The Role of Social Network Structure in Street-Level Innovation

Abstract

In considering how peer relationships can aid street-level bureaucrats in doing their jobs, existing literature has emphasized the importance of peers in providing the social and emotional support required to deal with uncertain and stressful working situations. By applying a social network perspective to examine the innovative behavior of a sample of teachers in a large urban high school, this article highlights the importance of an additional factor: the location of a frontline worker’s position in the larger structure of social connections within the organization. In particular, multilevel statistical models reveal a positive association between the extent to which an experienced teacher is located in a network position that bridges across different organizational subgroups and his or her level of innovation, suggesting that experienced frontline workers may benefit from the information diversity that comes from having multiple and diverse social contacts. More generally, the study highlights the value of complementing individual and organizational insights with network-level perspectives for understanding the discretionary behavior of frontline professionals.

Keywords

social networks frontline workers innovation

Cast in ways ranging from “responsible subversion” (Hutchinson, 1990) to “positive deviance” (Carey & Foster, 2011) and “creative insubordination” (Haynes & Licata, 1995), a substantial amount of literature in the public administration field highlights how street-level bureaucrats are capable of introducing significant organizational innovation particularly when attempting to fulfill their missions in the face of a myriad of obstacles and constraints (Borins, 2000). The decision to engage in innovative behaviors or, more generally, to exercise discretion of any sort, does not take place in isolation but instead through multiple types of interactions. This includes interactions with clients that form the basis of judgments of client worthiness (Maynard-Moody & Musheno, 2003) as well as interactions with the formal organizational environment that can shape attitudes and motivations (Oberfield, 2010, 2014). Importantly, a large number of case studies also highlight how frontline workers turn to one another for guidance on how to deal with the stress and uncertainty of street-level working conditions (Foldy & Buckley, 2010; Keiser, 2010; Sandfort, 1999).

Despite the well-acknowledged role of peer interactions in the scholarly discourse on street-level bureaucrats, however, attempts to use the ideas and tools of social network analysis to frame, map, and quantify frontline workplace relationships are surprisingly scarce (Kapucu, Hu, & Khosa, 2014). With the notable exception of a handful of recent studies examining the formation and consequences of street-level bureaucrats’ professional relations (Lee & Kim, 2011; Siciliano, 2013, 2014), very few social network analyses focusing on street-level bureaucrats exist, and none have specifically examined innovative behavior. This is particularly unexpected given the huge interest in using social network analysis to study inter-organizational networks in public administration (Feiock, Lee, & Park, 2012; Henry, Lubell, & McCoy, 2011; Robins, Bates, & Pattison, 2011; for review, see Kapucu, Hu, & Khosa, 2014).

This study addresses this gap in the literature by applying a social network perspective to examine the innovative behavior of a sample of teachers in a large urban high school. If one adopts the common conceptualization of innovation as pertaining to the development or adoption of new ideas and behaviors (Damanpour & Schneider, 2009), then a street-level bureaucrat’s social relations can facilitate or impede innovative behavior in one of two general ways. One way depends on the characteristics of those in his or her network, particularly as they relate to the types and amount of resources that peers are able to provide to support his or her decision making and actions (Siciliano, 2014). For example, a street-level bureaucrat with more knowledgeable or experienced peers may be able to make more informed decisions.

A less-often-mentioned way in which social relations can matter is through network structure, or, more specifically, how the location a street-level bureaucrat occupies in what one might consider the larger constellation of workplace connections within the organization may provide some kind of advantage. For example, given the tendency toward the uniformity of organizational subgroups (Turner & Pratkanis, 1998, for review), a street-level bureaucrat located in a social network position that “bridges” between organizational subgroups would have access to a greater diversity of ideas that can serve as the basis of his or her own experimentation (Burt, 1992). Alternatively, being located within a cohesive subgroup of peers (as opposed to primarily bridging between subgroups) can also influence frontline innovation by providing the support and comfort needed to try new ideas. Distinguishing between these two roles of network structure is not only theoretically but also practically important, as innovation advantages that arise from information diversity and those that arise from social cohesion imply the need for different managerial prescriptions for increasing frontline innovation.

This study addresses the issue by asking the following question: After accounting for individual characteristics, how and to what extent is a teacher’s location in a social network associated with his or her level of innovation? In doing so, this article makes two broad contributions to the existing scholarly literature. First, in considering how peer relationships can aid street-level bureaucrats in doing their jobs, the existing literature has emphasized the importance of peers in providing the social and emotional support required to deal with uncertain and stressful working situations. This article highlights the importance of an additional factor: the location of a frontline worker’s position in the larger structure of social connections within the organization. On average, network locations that bridge across cohesive subgroups within a school are associated with a greater level of innovation for experienced teachers. Second, the social network perspective and related methods have been very usefully applied in public administration to investigate inter-organizational relationships (Bogason & Toonen, 1998; Catlaw, 2007; Feiock et al., 2012; Henry et al., 2011; Rhodes, 1997; Robins et al., 2011; O’Toole, 1997). This article joins other recent articles in public administration (Lee & Kim, 2011; Siciliano, 2013, 2014; Maroulis & Wilensky, 2015) that highlight how the same framework and tools can be applied to gain insight into intra-organizational dynamics.

Street-Level Bureaucrats, Social Networks, and Innovation

Peer Relationships

Although little research directly examines the role of workplace relationships on street-level innovation, several related areas of research on bureaucrats from multiple levels of public organizations can help shed light on the important role peers can play in frontline innovation. At the managerial level, existing scholarship highlights the influence of managerial networks on productivity. For example, Shalala (1998) points to the importance of finding allies for increasing managerial productivity in large public organizations. Similarly, based on a survey of public managers, Berman, West, and Richter (2002) also found that workplace relationships were instrumental in improving communication and helping employees to get “things done.”

Discussions of organizational-level change and transformation also highlight the potential for peer relations to influence a bureaucrat’s ability to innovate. For example, in her work on guerrilla employees in public organizations, O’Leary (2013) writes that every guerilla described in her book “is part of, and used to his or her advantage, an extensive network” (p. 17). Similarly, in their analysis of factors influencing institutional change, de Vries and Balazs (1999) note, “[t]hose who decide to embark on a journey of transformation often seek out people in their environment who can give them the support they need, whether instrumental or emotional” (p. 663).

At the frontline of public organizations, particular attention has been drawn to the role that workplace relationships play in helping street-level bureaucrats to cope with the stress and uncertainty of their working environments (Foldy & Buckley, 2010; Keiser, 2010; Lipsky, 1980; Sandfort, 1999), in no small part because they perceive a disconnect between the policies of higher level decision makers and the realities of their day-to-day challenges (Maynard-Moody & Musheno, 2003). As Maynard-Moody and Musheno (2012) put it, this “mismatch between rules and problems encourages the street-level worker to improvise and innovate or, in the words of one social worker, to ‘be creative’” (p. S19). When engaging in such improvisational and innovative behavior, peer relationships can provide material and emotional resources that are an important prerequisite for innovative action. In particular, researchers have pointed to the importance of peer-level interactions in creating shared belief systems in street-level bureaucratic groups (Sandfort, 1999) as well as the impact of these interactions on the bureaucrats’ personal and administrative decisions in the workplace. As Gofen (2014) states, “Street-level workers . . . turn to their colleagues to inquire how they feel about the implemented policy, what they think about it, and whether they do anything different than what is expected from them . . . ” (p. 487). Keiser (2010) has also pointed out how the behaviors of similar actors under similar circumstances have an important influence on frontline workers’ decision making.

Finally, a small and recent set of studies has begun to analyze workplace relations by applying the ideas and tools of social network analysis (Freeman, 2004; Wasserman & Faust, 1994). Although a rich scholarly tradition of using the ideas and tools of social network analysis to understand contemporary governance has been established in public administration (Bogason & Toonen, 1998; Catlaw, 2007; Rhodes, 1997; O’Toole, 1997), only a handful of public administration studies have applied it to intra-organizational networks (e.g., Lee & Kim, 2011; Siciliano, 2013, 2014; Maroulis & Wilensky, 2015). One contribution of these intra-organizational network studies is that they have highlighted how individual-level outcomes are related to both the composition of a street-level bureaucrat’s network (i.e., the characteristics of those who are in it) as well as his or her location in the larger network structure. For example, Siciliano (2013) examined the network structure and network composition of a sample of schoolteachers in the United States and found that the level of organizational commitment among a teacher’s peers was associated with his or her work performance. Lee and Kim (2011) found a relationship between the affective commitment of employees in two urban governments in Korea and the “centrality” of the employee’s network location. They also found a relationship between affective commitment and an employee’s proximity to “structural holes” in the network. The next section describes how a similar perspective can be used to articulate and test specific hypotheses about street-level innovation.

Network Structure and Street-Level Innovation

From a social network perspective, the social structure of an organization can be conceived as a set of largely non-overlapping cohesive subgroups—small informal groups characterized by frequent interaction and positive sentiment, containing a high concentration of social relations (Blau, 1977; Homans, 1950). This is a long-standing and central image of the structure of a social network (for review, see Freeman, 1992), which, as illustrated in Figure 1, can serve as a useful starting point for operationalizing the potential benefits of an individual’s location in a larger social structure.

Figure 1.

The role of network structure.

Figure 1 depicts a hypothetical network structure consisting of four cohesive subgroups, which, in the case of a school, one can think of as representing small groups of teachers from different subject areas, grades, or workgroups. The nodes in the figure are teachers, and the links are workplace relations. From a structural perspective, the location of a teacher vis-à-vis the school’s cohesive subgroups can provide advantages that lead to more innovative behavior. One advantage is related to the “closure” of the teacher’s network location, a type of benefit from social relations most prominently put forth by the sociologist James Coleman (1988). According to Coleman (1988), network closure emerges when individuals share a significant number of mutual friends or close colleagues. In such a network structure, the behavior of any one individual is likely to be observed—and potentially sanctioned or rewarded—by others (Coleman, 1988, 1990). Moreover, this potential for a third-party sanction or reward fosters an environment of strong norms and trust, where individuals in the group are less likely to harm and, conversely, more likely to assist other members of the group. One particularly relevant benefit of high closure is the facilitation of the type of communication necessary to spread complex and tacit information (Uzzi, 1997), as such knowledge is often associated with the implementation of new ideas within organizations (Obstfeld, 2005).

The benefits of closure can be conceptualized and operationalized at multiple levels. At the subgroup level, the image most strongly associated with strong network closure is that of a densely connected clique, where a large fraction of the potential connections between individuals are realized as actual connections (Frank, 1995). That is, subgroups that exhibit high closure are ones in which the teacher’s close colleagues are likely to be close colleagues themselves. Although, by definition, all cohesive subgroups exhibit a greater density of ties than a random selection of connections does, network density can vary across subgroups. Consequently, teachers in different subgroups stand to benefit more or less with respect to which closure-related advantages might aid their ability to innovate. For example, in Figure 1, members of Subgroup 2 are hypothesized to have greater access to closure-related benefits than do members of the less densely connected Subgroup 1.

Hypothesis 1a: The greater the network density of a teacher’s subgroup, the greater his or her level of innovation.

The benefits of closure can vary across individuals within a subgroup as well. That is, a teacher’s location within a subgroup may also impact his or her access to the resources of the group, with better connected individuals disproportionately benefitting. For example, Teacher B in Figure 1 may be in a better position to reap the benefits of the subgroup than is Teacher C, who has a connection with only one member of the group.

Hypothesis 1b: The more connected a teacher is within his or her subgroup, the greater his or her level of innovation.

An alternative network mechanism that can influence the innovation of frontline workers is Burt’s (1992) notion of structural holes. The same norms that lead to increased trust within dense subgroups can also limit the diversity of information available to a subgroup and reduce the number of opportunities for novel recombination (Burt, 1992, 2004). Consequently, those who bridge “structural holes” between subgroups are hypothesized to have a greater diversity of information and a greater freedom to act. For example, in Figure 1, Teacher A occupies a network location theorized to hold an informational advantage, as her connections span all the other subgroups. Evidence for the association between individuals with networks that are rich in structural holes and individual performance among business professionals is plentiful (Burt, 1992, 2004), and it is easy to envision in a school setting. For example, Teacher A may be a 10th-grade mathematics teacher who learns that a small group of 12th-grade social studies teachers have had great success in improving homework completion rates by creating a call log for parents.

Hypothesis 2: The more a teacher’s network ties span different subgroups, the greater his or her level of innovation.

Previous research also suggests that individual traits, such as conformity (Zhou, Shin, Brass, Choi, & Zhang, 2009) and the uniqueness of one’s job (Burt, 1997), interact with an individual’s network location. In the school context, an individual characteristic of particular salience is a teacher’s level of experience. Although experience has not been specifically investigated with respect to network location, it has been associated with differences in social network characteristics (Kim, 2011) as well as a broad range of beneficial classroom behaviors, including teacher performance (Cohen, 2011; Darling-Hammond, 1999; Murnane & Phillips, 1981; Wayne & Youngs, 2003). The manner in which experience moderates each network mechanism is likely to be different. With respect to network closure, experienced teachers are less likely to benefit from the advantages of being well connected within a subgroup. They have had more time to build confidence in both their own abilities and their statuses in the school setting, and therefore, they are likely in less need of the social support required for risk-taking in the classroom as compared with inexperienced teachers (Cohen, 2011). However, experienced teachers, such as other frontline workers, are more likely to have become accustomed to the standard ways of doing things in the organization (Klein & Knight, 2005). Experienced frontline workers are also further removed from their formal education than are inexperienced teachers, and therefore, they are likely more dependent on their colleagues for information on pedagogical innovations. Therefore, experienced teachers are more likely to benefit from the advantages of bridging structural holes.

Hypothesis 3a: The relationship between the number of within-subgroup connections and innovation will be stronger for inexperienced teachers than for experienced teachers.

Hypothesis 3b: The relationship between the extent to which a teacher’s ties span different subgroups and innovation will be stronger for experienced teachers than for inexperienced teachers.

Summary

If network structure plays a role in street-level innovation above and beyond individual-level characteristics, then we would expect a significant association between one or more of the network variables (network density, within-subgroup connectivity, and spanning subgroups) and innovation. Furthermore, we can look for evidence with respect to the type of advantage related to the teacher’s location within an intra-organizational network structure: To the extent that the increased social cohesion that comes from network closure is a source of advantage, we would expect a positive and significant association between (a) the density of a teacher’s subgroup and his or her level of innovation and (b) the teacher’s greater level of connection within a subgroup and his or her level of innovation. However, to the extent that the diversity of information available to a teacher is a source of advantage, we would expect a positive and significant association between bridging network locations and a teacher’s level of innovation. Moreover, a teacher’s level of experience might limit or amplify the extent to which he or she can capitalize on the potential advantages of his or her network location. If that were the case, we would expect a significant interaction between network location and teacher experience.

Data and Method

The data in this article come from an in-depth investigation of a large public high school engaged in a “small schools” reform initiative. The reform placed teachers and students at the high school into smaller “schools-within-a-school” under the assumption that smaller subunits would promote stronger learning communities. The newly formed subunits were called “small schools” and had identities focused on different themes (e.g., performing arts, math, and science), but each subunit remained in the same building and shared the same principal, schedule, and support staff. An explicit aim of the reform was to increase innovation within the classroom, making the school a fertile context in which to study frontline innovation. In particular, teachers were encouraged to create, find, and implement classroom activities that would personalize the instruction and social experiences of the students.

As part of a larger research effort on the reform, the teachers were asked to complete a paper-based survey that included questions about their attitudes, behaviors, and their social networks. At the time the survey was administered, the formal organizational redesign was almost entirely complete. Six small “schools-within-a-school” had been created over the course of 2 years, and all but six of those teachers had been assigned exclusively to one of those six small schools.¹ Individual-level data gathered from the survey included information on teachers’ experience, conformity, self-efficacy, gender, subject taught, and social ties. Survey responses where obtained from 105 of a possible 125 classroom teachers.² A total of 99 of the 105 surveys contained no missing items. Table 1 provides the summary statistics for each variable.

Table 1.

Descriptive Statistics of Teachers (N = 99).

Variable	M	SD	Minimum	Maximum
Innovation	35.5	34.3	0	136
Conformity	2.10	0.36	1.0	3.0
Self-efficacy	3.88	0.65	2.3	5.0
Inexperienced	0.50	0.50	0	1
Special education	0.16	0.37	0	1
Female	0.62	0.49	0	1
Participation coefficient	0.65	0.21	0	0.90
Within-subgroup degree	0.047	0.95	2.25	2.4

In addition to the individual-level data, social network information was collected using the name-generator/name-interpreter method (Marsden, 1990), where individuals are first asked a question to elicit potential contacts and then subsequently asked questions that characterize or qualify those relationships. More specifically, teachers were asked the following question: “Who are the seven or eight people with whom you have had the most frequent and substantive work contact this school year?” This question was followed by two additional questions, one about the frequency of interaction and another about whether the cited person was inside or outside of the school. The responses to this question were transformed into a symmetric matrix of relations, with each cell in the matrix representing whether or not a relation between two teachers existed. The rows and columns of the network matrix included the survey respondents as well as the school colleagues whom the respondents named. Individuals cited with a frequency of “at least once a week” or more on any of the name-generating questions were included as relations in the final network used for the analysis.

Analysis of the data consisted of two broad sttif. First, to look for evidence of cohesive subgroups in the school, a general-purpose clustering procedure used to identify subgroups, KliqueFinder (Frank, 1995), was applied to the teacher network data. Originally developed to analyze social structures in schools, KliqueFinder has been used to identify subgroups in networks ranging from food webs (Krause, Frank, Mason, Ulanowicz, & Taylor, 2003) to the French financial elite (Frank & Yasumoto, 1998).³ Second, data from the survey were used to estimate multilevel models (Bryk & Radenbush, 2002; Gelman & Hill, 2006)—nesting teachers within cohesive subgroups—that predicted the level of frontline innovation as a function of individual-level attributes and the network variables of interest. All statistical analyses other than the identification of subgroups were carried out in the R statistical computing environment (R Core Team, 2012).

Dependent Variable

Innovation

Innovation captures the number of times in an academic year a teacher created or tried a new activity. It is the mean of the response to two questions asking teachers how often they have done the following: (a) “created a new classroom activity on my own” and (b) “tried an activity in class that I had not tried before.” In constructing the variable, the response categories of “not this year,” “less than once a month,” “once or twice a month,” “about once a week,” and “almost every day” were coded as 0, 4.5, 13.5, 36, and 136 times per year, respectively.⁴ The natural logarithm of innovation is used as the dependent variable in the analysis.

Note that this definition does not capture the degree to which a behavior deviates from the status quo. It does, however, focus on the creation and use of classroom activities, which was an objective of the larger reform at the school. Examples ranged from the introduction or creation of short “bell-ringer” activities that could be used at the beginning of class to settle-down and focus students, to classroom projects that emphasized longer term and interdisciplinary learning. This definition is also consistent with much prior research (Amabile, 1988; Damanpour & Wischnevsky, 2006; Walker, 2008; Zaltman, Duncan, & Holbek, 1973) defining innovation in organizations as “the development (generation) and/or use (adoption) of new ideas or behaviors” (Damanpour & Schneider, 2009, p. 496).

Network Variables

The network closure of a subgroup was captured by the subgroup-level characteristic, subgroup density, and was used to test Hypothesis 1a. At the individual level of the model, a teacher’s network location was characterized in two ways, following Guimera and Amaral’s (2005) operationalization of network roles. The first, within-subgroup degree, captured the extent to which the individual was connected (or isolated) within his or her own subgroup, and it was used to test Hypothesis 1b. The second, participation coefficient, captured the extent to which a teacher’s network ties spanned subgroups, and it was used to test Hypotheses 2. These variables are described in greater detail below.

Subgroup density

Subgroup density is a subgroup-level characteristic that serves as a proxy for the network closure of a subgroup. It is defined as the total number of ties that exist among members of a subgroup divided by the number of all possible ties that could exist among those contacts within the subgroup. For example, Subgroup 2 in Figure 1 has a subgroup density of 0.5 (5 of a possible 10 connections among the members of its group) in comparison with Subgroup 3, which has a subgroup density of 1 (6 of a possible 6 connections).

Within-subgroup degree

Within-subgroup degree is an individual-level variable that captures how well connected an individual was within a subgroup (and, therefore, how much he or she might benefit from closure-related advantages). It is defined relative to the subgroup. It is calculated as the number of ties an individual has to other members in his or her own subgroup minus the average number of ties for individuals in that subgroup (Guimera & Amaral, 2005). For example, in Figure 1, Person C has a lower within-subgroup degree than Person B does. The average number of ties in Subgroup 1 is 2.6, leading to a within-subgroup degree of 0.4 for Person B and −1.6 for Person C.

Participation coefficient

Participation is a measure of how evenly spread an individual’s ties are across subgroups. More precisely, the participation coefficient, P_i, for a network with N subgroups is defined as (Guimera & Amaral, 2005)

P_{i} = 1 - \sum_{s = 1}^{N} {(\frac{κ_{i s}}{κ_{i}})}^{2},

where $κ_{i s}$ is the number of ties individual i has to members in subgroup s, and $κ_{i}$ is the total number of ties for that person regardless of subgroup. A participation coefficient value of 0 means that all of an individual’s ties are to others within the same subgroup; a value of 1 means that the individual’s ties are uniformly distributed across all subgroups.

Note that this approach to capturing an individual’s proximity to structural holes first identifies the subgroups that exist in an organization and then defines network location vis-à-vis that characterization of the organization’s network (Frank et al., 2012; Guimera & Amaral, 2005). In this particular case, such an approach is more consistent with the underlying theory that is being tested versus other more commonly used “egocentric” measures. This is because even though the hypothesized advantages of closing structural holes depend on the existence of subgroups in the larger social structure, egocentric measures tend to be agnostic to the existence of the subgroups in a network. For example, the common egocentric measure of “network constraint” (Burt, 1992) only captures the redundancy and concentration of connections within two sttif of the focal individual (i.e., as far out as “a friend’s friend”). This is a very useful approximation when one does not have a complete network for an organization. When complete network data are available, however, using that data to first identify the subgroups is more consistent with the underlying conception of social structure in the theory (Frank & Yasumoto, 1996).

Control Variables

Conformity

Conformity is broadly defined as “[r]estraint of actions, inclinations, and impulses likely to upset or harm others and violate social expectations or norms” (Schwartz, 1994, p. 22). One might expect an individual’s conformity to influence his or her level of innovation, as trying something new implies a willingness to part ways with the past and the acceptance of the unknown consequences of the new behavior (cf. Antoncic & Hisrich, 2003; Hisrich & Peters, 1992). In addition, in a context where a reform that promotes frontline innovation is underway, an individual’s level of conformity might influence the extent to which he or she participates in reform-related activities. Conformity was measured using a nine-item version of a previously utilized scale that asked the respondents to indicate the level of their agreement or disagreement with particular statements related to their ability to relate to others and to emulate their peers (Mehrabian & Stefl, 1995). It was constructed by taking the mean of the non-missing items after dropping items with factor loadings of less than 0.3. This resulted in using six of the nine items (α = .65).

Self-efficacy

From the earliest empirical investigations of street-level bureaucrats’ decision making (Lipsky, 1980) to the more recent (Zhan, Lo, & Tang, 2013), the importance of frontline workers’ perceived ability to do their jobs well has been repeatedly noted. Unlike in managerial decision making (Bolton, 1993), a low level of job self-efficacy among frontline workers may have a positive association with innovative behavior. Indeed, Lipsky (1980) argues that we would expect the most deviation from the status quo when street-level bureaucrats cannot do their jobs well (for themselves or their clients). Self-efficacy was measured using a four-item scale where respondents indicated their opinions about their ability in the classroom to control disruptive behavior, motivate students, get students to believe they can do well in their schoolwork, and help students to value learning.⁵ Like conformity, it was constructed by taking the mean of all the non-missing items (α = .81).

Inexperienced

Inexperienced is a binary variable indicating whether a teacher had five or less years of experience. This formation is consistent with research indicating that for teachers, most of the benefits of experience have tapered off by year five (Darling-Hammond, 1999; Klitgaard & Hall, 1974). Half of the teachers in the sample had more than 5 years of experience.

The gender of the teacher (female = 1) and whether or not the teacher was designated as a special education teacher (special education = 1) were also included as control variables in the individual level of the model (Level 1). At the subgroup level of the model (Level 2), one might expect the network closure of a group to become more difficult to maintain as the size of the group increases. Therefore, the number of teachers in the subgroup, subgroup size, was included as a control.

Modeling Street-Level Innovation

Street-level innovation was modeled using a multilevel framework (Gelman & Hill, 2006; Raudenbush & Bryk, 2002) that nested teachers within the cohesive subgroups identified by KliqueFinder. More specifically, the innovation level of teacher i in subgroup s was modeled as a function of his or her individual-level attributes and network location:

i n n o v a t i o n_{i s} = β_{0 s} + β_{1 s} c o n f o r m i t y + β_{2 s} s e l f e f f i c a c y + β_{3 s} s p e c i a l e d u c a t i o n + β_{4 s} f e m a l e + β_{5 s} e x p e r i e n c e d + β_{6 s} w i t h i n s u b g r o u p d e g r e e + β_{7 s} p a r t i c i p a t i o n c o e f f i c i e n t + ε_{i s} .

The intercept, $β_{0 s}$ , was further modeled as a function of subgroup-level characteristics, giving the model its multilevel structure:

β_{0 s} = γ_{00} + γ_{01} s u b g r o u p s i z e + γ_{02} s u b g r o u p d e n s i t y + u_{0 s} .

The subgroup-level errors, $u_{0 s}$ , are assumed to be independent and normally distributed with a mean of zero and variance $τ_{00}$ (which is estimated in addition to the non-varying slopes in the model).

Results

Identifying Subgroups

KliqueFinder identified 25 non-overlapping subgroups ranging in size from 4 to 10 teachers, with an average size of 6.96. The results of the algorithm are graphically presented in Figure 2. A circle in Figure 2 represents a subgroup (teachers and ties within the subgroup are not shown), with the size of the circle corresponding to the number of teachers in the subgroup. A line between two subgroups indicates the existence of at least one link between members of each subgroup. The placement of the subgroups corresponds to the subgroups’ placement in “social space.” Subgroups that are close together either share a number of common ties with one another or common ties with similar others.

Figure 2.

Cohesive subgroups in the school.

There are two points to note about the results, which taken together confirm the importance of subgroups in the social network of the school. First, the teachers engaged in interactions within the subgroups displayed in Figure 2 at a rate that was unlikely to have occurred by chance alone. Comparing the actual results with a reference distribution of randomly generated results provided by KliqueFinder indicates that the probability of obtaining the concentration of ties within the subgroup boundaries in Figure 2 was less than one in a hundred. Second, the subgroups seem to be more aligned with small schools than with subject-based departments. More than three quarters of the 25 subgroups were either comprised of members who were all from the same small school, or had a majority of members from the same small school. Beyond the small school alignment, 5 of the 25 subgroups were also mostly comprised of special education teachers, who each predominately, but not exclusively, served a single small school.

Model Estimates

Table 2 presents the results of the analysis. A majority of the variance in innovation was attributable to the individual and not to the subgroup level: The fraction of the total variance that was between subgroups was small, approximately 5% in the estimated models.⁶ All models also included subgroup size and subgroup density as group-level predictors for the subgroup intercept, neither of which was significantly different from zero.

Table 2.

Multilevel Estimates of Frontline Innovation.

	Dependent variable: log(innovation)
	(1)	(2)	(3)	(4)	(5)
Subgroup-level covariates
Subgroup size	0.047	0.047	0.051	0.074	0.088
	(0.076)	(0.076)	(0.076)	(0.073)	(0.071)
Subgroup density	0.669	0.696	0.773	0.711	1.056
	(1.429)	(1.435)	(1.420)	(1.343)	(1.314)
Individual-level covariates
Conformity	−0.228	−0.214	−0.254	−0.256	−0.164
	(0.259)	(0.258)	(0.264)	(0.258)	(0.255)
Self-efficacy	−0.203	−0.198	−0.213	−0.185	−0.198
	(0.154)	(0.153)	(0.155)	(0.153)	(0.149)
Special education	0.355	0.375	0.382	0.364	0.382
	(0.282)	(0.281)	(0.281)	(0.273)	(0.266)
Female	0.183	0.205	0.185	0.183	0.180
	(0.205)	(0.204)	(0.206)	(0.203)	(0.199)
Inexperienced	0.712***	0.725***	0.728***	0.742***	0.795***
	(0.194)	(0.193)	(0.194)	(0.193)	(0.190)
Within-subgroup degree		−0.149	−0.065		−0.155
		(0.099)	(0.148)		(0.099)
Inexperienced × Within-subgroup degree			−0.159
Inexperienced × Within-subgroup degree			(0.209)
Participation coefficient				0.820*	1.940***
				(0.477)	(0.721)
Inexperienced × Participation coefficient					−1.994**
Inexperienced × Participation coefficient					(0.921)
Constant	3.346***	3.281***	3.380***	3.119***	2.751***
	(1.054)	(1.055)	(1.056)	(1.015)	(1.000)
Random effects
No. of individual observations	99	99	99	99	99
Individual-level residual SD	0.93	0.92	0.93	0.93	0.91
No. of subgroups	25	25	25	25	25
Subgroup-level residual SD	0.22	0.23	0.21	0.13	0.12

p < .10. **p < .05. ***p < .01.

Model 1 predicted a teacher’s level of innovation using individual attributes and no network variables. Of the individual attributes included in this baseline model, only the teacher’s experience was a significant predictor: The level of innovation of teachers with five or less years of experience was approximately twice as high (e^0.712) as that for teachers with more than 5 years of experience. Model 2 included a covariate for a teacher’s network location within a subgroup. The coefficient estimate was not significant at the 10% level. There was also little evidence of heterogeneity in the association between experienced and inexperienced teachers (interaction effect, Model 3). Model 4 replaced the within-subgroup location covariate with the participation coefficient, which captured the location of the teacher in relation to other subgroups, or, more specifically, the extent to which a teacher’s ties were evenly spread across all subgroups in the school. The coefficient on the participation coefficient covariate was positive and significant at the 10% level, indicating that occupying a position of “brokerage” in the network was, on average, associated with higher innovation. Interacting the participation coefficient with the experience level both increased the precision of the estimates and revealed an important difference between experienced and inexperienced teachers (Model 5). Figure 2 plots the estimated coefficients for the participation coefficient separately for experienced and inexperienced teachers over the observations used for the estimation (holding all other variables at their mean values). From Figure 3, we can see that the average positive association between the participation coefficient and innovation observed in Model 4 was driven largely by experienced teachers. Although the line for inexperienced teachers is essentially flat, the slope of the line for experienced teachers is substantially positive and statistically significant (p < .01 when inexperienced = 0). To give a sense of the strength of the association, moving from a value on the x axis that is at the 25th percentile of the participation coefficient distribution (0.5) to one at the 75th percentile of the distribution (0.8) corresponds with a 62% increase in innovation for experienced teachers.

Figure 3.

Participation coefficient versus log(innovation), by experience level.

Discussion and Conclusion

In considering how peer relationships can aid street-level bureaucrats in doing their jobs, the existing literature has emphasized the importance of peers in providing the social and emotional support required to deal with uncertain and stressful working situations (Foldy & Buckley, 2010; Keiser, 2010; Sandfort, 1999). By applying a social network perspective to better understand the innovative behavior of street-level bureaucrats, this article highlights the importance of an additional factor: the location of a frontline worker’s position in the larger structure of social connections within the organization. More specifically, the relationship between the extent to which a teacher’s collegial ties are spread across cohesive subgroups (participation coefficient) and his or her level of innovation is positive and statistically significant for experienced teachers.

This finding is consistent with two possible, and not mutually exclusive, explanations. The first is that network location, per se, provided an advantage and that experienced teachers were more sensitive to differences in their network locations. That is, access to new ideas through social interactions was more important for experienced teachers than for inexperienced teachers, who perhaps were already aware of innovations in classroom practice due to their more recently completed education. The second potential explanation reverses the causal story: Experienced frontline workers who were highly innovative were more likely to create collegial ties that spanned different subgroups of an organization. It was not possible to adjudicate between these two explanations given the cross-sectional data available. However, both explanations point to the importance of attending to the social interactions of frontline workers. If the former explanation was primarily responsible for the observed association, then from a managerial perspective, it is important for innovation to encourage and to aid experienced frontline workers in creating connections that go beyond their immediate subunits or departments. That is, emphasis should be placed on exposing frontline workers to a broader array of contacts and experiences within an organization. However, if the latter explanation was the primary cause of the association, then it is important to understand how innovative individuals find colleagues with whom to interact. Therefore, this study is consistent with the recommendation of other scholars who propose that public managers develop both formal and informal approaches for capturing workplace network structures (Lee & Kim, 2011).

In interpreting these findings, it is important to keep in mind two additional characteristics of the sample used in the study. First, the data come from a single school comprising 25 subgroups, making it difficult to reliably detect an association between subgroup-level network closure and innovation. Consequently, one should not over-interpret the insignificant results with respect to the subgroup-level hypotheses about network closure; the results were simply inconclusive. Second, the data come from a single organization and a single type of frontline worker. In particular, two features of the current context bear noting. One is that teachers were being asked to participate in a larger reform that asked them to alter their day-to-day activities. In this sense, this context is similar to many other public organization improvement efforts, such as community policing, that require the heavy involvement of frontline workers who must exercise significant discretion. The other is that the type of information being utilized to aid innovation is easily shared with others (e.g., ideas for classroom activities). In situations where that information is of a more sensitive, private, or tacit nature, such as information-sharing among caseworkers, it is possible that the support and trust stemming from network closure plays a larger role (Uzzi, 1997).

Moving beyond the interpretation and limitations of the specific findings, this study makes an important broader contribution to the existing public administration literature. The social network perspective has been very usefully applied in public administration to study inter-organizational networks (Bogason & Toonen, 1998; Catlaw, 2007; O’Toole, 1997; Rhodes, 1997). However, in a recent review of the public administration network literature, Kapucu, Hu, and Khosa (2014) reported that only 81 studies explicitly used social network analysis methods, and only 14% of those used the individual as the unit of analysis. Recent work in public administration is beginning to illustrate how the same framework that has yielded useful insight into inter-organizational dynamics can be fruitfully applied to investigate and draw inferences about the antecedents and consequences of workplace relationships within public organizations (Lee & Kim, 2011; Siciliano, 2013, 2014). This study contributes to this promising literature by providing a new way of operationalizing a street-level bureaucrat’s location within an organization’s social structure and providing evidence for its association with frontline innovation.

Footnotes

Acknowledgements

I would like to thank Azfar Nisar and Gabel Taggart, doctoral students at Arizona State University (ASU), for valuable research assistance. I am also grateful to Mary Feeney, Yushim Kim, and Justin Stritch for their helpful comments

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) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biography

Spiro Maroulis is an assistant professor at the School of Public Affairs, Arizona State University. His research addresses the theoretical and methodological problems involved with understanding the relationship between individual and collective behavior, with a particular focus on innovation implementation, social networks, and public policy. His work has appeared in Journal of Public Administration Research and Theory, Organization Science, Journal of Policy Analysis and Management, and Science.

References

Amabile

T. M.

(1988). A model of creativity and innovation in organizations. Research in Organizational Behavior, 10, 123-167.

Antoncic

Hisrich

R. D.

(2003). Clarifying the intrapreneurship concept. Journal of Small Business and Enterprise Development, 10, 7-24.

Berman

E. M.

West

J. P.

Richter

M. N.

Jr. (2002). Workplace relations: Friendship patterns and consequences (according to managers). Public Administration Review, 62, 217-230.

Blau

P. M.

(1977). Inequality and heterogeneity. New York, NY: Macmillan.

Bogason

Toonen

T. A.

(1998). Introduction: Networks in public administration. Public Administration, 76, 205-227.

Bolton

M. K.

(1993). Organizational innovation and substandard performance: When is necessity the mother of innovation? Organization Science, 4, 57-75.

Borins

(2000). Loose cannons and rule breakers, or enterprising leaders? Some evidence about innovative public managers. Public Administration Review, 60, 498-507.

Bryk

A. S.

Raudenbush

S. W.

(1992). Hierarchical linear models: Applications and data analysis methods. Newbury Park, CA: Sage.

Burt

R. S.

(1992). Structural holes: The social structure of competition. Cambridge, MA: Harvard University Press.

10.

Burt

R. S.

(1997). The contingent value of social capital. Administrative Science Quarterly, 42, 339-365.

11.

Burt

R. S.

(2004). Structural holes and good ideas. American Journal of Sociology, 110, 349-399.

12.

Carey

Foster

(2011). Introducing “deviant” social work: Contextualising the limits of radical social work whilst understanding (fragmented) resistance within the social work labour process. British Journal of Social Work, 41, 576-593.

13.

Catlaw

T. J.

(2007). Fabricating the people: Politics and administration in the biopolitical state. Tuscaloosa, AL: University of Alabama Press.

14.

Cohen

D. K.

(2011). Teaching and its predicaments. Cambridge, MA: Harvard University Press.

15.

Coleman

J. S.

(1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95-S120.

16.

Coleman

J. S.

(1990). Foundations of social theory. Cambridge, MA: Harvard University Press.

17.

Damanpour

Schneider

(2009). Characteristics of innovation and innovation adoption in public organizations: Assessing the role of managers. Journal of Public Administration Research and Theory, 19, 495-522.

18.

Damanpour

Wischnevsky

J. D.

(2006). Research on innovation in organizations: Distinguishing innovation-generating from innovation-adopting organizations. Journal of Engineering and Technology Management, 23, 269-291.

19.

Darling-Hammond

(1999). Teacher quality and student achievement: A review of state policy evidence. Seattle, WA: Center for the Study of Teaching and Policy, University of Washington.

20.

de Vries

M. F. K.

Balazs

(1999). Transforming the mind-set of the organization: A clinical perspective. Administration & Society, 30, 640-675.

21.

Feiock

R. C.

Lee

I. W.

Park

H. J.

(2012). Administrators’ and elected officials’ collaboration networks: Selecting partners to reduce risk in economic development. Public Administration Review, 72(s1), S58-S68.

22.

Foldy

E. G.

Buckley

T. R.

(2010). Re-creating street-level practice: The role of routines, work groups, and team learning. Journal of Public Administration Research and Theory, 20, 23-52.

23.

Frank

K. A.

(1995). Identifying cohesive subgroups. Social Networks, 17, 27-56.

24.

Frank

K. A.

Chen

Lee

Kalafatis

Chen

Y. J.

Lemos

M. C.

(2012). Network location and policy-oriented behavior: An analysis of two-mode networks of coauthored documents concerning climate change in the Great Lakes region. Policy Studies Journal, 40, 492-515.

25.

Frank

K. A.

Yasumoto

(1996). Embedding subgroups in the sociogram: Linking theory and image. Connections, 19(1), 43-57.

26.

Frank

K. A.

Yasumoto

J. Y.

(1998). Linking action to social structure within a system: Social capital within and between subgroups 1. American Journal of Sociology, 104(3), 642-686.

27.

Freeman

L. C.

(1992). The sociological concept of “group”: An empirical test of two models. American Journal of Sociology, 98, 152-166.

28.

Freeman

(2004). The development of social network analysis: A study in the sociology of science. Vancouver, Canada: Empirical Press.

29.

Gelman

Hill

(2006). Data analysis using regression and multilevel/hierarchical models. Cambridge, UK: Cambridge University Press.

30.

Gofen

(2014). Mind the gap: Dimensions and influence of street-level divergence. Journal of Public Administration Research and Theory, 24, 473-493.

31.

Guimera

Amaral

L. A. N.

(2005). Cartography of complex networks: Modules and universal roles. Journal of Statistical Mechanics: Theory and Experiment, 2005, P02001.

32.

Haynes

E. A.

Licata

J. W.

(1995). Creative insubordination of school principals and the legitimacy of the justifiable. Journal of Educational Administration, 33(4), 21-35.

33.

Henry

A. D.

Lubell

McCoy

(2011). Belief systems and social capital as drivers of policy network structure: The case of California regional planning. Journal of Public Administration Research and Theory, 21(3), 419-444.

34.

Hisrich

R. D.

Peters

M. P.

(1992). Entrepreneurship: Starting, developing, and managing a new enterprise. Homewood: McGraw-Hill/Irwin.

35.

Homans

(1950). The human group. New York, NY: Harcourt Brace.

36.

Hutchinson

S. A.

(1990). Responsible subversion: A study of rule-bending among nurses. Research and Theory for Nursing Practice, 4, 3-17.

37.

Kapucu

Khosa

(2014). The state of network research in public administration. Administration & Society. Advance online publication. doi:10.1177/0095399714555752

38.

Keiser

L. R.

(2010). Understanding street-level bureaucrats’ decision making: Determining eligibility in the Social Security Disability Program. Public Administration Review, 70, 247-257.

39.

Kim

(2011). The effect of teachers’ social networks on teaching practices and class composition (Unpublished doctoral dissertation). Michigan State University, East Lansing.

40.

Klein

K. J.

Knight

A. P.

(2005). Innovation implementation: Overcoming the challenge. Current Directions in Psychological Science, 14, 243-246.

41.

Klitgaard

R. E.

Hall

G. R.

(1974). Are there unusually effective schools? Journal of Human Resources, 10(3), 90-106.

42.

Krause

A. E.

Frank

K. A.

Mason

D. M.

Ulanowicz

R. E.

Taylor

W. W.

(2003). Compartments revealed in food-web structure. Nature, 426, 282-285.

43.

Lee

Kim

(2011). Exploring the role of social networks in affective organizational commitment: Network centrality, strength of ties, and structural holes. The American Review of Public Administration, 41, 205-223.

44.

Lipsky

(1980). Street-level bureaucracy: Dilemmas of the individual in public services. New York, NY: Russell Sage.

45.

Maroulis

Wilensky

(2015). Social and task interdependencies in the street-level implementation of innovation. Journal of Public Administration Research and Theory, 25, 721-750.

46.

Marsden

P. V.

(1990). Network data and measurement. Annual Review of Sociology, 16, 435-463.

47.

Maynard-Moody

S. W.

Musheno

M. C.

(2003). Cops, teachers, counselors: Stories from the front lines of public service. Ann Arbor, MI: University of Michigan Press.

48.

Maynard-Moody

S. W.

Musheno

M. C.

(2012). Social equities and inequities in practice: Street-level workers as agents and pragmatists. Public Administration Review, 72(Suppl. 1), S16-S23.

49.

Mehrabian

Stefl

C. A.

(1995). Basic temperament components of loneliness, shyness, and conformity. Social Behavior & Personality: An International Journal, 23, 253-263.

50.

Murnane

R. J.

Phillips

B. R.

(1981). Learning by doing, vintage, and selection: Three pieces of the puzzle relating teaching experience and teaching performance. Economics of Education Review, 1, 691-693.

51.

Oberfield

Z. W.

(2014). Motivation, change, and stability: Findings from an urban police department. The American Review of Public Administration, 44, 210-232.

52.

Oberfield

Z. W.

(2010). Rule following and discretion at government’s frontlines: Continuity and change during organization socialization. Journal of Public Administration Research and Theory, 20, 735-755.

53.

Obstfeld

(2005). Social networks, the tertius iungens orientation, and involvement in innovation. Administrative Science Quarterly, 50, 100-130.

54.

O’Leary

(2013). The ethics of dissent: Managing guerrilla government. Washington, DC: CQ Press.

55.

O’Toole

L. J.

Jr. (1997). Treating networks seriously: Practical and research-based agendas in public administration. Public Administration Review, 57, 45-52.

56.

R Core Team. (2012). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.R-project.org/

57.

Raudenbush

S. W.

Bryk

A. S.

(2002). Hierarchical linear models: Applications and data analysis methods (Vol. 1). Thousand Oaks, CA: Sage.

58.

Rhodes

R. A.

(1997). Understanding governance: Policy networks, governance, reflexivity and accountability. Buckingham, UK: Open University Press.

59.

Robins

Bates

Pattison

(2011). Network governance and environmental management: conflict and cooperation. Public Administration, 89(4), 1293-1313.

60.

Sandfort

(1999). The structural impediments to human service collaboration: Examining welfare reform at the front lines. Social Service Review, 73, 314-339.

61.

Schwartz

(1994). Are there universal aspects in the content and structure of values. Journal of Social Issues, 50, 19-45.

62.

Shalala

D. E.

(1998). Guest editorial: Are large public organizations manageable? Public Administration Review, 58, 284-289.

63.

Siciliano

M. D.

(2013, June 22). Competition and collaboration in public organizations: The case of urban schools. Public Management Research Conference, Madison, WI.

64.

Siciliano

M. D.

(2014, April 3). Linking network structure to objective performance measures in public organizations: How teacher networks affect student achievement. Midwest Political Science Association, Chicago, IL.

65.

Tschannen-Moran

Hoy

A. W.

(2001). Teacher efficacy: Capturing an elusive construct. Teaching and Teacher Education, 17(7), 783-805.

66.

Turner

M. E.

Pratkanis

A. R.

(1998). Twenty-five years of groupthink theory and research: Lessons from the evaluation of a theory. Organizational Behavior and Human Decision Processes, 73, 105-115.

67.

Uzzi

(1997). Social structure and competition in interfirm networks: The paradox of embeddedness. Administrative Science Quarterly, 42(1), 35-67.

68.

Walker

R. M.

(2008). An empirical evaluation of innovation types and organizational and environmental characteristics: Towards a configuration framework. Journal of Public Administration Research and Theory, 18, 591-615.

69.

Wasserman

Faust

(1994). Social network analysis: Methods and applications. Cambridge, UK: Cambridge University Press.

70.

Wayne

A. J.

Youngs

(2003). Teacher characteristics and student achievement gains: A review. Review of Educational research, 73, 89-122.

71.

Zaltman

Duncan

Holbek

(1973). Innovations and organizations. New York, NY: Wiley.

72.

Zhan

C. W. H.

Tang

S. Y.

(2014). Contextual changes and environmental policy implementation: A longitudinal study of street-level bureaucrats in Guangzhou, China. Journal of Public Administration Research and Theory, 24(4), 1005-1035.

73.

Zhou

Shin

S. J.

Brass

D. J.

Choi

Zhang

Z. X.

(2009). Social networks, personal values, and creativity: Evidence for curvilinear and interaction effects. Journal of Applied Psychology, 94, 1544-1552.