thus a good relationship, with a student, patient, or inmate, employees are unable to reach them, make contact, and change them (Hasenfeld, 1972, 1983). This argument leads to our first hypothesis:
In people-processing service providers, such as many functions within a city hall, the focus is on regulating services and the product is a changed status of the user. This requires objective classification and often entails short, one-off interactions (Hasenfeld, 1972, 1983). The primary process emphasizes the community at large rather than the individual user. For instance, a city hall employee can change the status of a citizen by granting a residence permit, but will then move on to the next case (Kjeldsen, 2014). In people-processing service providers, responsiveness may be of less importance. Given that the core tasks are regulating and redistributing public goods, public service motivated employees can be expected to focus more on fair treatment, due process, and the value for society as a whole—thus service and democratic outcome. This leads to our second hypothesis:
Method
In this section, the data collection, the measures employed, and the data analysis are explained.
Data Collection
An online survey was sent out in 2012 to several organizations with a public function, including schools, municipalities (city hall), police, prisons, and a hospital. All the employees of the selected organizations were invited to participate through email (except in the hospital, where the survey was posted on the internal network). Although the distinction between people-processing and people-changing logics was derived from interviews prior to the survey (see Van Loon et al., 2013), all the job descriptions as provided by the respondents were also independently coded by three researchers as management, supportive, people-processing, people-changing, or mixed. The job coding confirmed the division of the organizations based on the earlier interviews. Two groups were made: Schools, prisons, and hospitals were identified as people-changing services, and city hall administration (i.e. employees in policy related-functions, excluding mayors, politicians and social services) and the police as people-processing services.
Although online surveys have many advantages, such as low costs and ease of use, a drawback is the associated response rates (Crawford, Couper, & Lamias, 2001). Therefore, great care was taken in designing the survey (from use of color and typeface to ease of navigation). Furthermore, the survey was personalized by including a photograph of the researcher, anonymity was guaranteed, a chance to win a voucher was offered and several reminders were sent (Couper, 2008). Finally, it was not compulsory to answer all the questions. All these actions were designed to increase the response rate and reduce social bias (Couper, 2008).
In total, 1,138 surveys were returned (38.7%). Of the respondents, 40.1% were male and 51.4% were female (8.5% did not say). The average age was 43.4, the average tenure was 11 years, and 14.7% held a supervisory position. Additional analyses were conducted to check how representative the samples were of the wider population based on demographic characteristics. The samples, based on national statistics on gender division and average age, were representative for all the types of organizations except for the average age of the police (although the sample was representative of the region from where it was drawn) and there being a slight overrepresentation of women in the school sample. After checking for missing data on the key variables, 1,031 respondents filled in the questionnaire at least partly. However, for the regression analysis responses from 459 employees working in people-changing and 461 employees working in people-processing services were usable due to missing responses on key variables.
Measures
Performance-related behavior was measured with items developed by the author that drew on Boyne’s (2002) five dimensions of performance. They were formulated to refer to specific behaviors employees performed as part of their job. Responses from employees to the survey revealed that one item did not measure what had been intended (the aim was to investigate the provision of equal treatment; but the respondents saw it as referring to distinguishing between citizens to provide good services). Furthermore, the item for efficiency did not actually measure efficiency but output. Four types of behaviors could be distinguished: output, service outcome, responsiveness, and democratic outcome (see the appendix for full list of items).
PSM was measured with items from the international scale developed by Kim et al. (2013) with four items for each of the four dimensions (attraction to public service, commitment to public values, compassion, and self-sacrifice). However, the dimensional structure was not supported by the data: The overlap between dimensions was too high for the individual dimensions to be distinguished. Therefore, using two items from each dimension, a global PSM scale was tested. Although each dimension may have different effects on work outcomes, a global scale reflects the general motivation to contribute to society (B. E. Wright, Christensen, & Pandey, 2013).
Finally, several control variables that have been found elsewhere to be related to performance were included in the structural equation modeling. Gender, job tenure, and supervisory position were included as these may be related to performance (Bright, 2007).
Data Analysis
Full structural equation modeling, using Mplus v7 (Muthén & Muthén, 2010-2013), was applied to test the hypothesized constructs and relationships. By using structural equation modeling, it is possible to simultaneously test for multiple dependent variables (Byrne, 2012; Kline, 2010). Moreover, as the measurement model is also included, it can partially control for measurement error. A two-step approach is used, in which the measurement model (i.e., only the structure of the constructs) is first tested, and only if this fits the data are the regression paths added in a second step (Anderson & Gerbing, 1988; Byrne, 2012; Kline, 2010).
To confirm if the structure of the measures was acceptable, a confirmatory factor analysis (CFA) with a robust maximum likelihood (MLR) estimator was used. This estimator corrects for the skewness of the parameters and non-normality of the items (Kline, 2010). Three fit indices are used to assess the fit of the measures to the data. As the commonly used chi-square index is known to be inflated when the sample size exceeds 200, the comparative fit index (CFI), the Tucker–Lewis Index (TLI), and the root mean square error of approximation (RMSEA) were used. CFI and TLI values above .90 are indicative of acceptable fit, and values above .95 are indicative of an excellent one; similarly, a RMSEA below .10 reflects acceptable fit, and below .08 reflects an excellent one (Byrne, 2012; Hu & Bentler, 1999; Kline, 2010). Reliability was assessed using Raykov’s rho (Raykov, 2009), which is considered more appropriate than Cronbach’s alpha in structural equation modeling as it is based on factor loadings.
Invariance between the people-changing and people-processing groups was also tested. Configural invariance tests whether the construct has the same factor structure across groups and, in this multigroup model, all loadings and variances are allowed to differ. In testing for metric invariance, all the factor loadings are constrained, and for scalar invariance, factor loadings and intercepts are constrained to be equal.
To test the hypotheses, correlations are first analyzed. Following this, a full structural equation model, including paths from PSM to each type of performance-related behavior and the control variables, is tested.
Results
In this section, first the measurement model is discussed, followed by the correlations and structural equation models.
Testing the Performance-Behavior Measurement Scale
Each construct was first tested separately, followed by a full measurement model that included all the constructs. A CFA using the four types of performance-related behaviors (Model 1: outputs [three items], service outcomes [four items], responsiveness [three items], and democratic outcome [two items]) indicated that there was room for improvement (see Table 1). Therefore, items with low loadings were removed. For the resulting Model 2 (with three items for outputs, and two each for service, responsiveness, and democratic outcome), all fit indices indicated a good fit.
Measurement Models for Performance-Related Behavior.
Note. CFI = comparative fit index; TLI = Tucker–Lewis Index; RMSEA = root mean square error of approximation.
Measurement Model
The global PSM scale and the performance-behavior scale were tested with the people-changing and the people-processing samples, as shown in Table 2. For PSM one item, on self-sacrifice, had to be deleted to achieve a good fit on all three indicators. Table 2 shows that the seven-item model fitted both samples well. The performance-behavior measure was also tested separately within the people-changing sample and the people-processing sample. Table 2 shows that the model fits both groups well. A full measurement model was tested that included both PSM and the various types of performance-related behaviors. This model fitted both groups well: people-changing: CFI = .959, TLI = .948, RMSEA = .033; people-processing: CFI = .957, TLI = .946, RMSEA = .032. All items and factor loadings are shown in the appendix.
Fit Indices for PSM and Performance-Related Behavior in PC and PP Groups.
Note. PSM = public service motivation; PC = people-changing; PP = people-processing; CFI = comparative fit index; TLI = Tucker–Lewis Index; RMSEA = root mean square error of approximation.
The full measurement model’s invariance was tested, and comparisons were made between three levels of invariance. When comparing the configural and the metric models, there was no significant difference in their chi-square values (Δχ2 = 14.588, df = 11, p = .202). As chi-square does not always accurately reflect the invariance, the difference in the CFIs of the two models was also examined (Cheung & Rensvold, 2002). The difference in the CFIs was .001 (.959 and .958), which is just on the threshold of demonstrating metric invariance. However, as the factor loadings differed between the groups and the fit decreased significantly when further constraints were placed on the model, it would be dangerous to assume invariance. Testing for scalar invariance, the difference between the chi-square values of the configural and the metric models was significant (Δχ2 = 78.627, p < .0001), and the model fit decreased significantly (CFI = .920, TLI = .912, RMSEA = .042). The people-changing and people-processing groups are thus analyzed separately.
Harman’s single-factor test, in which all items are loaded onto one dimension, was performed to test for common method bias within each group. These models had significantly worse fits (people-changing: CFI = .482, TLI = .402, RMSEA = .111; people-processing: CFI = .442, TLI = .356, RMSEA = .110), indicating that common method bias is unlikely to influence the results (Podsakoff & Organ, 1986).
Structural Model
The first step in analyzing the relationship between PSM and perceived behavior is to consider the correlations between the main concepts and with the control variables. Table 3 shows that, in the people-changing sample, PSM is related to all types of performance-related behaviors, whereas in the people-processing sample, it is not significantly related to output or to responsiveness behaviors (Table 4). The tables also show that the different types of behaviors are related, but not so strongly as to create analytical problems. Testing a structural equation model enables the relationship between PSM and behavior to be better understood.
Correlation Table for the People-Changing Group.
Note. PSM = public service motivation.
p < .05. **p < .01. ***p < .001.
Correlation Table for the People-Processing Group.
PSM = public service motivation.
p < .05. **p < .01. ***p < .001.
The full structural equation model for the people-changing group of employees (controlling for gender, job tenure, and supervisory position) is shown in Figure 1. The overall model fitted well (CFI = .953, TLI = .939, RMSEA = .032, df = 130, n = 459). In terms of the various behaviors, the model explained 3.2% of the variance in output, 4.6% in service outcome, 7.2% in responsiveness, and 4.3% in democratic outcome. PSM is positively related to all types of performance-related behaviors. PSM is most strongly related to responsiveness (β = .211, p < .01) and democratic outcome behaviors (β = .267, p < .001). As such, the results fully support Hypothesis 1 in that PSM is related to all types of performance-related behaviors and most strongly to democratic outcome and responsiveness behavior.
Structural equation model, people-changing group.
Figure 2 shows the full structural equation model for the people-processing group. The model fitted the data (CFI = .935, TLI = .916, RMSEA = .035, df = 130, n = 461). The explained variances were 4.8% in terms of output behavior, 6.1% for service outcome, 3.0% for responsiveness, and 15.8% for democratic outcome. For this group, higher PSM was significantly related to democratic outcome (β = .362, p < .001) and service outcome (β = .202, p < .001) behaviors but not to output and responsiveness behaviors. As such, the results only partially support Hypothesis 2 in that the expected relationships between PSM and output and responsiveness types of performance-related behaviors were not found, although the strongest relationships were indeed with democratic outcome and service outcome behaviors.
Structural equation model, people-processing group.
The results of the analysis in the people-changing and people-processing groups should not be compared one-on-one as there was no scalar invariance. Therefore, this study is only able to say that PSM was related to all types of performance-related behaviors in the people-changing group but not in the people-processing group. Possible explanations for these results are discussed in the final section.
Discussion
Unlike the findings of earlier studies, the results of this study indicate that PSM is not always related to all types of self-reported performance-related behaviors. In a context with a people-changing logic, such as schools and hospitals, it appears that high PSM is related to all behavioral types but mostly to democratic outcome and to responsiveness toward users. Considering the primary process, this seems logical: Responding to the needs of the users, responsiveness, can be seen as crucial to delivering the service. Without satisfactory contact, employees will be unable to teach or change the users.
Moreover, if treated unfairly, students, inmates, and patients may be unwilling to cooperate, as their cooperation is needed to deliver good services (Hasenfeld, 1972; Lipsky, 1980). Employees with high PSM also strongly perceive themselves as delivering more high-quality work and giving society good value-for-money than do those with a lower PSM. The context influences the relationship between PSM and performance in such a way that it, first, leads to differences in what performance means—shown by the measurement invariance—and second, that PSM had a role in all dimensions of performance in one context and not in the other.
The finding that PSM was related to all types of performance-related behaviors in the people-changing group corroborates with other studies in such settings, which similarly found that PSM was positively related to performance (Andersen et al., 2014; Bellé, 2013). This study shows that employees with a high PSM in such organizations may perform well because they deliver high-quality work, work efficiently, pay attention to equity, account for what they do, and are responsive to their users. Often, the distinction between in-role and extra-role behavior of employees is made (Kim, 2006; Podsakoff et al., 1997; Williams & Anderson, 1991), but that distinction does not refer to the specific context in which the work is done. This study shows that the perceptions of “in-role” may differ substantially between various types of service providers.
For employees in people-processing service providers (city hall and police), highly significant relationships were found between PSM and democratic outcome (accountability and equity) and service outcome but not with output or responsiveness behaviors. In people-processing organizations, such as city halls, employees may be asked or socialized to refrain from being overresponsive. When redistributing services or regulating, it may be best to keep a distance to ensure you cannot be accused of unfair treatment or processes (Hasenfeld, 1972, 1983; Kjeldsen, 2014). Moreover, PSM was unrelated to output. It may be that new public management has not (yet) made output an important aspect, or that, due to budget cuts highly public service motivated employees were unable to reach high output. This is however not in line with most studies in such a context that have mainly found a positive relationship between PSM and performance.
Although some have stated that the evidence of a relationship between PSM and performance is now overwhelming (Andersen et al., 2014), this study shows that this evidence may be context dependent, and that it might be premature to conclude that PSM is positively related to performance in every setting (see also Petrovsky & Ritz, 2014). Institutional context plays an important role, and this supports those authors who have stated that a fit with the environment is an essential mechanism to consider when studying the relationship between PSM and work outcomes (Bright, 2007; Ryu, 2014; B. E. Wright, 2007).
This study has shown that, when distinguishing between different types of performance-related behaviors (Boyne, 2002), PSM is strongly associated with democratic outcome behaviors. Previous studies have shown that those who have a strong PSM will show more ethical behavior such as whistle-blowing (Brewer & Selden, 1998), and this study underlines those findings. The results also show that some behaviors may be more appropriate in one setting than another (March & Olsen, 1989), and it appears that public service motivated employees respond in ways that are seen as appropriate in their specific institutional context. They are therefore neither runaway agents nor fully controlled by the institutional context but appear to combine their own motivation with what is asked of them by their environment.
Could PSM then be an important way to safeguard public values? In a time when public organizations are pressured to become more businesslike and to focus on efficiency, democratic values are downplayed as a part of the performance of public service providers and less measured (Boyne, 2002; Moynihan et al., 2011). These more general institutional pressures have probably found their way to the employees studied here. The people-changing service providers, who are more on the fringes of the public sector, may have felt strong pressures to become more businesslike. The financial incentives offered to, for instance, hospitals have changed from input to output subsidies, and this may well be visible in the results as a significant relationship between PSM and output was found for the group of employees of whom hospital workers are a significant proportion. However, the results also indicate that institutional pressures do not fully determine what public service motivated employees do. Employees with high PSM in both groups reported higher democratic-outcome-related behavior such as treating citizens fairly and working transparently. PSM may thus be a buffer against reduced attention to these aspects of performance.
This study has, by distinguishing between people-changing and people-processing, only addressed one aspect on which public organizations differ. The found differences may however also be due to variance on other institutional settings and other levels. For instance, variation in resources may influence how responsive employees can be; if they have sufficient resources and time, they may be more responsive. Moreover, variation between jobs in their tasks and client contact may matter. The distinction in organizational logic made here was previously found to be relevant in interviews (Van Loon et al., 2013) and can be said to be a major distinction between organizations that encompasses to a certain degree the mission and aim of an organization and the nature of contact with clients. Nevertheless, future research should dive deeper in whether other variation in context matters and how.
An important limitation of this study is its cross-sectional nature that does not allow one to draw cause-and-effect conclusions. That is, from this study, it cannot be concluded whether behavior influenced PSM or vice versa. However, the theoretically argued relationship is that PSM influences performance-related behavior (Perry & Wise, 1990). Furthermore, as the questions on PSM and performance were asked in the same survey and were thus subjective, the data could be subject to common method and social desirability bias. To limit possible bias, several actions were taken, such as providing full anonymity in completing the survey and separating motivation and performance-related behavior in the survey. Additional tests were also conducted, which indicated that the results were unlikely to be due to bias (Podsakoff & Organ, 1986). Moreover, by using self-reported data, it was possible to study different jobs and organizations (Brewer, 2006). The results here offer one view on performance, that from the employee perspective. Nevertheless, this study is valuable in that it is the first study to show how PSM relates to several types of performance-related behaviors studied simultaneously.
This study was based on data from providers of public services in the Netherlands, and therefore its generalizability to other contexts may be limited (Giauque, Ritz, Varone, Anderfuhren-Biget, & Waldner, 2011). It is quite possible that, in other countries, other institutional logics matter in the relationship between PSM and behavior. For instance, in the Netherlands the public ethos to which employees in municipalities have to ascribe still very much resembles the distant, objective civil servant. Moreover, schools and hospitals are almost all publicly funded but privately organized. These specific country settings may influence the role of PSM in performance. Future research could try to compare countries to address the influence of national context; this study shows how organizational logics may matter. Finally, in terms of the measures, a global measure of PSM was used, and it could be that the various dimensions of PSM are related to behaviors in different ways. However, a global measure is more comparable between groups as these might differ in the type of motive emphasized. Finally, in measuring performance, Boyne’s (2002) five-dimensional view on public service performance was adjusted to the employee level but only four of the dimensions (output, responsiveness, service impact, and democratic outcome) were analyzed.
This research makes two major contributions. First, by distinguishing between several types of performance-related behaviors in testing the relationship with PSM, it shows that, for the people-processing group of employees, there is no significant link between PSM and either output or responsiveness behaviors. Second, this study adds to the knowledge on how the relationship between PSM and performance-related behavior is dependent on the context. PSM’s influence on performance cannot be seen separately from the influence of the institutional context in which employees work.
What does this mean for practice and future research? It seems that organizations are able to steer their public service motivated employees toward certain types of performance through the organizational logic. It appears that public service motivated employees are not “runaway agents” who do whatever they think is best for the society, but at least to an extent behave appropriately and observe the organizational logic (DiIulio, 1994; Perry & Wise, 1990). Employees are driven by many factors, and their behavior is determined not only by their (public service) motivation but also by what is seen as appropriate. As opposed to the structural view on institutions, which focuses on rules and control systems, this article is focused on institutional logics and differentiates between people-changing and people-processing organizations. Vandenabeele (2011) argued that value aspects of institutions may be more important than structural elements. This study shows that structural arrangements are not the only way to steer on performance; it is also possible for employees to follow an institutional logic by internalizing norms and adopting appropriate behaviors (Scott, 2001; Thornton & Ocasio, 2008). Public service providers can communicate such logics through transformational leadership, training, and communicating their mission (Paarlberg & Lavigna, 2010; Vandenabeele, 2014).
This study also leads to new questions that could not be answered with the available data. First, greater insight is warranted into which dimensions of performance-related behaviors are related to PSM. To what dimensions of performance is PSM related in other contexts? Second, in this study employees’ perceptions of their own behavior were used. Other stakeholders such as citizens, inspectors, or supervisors may have different views (Andrews et al., 2011). Future research could usefully investigate whether these findings are repeated if other data sources such as client or supervisor ratings are used.
Third, another important question is how institutional logics matter in the relationship between PSM and performance. Studies including person–environment fit have shown that the congruence between individual values and perceived organizational values has a role in the relationship between PSM and performance (Bright, 2007; Gould-Williams, Mostafa, & Bottomley, 2013). This study suggests that the environment may also matter in a different way, apart from the perceptions of the individual, by determining what public service motivated employees view as appropriate behavior (March & Olsen, 1989). Other distinctions may also be relevant. Kjeldsen (2012), for instance, found that the degree of professionalization mattered, whereas Bellé (2013) found that the degree of prosocial impact mattered. More research on how differences between public organizations influence the relationship between PSM and performance is necessary to gain insight into the context dependency of this relationship. More survey research can show whether the findings presented here hold up in a different context, but interviews may also be useful as they can provide insight into how organizational logics matter and why public service motivated employees in one organization focus or prioritize some dimensions of performance and other dimensions in a different context. Although this research has been able to provide a first step in unraveling the relationship between PSM and different types of performance-related behaviors, there is more to discover.
Footnotes
Appendix
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was financed by the Netherlands Organization for Scientific Research (NWO) Grant 404-10-092.