Entrepreneur: Start & Grow Your Business

Internal and External Communication, Boundary Spanning, and Innovation Adoption: An Over-Time Comparison of Three Explanations of Internal and External Innovation Communication in a New Organizational Form.

This research report compares three differing explanations of the dynamic interrelationships between internal and external innovation-related communication in a new organizational form. In the functional specialization explanation, individuals are said to focus on the mix of internal and/or external communication dictated by their formal positions. The communication stars explanation suggests that individuals maintain similar levels of communication in both networks. The cyclical model posits a more dynamic pattern that shifts back and forth between internal and external communication, depending on the consequences of their prior communication behavior. The new organizational form examined for three years was the Cancer Information Service, a geographically dispersed federal government health information program. Our results indicated that there was a lagged effect for the communication stars explanation.

Keywords: Boundary Spanning, Communication Stars, Health information, Innovation, Networks

External communication links, which are often associated with boundary spanning, are critical to enhancing innovations since they provide opportunities for learning and for securing needed resources (Goes & Park, 1997) and for the diffusion of ideas between and within organizations (Cziepel, 1975; Daft, 1978; Ghosal & Bartlett, 1987: Kimberly, 1978; Robertson & Wind, 1983). Such links are the mechanism that operationalizes environmental cues to the internal organizational structure (Corwin, 1972; Lozada & Calantone, 1996; Spekman, 1979). The present study examines external communication longitudinally in the more voluntary communication environment of a new organizational form. Most of the prior literature on innovaton-related communication has emphasized the constraints posed by a person's formal position; however, more recently it has been suggested that new, emerging designs provide opportunities for individuals to shape their own innovation related communication patterns.

We will directly contrast explanations derived from formally prescribed (functional specialization) and emergent (communication stars) theoretical positions (Monge & Eisenberg, 1987; Johnson, 1993). In the functional specialization model, individuals are predicted to focus on either internal or external communication depending on their formal functional positions. The communication stars explanation argues that individuals are disposed to the same levels of communication in both internal and external networks. Yet a third model offers a cyclical explanation, positing that individuals rotate their internal and external communication in a dynamic pattern depending on organizational requirements.

Most of the current organizational literature tends to favor virtual designs. Virtual designs are based on market assumptions (Galbraith, 1995) and place an increasing burden on individuals to find their way amidst chaos (Miles, Snow, Matthews, Miles, & Coleman, 1997). However, other studies have suggested that the increasing complexity of these forms needs to be balanced by a concomitant interest in formalization (Johnson, LaFrance, Meyer, Speyer, & Cox, 1998; Johnson, Meyer, Berkowitz, Ethington, & Miller, 1997), which reduces uncertainties arising in these new forms. A host of environmental factors contribute to the development of new organizational forms: concerns about personnel costs (e.g., pensions, health costs); external pressures to keep the number of members on their permanent staff low; uncertainty reduction; needs to pool knowledge and information or to create it in the case of research and development (R&D) firms (Gibson & Rogers, 1994); increasing access to information by reducing institutiona l barriers (DeBresson & Amesse, 1991); affiliation (e.g., with a more credible national organization); and building mutually supportive power bases to lobby various stakeholders. Fundamentally, consortiums are formed so that their members can accomplish more than they could do on their own. Increasingly, organizations find that they are either strapped for resources or are pursuing such large projects that they must pool their resources to pursue innovations (Browning, Beyer, & Shelter, 1995; DeBresson & Amesse, 1991; Hakansson & Sharma, 1996).

Creating new organizational forms is difficult, particularly in health care (Arnold & Hink, 1968; Farace, Monge, Bettinghaus, Eisenberg, White, Kurchner-Hawkins, & Williams, 1982; Kaluzny, Lacey, Warnecke, Hynes, Morrissey, Ford, & Sondik, 1993; Kaluzny & Warnecke, 1996; Luke, Begun, & Pointer, 1989). At least five barriers have been identified: (a) Cooperating agencies often have different missions (e.g., providing social support vs. treatment for cancer patients); (b) outcome and effectiveness measures differ among agencies; (c) the coordination costs are too heavy (DeBresson & Amesse, 1991) to truly integrate the efforts of diverse organizations (Arnold & Hink, 1968); (d) members of coalitions may have multiple goals (Stevenson, Pearce, & Porter, 1985), and may resent the loss of decision-making latitude; and (e) the cost of managing their linkages increases (Oliver, 1990). In the face of these obstacles, there is an increasing need to develop new theories and fresh perspectives based on empirical data ab out the operation of these new organizational forms (Luke et al., 1989). Indeed, the ability of a society to create new organizational forms may directly affect its ability to adapt to new environmental circumstances (Romanelli, 1991), such as an increasingly competitive global environment.

The success of new organizational forms depends on managing interorganizational relationships through external communication. Consequently, more empirical study is necessary to determine the ideal balance between formalized structure and emergent communication networks in these new organizational forms. Specifically, research must give more attention to individual patterns of internal and external communication. Research is particularly important now given the increased attention paid to new organizational forms and an increased recognition that within these new forms there are not clear boundaries, but rather gradations of affiliation between entities (Sheppard & Tuchinsky, 1996).

Three Explanations of the Interrelationships Between Internal and External Communication

Figure 1 contains a classic panel representation of the interrelationships between internal and external communication over three points in time (see Finkel, 1995; Williams & Podsakoff, 1989). In this research report we will examine possible configurations of relationships, or paths in a classic path analytic sense, for three contrasting models: functional specialization, communication stars, and cyclical. Each explanation predicts different key paths in Figure 1, where a and b paths represent the stability of internal or external communication at consecutive or lagged periods respectively, a paths represent the cross-sectional interrelationships between internal and external communication at any one time, d paths specify relationships between internal and external communication across consecutive points in time, and e paths represent the lagged interrelationships between internal and external innovation communication from Time 1 to 3.

Functional Specialization

Since organizations must adapt to their environments, a number of formal structures and associated functional roles are created explicitly to deal with them (Galbraith, 1974). For example, boundary spanners (e.g., department heads, customer service representatives) maintain external communication contacts because of their formally assigned roles At-Twaijri & Montanari, 1987; Burk, 1994; Friedman & Podolny, 1992; Grover, Jeong, Kettinger, & Lee, 1993; Keller, Szilagyi, & Holland, 1976; Lysonski & Johnson, 1983; Schwab, Ungson, & Brown, 1985; Singh, Goolsby, & Rhoads, 1994; Stevenson, 1990). Boundary spanners are responsible for making communication contacts with external information sources and supplying their colleagues with information concerning the outside environment, all while maintaining an organization's autonomy (Adams, 1976; Aldrich & Herker, 1977).

Boundary spanners play an important role in the diffusion of ideas between and within organizations (Albrecht & Ropp, 1984; Cziepel, 1975; Daft, 1978; Ghosal & Bartlett, 1987; Schwab, Ungson, & Brown, 1985). Nowhere is this more true than in the health care environment (Robertson & Wind, 1983). Boundary spanners are the mechanism that operationalizes environmental cues to the internal organizational structure (Jemison, 1984; Lozada & Calantine, 1996; Spekman, 1979). A substantial proportion of the boundary spanning literature has implicitly adopted a two-step communication process (e.g., Katz & Lazarsfeld, 1955), with an emphasis on information flowing through boundary spanners who act as opinion leaders in their organizations. However, in new organizational forms most individuals engage in some boundary spanning behavior, rendering more traditional organizational boundaries increasingly arbitrary (Starbuck, 1976). While the literature suggests various types of boundary spanning communication activities, few studies simultaneously have examined internal (between organizational units) and external (with other organizations) communication patterns over time, especially in relation to innovation processes (Goes & Park, 1997). Inherent in functional specialization is that individuals will concentrate on either internal or external communication, depending on their formal position. A more structured way to represent the functional specialization approach is as follows:

H1: A functional specialization model posits

A: Positive relationships between internal communication at all three times (the a and b paths in Figure 1).

B. Positive relationships between external communication at all three times (the a and b paths in Figure 1).

C. Negative relationships between internal and external communication at each time (c paths) and in their cross-lagged relationships (d and e paths).

Communication Stars

Disputing the functional specialization explanation, some boundary spanning literature suggests that the two distinctive external and internal communication roles can be played by the same person (Aldrich & Herker, 1978; Allen, 1989; Friedman & Podolny, 1992; Katz & Tushman, 1981; Tushman & Scanlan, 1981a, 1981b). Consequently, research has also focused on boundary spanners who communicate externally as well as internally. For example, Nagpaul and Pruthi (1979) reported that technical R&D gatekeepers used external contacts for idea-generation and internal networks for problem solving. Similar findings are reported in Tushman and Scanlan's (1981b) investigation in a high-tech R&D facility. These researchers found that boundary spanners were likely to be identified as a valuable internal source of new information. Thus, when a person's communication spans both internal and external networks, it would appear that the two networks are mutually reinforcing. Some management research also recognizes boundary spanni ng activities both inside and outside the organization (Mintzberg, 1973). For instance, middle management sometimes requires individuals to be both internal and external stars. Recent research has suggested that balancing internal and external ties establishes individual influence (Manev & Stevenson, 1996).

While it would seem obvious that there are finite limits to the amount of communication in which one can engage (Baker, 1992), several studies suggest that individuals who are high communicators in one setting are also high in others. That is, heavy users of one information medium related to work are likely to be users of other media that also carry this same information (Blau & Alba, 1982; Caroll & Teo, 1996; Paisley, 1980; Weedman, 1992), which is also a finding of more general media use studies (Berelson & Steiner, 1964).

Based on the preceding discussion, it seems reasonable to suggest that boundary spanners focus on both internal and external activities simultaneously. Moreover, boundary spanners acquire relevant information from their extensive external contacts and filter and feed the information into the organization. Boundary spanners also often seek out information from each other, and peers tend to rate them as influential (Paisley, 1980; Reynolds & Johnson, 1982). A more structured way to represent the communication stars approach is our second hypothesis:

H2: A communication star model posits

A: Positive relationships between internal communication at all three points in time (the a and b paths in Figure 1).

B. Positive relationships between external communication at all three times (a and b).

C. Positive relationships between internal and external communication at each point in time (c paths) and in their cross-lagged relationships (d and e paths).

Cyclical Model

A melding of both of the functional specialization and communication stars literatures suggests a third model which accounts for the interrelationships between internal and external communication patterns. This alternative model suggests that communication relationships may shift due to the systemic consequences resulting from the boundary spanning activities and dynamic organizational requirements. For instance, to avoid role conflict, boundary spanners might focus their efforts in one network (internal or external). As the R&D literature suggests, importing external ideas might result in considerable internal communication generating internal innovations, which, in turn, are then exported to other organizations though external communication. A more structured way to represent the cyclical approach is our third hypothesis:

H3: A cyclical model posits

A. Negative relationships between internal communication at consecutive time points (the a paths in Figure 1).

B. Positive relationships between internal communication at Time 1 and Time 3 (the b paths).

C. Negative relationships between external communication at consecutive time points (the a paths).

D. Positive relationships between external communication at Time 1 and Time 3 (the b paths).

E. Negative relationships between internal and external communication at each point in time (c paths) and in their lagged relationships (e paths).

F. Positive relationships between internal and external communication at consecutive points in time (d paths)

Methods

The present study evaluated the three competing boundary spanning explanations in a unique setting, focusing on innovation-related communication in a new organizational form. In this section we describe the unique features of the organization we studied (Cancer Information Service). We then focus on the approach adopted to define the boundaries of the organization, the participants in the research studies, and the specific questions asked of the study participants.

The Cancer Information Service as a New Organizational Form

This study examines a confederation of contractors who provided services to the Cancer Information Service (CIS). The CIS was established in 1975 by the National Cancer Institute (NCI) to disseminate accurate, up-to-date information about cancer to cancer patients, their relatives and friends, health care professionals, and the general public (Morra, Bettinghaus, & Marcus, 1993; Morra, Van Nevel, Nealon, Mazan, & Thomsen, 1993). In response to this mandate, the CIS currently maintains a network of 19 Regional Offices (ROs) that are typically linked to NCI-funded regional cancer centers. The activities of the CIS network are coordinated and supervised by the Office of Cancer Communications (OCC) at the NCI. These activities fall into two broad categories: (a) responding to requests for information over the telephone (the CIS operates a toll-free telephone number in which callers are automatically triaged to their regional office for response from a trained and certified Cancer Information Specialist), and (b) conducting community outreach activities. The outreach program of the CIS serves as a catalyst and focal point for cancer education at the state and regional level and is the focus of much of the external communication examined here.

The ROs are brought together by a classic fee-for-services contract. In effect, the contract hires temporary organizations for five years to conduct a specified scope of work for the NOI. The unique characteristics of the CIS become apparent when contrasted with more conventional organizational forms. For instance, although the ROs are formally members of other organizations, the agency itself has many of the characteristics of unitary organizations, such as centrally determined goals, a formal bureaucratic structure of authority, a division of labor, formal plans for coordination (e.g., sharing calls), a high normative commitment to providing service to callers, and targeted outreach activities to priority audiences. Performance standards are set nationally and are monitored by an extensive formal evaluation effort (Kessler, Fintro, Muha, Wun, Annett, & Mazen, 1993). However, important personnel issues such as salaries and fringe benefits are determined at the RO level. Thus, regions internally boundary spa n with each other and OCC, while maintaining external communication contacts with other organizations in their communities. External communication primarily developed community-based initiatives (e.g., breast cancer awareness month), while internal communication focused on implementing new techniques (e.g., making outcalls to encourage mammography screening) for reaching the public.

Boundaries of the CIS

For the purpose of this research project, the composition of the CIS internal network was based on nominalist views of network boundaries (Lauman, Marsden, & Prensky, 1983). From a nominalist perspective, five major functional roles, representing key decision-makers within the CIS, were examined: (a) Office of Cancer Communications (OCCs) staff at NCI and (b) Project Directors (PDs), (c) Outreach Managers (OMs), (d) Telephone Service Managers (TSMs), and (e) Principal Investigators (PIs) at ROs. OCCs are in charge of coordinating and supervising the activities of the regional CIS network. PDs engage in a mixture of internal and external communication coordinating work with OCCs, other ROs, or their local cancer centers. TSMs primarily focus on the internal telephone communication and referral services. OMs are active in the external network since they are responsible for developing relationships with community organizations. PIs are the principal investigators of the CIS contract. Thus, we examined formal ro les with a mix of internal and external communication responsibilities.

Sampling Interval

This study was part of a larger project designed to evaluate the impact of three planned innovations over four years (see Johnson, Berkowitz, Ethington, & Meyer, 1994a; Johnson, Bettinghaus, Woodworth, Fleisher, Ward, & Meyer, 1997; Meyer, Johnson, & Ethington, 1997). Consequently, selection of the sampling interval was particularly critical given the complexity of the overall investigation. As a result of extensive pretesting and discussions with members of the network about internal communication reports, we decided to focus on a three-day period every three months. In addition, data were collected by rotating days of the week and weeks of the month throughout the duration of the project. Internal communication network data were regularly collected at each of 14 scheduled sampling periods (see Johnson et al., 1994a). Unlike the internal communication data gathered at relatively frequent intervals, external communication data (i.e., radial communication network data representing individual reports of their linkages to other organizations) were collected once a year for three years. The sampling was frequent enough to detect major cycles of activities within the CIS system, while recognizing limitations of respondent memory and the vast volumes of data that might be generated.

Data Collection

Communication data on external contacts were collected at three points of time: T1, T2, and T3 (c.f., Johnson, Chang, Ethington, Meyer, & LaFrance, 1994; Johnson, Chang, La France, et al., 1996a, b). [1] At each data collection period, a package was sent to respondents containing a communication log and a battery of questions relating to their external communication contacts. To help ensure completion, the self-report questionnaires were mailed to the respondents approximately 10 days before the sample time period. A personalized letter explained the issues that would be examined and urged participation in the project. In addition, participants also received an e-mail to notify them that they would soon receive the questionnaires. A second e-mail was sent the day before the sample time period, reminding participants that they should begin recording their communication contacts for the next three days. A third e-mail was sent the day after the sample time period concluded, reminding participants to return the ir questionnaires in the provided stamped, self-addressed envelope. In fact, many follow-up steps (e.g., letters, faxes, emails) recommended by the literature (e.g., Dillman, 1978, 1991) were used in these recurring data collections. Perhaps because of the extensive follow-up efforts, we achieved a satisfactory response rate of 93 percent, 93 percent, and 95 percent at T1, T2, and T3, respectively.

Respondents and Level of Analysis

The sample sizes were 110, 103, 121 at T1, T2, and T3, respectively, with the core cohort remaining essentially stable. Study participants were highly educated: 92 percent of the respondents had earned college degrees, 51 percent of which were graduate degrees. Interestingly, fewer than one-third of respondents had worked for the CIS for five years or more.

Many prior studies have used organizations as the unit of analysis, distinguishing between internal and external communication on the level of intra-organization and inter-organization respectively (Allen, 1989; Zoch, 1993). Yet, others have used alternative levels of analysis. For example, Tushman and Scanlan (1981a) chose the department as the unit of analysis in an R&D setting. These researchers defined internal communication as the communication occurring within the department, while external communication reflected activities taking place on an intra-organizational and extra-organizational level. As a result, external activities in one study may be treated as internal communication in others, and internal communication stars on an intra-organizational level may be defined as external communication stars on a group level.

CIS is the unit of analysis in the present study. Thus, internal communication refers to the communication occurring among and between the 19 ROs and OCC, while external communication denotes the communication contacts that occurred with the organizations outside the CIS network (e.g., American Cancer Society, Health Department, and so on). This study focused on the boundary spanning communication of all individuals, given the assumption that communication is widely distributed in new organizational forms.

Internal Communication

Respondents were asked to record the interpersonal communication contacts they initiated or received within CIS network for a three-day period. [2] They were instructed to record the inter-regional communication on the national level. [3] For the respondents' convenience, a directory of individuals within the CIS network and pre-dated pages of the log were provided. Respondents were asked to record their innovation-related communication concerning intervention strategies within the network. These contacts included initiatives that related to the development or implementation of programs which focused on reaching various target populations, such as counseling protocols for special target populations, targeted outreach activities using the telephone, and responses to calls associated with communication campaigns (see Meyer et al., 1997 for more detail). [4]

External Communication

Respondents were asked to record or estimate the number of times they initiated or received contacts with a member representing outside organizations about intervention strategies. Respondents were also instructed to count each individual only once. For respondents' convenience and for reliability purposes, participants were also provided with an operational definition of intervention strategies. The list of the outside organizations was developed after considerable collaboration with the CIS staff, and the list was finalized after several pretests within the CIS network (see Johnson et al., 1994a). Separate questionnaires were developed for OCCs and other functional groups because of their job requirements.

Results

In this section of the report we first provide descriptive data, followed by a description of the analysis plan. We then compare each of the three models used to study the CIS.

Descriptive Statistics

The means, standard deviations, ranges, and censored correlations for the variables are presented in Table 1. For external communication, approximately one-sixth of respondents reported zero contacts across the three time points (n = 22, 21, & 27, respectively). Over one half of the respondents reported having 0 contacts for internal communication at T1 and T2 (n = 71 and 64 respectively) and even more reported no contact at T3 (n=98). These data led us to adopt a censored variable approach, normalizing the variables at the lower ranges of observed values. The data were subjected to LISREL analysis by means of the PRELIS computer software (see SPSS, 1993).

In general, respondents had considerably more external than internal contacts. Comparable numbers of communication contacts were observed for internal communication at T1 and T2 (mean = .65 and .68, respectively), and for external communication at T2 and T3 (mean = 10.97 and 10.82, respectively). Standard deviations for both internal and external communication across all three time points were relatively high (ranging from .60 to 1.17 and from 13.45 to 14.68, respectively). Extreme outliers (those with scores above 80 for external communication) were removed before censored correlation coefficients were calculated for the model. The highest correlation was observed between external contacts at T2 and T3 r = .45), while the lowest correlation was observed between internal communication and external communication at T3 r = .02). There were no obvious problems with multicollinearity present in the censored correlation matrix.

Analysis Plan

Analysis of panel data has been a source of controversy because of the intractable difficulties that a researcher often confronts, such as an inability to account for random measurement error, correlated disturbances attributable to unspecified third variables, and identification problems (see Williams & Podsakoff, 1989, for an exhaustive discussion). In this research we used an iterative approach for estimating an optimal model from a nested series of models for evaluating structural parameter estimates by means of LISREL. Sometimes referred to as path analysis, LISREL is a general analytic technique for estimating a linear structural equation system. One of the unique advantages of LISREL is that it provides estimates of fit for the entire model to the data. Consequently, LISREL ameliorates the problems inherent in a classic approach to analysis of panel models (Finkel, 1995; Williams & Podsakoff, 1989). The approach also permits the analysis of residuals, reciprocal relationships, and the equality of caus al parameters over-time (Finkel, 1995; Williams & Podsakoff, 1989).

To determine superior parameter estimates a series of nested models using the Normed Fit Index (NFI), which is (Fn - Fm)/Fn, where Fn = [[chi].sup.2]/n for the null model and Fm = [[chi].sup.2]/n for the model under examination, were evaluated. NFI can range from 0 to 1.00, is independent of sample size, and reflects the goodness of fit of competing models (Bentler & Bonett, 1980; Williams & Podsakoff, 1989). Model comparisons can be used to assess the adequacy of causal lags and the relative importance of different variable groups (Williams & Podsakoff, 1989). The question of fit is becoming increasingly complicated (see Bollen & Long, 1993). We therefore also report the Akaike Information Criterion (AIC) because it differs from the NFI in several important ways: It does not range from 01, it does not apply just to nested models, and it favors simple models (Tanaka, 1993). The AIC was the test statistic, (chi-square), + 2 times the number of free or estimated parameters.

Following the iterative procedure outlined by Finkel (1995), we tested a number of models. In the null model (see Table 2) none of the parameters were estimated, except that the on diagonal elements of psi were constrained to equal the same value of one (consult Anderson & Gerbing, 1988). This model became the baseline to which subsequent models were nested. The model yielded a chi-square value of 65.01 with 20 df. In model B through H, differing combinations of substantive (beta) paths, lagged paths over two points in time, psi, or covariances of endogenous variables, and theta epsilon, or measurement error, paths were estimated and/or constrained to equal each other. Given roughly equal time lags, similar paths during different time intervals should be equal. Hence, the rough equivalence of these paths would be one indication of the appropriateness of these causal intervals (Finkel, 1995). In model G, for example, paths between T1 and T3 were estimated, the mirror gamma paths and psi correlations at T2 and T3 were constrained to equal each other (e.g., between internal communication at T1 and external communication at T2 and between internal communication at T2 and external communication at T3), and the diagonal elements of psi were constrained to equal each other. Model G had a chi-square of 5.99 with 4 df, a NFI of .91, and an AIC of 39.99.

Models A through H exhibited a narrow range of scores on both the AIC and NFI. Model H, with a score of .91, had the superior NFI, whereas Model E yielded the best AIC score (28.20). Accordingly, we looked at other criteria to determine which model provided the optimal baseline on which to make comparisons of the three explanations. We chose Model D because it was the only one of the models that did not generate a warning that the measurement error matrix theta epsilon was positive definite. Nor did Model D have problems with q plots or the correlation of estimates. Model D was also in the middle of the range of competing models for both the NFI and AIC. Moreover, Model D's similar substantive parameters were constrained to equal each other in value, diagonal elements of psi were constrained, and theta epsilon was free. Model D had an excellent fit to the data with a chi-square value of 10.26, with 8 df, a probability level of .25, goodness of fit index of .96, an adjusted goodness of fit index of .90, a roo t mean square residual of .08, and a coefficient of determination of .998 for the y variables. The estimates of errors of measurement were all nonsignificant for the single item indicants, ranging from .22 for T1 internal to .48 for T3 internal communication. The diagonal elements of psi were all constrained to .51, a significant value, indicating a moderately high level of explained variance. The normalized residuals fell within acceptable ranges. The standard errors for the parameters were generally low to moderate. No obvious problems with identification and/or estimation were present in the results, and there was also no indication of colinearity in the estimates.

Comparing the Three Models

Figure 2 contains the parameter estimates for the substantive (beta) paths used to compare the models. The a paths were significant and high in value for both internal and external communication, suggesting a fair amount of stability in these variables over consecutive time points. The b paths were low and nonsignificant when estimated in Model E, indicating a mediating impact for intervening time points, which also provided some support for our measurement procedures. The c paths indicated little contemporaneous association between internal and external communication. The d paths were low to moderate in value, with the ones from internal communication to external communication significant. The e paths between internal and external indicated little lagged effect over two distant time periods.

In sum, the results for internal and external communication at consecutive time periods provided partial support for both H1A and B and H2A and B; that i