Description
In the broadest sense knowledge environments are social practices, technological and physical arrangements intended to facilitate collaborative knowledge building, decision making, inference or discovery, depending on the epistemological premises and goals.
Research Environment on Innovation, Networks in Examining the Linkages
Abstract: While the performance and productivity benefits of social networks in science have been generally demonstrated, there is still little understanding of how networks might affect scientists' perceptions of their research environment. Because the work environment has also been identified as a key factor for research success, it is essential to understand the interaction between social networks and the organisational research environment. This paper examines the relationship between the social network positions of scientists and the perceptions of their organisation's research environment. The analysis utilises three related datasets that were gathered with a comprehensive survey instrument administered in Spring 2005 to a small, publicly funded research organisation consisting primarily of oceanographers and atmospheric scientists. Results are discussed in terms of their implications for managerial practice and future research. Keywords: social networks; work environment; organisations; innovation; Research and Development; R&D; R&D management.
Introduction
As Jordan (2005) highlights, an important question in the Research and Development (R&D) literature focuses on how organisations can support and encourage high performance. The growing literature on social networks has highlighted the important role played by these social structures in research and R&D performance. For example, a number of studies have examined the important role played by intra-organisational dynamics in R&D performance, including communication networks (Allen, 1970), knowledge flows (Almeida and Kogut, 1999), diversity (Reagans and Zuckerman, 2001), idea innovation chains (Hage and Hollingsworth, 2000), interorganisational networks (Powell et al., 1996) and complexity (Mote, 2005). However, these studies typically look at social networks in isolation, without taking into account the interaction between the network and the organisational environment, particularly how network position might affect scientists' perception of the research environment. Because the work environment has also been identified as a key factor for creativity and innovation (Cummings, 1965; Balachandra and Friar, 1997), it is essential to understand the interaction between social networks and the organisational research environment and how the latter might facilitate or inhibit network behaviour and patterns that are conducive for higher performance. While a great deal of work has been focused on identifying optimal work environments for R&D (Jordan, 2005; Jordan et al., 2003), very little attention has been paid to identifying the optimal network structures. Recently, however, Borgatti (2005) has proposed a tentative framework for understanding the relationship between network structure and types of innovation, although such work is still preliminary. This paper examines the relationship between the social network positions of scientists and their perceptions of the research environment within their respective research organisation. Specifically, this paper utilises data from a research environment survey administered to a small research organisation consisting of approximately 80 physical scientists. The research organisation in the study is a subunit within a large, mission-oriented agency and conducts basic and applied research on the use of satellite data for monitoring meteorological, climatological and oceanographic environmental characteristics. The research survey utilised in this analysis has been administered and tested in a number of R&D settings (Jordan and Streit, 2003; Jordan et al., 2003; Jordan, 2005; Jordan et al., 2005), including similar mission-oriented research organisations. In the survey, the research environment has been characterised as a set of specific organisational attributes previously identified by researchers as important for conducting
248 high-quality and relevant research, including such things as sufficient and stable research funding, internal communication, cross-fertilisation of ideas and career advancement, to name only a few (Jordan, 2005). In addition to questions about the research environment, the survey also collected information about researchers' networks. To gather network data, the survey included a name generator and a question that identified project affiliations. The name generator sought to identify the ego networks of scientists based on frequent internal (within the organisation) and external (outside the organisation) contacts regarding their current research. This dataset does not allow for an exploration of the weak/strong tie distinction, but it does allow a rough glimpse into the diversity of network ties. As Ruef (2002) demonstrates, actors in a nonredundant, diverse set of network ties are more likely to be innovative than actors embedded in a dense network, with the former allowing for greater heterogeneity in role relationships. With regard to the project affiliation data, it is argued that this offers another way of looking at the social structure of an organisation (Mote, 2005). Unlike a straightforward communication network, a project affiliation network looks at how individuals are connected through project affiliations (work ties) and shows the potential for network interaction. As we will illustrate, a project affiliation network offers an approach that seeks to "explore interdependencies between projects and the firms as well as the personal relations, localities and corporate networks on and around which projects are built" (Grabher, 2002, p.246). In general, this paper is an effort to move social network analysis into new directions, principally the analysis of the interrelationship between network structures and perceptions of the research environment. One of the goals of such an effort is to begin to identify optimal network structures for R&D. After an overview of the relevant literature to frame our question, we discuss in greater detail the data and methods utilised in the analysis. We then look at the survey and network data in turn and explore the relationship between social networks and perceptions of the research environment. The paper concludes with a discussion of the results and implications for further research on social networks on research, research productivity and R&D management.
2
Work environment and social networks
As our concern in this paper is the effects of physical proximity, such as the similarity of beliefs and attitudes and the amount of interaction, the analysis undertaken largely resides in the category of social networks and group processes (Borgatti and Foster, 2003). Specifically, the analysis focuses on the interaction between network position and organisational status on perceptions of the work environment. In this section, we briefly discuss two relevant literature for our analysis. First, we provide a summary of research on the reliability and validity of work environment studies. Next, we identify and discuss a handful of studies that have attempted to study the linkage between social network and perceptions of the work environment. Work environment plays an important role in any organisation, and it has been shown that the perception of the work environment, in particular, impacts creativity and innovation in organisations (Amabile et al., 1996). However, there has been less certainty about the measurement of the work environment, particularly at the level of smaller organisational groupings. Typically, the work environment has been measured as the aggregate of individuals' perceptions of the work environment (see Pierce et al., 1989).
249 While it is generally agreed that such measures provide an adequate assessment of the overall work environment (Joyce and Slocum, 1984), there has been less agreement about the use of smaller aggregates at the level of the group or organisational subunit (Young and Parker, 1999). For example, in a study of the success and failure of work teams in an organisation, Gersick (1988) suggested differences in the perception of work organisation from one team to another. Assuming that the perception of the work environment differs depending on one's activities in subgroups, assessing meaningful intraorganisational differences in the work organisation perception is a key to understanding how to promote creativity and innovation in an organisation. Setting aside the question of measurement, a number of studies have demonstrated some linkages between organisational groups and perceptions of the work environment. For instance, in a study on interaction groups and organisational perceptions, Rentsch (1990) found a significant relationship between membership in interaction groups and the meanings individuals attach to organisational events. Similarly, Young and Parker (1999) demonstrated that employee interaction is a key predictor of shared perceptions of the work environment. Krackhardt and Kilduff (2002) show how clusters of individuals develop cultural agreements. Most recently, Lawrence (2006) explored the concept of the organisational reference group, which she defined as the group of people that individuals perceive as central to a person's work environment. Although her data was limited to a single organisation, she demonstrated that the organisational reference group was a better indicator of collective perceptions of the work environment than other formal or informal organisational groupings such as social networks. Despite Lawrence's finding, the importance of social networks in organisational settings is readily acknowledged (Cross and Parker, 2004). But, while the impact of networks on such things as communication and workflow (see, e.g., Allen, 1977; Cross et al., 2001)and performance (see Reagans and Zuckerman, 2001; Mote, 2005) have been amply studied and documented, there has been considerably less work done on the relationship between networks and perceptions of the broader aspects of the work environment. In a review of the literature on intraorganisational networks, Krackhardt and Brass (1994) highlight a handful of studies that found some connection between networks and attitude formation, but nonetheless conclude that the evidence is not clear and a great many questions still remain. For example, Dean and Brass (1985) found that centrality played some role in the convergence of work attitudes, while Galaskiewicz and Burt (1991) found that structural equivalence helped to explain a convergence of attitudes across non-profit organisations. Finally, as Smith-Doerr et al. (2004) concluded in their study of the relationship between the network position of managers and the different meanings they attach to the outcomes in an R&D organisation, it is not clear to what extent network position is a better predictor of convergence of attitude than other factors such as work experience. While the literature still remains somewhat inconclusive, the findings are suggestive that networks and other group dynamics, such as repeated interactions in projects, serve to allow for a convergence of attitudes and perceptions among workers. We would suggest that the two approaches, networks and organisational groups, working in tandem could tease out the relative impact of each. In the next section, we discuss the data and methodology we utilise to pursue this line of investigation.
250
3
Data and methodology
The primary data used in this study comes from a survey administered to scientific researchers at a small research organisation focused on atmospheric science. The survey covers key attributes of organisational structure and management practices within the research environment. The survey was developed through an extensive literature review and input from 15 focus groups that included bench scientists, engineers and technologists, as well as their managers, across various R&D tasks, and it has been fieldtested in a number of research organisations (Jordan and Streit, 2003; Jordan et al., 2003; Jordan, 2005). In total, the survey encompasses 36 attributes in four discrete categories as critical for creating an environment that fosters excellent research. These categories represent different approaches to achieving good performance: Development of human resources, Stimulating creativity and cross-fertilisation, Providing internal support systems, and Setting and achieving relevant goals (see Appendix). All of the approaches are important in R&D organisations, but one or another may be given greater emphasis in a management intervention to encourage high performance. The research organisation studied in this analysis is the Center for Satellite Applications and Research (STAR), a subunit of the National Oceanic and Atmospheric Administration (NOAA) - a large, mission-oriented agency. STAR consists of approximately 80 physical scientists who are organised into three divisions that encompass satellite meteorology, oceanography, climatology and cooperative research with academic institutions. In addition, STAR has a complex physical structure, consisting of one primary office, a nearby secondary office and several smaller offices scattered around the country but typically located within research centres located at major universities. STAR scientists are the primary developers of satellite-derived land, ice, ocean and atmospheric environmental data products that support all of NOAA's mission goals. In addition to actively developing new data products, the scientists also provide support, typically Calibration and Validation (Cal/Val) support, to nearly 400 current satellite-derived products on a routine basis. In addition, the scientists actively work with the numerical weather modelling community within NOAA to support the development of new methods for the assimilation of satellite data. Finally, much of the work of these scientists is conducted in collaboration and partnership with other agencies, academic institutes and industry. The research environment survey was administered to all scientific and technical staff of STAR. Out of 81 potential respondents, 64 completed surveys, yielding a response rate of 79%. Of the 64 respondents, 58 were scientists and 6 were technical staff. With regard to the network questions, we collected the name-generator data from 39 respondents and the project affiliation question from 63 respondents. To protect respondents' privacy, the namegenerator did not require respondents to identify full names of their contacts. As a result, we quantified the data in terms of the number of internal and external contacts. The project affiliation data was gathered by asking them to identify all of the projects that respondents were affiliated with as a project member in a list of projects. The list was initially presented in a 2-mode format, which was then transformed into 1-mode format in the coding process to facilitate the derivation of network measures, principally centrality.
251
4
Project networks and the research environment
Typically, network studies focus on networks derived from advice networks (see Cross et al., 2001), friendship networks (see Zeggelink, 1995), or, more predominantly, communication networks (see Krackhardt and Porter, 1986). However, one type of intraorganisational network that has not been explored in any detail is the project affiliation network. While this represents a novel perspective on organisational networks (Grabher, 2002; Mote, 2005), it is suggested that this type of network captures a highly relevant level of social structure within a project-based or matrix organisation, particularly within the context of a research organisation (Hobday, 2000). While the ties contained in this network do not represent actual communication linkages, these networks suggest the potential for communication and interaction, as well as depicting the level of knowledge complexity, defined as the diversity of competencies and skills, contained within the project (Mote, 2005). Within STAR, it is highly relevant to look at the network of project affiliations. The organisation resembles a matrix-based organisation, where the development of most projects, including the selection of personnel, is often at the initiative of individual researchers who act as project managers. In this manner, the project network captures the fluid nature of cross-functional interaction within the organisation. In this context, however, the usual challenge of boundary specification is exacerbated due to the fact that a great deal of the research is conducted in collaboration with the actors outside of the organisation. Of course, this is a fact of scientific research that has been long recognised in previous network studies such as Zuckerman's (1967) examination of collaboration among Nobel laureates and Crane's (1969) exploration of the invisible college hypothesis, to name only two studies. Despite the multiplex nature of social networks, however, it is argued that the project network nonetheless captures an important intraorganisational social structure within this organisation, one that has been overlooked in intraorganisational network studies. Initially, we developed a number of network visualisations (sociograms) data to explore the interrelationship between network structure, project affiliations and other potential attributes that could lead to organisational clusters. In addition, we conducted Analyses of Variance (ANOVA) using the survey data to determine if there were group differences based on these organisational aspects. Figure 1 represents a multidimensional scaling of the network of connections by project affiliation. As such, the figure allows for visual identification of the structure of social relations among the projects, as well as the key players within this intraorganisational network field. In Figure 1, it is possible to identify one, and possibly two, distinct clusters of scientists by project affiliation. To better characterise this network representation, a number of other visualisations using this basic figure will also be presented which examine specific types of organisational categories. In addition to identifying the network location of scientists, we can use this visualisation to investigate the relationship with physical location. In this manner, we are interested in whether there is a correlation, at least on a visual basis, between physical location and network location. In this respect, STAR represents an interesting organisation to study. As mentioned earlier, STAR scientists are dispersed geographically. In Figure 2, respondents at the main location are indicated in red, the secondary location in yellow and the other locations in blue. As we see, the respondents are broadly distributed within the network structure, although respondents from other
252 locations do exhibit some clustering. The results of an ANOVA found no significant differences in work environment perceptions, which suggests that respondents did not face significantly different work environments despite residing in different physical locations.
Figure 1 The network of connections by project affiliation (for colour see online version)
Figure 2
Physical location of scientists and project network (for colours see online version)
Because managers play a key role in any organisation, it is important to explore where they might be located within the project network. Intuitively, one might assume that managers would be clustered in the middle of the diagram, to coincide with a more central location within the internal project network. However, we actually find that many
253 STAR managers reside on the periphery of the network (see Figure 3). It would be an error to assume that this represents a significant managerial issue without further investigation, as managers on the periphery could play a boundary-spanning role depending on their network ties external to the project network. Table 1 displays the results of an ANOVA highlighting those items with significant differences by managerial status. The results showed a number of sharp differences in the work environment perception between managers and non-managers, such as significantly lower perceptions of time to think and explore, as well as having resources and freedom to pursue new ideas. Utilising both of these analyses, we see some interesting differences in structural position and job characteristics between managers and non-managers. In general, the lower perceptions of managers on these key organisational attributes could be a cause of concern and merit further exploration by senior management.
Figure 3 Managers in project network (for colours see online version)
An extremely important component of the work of STAR is the Cal/Val of satellite instrumentation and data. Understandably, the largest project within the organisation is devoted specifically to this work, although project members do not necessarily devote all of their time to these tasks. Much of this work can be categorised as 'routine work', and a significant body of literature has identified numerous organisational implications associated with this type of work (see, e.g., Hage and Aiken, 1969). As we see in Figure 4, the Cal/Val project members (represented in blue) form a slight cluster in the middle of the project network, indicating that the project might potentially exert a significant amount of influence throughout the organisation if it operates as an organisational reference group (Lawrence, 2006). ANOVA results conducted on the differences in responses to the survey by affiliation with the Cal/Val project are presented in Table 2. Interestingly, the Cal/Val project members showed significantly more positive perceptions regarding the level of qualified staff, adequate salaries and stability of project funding compared to non-Cal/Val project members. Despite the level of routineness associated with Cal/Val, this finding seems to reflect a positive level of organisational support among project members on a number of key organisational attributes.
254
Table 1 ANOVA table for the items with significant differences in perceptions by managerial status Environmental attributes People have time to think creatively and explore There are resources and freedom to pursue new ideas The project-level measures of success that are applied to my project are good Senior management does a good job of allocating internal research funds Senior management champions long-term foundational research NOAA/NESDIS systems and processes work efficiently There are good research competencies and knowledge base in key technical areas My management maintains an integrated and relevant research portfolio Note: * p< .05; ** p< .001; *** p< .0001. Figure 4 The Cal/Val project members in the network (for colours see online version) Managers (n = 12) 1.50 1.75 Not managers (n = 44) 2.84 3.00 Significance 0.001*** 0.002**
2.75 2.60 2.08 2.33
3.49 3.62 3.00 3.08
0.038* 0.004** 0.040* 0.046*
2.83
3.51
0.038*
2.83
3.62
0.022*
NESDIS - National Environmental Satellite, Data, and Information Service.
255
Table 2 ANOVA table for the items with significant differences in perceptions by Cal/Val status Others (n = 40) 2.22 4.00 3.25 4.29 2.98 Cal/Val (n = 16) 2.92 3.25 2.38 3.50 3.69 Significance 0.055§ 0.027* 0.027* 0.005** 0.059§
Environmental attributes Internal project communication between my management and senior management There is an abundance of high quality technical staff There is sufficient, stable project funding NOAA/NESDIS provides good salaries and benefits My management adds value to my work
§
p< .1; * p< .05; ** p< .001.
In this preliminary analysis of network structure, we explored the impact of key organisational attributes - physical location, managerial status and primary work type - on perceptions of the work environment. While each of these showed some interesting results, the impact on perceptions of the work environment was not demonstrative. In the next section, we turn to a more direct examination of network structure and position.
5
Network measures: centrality
In exploring the impact of network structure, we principally examined the role of network centrality on perceptions. The focus on centrality is important, as it is a primary measure of the structural position of actors within a network. In general, centrality details the prominence of actors and the nature of their relation to the rest of the network primarily through calculating the number and distance of ties a network actor has with other members of the network (Scott, 2000). In this manner, the use of centrality measures gives us a more direct indication of the potential flow of knowledge and communication between projects. Although centrality is typically treated as a single type of network property, Freeman (1978, p.238) argues that the three primary measures of centrality - degree, betweenness and closeness - imply 'three competing "theories" of how centrality might affect group processes ? centrality as control, centrality as independence or centrality as activity'. The fourth measure of centrality - eigenvector centrality - can be considered an extension of degree centrality, reflecting that centrality is not simply a matter of one's own network ties, but also the network ties of those to which you are connected (Bonacich, 1987). As we discuss in greater detail below, the measures of centrality offer four different ways of identifying how network structure might affect not only perceptions, but also the flow of information and, potentially, creativity and innovation. Most simply, degree centrality is the number of nodes to which an actor is adjacent, and it offers an idea about the potential communication activity of an actor, that is, the higher the measure, the greater potential for activity within the flow of communication (Freeman, 1978). Typically, actors with high degree centrality are assumed to have a great deal of influence over the network. In contrast, closeness indicates the potential independence of an actor from the flow of communication, in the sense that an actor is not dependent on a single source of information. As Scott indicates, the simplest notion of closeness is calculated from the sum of the geodesic distance to all other points in the
256 graph, and a node is 'close' if it lies at short distance from many other points (Scott, 2000). In this manner, an actor is centrally located but is not dependent on others as 'intermediaries' or 'relayers' of information (Freeman, 1979, p.224) and generally means that these actors have the quickest access to the highest number of other actors in the network. Betweenness is defined as the extent to which a node is 'between' two other modes (Scott, 2000), and it captures the capacity for an actor to play the role of intermediary in the network, connecting two actors that are not otherwise connected. In other words, the betweenness of an actor is a function of paths from project to project and is typically considered a measure of the extent that an actor can control the flow of information, often referred to as 'gatekeepers'. Finally, eigenvector centrality is a variant of degree centrality and is proportional to the sum of centrality of the nodes to which the node is attached (Borgatti and Everett, 1997). In other words, eigenvector centrality captures not only how many actors you 'know', but how many actors they 'know' as well. In this manner, if an actor who is connected to many actors (high degree centrality) who are themselves well-connected (also with high degree centrality), the actor has a high level of eigenvector centrality. Conversely, an actor who is connected only to actors who are less connected (isolates or near isolates) does not have a high level of eigenvector centrality, even if the actor has a high measure of degree centrality. In a sense, eigenvector centrality offers a measure of the diversity of an actor's network. After deriving measures of the centrality for all the actors in the STAR project network, we conducted ANOVAs with each measure to determine if there were any group differences based on network position. For each measure, actors were grouped based on whether they had a high or low measure of centrality. Of the four measures of centrality, closeness was found to have the greatest impact on perceptions. After calculating the measures of closeness for each person in the network, we grouped the respondents into two categories (low closeness and high closeness) using mean closeness. Comparing the two groups of scientists on the 36 items on the research environment survey, we found that there were statistically significant differences on six of the items (see Table 3). On an additional four times, the differences were approaching significance. However, it is interesting to point out that on all of these items, the high closeness group reported a significantly higher mean percent time true. The statistically significant items are listed before the two near significant items, as shown in Table 3. In addition, those in the 'high closeness' category also reported higher ratings on two overall questions regarding the research environment, which were approaching significance. In one respect, this kind of finding should be reassuring to management because it means that those who have more visibility of what is occurring across the organisation are most likely to perceive a higher percent time true on a number of desirable attributes. In addition, these results strongly suggest that network position, in this case closeness, is associated with positive perceptions on a number of key organisational attributes.
257
Table 3 ANOVA table for items with significant differences in perceptions by closeness Low closeness 3.35 3.42 3.18 3.39 3.88 2.82 3.76 3.27 3.06 3.28 4.79 3.76 High closeness 4.17 4.04 3.78 3.96 4.43 3.74 4.43 3.83 3.70 3.83 5.43 4.13 Significance 0.01* 0.02* 0.06§ 0.08§ 0.06§ 0.01** 0.00*** 0.03* 0.03* 0.07§ 0.06§ 0.08§
Environmental attributes People show a commitment to critical thinking There is teamwork and collaboration External collaborations and interactions occur frequently for this project My management rewards and recognises merit People are treated with respect as individuals My management adds value to my work People are given the authority to make decisions about how to do their jobs There is good planning and execution of research projects My management has a clear research vision and strategies My management maintains an integrated and relevant research portfolio Overall, I would rate my research/work environment as... The organisation is a great place to work
§
p< .1; * p< .05; ** p< .001; *** p< .0001.
Because individuals with high closeness are in a relatively better position than others to monitor the information flow in the network and have the best visibility into what is happening in the overall network, we wanted to explore what type of research is pursued by those with high closeness. Initially, we focused on the research goal for each project, simply defined as the orientation of the project towards current products or new product development. The cross-tabulation in Table 4 shows the association between projects based on closeness composition and product orientation. The closeness composition of projects was based on the closeness measures of the individuals involved with the project. If more than 60% of the individuals in a project had low closeness (based on mean closeness), the project was categorised as 'low closeness'. Conversely, if more than 60% of the individuals in a project had high closeness, the project was categorised as 'high closeness'. For projects that had a more or less equal number of individuals with low and high closeness, more or less, the project was categorised as 'mixed closeness'. The product orientation of projects was based on the orientation of the project's goal, either maintaining or improving current products or developing new products.
Table 4 Projects by closeness composition and product orientation Product orientation New product Current product development 5 5 9 7 21 9 33 23
Closeness composition Low closeness Mixed closeness High closeness Total
Total 10 16 30 56
258 As Table 4 illustrates, a large number of projects oriented towards new product development also had a large number of individuals with high closeness. In testing the strength of this association, we used several symmetric measures of association for ordinal by ordinal tables. While closeness composition neatly fits the definition forordinal variable (as it is ranked from low to high), the variable for product orientation is less clear as an ordinal variable (and seems more akin to a nominal variable). However, one could argue that if the interest is in innovation, product orientation is more clearly an ordinal variable, as new product development would be ranked higher for that purpose. Nonetheless, the cross-tabulation showed a slightly significant association between closeness composition and product orientation. We discuss this result in greater detail in the conclusion.
6
Ego networks and the research environment
Consistent research findings in the organisational literature have suggested that internal communication and external communication networks are uniquely different in terms of their effects on performance due to differential access to various resources (Borgatti and Cross, 2003). This finding is supported for R&D contexts as well (Helble and Chong, 2004). Typically, no primacy is given to either internal or external ties; rather, emphasis is generally placed on the value of the diversity of ties. One variant of this argument is Burt's (1992) argument regarding structural holes, where it is not the so much diversity of ties that is important but the strategic placement of ties (or lack thereof). Due to the quality of the responses to the name-generator question, we were unable to fully explore the diversity of ties in terms of other organisational attachments. Rather, we simply quantified the number of self-reported internal and external contacts. In Table 5, we provide a summary of the network ties of STAR scientists. In total, the namegenerator yielded 39 respondents who indicated mean internal contacts of 4.62 and mean external contacts of 3.67.
Table 5 Mean and median of network contacts Internal contacts 4.62 5.00 External contacts 3.67 3.00
Mean Median
To differentiate the types of networks that researchers had, we categorised respondents according to the amount of contacts they reported. Given the considerable upside bias of the mean for external contacts, we determined to use the median as the cutoff point for both types of contacts: for internal contacts, it is 5.00; for external contacts, it is 3.00. As shown in Table 6, internally well-connected scientists are likely to be well-connected externally as well.
Table 6 Combinations of external and internal contacts among scientists Internal contacts External contacts Low Low High High 12 7 19 7 13 20 Total 19 20 39
Total
259 Table 7 indicates where each respondent resides according to the type of contact and physical location. The largest group of respondents was located at headquarters, and scientists at headquarters show a mixture of low and high connectedness in internal and external networks. On the other hand, approximately three quarters of scientists in nonheadquarter locations showed equal or higher levels of contacts outside the organisation compared to within the organisation. This is nearly double of the percentage of scientists at headquarters with a high level of external contacts.
Table 7 Physical location and ego network composition (percent in parentheses) Low internal/low external Headquarter Metro DC area Other location Total
1
High internal/low external 7 (26%) 0 (0%) 0 (0%) 7 (18%)
High external/low internal 3 (11%) 1 (25%) 3 (38%) 7 (18%)
High external/high internal 8 (30%) 2 (50%) 3 (38%) 13 (33%)
Total 27 (100%) 4 (100%) 8 (101%)1 39 (100%)
9 (33%) 1 (25%) 2 (25%) 12 (31%)
The total percentage exceeds 100 due to rounding numbers at the first decimal place.
Table 8 displays a cross-tabulation of funding sources and network contacts, and it is clear that the level of internal and external contacts are strongly related to funding sources of the projects in which individual scientists are involved. Approximately 57% of scientists with low levels of internal and external contacts work on projects directly funded by STAR, hence there is little organisational imperative for external contacts. In contrast, those scientists with a large amount of funding from sources outside the immediate organisational context of STAR exhibited a higher number of external contacts.
Table 8 Funding composition and ego network composition (in percent) Low internal/low external (n = 12) 57.1% 15.0% 4.2% 7.5% 12.1% 5.0% 0.0% High internal/low external (n = 7) 34.3% 15.9% 13.3% 22.0% 0.1% 6.6% 0.0% High external/low internal (n = 7) 30.7% 49.3% 10.0% 2.9% 0.0% 7.1% 0.0% High external/high internal (n = 12) 20.0% 49.6% 8.1% 14.2% 0.0% 8.1% 0.0%
Funding source Internal source From other division sources From other parent sources From NASA From other Federal agencies From other sources Do not know
It is clear that physical location and funding source help to explain the type of network contacts, but is the composition of ego network ties related to perceptions of the research environment? To explore this question, differences by category of network contact were examined with an ANOVA using survey responses. However, the analysis of ego
260 network composition did not show any significant, or even substantive, differences on perception of the work environment. Despite the lack of significant results with ego network composition and perceptions of the work environment, we explored the interrelationship between network position, principally closeness, and ego network composition. As Table 9 illustrates, the individuals with high closeness tended to have a lower number of external contacts than individuals with low closeness. Indeed, the mean number of external contacts was 4.3 and 2.9 for individuals with low closeness and high closeness, respectively. Such a result raises a number of intriguing questions. For instance, to what extent is high (internal) closeness related to a limited number of external ties? Does the number of lower external ties among those with high closeness represent a more strategic placement of structural holes? Unfortunately, the quality of the ego network data is such that a more in-depth examination of these results is limited.
Table 9 Closeness and ego network composition Ego network composition Closeness composition Low closeness High closeness Total Low internal/lo w external 5 7 12 High internal/lo w external 4 2 6 High external/low internal 4 3 7 High external/high internal 9 4 13
Total 22 16 38
7
Concluding discussion
As our analysis suggests, network position and type of network contact are related to perceptions of the research environment in a way that is different from other types of organisational clustering such as managerial status or project affiliation. In this respect, the combination of network analysis with an organisational survey, such as the research environment survey, offers a path for better identifying optimal intraorganisational network structures. While network theory suggests that particular network structures may be better suited for particular organisations and types of work, studies that have provided some empirical verification tend to be limited in scope. Although marked by some shortcomings, the analysis offers some interesting avenues for continued research on networks in R&D, particularly with regard to the topic of organisational learning. For instance, in the industrial organisation literature, as distinct from the scientific research literature, learning has usually been measured by an improvement in performance, most typically productivity. If one accepts this measure, then the larger amount of involvement of projects with higher degrees of closeness suggests more improvement for a new kind of performance, namely, product development and new ideas. This is a performance of much more interest in science and technology and should be of more interest to sociologists who want to avoid the productivity trap laid by economists. In addition, the relationship between closeness, new product development and ego network composition suggests that organisational learning might be a combination of matching an appropriate intraorganisational network structure with an appropriate
261 interorganisational network structure. While several prominent studies have identified that boundary-spanning networks increase organisational learning in the biotechnology industry (Liebeskind et al., 1996; Powell et al., 1996), it is not clear what kind of impact this might have on the intraorganisational network. Specifically, who are the boundary spanners within the organisation, how many external network contacts do they have and where do these individuals fit within the intraorganisational network? - these questions are further complicated by some of the issues involved with organisational learning such as the diversity of labour needed for innovation (Hage, 1999), the 'ecologies of learning' discussed by Levitt and March (1988) and the choice of innovation to be pursued (Borgatti, 2005), to name only a few. While intraorganisational networks may have less importance to the small organisations in the Powell and Liebeskind studies, they are extremely important in the context of larger, more bureaucratic research organisations. The analysis pursued in this paper represents only a very early exploration, and the analysis raises more questions than it answers. Further, the analysis has a number of limitations. First, the project affiliation network is a novel approach, and we recognise the validity of the network derived in terms of actual patterns of interaction has not been established. Second, the differences found between network position and organisation clustering might represent the impact of multiplexity, and these differences, rather than being viewed as negative, might be integral to the operation of various types of networks. Finally, we are mindful of the issue of causality between network characteristics and the work environment perception. Nonetheless, our data showed significant associations arose between these groupings and perceptions. However, we need to continue to explore these relationships and conduct additional analyses such as an exploration of cliques and other network groups and follow-up interviews with project members, in order to have more conclusive findings of the causal relationship between network structure, the research environment and innovation in R&D.
Acknowledgements
The authors gratefully acknowledge the support of the National Oceanic and Atmospheric Administration (NOAA), Marie Colton and Al Powell. This research has been performed under contract with NOAA. The opinions expressed are those of the authors and not NOAA. Previous versions of this paper were presented at the 2006 Annual Meeting of the American Sociological Association and the 2006 International Sunbelt Social Network Conference.
References
Allen, T.J. (1970) 'Communication networks in R&D laboratories', R&D Management, Vol. 1, pp.14-21. Allen, T.J. (1977) 'Managing the Flow of Technology', MIT Press, Cambridge, MA. Almeida, P. and Kogut, B. (1999) 'Localization of knowledge and the mobility of engineers in regional networks', Management Science, Vol. 45, pp.905-917. Amabile, T., Conti, R., Coon, H., Lazenby, J. and Herron, M. (1996) 'Assessing the work environment for creativity', Academy of Management Journal, Vol. 39, pp.1154-1184.
262
Balachandra, R. and Friar, J.H. (1997) 'Factors for success in R&D projects and new product development', IEEE Transactions on Engineering Management, Vol. 44, pp.276-287. Bonacich, P. (1987) 'Power and centrality: a family of measures', American Journal of Sociology, Vol. 92, pp.1170-1182. Borgatti, S.P. (2005) Creating knowledge: network structure and innovation. Available online at: http://www.socialnetworkanalysis.com/knowledge_creation.htm (accessed on 3 July 2007). Borgatti, S.P. and Everett, M. (1997) 'Network analysis of 2-mode data', Social Networks, Vol. 19, pp.243-269. Borgatti, S.P. and Cross, R. (2003) 'A relational view of information seeking and learning in social networks', Management Science, Vol. 49, pp.432-445. Borgatti, S.P. and Foster, P.C. (2003) 'The network paradigm in organizational research: a review and typology', Journal of Management, Vol. 29, pp.991-1013. Burt, R.S. (1992) Structural Holes: The Social Structure of Competition, Harvard University Press, Cambrige. Crane, D. (1969) 'Social structure in a group of scientists: a test of the "Invisible college" Hypothesis', American Sociological Review, Vol. 34, pp.335-352. Cross, R., Borgatti, S. and Parker, A. (2001) 'Beyond answers: dimensions of the advice network', Social Networks, Vol. 23, pp.215-235. Cross, R. and Parker, A. (2004) 'The Hidden Power of Social Networks: Understanding How Work Really Gets Done in Organizations', Harvard University Press, Boston, MA. Cummings, L.L. (1965) 'Organizational climates for creativity', Academy of Management Journal, Vol. 8, pp.220-227. Dean, J.W. and Brass, D.J. (1985) 'Social interaction and the perception of job characteristics in an organization', Human Relations, Vol. 38, pp.571-582. Freeman, L.C. (1979) 'Centrality in social networks conceptual clarification', Social Networks, Vol. 1, pp.215-239. Galaskiewicz, J. and Burt, R.S. (1991) 'Interorganization contagion in corporate philanthropy', Administrative Science Quarterly, Vol. 36, pp.88-105. Gersick, C.J.G. (1988) 'Time and transition in work teams: toward a new model of group development', Academy of Management Journal, Vol. 31, pp.9-41. Grabher, G. (2002) 'The project ecology of advertising: tasks, talents and teams', Regional Studies, Vol. 36, pp.245-262. Hage, J. (1999) 'Organizational innovation and organizational change', Annual Review of Sociology, Vol. 25, pp.597-622. Hage, J. and Aiken, M. (1969) 'Routine technology, social structure, and organization goals', Administrative Science Quarterly, Vol. 14, pp.366-376. Hage, J. and Hollingsworth, J.R. (2000) 'A strategy for the analysis of idea innovation networks and institutions', Organization Studies, Vol. 21, pp.971-1004. Helble, Y. and Chong, L.C. (2004) 'The importance of internal and external R&D network linkages for R&D organizations: evidence from Singapore', R&D Management, Vol. 34, pp.605-612. Hobday, M. (2000) 'The project-based organisation: an ideal form for managing complex products and systems?', Research Policy, Vol. 29, pp.871-893. Jordan, G.B. (2005) 'What matters to R&D workers', Research Technology Management, Vol. 48, pp.23-32. Jordan, G.B., Hage, J., Mote, J. and Hepler, B. (2005) 'Investigating differences among research projects and implications for managers', R&D Management, Vol. 35, pp.501-511. Jordan, G.B. and Streit, L.D. (2003) 'Recognizing the competing values in science and technology organizations: implications for evaluation', in Shapira, P. and Kuhlmann, S. (Eds): Learning from Science and Technology Policy Evaluations, Edward Elgar, Northhampton, MA.
263
Jordan, G.B., Streit, L.D. and Binkley, J.S. (2003) 'Assessing and improving the effectiveness of national research laboratories', IEEE Transactions on Engineering Management, Vol. 50, pp.228-235. Joyce, W.F. and Slocum, J.W. (1984) 'Collective climate: agreement as a basis for defining aggregate climates in organizations', Academy of Management Journal, Vol. 27, pp.721-742. Krackhardt, D. and Brass, D.J. (1994) 'Intra-organizational networks: the micro side', in Wasserman, S. and Galaskiewicz, J. (Eds) Advance in Social Network Analysis: Research in the Social and Behavioral Sciences, Sage, Thousand Oaks, CA. Krackhardt, D. and Kilduff, M. (2002) 'Structure, culture and simmelian ties in entrepreneurial firms', Social Networks, Vol. 24, pp.279-290. Krackhardt, D. and Porter, L.W. (1986) 'The snowball effect: turnover embedded in communication networks', Journal of Applied Psychology, Vol. 71, pp.50-55. Lawrence, B.S. (2006) 'Organizational reference groups: a missing perspective on social context', Organization Science, Vol. 17, pp.80-100. Levitt, B. and March, J.G. (1988) 'Organizational learning', Annual Review of Sociology, Vol. 14, pp.319-340. Liebeskind, J.P., Oliver, A.L., Zucker, L. and Brewer, M. (1996) 'Social networks, learning, and flexibility: sourcing scientific knowledge in new biotechnology firms', Organization Science, Vol. 7, pp.428-443. Mote, J.E. (2005) 'R&D ecology: using 2-mode network analysis to explore complexity in R&D environments', Journal of Engineering & Technology Management, Vol. 22, pp.93-111. Pierce, J.L., Gardner, D.G., Cummings, L.L. and Dunham, R.B. (1989) 'Organization-based selfesteem: construct definition, measurement and validation', Academy of Management Journal, Vol. 32, pp.622-648. Powell, W.W., Koput, K.W. and Smith-Doerr, L. (1996) 'Interorganizational collaboration and the locus of innovation: networks of learning in biotechnology', Administrative Science Quarterly, Vol. 41, pp.116-145. Reagans, R. and Zuckerman, E.W. (2001) 'Networks, diversity, and productivity: the social capital of corporate R&D teams', Organization Science, Vol. 12, pp.502-517. Rentsch, J.R. (1990) 'Climate and culture: interaction and qualitative differences in meanings', Journal of Applied Psychology, Vol. 75, pp.668-681. Ruef, M. (2002) 'Strong ties, weak ties and islands: structural and cultural predictors of organizational innovation', Industrial & Corporate Change, Vol. 11, pp.427-449. Scott, J. (2000) Social Network Analysis: A Handbook, Sage, Thousand Oaks, CA. Smith-Doerr, L., Manev, I.M. and Rizova, P. (2004) 'The meaning of success: network position and the social construction of project outcomes in an R&D lab', Journal of Engineering &Technology Management, Vol. 21, pp.51-81. Young, S.A. and Parker, C.P. (1999) 'Predicting collective climates: assessing the role of shared work values, needs, employee', Journal of Organizational Behavior, Vol. 20, pp.11991218. Zeggelink, E. (1995) 'Evolving friendship networks: an individual-oriented approach implementing similarity', Social Networks, Vol. 17, pp.83-110. Zuckerman, H.H. (1967) 'Nobel laureates in science: patterns of productivity, collaboration, and authorship', American Sociological Review, Vol. 32, pp.391-403.
264
Appendix: 36 key attributes in the work environment survey (4 categories)
Development of human resources People treated with respect Optimal mix of staff Creativity and crossfertilisation Time to think and explore Resources/freedom to pursue new ideas Autonomy to make decisions Cross-fertilisation of ideas Frequent external collaborations Relevant research portfolio Commitment to critical thinking Identification of new opportunities Sense of challenge and enthusiasm Internal support systems Good research competencies Good equipment/physic al environment Good salaries and benefits Good allocation of internal funds Informed and decisive management Rewards and recognises merit Efficient laboratory systems Laboratory services meet needs Overhead rates not burdensome Set and achieve relevant goals Sufficient, stable project funding Good planning and execution of projects Good project-level measures of success Good relationship with sponsors Reputation for excellence
Management integrity Teamwork and collaboration Good internal project communication Management adds value to work High-quality technical staff Good professional development Good career advancement opportunities
Management champions foundational research Good lab-wide measures of success Clear research vision and strategy Invests in future capabilities
doc_365867054.docx
In the broadest sense knowledge environments are social practices, technological and physical arrangements intended to facilitate collaborative knowledge building, decision making, inference or discovery, depending on the epistemological premises and goals.
Research Environment on Innovation, Networks in Examining the Linkages
Abstract: While the performance and productivity benefits of social networks in science have been generally demonstrated, there is still little understanding of how networks might affect scientists' perceptions of their research environment. Because the work environment has also been identified as a key factor for research success, it is essential to understand the interaction between social networks and the organisational research environment. This paper examines the relationship between the social network positions of scientists and the perceptions of their organisation's research environment. The analysis utilises three related datasets that were gathered with a comprehensive survey instrument administered in Spring 2005 to a small, publicly funded research organisation consisting primarily of oceanographers and atmospheric scientists. Results are discussed in terms of their implications for managerial practice and future research. Keywords: social networks; work environment; organisations; innovation; Research and Development; R&D; R&D management.
Introduction
As Jordan (2005) highlights, an important question in the Research and Development (R&D) literature focuses on how organisations can support and encourage high performance. The growing literature on social networks has highlighted the important role played by these social structures in research and R&D performance. For example, a number of studies have examined the important role played by intra-organisational dynamics in R&D performance, including communication networks (Allen, 1970), knowledge flows (Almeida and Kogut, 1999), diversity (Reagans and Zuckerman, 2001), idea innovation chains (Hage and Hollingsworth, 2000), interorganisational networks (Powell et al., 1996) and complexity (Mote, 2005). However, these studies typically look at social networks in isolation, without taking into account the interaction between the network and the organisational environment, particularly how network position might affect scientists' perception of the research environment. Because the work environment has also been identified as a key factor for creativity and innovation (Cummings, 1965; Balachandra and Friar, 1997), it is essential to understand the interaction between social networks and the organisational research environment and how the latter might facilitate or inhibit network behaviour and patterns that are conducive for higher performance. While a great deal of work has been focused on identifying optimal work environments for R&D (Jordan, 2005; Jordan et al., 2003), very little attention has been paid to identifying the optimal network structures. Recently, however, Borgatti (2005) has proposed a tentative framework for understanding the relationship between network structure and types of innovation, although such work is still preliminary. This paper examines the relationship between the social network positions of scientists and their perceptions of the research environment within their respective research organisation. Specifically, this paper utilises data from a research environment survey administered to a small research organisation consisting of approximately 80 physical scientists. The research organisation in the study is a subunit within a large, mission-oriented agency and conducts basic and applied research on the use of satellite data for monitoring meteorological, climatological and oceanographic environmental characteristics. The research survey utilised in this analysis has been administered and tested in a number of R&D settings (Jordan and Streit, 2003; Jordan et al., 2003; Jordan, 2005; Jordan et al., 2005), including similar mission-oriented research organisations. In the survey, the research environment has been characterised as a set of specific organisational attributes previously identified by researchers as important for conducting
248 high-quality and relevant research, including such things as sufficient and stable research funding, internal communication, cross-fertilisation of ideas and career advancement, to name only a few (Jordan, 2005). In addition to questions about the research environment, the survey also collected information about researchers' networks. To gather network data, the survey included a name generator and a question that identified project affiliations. The name generator sought to identify the ego networks of scientists based on frequent internal (within the organisation) and external (outside the organisation) contacts regarding their current research. This dataset does not allow for an exploration of the weak/strong tie distinction, but it does allow a rough glimpse into the diversity of network ties. As Ruef (2002) demonstrates, actors in a nonredundant, diverse set of network ties are more likely to be innovative than actors embedded in a dense network, with the former allowing for greater heterogeneity in role relationships. With regard to the project affiliation data, it is argued that this offers another way of looking at the social structure of an organisation (Mote, 2005). Unlike a straightforward communication network, a project affiliation network looks at how individuals are connected through project affiliations (work ties) and shows the potential for network interaction. As we will illustrate, a project affiliation network offers an approach that seeks to "explore interdependencies between projects and the firms as well as the personal relations, localities and corporate networks on and around which projects are built" (Grabher, 2002, p.246). In general, this paper is an effort to move social network analysis into new directions, principally the analysis of the interrelationship between network structures and perceptions of the research environment. One of the goals of such an effort is to begin to identify optimal network structures for R&D. After an overview of the relevant literature to frame our question, we discuss in greater detail the data and methods utilised in the analysis. We then look at the survey and network data in turn and explore the relationship between social networks and perceptions of the research environment. The paper concludes with a discussion of the results and implications for further research on social networks on research, research productivity and R&D management.
2
Work environment and social networks
As our concern in this paper is the effects of physical proximity, such as the similarity of beliefs and attitudes and the amount of interaction, the analysis undertaken largely resides in the category of social networks and group processes (Borgatti and Foster, 2003). Specifically, the analysis focuses on the interaction between network position and organisational status on perceptions of the work environment. In this section, we briefly discuss two relevant literature for our analysis. First, we provide a summary of research on the reliability and validity of work environment studies. Next, we identify and discuss a handful of studies that have attempted to study the linkage between social network and perceptions of the work environment. Work environment plays an important role in any organisation, and it has been shown that the perception of the work environment, in particular, impacts creativity and innovation in organisations (Amabile et al., 1996). However, there has been less certainty about the measurement of the work environment, particularly at the level of smaller organisational groupings. Typically, the work environment has been measured as the aggregate of individuals' perceptions of the work environment (see Pierce et al., 1989).
249 While it is generally agreed that such measures provide an adequate assessment of the overall work environment (Joyce and Slocum, 1984), there has been less agreement about the use of smaller aggregates at the level of the group or organisational subunit (Young and Parker, 1999). For example, in a study of the success and failure of work teams in an organisation, Gersick (1988) suggested differences in the perception of work organisation from one team to another. Assuming that the perception of the work environment differs depending on one's activities in subgroups, assessing meaningful intraorganisational differences in the work organisation perception is a key to understanding how to promote creativity and innovation in an organisation. Setting aside the question of measurement, a number of studies have demonstrated some linkages between organisational groups and perceptions of the work environment. For instance, in a study on interaction groups and organisational perceptions, Rentsch (1990) found a significant relationship between membership in interaction groups and the meanings individuals attach to organisational events. Similarly, Young and Parker (1999) demonstrated that employee interaction is a key predictor of shared perceptions of the work environment. Krackhardt and Kilduff (2002) show how clusters of individuals develop cultural agreements. Most recently, Lawrence (2006) explored the concept of the organisational reference group, which she defined as the group of people that individuals perceive as central to a person's work environment. Although her data was limited to a single organisation, she demonstrated that the organisational reference group was a better indicator of collective perceptions of the work environment than other formal or informal organisational groupings such as social networks. Despite Lawrence's finding, the importance of social networks in organisational settings is readily acknowledged (Cross and Parker, 2004). But, while the impact of networks on such things as communication and workflow (see, e.g., Allen, 1977; Cross et al., 2001)and performance (see Reagans and Zuckerman, 2001; Mote, 2005) have been amply studied and documented, there has been considerably less work done on the relationship between networks and perceptions of the broader aspects of the work environment. In a review of the literature on intraorganisational networks, Krackhardt and Brass (1994) highlight a handful of studies that found some connection between networks and attitude formation, but nonetheless conclude that the evidence is not clear and a great many questions still remain. For example, Dean and Brass (1985) found that centrality played some role in the convergence of work attitudes, while Galaskiewicz and Burt (1991) found that structural equivalence helped to explain a convergence of attitudes across non-profit organisations. Finally, as Smith-Doerr et al. (2004) concluded in their study of the relationship between the network position of managers and the different meanings they attach to the outcomes in an R&D organisation, it is not clear to what extent network position is a better predictor of convergence of attitude than other factors such as work experience. While the literature still remains somewhat inconclusive, the findings are suggestive that networks and other group dynamics, such as repeated interactions in projects, serve to allow for a convergence of attitudes and perceptions among workers. We would suggest that the two approaches, networks and organisational groups, working in tandem could tease out the relative impact of each. In the next section, we discuss the data and methodology we utilise to pursue this line of investigation.
250
3
Data and methodology
The primary data used in this study comes from a survey administered to scientific researchers at a small research organisation focused on atmospheric science. The survey covers key attributes of organisational structure and management practices within the research environment. The survey was developed through an extensive literature review and input from 15 focus groups that included bench scientists, engineers and technologists, as well as their managers, across various R&D tasks, and it has been fieldtested in a number of research organisations (Jordan and Streit, 2003; Jordan et al., 2003; Jordan, 2005). In total, the survey encompasses 36 attributes in four discrete categories as critical for creating an environment that fosters excellent research. These categories represent different approaches to achieving good performance: Development of human resources, Stimulating creativity and cross-fertilisation, Providing internal support systems, and Setting and achieving relevant goals (see Appendix). All of the approaches are important in R&D organisations, but one or another may be given greater emphasis in a management intervention to encourage high performance. The research organisation studied in this analysis is the Center for Satellite Applications and Research (STAR), a subunit of the National Oceanic and Atmospheric Administration (NOAA) - a large, mission-oriented agency. STAR consists of approximately 80 physical scientists who are organised into three divisions that encompass satellite meteorology, oceanography, climatology and cooperative research with academic institutions. In addition, STAR has a complex physical structure, consisting of one primary office, a nearby secondary office and several smaller offices scattered around the country but typically located within research centres located at major universities. STAR scientists are the primary developers of satellite-derived land, ice, ocean and atmospheric environmental data products that support all of NOAA's mission goals. In addition to actively developing new data products, the scientists also provide support, typically Calibration and Validation (Cal/Val) support, to nearly 400 current satellite-derived products on a routine basis. In addition, the scientists actively work with the numerical weather modelling community within NOAA to support the development of new methods for the assimilation of satellite data. Finally, much of the work of these scientists is conducted in collaboration and partnership with other agencies, academic institutes and industry. The research environment survey was administered to all scientific and technical staff of STAR. Out of 81 potential respondents, 64 completed surveys, yielding a response rate of 79%. Of the 64 respondents, 58 were scientists and 6 were technical staff. With regard to the network questions, we collected the name-generator data from 39 respondents and the project affiliation question from 63 respondents. To protect respondents' privacy, the namegenerator did not require respondents to identify full names of their contacts. As a result, we quantified the data in terms of the number of internal and external contacts. The project affiliation data was gathered by asking them to identify all of the projects that respondents were affiliated with as a project member in a list of projects. The list was initially presented in a 2-mode format, which was then transformed into 1-mode format in the coding process to facilitate the derivation of network measures, principally centrality.
251
4
Project networks and the research environment
Typically, network studies focus on networks derived from advice networks (see Cross et al., 2001), friendship networks (see Zeggelink, 1995), or, more predominantly, communication networks (see Krackhardt and Porter, 1986). However, one type of intraorganisational network that has not been explored in any detail is the project affiliation network. While this represents a novel perspective on organisational networks (Grabher, 2002; Mote, 2005), it is suggested that this type of network captures a highly relevant level of social structure within a project-based or matrix organisation, particularly within the context of a research organisation (Hobday, 2000). While the ties contained in this network do not represent actual communication linkages, these networks suggest the potential for communication and interaction, as well as depicting the level of knowledge complexity, defined as the diversity of competencies and skills, contained within the project (Mote, 2005). Within STAR, it is highly relevant to look at the network of project affiliations. The organisation resembles a matrix-based organisation, where the development of most projects, including the selection of personnel, is often at the initiative of individual researchers who act as project managers. In this manner, the project network captures the fluid nature of cross-functional interaction within the organisation. In this context, however, the usual challenge of boundary specification is exacerbated due to the fact that a great deal of the research is conducted in collaboration with the actors outside of the organisation. Of course, this is a fact of scientific research that has been long recognised in previous network studies such as Zuckerman's (1967) examination of collaboration among Nobel laureates and Crane's (1969) exploration of the invisible college hypothesis, to name only two studies. Despite the multiplex nature of social networks, however, it is argued that the project network nonetheless captures an important intraorganisational social structure within this organisation, one that has been overlooked in intraorganisational network studies. Initially, we developed a number of network visualisations (sociograms) data to explore the interrelationship between network structure, project affiliations and other potential attributes that could lead to organisational clusters. In addition, we conducted Analyses of Variance (ANOVA) using the survey data to determine if there were group differences based on these organisational aspects. Figure 1 represents a multidimensional scaling of the network of connections by project affiliation. As such, the figure allows for visual identification of the structure of social relations among the projects, as well as the key players within this intraorganisational network field. In Figure 1, it is possible to identify one, and possibly two, distinct clusters of scientists by project affiliation. To better characterise this network representation, a number of other visualisations using this basic figure will also be presented which examine specific types of organisational categories. In addition to identifying the network location of scientists, we can use this visualisation to investigate the relationship with physical location. In this manner, we are interested in whether there is a correlation, at least on a visual basis, between physical location and network location. In this respect, STAR represents an interesting organisation to study. As mentioned earlier, STAR scientists are dispersed geographically. In Figure 2, respondents at the main location are indicated in red, the secondary location in yellow and the other locations in blue. As we see, the respondents are broadly distributed within the network structure, although respondents from other
252 locations do exhibit some clustering. The results of an ANOVA found no significant differences in work environment perceptions, which suggests that respondents did not face significantly different work environments despite residing in different physical locations.
Figure 1 The network of connections by project affiliation (for colour see online version)
Figure 2
Physical location of scientists and project network (for colours see online version)
Because managers play a key role in any organisation, it is important to explore where they might be located within the project network. Intuitively, one might assume that managers would be clustered in the middle of the diagram, to coincide with a more central location within the internal project network. However, we actually find that many
253 STAR managers reside on the periphery of the network (see Figure 3). It would be an error to assume that this represents a significant managerial issue without further investigation, as managers on the periphery could play a boundary-spanning role depending on their network ties external to the project network. Table 1 displays the results of an ANOVA highlighting those items with significant differences by managerial status. The results showed a number of sharp differences in the work environment perception between managers and non-managers, such as significantly lower perceptions of time to think and explore, as well as having resources and freedom to pursue new ideas. Utilising both of these analyses, we see some interesting differences in structural position and job characteristics between managers and non-managers. In general, the lower perceptions of managers on these key organisational attributes could be a cause of concern and merit further exploration by senior management.
Figure 3 Managers in project network (for colours see online version)
An extremely important component of the work of STAR is the Cal/Val of satellite instrumentation and data. Understandably, the largest project within the organisation is devoted specifically to this work, although project members do not necessarily devote all of their time to these tasks. Much of this work can be categorised as 'routine work', and a significant body of literature has identified numerous organisational implications associated with this type of work (see, e.g., Hage and Aiken, 1969). As we see in Figure 4, the Cal/Val project members (represented in blue) form a slight cluster in the middle of the project network, indicating that the project might potentially exert a significant amount of influence throughout the organisation if it operates as an organisational reference group (Lawrence, 2006). ANOVA results conducted on the differences in responses to the survey by affiliation with the Cal/Val project are presented in Table 2. Interestingly, the Cal/Val project members showed significantly more positive perceptions regarding the level of qualified staff, adequate salaries and stability of project funding compared to non-Cal/Val project members. Despite the level of routineness associated with Cal/Val, this finding seems to reflect a positive level of organisational support among project members on a number of key organisational attributes.
254
Table 1 ANOVA table for the items with significant differences in perceptions by managerial status Environmental attributes People have time to think creatively and explore There are resources and freedom to pursue new ideas The project-level measures of success that are applied to my project are good Senior management does a good job of allocating internal research funds Senior management champions long-term foundational research NOAA/NESDIS systems and processes work efficiently There are good research competencies and knowledge base in key technical areas My management maintains an integrated and relevant research portfolio Note: * p< .05; ** p< .001; *** p< .0001. Figure 4 The Cal/Val project members in the network (for colours see online version) Managers (n = 12) 1.50 1.75 Not managers (n = 44) 2.84 3.00 Significance 0.001*** 0.002**
2.75 2.60 2.08 2.33
3.49 3.62 3.00 3.08
0.038* 0.004** 0.040* 0.046*
2.83
3.51
0.038*
2.83
3.62
0.022*
NESDIS - National Environmental Satellite, Data, and Information Service.
255
Table 2 ANOVA table for the items with significant differences in perceptions by Cal/Val status Others (n = 40) 2.22 4.00 3.25 4.29 2.98 Cal/Val (n = 16) 2.92 3.25 2.38 3.50 3.69 Significance 0.055§ 0.027* 0.027* 0.005** 0.059§
Environmental attributes Internal project communication between my management and senior management There is an abundance of high quality technical staff There is sufficient, stable project funding NOAA/NESDIS provides good salaries and benefits My management adds value to my work
§
p< .1; * p< .05; ** p< .001.
In this preliminary analysis of network structure, we explored the impact of key organisational attributes - physical location, managerial status and primary work type - on perceptions of the work environment. While each of these showed some interesting results, the impact on perceptions of the work environment was not demonstrative. In the next section, we turn to a more direct examination of network structure and position.
5
Network measures: centrality
In exploring the impact of network structure, we principally examined the role of network centrality on perceptions. The focus on centrality is important, as it is a primary measure of the structural position of actors within a network. In general, centrality details the prominence of actors and the nature of their relation to the rest of the network primarily through calculating the number and distance of ties a network actor has with other members of the network (Scott, 2000). In this manner, the use of centrality measures gives us a more direct indication of the potential flow of knowledge and communication between projects. Although centrality is typically treated as a single type of network property, Freeman (1978, p.238) argues that the three primary measures of centrality - degree, betweenness and closeness - imply 'three competing "theories" of how centrality might affect group processes ? centrality as control, centrality as independence or centrality as activity'. The fourth measure of centrality - eigenvector centrality - can be considered an extension of degree centrality, reflecting that centrality is not simply a matter of one's own network ties, but also the network ties of those to which you are connected (Bonacich, 1987). As we discuss in greater detail below, the measures of centrality offer four different ways of identifying how network structure might affect not only perceptions, but also the flow of information and, potentially, creativity and innovation. Most simply, degree centrality is the number of nodes to which an actor is adjacent, and it offers an idea about the potential communication activity of an actor, that is, the higher the measure, the greater potential for activity within the flow of communication (Freeman, 1978). Typically, actors with high degree centrality are assumed to have a great deal of influence over the network. In contrast, closeness indicates the potential independence of an actor from the flow of communication, in the sense that an actor is not dependent on a single source of information. As Scott indicates, the simplest notion of closeness is calculated from the sum of the geodesic distance to all other points in the
256 graph, and a node is 'close' if it lies at short distance from many other points (Scott, 2000). In this manner, an actor is centrally located but is not dependent on others as 'intermediaries' or 'relayers' of information (Freeman, 1979, p.224) and generally means that these actors have the quickest access to the highest number of other actors in the network. Betweenness is defined as the extent to which a node is 'between' two other modes (Scott, 2000), and it captures the capacity for an actor to play the role of intermediary in the network, connecting two actors that are not otherwise connected. In other words, the betweenness of an actor is a function of paths from project to project and is typically considered a measure of the extent that an actor can control the flow of information, often referred to as 'gatekeepers'. Finally, eigenvector centrality is a variant of degree centrality and is proportional to the sum of centrality of the nodes to which the node is attached (Borgatti and Everett, 1997). In other words, eigenvector centrality captures not only how many actors you 'know', but how many actors they 'know' as well. In this manner, if an actor who is connected to many actors (high degree centrality) who are themselves well-connected (also with high degree centrality), the actor has a high level of eigenvector centrality. Conversely, an actor who is connected only to actors who are less connected (isolates or near isolates) does not have a high level of eigenvector centrality, even if the actor has a high measure of degree centrality. In a sense, eigenvector centrality offers a measure of the diversity of an actor's network. After deriving measures of the centrality for all the actors in the STAR project network, we conducted ANOVAs with each measure to determine if there were any group differences based on network position. For each measure, actors were grouped based on whether they had a high or low measure of centrality. Of the four measures of centrality, closeness was found to have the greatest impact on perceptions. After calculating the measures of closeness for each person in the network, we grouped the respondents into two categories (low closeness and high closeness) using mean closeness. Comparing the two groups of scientists on the 36 items on the research environment survey, we found that there were statistically significant differences on six of the items (see Table 3). On an additional four times, the differences were approaching significance. However, it is interesting to point out that on all of these items, the high closeness group reported a significantly higher mean percent time true. The statistically significant items are listed before the two near significant items, as shown in Table 3. In addition, those in the 'high closeness' category also reported higher ratings on two overall questions regarding the research environment, which were approaching significance. In one respect, this kind of finding should be reassuring to management because it means that those who have more visibility of what is occurring across the organisation are most likely to perceive a higher percent time true on a number of desirable attributes. In addition, these results strongly suggest that network position, in this case closeness, is associated with positive perceptions on a number of key organisational attributes.
257
Table 3 ANOVA table for items with significant differences in perceptions by closeness Low closeness 3.35 3.42 3.18 3.39 3.88 2.82 3.76 3.27 3.06 3.28 4.79 3.76 High closeness 4.17 4.04 3.78 3.96 4.43 3.74 4.43 3.83 3.70 3.83 5.43 4.13 Significance 0.01* 0.02* 0.06§ 0.08§ 0.06§ 0.01** 0.00*** 0.03* 0.03* 0.07§ 0.06§ 0.08§
Environmental attributes People show a commitment to critical thinking There is teamwork and collaboration External collaborations and interactions occur frequently for this project My management rewards and recognises merit People are treated with respect as individuals My management adds value to my work People are given the authority to make decisions about how to do their jobs There is good planning and execution of research projects My management has a clear research vision and strategies My management maintains an integrated and relevant research portfolio Overall, I would rate my research/work environment as... The organisation is a great place to work
§
p< .1; * p< .05; ** p< .001; *** p< .0001.
Because individuals with high closeness are in a relatively better position than others to monitor the information flow in the network and have the best visibility into what is happening in the overall network, we wanted to explore what type of research is pursued by those with high closeness. Initially, we focused on the research goal for each project, simply defined as the orientation of the project towards current products or new product development. The cross-tabulation in Table 4 shows the association between projects based on closeness composition and product orientation. The closeness composition of projects was based on the closeness measures of the individuals involved with the project. If more than 60% of the individuals in a project had low closeness (based on mean closeness), the project was categorised as 'low closeness'. Conversely, if more than 60% of the individuals in a project had high closeness, the project was categorised as 'high closeness'. For projects that had a more or less equal number of individuals with low and high closeness, more or less, the project was categorised as 'mixed closeness'. The product orientation of projects was based on the orientation of the project's goal, either maintaining or improving current products or developing new products.
Table 4 Projects by closeness composition and product orientation Product orientation New product Current product development 5 5 9 7 21 9 33 23
Closeness composition Low closeness Mixed closeness High closeness Total
Total 10 16 30 56
258 As Table 4 illustrates, a large number of projects oriented towards new product development also had a large number of individuals with high closeness. In testing the strength of this association, we used several symmetric measures of association for ordinal by ordinal tables. While closeness composition neatly fits the definition forordinal variable (as it is ranked from low to high), the variable for product orientation is less clear as an ordinal variable (and seems more akin to a nominal variable). However, one could argue that if the interest is in innovation, product orientation is more clearly an ordinal variable, as new product development would be ranked higher for that purpose. Nonetheless, the cross-tabulation showed a slightly significant association between closeness composition and product orientation. We discuss this result in greater detail in the conclusion.
6
Ego networks and the research environment
Consistent research findings in the organisational literature have suggested that internal communication and external communication networks are uniquely different in terms of their effects on performance due to differential access to various resources (Borgatti and Cross, 2003). This finding is supported for R&D contexts as well (Helble and Chong, 2004). Typically, no primacy is given to either internal or external ties; rather, emphasis is generally placed on the value of the diversity of ties. One variant of this argument is Burt's (1992) argument regarding structural holes, where it is not the so much diversity of ties that is important but the strategic placement of ties (or lack thereof). Due to the quality of the responses to the name-generator question, we were unable to fully explore the diversity of ties in terms of other organisational attachments. Rather, we simply quantified the number of self-reported internal and external contacts. In Table 5, we provide a summary of the network ties of STAR scientists. In total, the namegenerator yielded 39 respondents who indicated mean internal contacts of 4.62 and mean external contacts of 3.67.
Table 5 Mean and median of network contacts Internal contacts 4.62 5.00 External contacts 3.67 3.00
Mean Median
To differentiate the types of networks that researchers had, we categorised respondents according to the amount of contacts they reported. Given the considerable upside bias of the mean for external contacts, we determined to use the median as the cutoff point for both types of contacts: for internal contacts, it is 5.00; for external contacts, it is 3.00. As shown in Table 6, internally well-connected scientists are likely to be well-connected externally as well.
Table 6 Combinations of external and internal contacts among scientists Internal contacts External contacts Low Low High High 12 7 19 7 13 20 Total 19 20 39
Total
259 Table 7 indicates where each respondent resides according to the type of contact and physical location. The largest group of respondents was located at headquarters, and scientists at headquarters show a mixture of low and high connectedness in internal and external networks. On the other hand, approximately three quarters of scientists in nonheadquarter locations showed equal or higher levels of contacts outside the organisation compared to within the organisation. This is nearly double of the percentage of scientists at headquarters with a high level of external contacts.
Table 7 Physical location and ego network composition (percent in parentheses) Low internal/low external Headquarter Metro DC area Other location Total
1
High internal/low external 7 (26%) 0 (0%) 0 (0%) 7 (18%)
High external/low internal 3 (11%) 1 (25%) 3 (38%) 7 (18%)
High external/high internal 8 (30%) 2 (50%) 3 (38%) 13 (33%)
Total 27 (100%) 4 (100%) 8 (101%)1 39 (100%)
9 (33%) 1 (25%) 2 (25%) 12 (31%)
The total percentage exceeds 100 due to rounding numbers at the first decimal place.
Table 8 displays a cross-tabulation of funding sources and network contacts, and it is clear that the level of internal and external contacts are strongly related to funding sources of the projects in which individual scientists are involved. Approximately 57% of scientists with low levels of internal and external contacts work on projects directly funded by STAR, hence there is little organisational imperative for external contacts. In contrast, those scientists with a large amount of funding from sources outside the immediate organisational context of STAR exhibited a higher number of external contacts.
Table 8 Funding composition and ego network composition (in percent) Low internal/low external (n = 12) 57.1% 15.0% 4.2% 7.5% 12.1% 5.0% 0.0% High internal/low external (n = 7) 34.3% 15.9% 13.3% 22.0% 0.1% 6.6% 0.0% High external/low internal (n = 7) 30.7% 49.3% 10.0% 2.9% 0.0% 7.1% 0.0% High external/high internal (n = 12) 20.0% 49.6% 8.1% 14.2% 0.0% 8.1% 0.0%
Funding source Internal source From other division sources From other parent sources From NASA From other Federal agencies From other sources Do not know
It is clear that physical location and funding source help to explain the type of network contacts, but is the composition of ego network ties related to perceptions of the research environment? To explore this question, differences by category of network contact were examined with an ANOVA using survey responses. However, the analysis of ego
260 network composition did not show any significant, or even substantive, differences on perception of the work environment. Despite the lack of significant results with ego network composition and perceptions of the work environment, we explored the interrelationship between network position, principally closeness, and ego network composition. As Table 9 illustrates, the individuals with high closeness tended to have a lower number of external contacts than individuals with low closeness. Indeed, the mean number of external contacts was 4.3 and 2.9 for individuals with low closeness and high closeness, respectively. Such a result raises a number of intriguing questions. For instance, to what extent is high (internal) closeness related to a limited number of external ties? Does the number of lower external ties among those with high closeness represent a more strategic placement of structural holes? Unfortunately, the quality of the ego network data is such that a more in-depth examination of these results is limited.
Table 9 Closeness and ego network composition Ego network composition Closeness composition Low closeness High closeness Total Low internal/lo w external 5 7 12 High internal/lo w external 4 2 6 High external/low internal 4 3 7 High external/high internal 9 4 13
Total 22 16 38
7
Concluding discussion
As our analysis suggests, network position and type of network contact are related to perceptions of the research environment in a way that is different from other types of organisational clustering such as managerial status or project affiliation. In this respect, the combination of network analysis with an organisational survey, such as the research environment survey, offers a path for better identifying optimal intraorganisational network structures. While network theory suggests that particular network structures may be better suited for particular organisations and types of work, studies that have provided some empirical verification tend to be limited in scope. Although marked by some shortcomings, the analysis offers some interesting avenues for continued research on networks in R&D, particularly with regard to the topic of organisational learning. For instance, in the industrial organisation literature, as distinct from the scientific research literature, learning has usually been measured by an improvement in performance, most typically productivity. If one accepts this measure, then the larger amount of involvement of projects with higher degrees of closeness suggests more improvement for a new kind of performance, namely, product development and new ideas. This is a performance of much more interest in science and technology and should be of more interest to sociologists who want to avoid the productivity trap laid by economists. In addition, the relationship between closeness, new product development and ego network composition suggests that organisational learning might be a combination of matching an appropriate intraorganisational network structure with an appropriate
261 interorganisational network structure. While several prominent studies have identified that boundary-spanning networks increase organisational learning in the biotechnology industry (Liebeskind et al., 1996; Powell et al., 1996), it is not clear what kind of impact this might have on the intraorganisational network. Specifically, who are the boundary spanners within the organisation, how many external network contacts do they have and where do these individuals fit within the intraorganisational network? - these questions are further complicated by some of the issues involved with organisational learning such as the diversity of labour needed for innovation (Hage, 1999), the 'ecologies of learning' discussed by Levitt and March (1988) and the choice of innovation to be pursued (Borgatti, 2005), to name only a few. While intraorganisational networks may have less importance to the small organisations in the Powell and Liebeskind studies, they are extremely important in the context of larger, more bureaucratic research organisations. The analysis pursued in this paper represents only a very early exploration, and the analysis raises more questions than it answers. Further, the analysis has a number of limitations. First, the project affiliation network is a novel approach, and we recognise the validity of the network derived in terms of actual patterns of interaction has not been established. Second, the differences found between network position and organisation clustering might represent the impact of multiplexity, and these differences, rather than being viewed as negative, might be integral to the operation of various types of networks. Finally, we are mindful of the issue of causality between network characteristics and the work environment perception. Nonetheless, our data showed significant associations arose between these groupings and perceptions. However, we need to continue to explore these relationships and conduct additional analyses such as an exploration of cliques and other network groups and follow-up interviews with project members, in order to have more conclusive findings of the causal relationship between network structure, the research environment and innovation in R&D.
Acknowledgements
The authors gratefully acknowledge the support of the National Oceanic and Atmospheric Administration (NOAA), Marie Colton and Al Powell. This research has been performed under contract with NOAA. The opinions expressed are those of the authors and not NOAA. Previous versions of this paper were presented at the 2006 Annual Meeting of the American Sociological Association and the 2006 International Sunbelt Social Network Conference.
References
Allen, T.J. (1970) 'Communication networks in R&D laboratories', R&D Management, Vol. 1, pp.14-21. Allen, T.J. (1977) 'Managing the Flow of Technology', MIT Press, Cambridge, MA. Almeida, P. and Kogut, B. (1999) 'Localization of knowledge and the mobility of engineers in regional networks', Management Science, Vol. 45, pp.905-917. Amabile, T., Conti, R., Coon, H., Lazenby, J. and Herron, M. (1996) 'Assessing the work environment for creativity', Academy of Management Journal, Vol. 39, pp.1154-1184.
262
Balachandra, R. and Friar, J.H. (1997) 'Factors for success in R&D projects and new product development', IEEE Transactions on Engineering Management, Vol. 44, pp.276-287. Bonacich, P. (1987) 'Power and centrality: a family of measures', American Journal of Sociology, Vol. 92, pp.1170-1182. Borgatti, S.P. (2005) Creating knowledge: network structure and innovation. Available online at: http://www.socialnetworkanalysis.com/knowledge_creation.htm (accessed on 3 July 2007). Borgatti, S.P. and Everett, M. (1997) 'Network analysis of 2-mode data', Social Networks, Vol. 19, pp.243-269. Borgatti, S.P. and Cross, R. (2003) 'A relational view of information seeking and learning in social networks', Management Science, Vol. 49, pp.432-445. Borgatti, S.P. and Foster, P.C. (2003) 'The network paradigm in organizational research: a review and typology', Journal of Management, Vol. 29, pp.991-1013. Burt, R.S. (1992) Structural Holes: The Social Structure of Competition, Harvard University Press, Cambrige. Crane, D. (1969) 'Social structure in a group of scientists: a test of the "Invisible college" Hypothesis', American Sociological Review, Vol. 34, pp.335-352. Cross, R., Borgatti, S. and Parker, A. (2001) 'Beyond answers: dimensions of the advice network', Social Networks, Vol. 23, pp.215-235. Cross, R. and Parker, A. (2004) 'The Hidden Power of Social Networks: Understanding How Work Really Gets Done in Organizations', Harvard University Press, Boston, MA. Cummings, L.L. (1965) 'Organizational climates for creativity', Academy of Management Journal, Vol. 8, pp.220-227. Dean, J.W. and Brass, D.J. (1985) 'Social interaction and the perception of job characteristics in an organization', Human Relations, Vol. 38, pp.571-582. Freeman, L.C. (1979) 'Centrality in social networks conceptual clarification', Social Networks, Vol. 1, pp.215-239. Galaskiewicz, J. and Burt, R.S. (1991) 'Interorganization contagion in corporate philanthropy', Administrative Science Quarterly, Vol. 36, pp.88-105. Gersick, C.J.G. (1988) 'Time and transition in work teams: toward a new model of group development', Academy of Management Journal, Vol. 31, pp.9-41. Grabher, G. (2002) 'The project ecology of advertising: tasks, talents and teams', Regional Studies, Vol. 36, pp.245-262. Hage, J. (1999) 'Organizational innovation and organizational change', Annual Review of Sociology, Vol. 25, pp.597-622. Hage, J. and Aiken, M. (1969) 'Routine technology, social structure, and organization goals', Administrative Science Quarterly, Vol. 14, pp.366-376. Hage, J. and Hollingsworth, J.R. (2000) 'A strategy for the analysis of idea innovation networks and institutions', Organization Studies, Vol. 21, pp.971-1004. Helble, Y. and Chong, L.C. (2004) 'The importance of internal and external R&D network linkages for R&D organizations: evidence from Singapore', R&D Management, Vol. 34, pp.605-612. Hobday, M. (2000) 'The project-based organisation: an ideal form for managing complex products and systems?', Research Policy, Vol. 29, pp.871-893. Jordan, G.B. (2005) 'What matters to R&D workers', Research Technology Management, Vol. 48, pp.23-32. Jordan, G.B., Hage, J., Mote, J. and Hepler, B. (2005) 'Investigating differences among research projects and implications for managers', R&D Management, Vol. 35, pp.501-511. Jordan, G.B. and Streit, L.D. (2003) 'Recognizing the competing values in science and technology organizations: implications for evaluation', in Shapira, P. and Kuhlmann, S. (Eds): Learning from Science and Technology Policy Evaluations, Edward Elgar, Northhampton, MA.
263
Jordan, G.B., Streit, L.D. and Binkley, J.S. (2003) 'Assessing and improving the effectiveness of national research laboratories', IEEE Transactions on Engineering Management, Vol. 50, pp.228-235. Joyce, W.F. and Slocum, J.W. (1984) 'Collective climate: agreement as a basis for defining aggregate climates in organizations', Academy of Management Journal, Vol. 27, pp.721-742. Krackhardt, D. and Brass, D.J. (1994) 'Intra-organizational networks: the micro side', in Wasserman, S. and Galaskiewicz, J. (Eds) Advance in Social Network Analysis: Research in the Social and Behavioral Sciences, Sage, Thousand Oaks, CA. Krackhardt, D. and Kilduff, M. (2002) 'Structure, culture and simmelian ties in entrepreneurial firms', Social Networks, Vol. 24, pp.279-290. Krackhardt, D. and Porter, L.W. (1986) 'The snowball effect: turnover embedded in communication networks', Journal of Applied Psychology, Vol. 71, pp.50-55. Lawrence, B.S. (2006) 'Organizational reference groups: a missing perspective on social context', Organization Science, Vol. 17, pp.80-100. Levitt, B. and March, J.G. (1988) 'Organizational learning', Annual Review of Sociology, Vol. 14, pp.319-340. Liebeskind, J.P., Oliver, A.L., Zucker, L. and Brewer, M. (1996) 'Social networks, learning, and flexibility: sourcing scientific knowledge in new biotechnology firms', Organization Science, Vol. 7, pp.428-443. Mote, J.E. (2005) 'R&D ecology: using 2-mode network analysis to explore complexity in R&D environments', Journal of Engineering & Technology Management, Vol. 22, pp.93-111. Pierce, J.L., Gardner, D.G., Cummings, L.L. and Dunham, R.B. (1989) 'Organization-based selfesteem: construct definition, measurement and validation', Academy of Management Journal, Vol. 32, pp.622-648. Powell, W.W., Koput, K.W. and Smith-Doerr, L. (1996) 'Interorganizational collaboration and the locus of innovation: networks of learning in biotechnology', Administrative Science Quarterly, Vol. 41, pp.116-145. Reagans, R. and Zuckerman, E.W. (2001) 'Networks, diversity, and productivity: the social capital of corporate R&D teams', Organization Science, Vol. 12, pp.502-517. Rentsch, J.R. (1990) 'Climate and culture: interaction and qualitative differences in meanings', Journal of Applied Psychology, Vol. 75, pp.668-681. Ruef, M. (2002) 'Strong ties, weak ties and islands: structural and cultural predictors of organizational innovation', Industrial & Corporate Change, Vol. 11, pp.427-449. Scott, J. (2000) Social Network Analysis: A Handbook, Sage, Thousand Oaks, CA. Smith-Doerr, L., Manev, I.M. and Rizova, P. (2004) 'The meaning of success: network position and the social construction of project outcomes in an R&D lab', Journal of Engineering &Technology Management, Vol. 21, pp.51-81. Young, S.A. and Parker, C.P. (1999) 'Predicting collective climates: assessing the role of shared work values, needs, employee', Journal of Organizational Behavior, Vol. 20, pp.11991218. Zeggelink, E. (1995) 'Evolving friendship networks: an individual-oriented approach implementing similarity', Social Networks, Vol. 17, pp.83-110. Zuckerman, H.H. (1967) 'Nobel laureates in science: patterns of productivity, collaboration, and authorship', American Sociological Review, Vol. 32, pp.391-403.
264
Appendix: 36 key attributes in the work environment survey (4 categories)
Development of human resources People treated with respect Optimal mix of staff Creativity and crossfertilisation Time to think and explore Resources/freedom to pursue new ideas Autonomy to make decisions Cross-fertilisation of ideas Frequent external collaborations Relevant research portfolio Commitment to critical thinking Identification of new opportunities Sense of challenge and enthusiasm Internal support systems Good research competencies Good equipment/physic al environment Good salaries and benefits Good allocation of internal funds Informed and decisive management Rewards and recognises merit Efficient laboratory systems Laboratory services meet needs Overhead rates not burdensome Set and achieve relevant goals Sufficient, stable project funding Good planning and execution of projects Good project-level measures of success Good relationship with sponsors Reputation for excellence
Management integrity Teamwork and collaboration Good internal project communication Management adds value to work High-quality technical staff Good professional development Good career advancement opportunities
Management champions foundational research Good lab-wide measures of success Clear research vision and strategy Invests in future capabilities
doc_365867054.docx